Walden University

ScholarWorks
Harold L. Hodgkinson Award for Outstanding
University Awards
Dissertation

2012

A Multiple Case Study Analysis of the Positive

Deviance Approach in Community Health
Piroska Bisits Bullen

Follow this and additional works at: http://scholarworks.waldenu.edu/hodgkinson

This Dissertation is brought to you for free and open access by the University Awards at ScholarWorks. It has been accepted for inclusion in Harold L.
Hodgkinson Award for Outstanding Dissertation by an authorized administrator of ScholarWorks. For more information, please contact
[email protected].
 Walden University

COLLEGE OF HEALTH SCIENCES

This is to certify that the doctoral dissertation by

Piroska Bisits Bullen

has been found to be complete and satisfactory in all respects,

and that any and all revisions required by
the review committee have been made.

Review Committee
Dr. Michael Schwab, Committee Chairperson, Public Health Faculty
Dr. Jacqueline Fraser, Committee Member, Public Health Faculty
Dr. Marydee Spillett, University Reviewer, Public Health Faculty

Chief Academic Officer

Eric Riedel, Ph.D.

Walden University
2012
 Abstract

A Multiple Case Study Analysis of the Positive Deviance Approach

in Community Health

by

Piroska Bisits Bullen

B.Sc. (Hons), University of New South Wales, 2005

Dissertation Submitted in Fulfillment

of the Requirements for the Degree of

Doctor of Philosophy

Public Health

Walden University

April 2012
 Abstract

The positive deviance (PD) approach involves finding individuals who have solved a

problem and spreads their unique solutions to others. While there have been calls for PD

to become a standard tool in community health, there has been little research on the

approach. This study investigated how PD is used in practice and evidence of its

effectiveness by analyzing case studies of 40 PD programs and 32 PD inquiries

implemented in a range of high, middle, and low income countries by both national and

international organizations. Case studies were developed using data from publicly

available documents. Qualitative within-case and cross-case analyses were used to

identify common themes and trends using the theory of diffusion of innovations. Results

show that the first large scale applications of the PD approach were in child malnutrition

in the 1990s. Since then the approach has been applied to other issues in individual

behavior change (e.g., HIV/AIDS), organizational change (e.g., health services), and

sociocultural change (e.g., female genital mutilation). Current PD approaches can be

classified by the level of intervention, and the methods used to identify positive deviants,

discover their behaviors, and spread the behaviors to others. Most programs do not fully

involve the community at all stages. While there is substantial evidence for the

effectiveness of the PD approach in child malnutrition, few high quality outcome

evaluations have been conducted in other areas. Implications for positive social change

include providing data to encourage practitioners to use the PD approach as a standard

tool for child malnutrition, where it has the potential to improve nutritional status and

thus contribute to long term outcomes in child health, education and social development.
A Multiple Case Study Analysis of the Positive Deviance Approach

in Community Health

by

Piroska Bisits Bullen

B.Sc. (Hons), University of New South Wales, 2005

Dissertation Submitted in Fulfillment

of the Requirements for the Degree of

Doctor of Philosophy

Public Health

Walden University

April 2012
 UMI Number: 3503498

All rights reserved

INFORMATION TO ALL USERS

The quality of this reproduction is dependent on the quality of the copy submitted.

In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.

UMI 3503498
Copyright 2012 by ProQuest LLC.
All rights reserved. This edition of the work is protected against
unauthorized copying under Title 17, United States Code.

ProQuest LLC.
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, MI 48106 - 1346
 Table of Contents

List of Tables .......................................................................................................................v

List of Figures ................................................................................................................... vii

Chapter 1: Introduction to the Study....................................................................................1

Background of the Study ...............................................................................................1

Variation in PD approaches .....................................................................................6

Related Approaches and Theories in Community Health........................................7

Problem Statement .........................................................................................................8

Purpose of the Study ......................................................................................................8

Nature of the Study ........................................................................................................9

Research Questions ......................................................................................................10

Theoretical Base...........................................................................................................11

Definition of Terms......................................................................................................12

Assumptions.................................................................................................................13

Limitations ...................................................................................................................13

Delimitations ................................................................................................................13

Significance of the Study .............................................................................................14

Summary ......................................................................................................................14

Chapter 2: Literature Review .............................................................................................16

Introduction ..................................................................................................................16

Research Strategy.........................................................................................................16

Positive Deviance.........................................................................................................17

History of Positive Deviance .................................................................................17

i
 Benefits of the Positive Deviance Approach .........................................................22

Limitations of the Positive Deviance Approach ....................................................25

Positive Deviance Methodology ............................................................................27

Related Approaches and Theories in Community Health............................................31

Participatory Approaches .......................................................................................32

Assets Based Approaches ......................................................................................34

Theory of Diffusion of Innovations .............................................................................37

Qualitative Case Study Research .................................................................................39

Alternative Research Methods .....................................................................................41

Chapter 3: Research Method ..............................................................................................43

Introduction ..................................................................................................................43

Research Design and Approach ...................................................................................43

Role of the Researcher .................................................................................................45

Research Questions ......................................................................................................45

Setting and Sample ......................................................................................................46

Inclusion Criteria ...................................................................................................47

Sampling Methods .................................................................................................47

Data Collection and Analysis.......................................................................................49

Data Collection ......................................................................................................49

Validity and Reliability ..........................................................................................51

Data Analysis .........................................................................................................53

Instrumentation and Materials .....................................................................................55

Protection of Human Participants ................................................................................57

ii
 Dissemination of Findings ...........................................................................................57

Chapter 4: Results ..............................................................................................................59

Sampling and Data Collection .....................................................................................59

Data Management ........................................................................................................64

Findings........................................................................................................................64

Research Question 1 ..............................................................................................64

Research Question 2 ..............................................................................................79

Research Question 3 ..............................................................................................86

Research Question 4 ..............................................................................................99

Evidence of Quality ...................................................................................................110

Chapter 5: Discussion, Conclusions, and Recommendations ..........................................114

Overview ....................................................................................................................114

Current Use of PD Approach in Community Health Practice ...................................115

Local Involvement and Community Participation ...............................................117

PD Steps ...............................................................................................................120

Diffusion of PD Approach Through Community Health Practice.............................122

Possibility of Grouping PD Approaches in Typology ...............................................124

Limitations .................................................................................................................137

Implications for Positive Social Change ....................................................................138

Recommendations for Action ....................................................................................139

Recommendations for Further Study .........................................................................141

Reflection on the Researcher’s Experience ...............................................................144

Conclusion .................................................................................................................146

iii
References ........................................................................................................................147

Appendix A: Document Data Collection Form ...............................................................165

Appendix B: Program / Inquiry Data Collection Form ...................................................166

Appendix C: Databases ....................................................................................................171

Appendix D: Gray Literature Sources .............................................................................172

Appendix E: Quality Assessment Tool For Quantitative Studies ....................................173

Appendix F: Quality Assessment Tool For Qualitative Studies ......................................181

Appendix G: Peer Debriefing Meeting Miuntes ..............................................................187

Appendix H: Search Results ............................................................................................204

Appendix I: Final Sample Of PD Programs And Inquiries .............................................206

Appendix J: Examples Of Completed Data Collection Forms ........................................140

Appendix K: Audit Log ...................................................................................................151

Appendix L: Quality Assessment Log .............................................................................157

Curriculum Vitae .............................................................................................................161

iv
 List of Tables

Table 1. Type of Documents Describing PD Programs and PD Inquiries.........................64

Table 2. PD Programs and PD Inquiries by Topic.............................................................65

Table 3. Number of PD Programs and PD Inquiries by Geographic Region ....................66

Table 4. Examples of Positive Deviant Behaviors Identified ............................................72

Table 5. Comparison of Three PD Programs with Different Levels of Community

Participation ...........................................................................................................76

Table 6. Number of PD Programs and Inquiries for Each Topic and Level of Intervention

................................................................................................................................89

Table 7. Methods Used to Identify Positive Deviants and Their Behaviors by Level of

Intervention for PD Programs ................................................................................94

Table 8. Methods Used to Identify Positive Deviants and Their Behaviors by Level of

Intervention for PD Inquiries .................................................................................94

Table 9. Method Used to Spread PD Behaviors by Level of Intervention ........................97

Table 10. Method Used to Identify, Discover, and Spread PD Behaviors by Level of

Intervention ............................................................................................................98

Table 11. Number and Type of PD Program Evaluations by Topic and Level of

Intervention ..........................................................................................................100

Table 12. Number and Type of PD Program Evaluations by Method Used to Identify,

Discover, and Spread Positive Deviant Behaviors ..............................................101

Table 13. Examples of Positive Outcomes Reported from Quantitative Evaluations of PD

Programs ..............................................................................................................103

Table 14. Number of Evaluations with Complete, Incomplete, and Missing Report

v
 Sections ................................................................................................................107

Table 15. Quantitative and Qualitative Assessment Ratings for Evaluations with Enough

Information to Complete the Quality Assessment Tools .....................................109

vi
 List of Figures

Figure 1. Sampling process ................................................................................................61

Figure 2. Percentage of PD programs and PD inquiries conducted in low, middle, and

high income countries ............................................................................................67

Figure 3. Percentage of PD programs and inquiries conducted in various settings ...........68

Figure 4. Percentage of PD programs that included community participation in each PD

step .........................................................................................................................74

Figure 5. Percentage of PD inquiries that included community participation in each PD

step .........................................................................................................................77

Figure 6. Percentage of PD programs that were integrated with other program activities or

run as stand-alone programs ..................................................................................78

Figure 7.Timeline of PD programs and PD inquiries ........................................................80

Figure 8. Number and topic of PD programs/inquiries starting each year ........................81

Figure 9. Levels of intervention .........................................................................................87

Figure 10. Timeline of PD programs and PD inquiries grouped by level of intervention ....

................................................................................................................................90

Figure 11. PD approaches used in practice ......................................................................134

vii
 1
Chapter 1: Introduction to the Study

Background of the Study

Positive deviance (PD) is the concept that "in every community or organization,

there are a few individuals who have found uncommon practices and behaviors that

enable them to achieve better solutions to problems than their neighbors who face the

same challenges and barriers" (Pascale, Sternin, & Sternin, 2010, p. 206). The PD

approach aims to solve community problems by focusing on positive deviance within the

community, rather than by focusing on the community’s needs. The approach seeks out

“positive deviants” in the community and uses their existing solutions to bring about

sustainable behavioral and social change. (Pascale et al., 2010, p. 206). It is important to

note that in popular culture the word “deviant” usually has negative connotations and is

applied to people, such as criminals and psychopaths, who deviate from the social norms

in a negative way. In the context of this work, however, a positive deviant refers to an

individual who deviates from the social norms in a positive way (Babalola, Ouedraogo, &

Vondrasek, 2006). Thus, the term is intended to have positive connotations.

The concept of PD entered the community health field in the 1970s in the field of

child nutrition. Wishik and Van Der Vynckt (1976) were among the first researchers to

suggest that the PD approach could be used to find solutions to childhood malnutrition by

identifying mothers whose children were well nourished, despite living in poverty. The

aim was to find out what the mothers of these children were doing that allowed their

children to be well nourished, for example, feeding them special foods. Community

health practitioners could then incorporate these behaviors into a nutrition program,

knowing that the behaviors were both affordable and culturally acceptable because they
 2

were already being practiced by members of the community. The term positive deviance

was first used to describe this approach by Marian Zeitlin, as part of her pioneering work

in the 1990s documenting positive deviance in child nutrition (Zeitlin, Ghassemi, &

Mansour, 1990). Since then, the PD approach has been used to find solutions to child

malnutrition in many countries (Bolles, Speraw, Berggren, & Lafontant, 2002; Dearden,

Quan, Do, Marsh, Pachón et al., 2002; Guldan et al., 1993; Lapping, Schroeder, Marsh,

Albalak, & Zahir Jabarkhil, 2002; Levinson, Barney, Bassett, & Schultink, 2007; Sethi,

Kashyap, Aggarwal, Pandey, & Kondal, 2007; Sethi, Kashyap, Seth, & Agarwal, 2003).

The PD approach has also been applied to other issues, such as nurse-patient

communication in Indonesia (Kim, Heerey, & Kols, 2008), smoking cessation in

Australian prisons (Awofeso, Irwin, & Forrest, 2008), and Methicillin-resistant

Staphylococcus aureus (MRSA) prevention in U.S. hospitals (Singhal & Greiner, 2007).

Examples of innovative solutions practiced by positive deviants include mothers feeding

children shrimp in Vietnam to prevent malnutrition (Sternin, 2002), an orderly sealing an

MRSA exposed gown inside a glove to prevent transmission in a hospital (Gertner,

2008), and an HIV-negative sex worker telling clients she “has the disease” if they refuse

to use a condom, which scares them so much they agree to wear it (Positive Deviance

Initiative, 2010d).

The PD approach has been incorporated into PD informed programs that aim to

identify and promote PD behaviors. A small number of these programs have been

evaluated, with some showing very positive results. For example, a PD informed program

in Vietnam evaluated by Marsh et al. (2004) resulted in a 74% reduction in child

malnutrition, that was maintained 3-4 years after the end of the program (Trinh
 3

Mackintosh et al., 2002). This program in Vietnam was the first PD program run by Jerry

Sternin and Monique Sternin, both of whom have been active in promoting the PD

approach and set up the Positive Deviance Initiative at Tufts University. Jerry Sternin

was a visiting scholar at Tufts University for many years prior to his death, and Monique

Sternin is now a Senior Consultant to the Positive Deviance Initiative, which collaborates

with a range of organizations including UNICEF, Peace Corps, USAID, and the World

Bank to promote PD (Positive Deviance Initiative, 2010e). Their recent book, “The

Power of Positive Deviance,” describes several of the PD programs they have worked on

and provides guidelines for practitioners wanting to use their PD approach (Pascale et al.,

2010).

Another area where the PD approach has been applied and evaluated is the

prevention of MRSA transmission in U.S. hospitals (Singhal & Greiner, 2007). The use

of PD in this field was also initiated by Jerry Sternin (Pascale et al., 2010) and has

continued after his death in 2008. In these programs the PD approach has been used to

identify staff members within hospitals who practice behaviors that prevent MRSA

transmission. In one hospital MRSA infections declined 55% when a PD initiative was

implemented that required no extra resources; this rate was compared to a 35% reduction

after using an external quality management system that was unsustainable in the long

term because it required large funding (Singhal & Greiner, 2007).

Although examples of PD programs have been highly successful and it has been

used in practice, there has been little scientific research into the PD approach itself. For

example, there has been no study into the wide variation of PD approaches currently in

use, the characteristics of programs using the PD approach, or the effectiveness of the PD
 4

approach in a range of different contexts.

The PD approach is an “assets based” approach, similar to appreciative inquiry

and assets based community development (ABCD; ACCESS, 2009), but more specific in

relating community resources or ideas to a predefined problem. Proponents of the PD

approach consider it more effective than typical “needs based” approaches for difficult,

ingrained problems because it identifies existing solutions from within the community,

which means that the solutions are more likely to be affordable, acceptable, and

sustainable in the long term (Marsh & Schroeder, 2002; Sternin, 2002; Trinh Mackintosh

et al., 2002). These full or partial solutions can also be implemented immediately,

without waiting for all underlying determinants of the problem to be addressed (Sternin,

2002). For example, in the case of the sex worker confronted by a client who refuses to

wear a condom, there are many underlying determinants that contribute to the problem,

including issues of gender equity, lack of alternative employment for women, and poor

regulation of the sex industry. Addressing these underlying determinants will take

decades of advocacy and long term programming to have a real effect on the life of the

sex worker. In the meantime, the sex worker still has a need to protect herself. To do this

she can use the positive deviant behavior described previously to tell her clients she “has

the disease” if they refuse to use a condom. This tactic does not require the underlying

determinants of the problem to be addressed, and can be implemented immediately.

Anecdotal evidence from studies that use the PD approach also suggests that

using a PD approach can lead to intangible benefits, such as community mobilization

(Marsh et al, 2004; Save the Children, n.d.; Singhal & Greiner, 2007), reduced aid

dependency (Milton & Ochieng, 2007; Schooley & Morales, 2007), community
 5

empowerment (Hendrickson et al., 2002; Schooley & Morales, 2007), and improved

advocacy (Awofeso et al., 2008; Lapping, Marsh et al., 2002; Sternin, 2002), among

others. However, there has only been one study that rigorously evaluated these intangible

benefits (Hendrickson et al., 2002), and more research is needed to fully understand these

reported effects (Marsh et al, 2004; Schooley & Morales, 2007).

While the PD approach appears to have many benefits, it also has limitations. By

definition, solutions found through the PD approach can be implemented immediately

with existing resources (Sternin, 2002). As a result, they do not usually address the

underlying causes of the problem, which requires a long term strategy. They are also

highly context specific and cannot be transferred to other communities or even between

different seasons of the year (Berggren & Wray, 2002). The Positive Deviance Initiative

(2009) defined several criteria that can be used to determine whether a PD approach is

suitable for the problem being addressed. Those criteria include the following: (a) The

problem is not exclusively technical and requires behavioral or/and social change; (b) the

problem is “intractable”—that is, other solutions have not worked; (c) positive deviants

exist—that is, solutions are possible; and, (d) there is leadership commitment to address

the issue or, in other words, PD champions exist. Devane (2009) also identified a range of

situations when the PD approach is not suitable, includes situations in which (a) the

problem is primarily technical and requires minimal behavior change; (b) complex

analysis, data collection, or truly special skills are required to determine proper behaviors

and next steps; or, (c) there is little agreement about what needs to be done and/or there is

a low certainty of a reliable outcome.

Authors have presented different versions of the steps required to implement the
 6

PD approach (Bradley et al., 2009; Devane, 2009; Marsh et al, 2004; Marsh, Sternin et

al., 2002; Positive Deviance Initiative, 2009; Save the Children, n.d.; Schooley &

Morales, 2007; Walker et al., 2007), although most include variations on the “four Ds”

presented by Pascale et al.(2010): (a) Define the problem and desired outcome, (b)

determine common practices, (c) discover uncommon but successful behaviors and

strategies through inquiry and observation, and (d) design an action learning initiative

based on the findings (p. 202). Action learning initiatives typically involve community

members learning the new behaviors by practicing them during group education sessions.

According to Pascale et al (2010), the most critical part of the PD approach is that the

community must own the entire process from start to finish. The PD approach should be

highly participatory in nature, and the experts should remain as facilitators only, allowing

the community to decide on the problem to be solved and to discover and disseminate the

PD behaviors themselves (Pascale et al., 2010). However, not all studies that used a PD

approach have been participatory, so the question remains as to whether the PD approach

is actually participatory by definition.

Variation in PD approaches

A literature review of the PD approach was conducted to develop this proposal.

The review identified several issues that will need to be investigated in more detail as

part of this study. During the literature review, 24 programs that used a PD approach

were identified, as well as four articles that gave instructions for conducting a PD study

or program. The PD approach used varied widely between studies. For example, methods

used to identify PD behaviors included qualitative (Awofeso et al., 2008; Friedman et al.,

2008; Kim et al., 2008;), quantitative (Aruna, Vazir, & Vidyasagar, 2001; Vossenaar et
 7

al., 2009; Walker et al., 2007), mixed methods exploratory (Babalola et al., 2006; Bradley

et al., 2009; The United Nations University, 1990), and mixed methods explanatory

(Ahrari et al, 2002; Dearden, Quan, Do, Marsh, Pachón et al., 2002; Wishik & Van Der

Vynckt, 1976) designs, as defined by Creswell (2009). In addition, many of the studies

did not appear to adhere to the guidelines published by Pascale et al. (2010), such as the

requirement that all steps be fully participatory. There were no articles that attempted to

compare or consolidate the different designs. Only one study attempted to validate the PD

approach by comparing a quantitative case-control study using a sample of 50 families,

with a small qualitative phenomenological PD study using a sample of eight families to

see which method would be more successful in identifying behaviors related to positive

health outcomes (Lapping, Schroeder et al., 2002). The authors concluded that the small-

scale qualitative PD study is an affordable and valid rapid assessment tool, although

using both methods would be optimal if resources allowed.

In addition to the overall methods used, there are a number of other key areas

where studies differ in their PD approach. These areas include how success is defined to

identify positive deviants, what is considered a PD behavior, and whether distinctions are

made between PD behaviors that can be transferred to others and those that are “True But

Useless (TBU)” (Save the Children, n.d., p.4).

Related Approaches and Theories in Community Health

The PD approach shares many similarities with other approaches in community

health, including participatory approaches (Minkler & Wallerstein, 2008), appreciative

inquiry (AI; Cooperrider, Whitney, & Stavros, 2008), assets based community

development (ABCD; Mathie and Cunningham, 2003), and resiliency (Zeller, 1991). A
 8

clearly defined distinction between the PD approach and these other approaches is yet to

be developed. None of the studies identified during the literature review presented any

analysis of the theoretical underpinnings of the approach, or its relationship to other

similar approaches.

Problem Statement

While the PD approach has now become a relatively well established technique in

specific areas of practice such as nutrition and MRSA prevention, there have been calls

for it to be used more widely (Devane, 2009; Lapping, Marsh et al., 2002; Marsh et al,

2004; Marsh & Schroeder, 2002) and for it to become a standard part of the community

health worker’s toolbox (Lapping, Marsh et al. 2002; Marsh et al, 2004;). The PD

approach is still not included as a technique in many health promotion and community

health textbooks, such as those by McKenzie, Neiger, and Thackery (2009), Minkler and

Wallerstein (2008), Issel (2004), and McKenzie, Pinger, and Kotecki (2008), while other

participatory approaches are included.

Before the PD approach can be formalized as a standard part of community health

practice, several gaps in understanding of the approach need to be filled. There have been

many reports of the PD approach being used in community health practice. However,

there has been little scientific research into the wide variation of PD approaches currently

in use, the characteristics of programs using the PD approach, or the effectiveness of the

PD approach in a range of different contexts.

Purpose of the Study

The purpose of the study was to identify, describe, and analyze multiple case

studies of past and present community health programs that use the PD approach. This
 9

analysis provides an overall picture of how, where, and when the PD approach is

currently being used and its diffusion through community health practice. The study also

helps better define the spectrum of PD approaches currently in use and attempts to group

them into a typology of PD approaches. Finally, the study assesses both qualitative and

qualitative evidence for the effectiveness of the PD approach in a range of community

health contexts. As a result of better defining the PD approach and its current use and

potential effectiveness, it is possible to provide evidence for and against its wider use in

mainstream community health practice.

Nature of the Study

This study uses qualitative case study research with a focus on document review

to identify and analyze community health programs that have used the PD approach.

Forty PD program case studies and 32 PD inquiry case studies were developed using 226

publicly available program documents, including program reports, training manuals,

presentations, workshop notes, web pages, evaluation papers, online audiovisual

materials, and published peer reviewed studies. PD programs included all stages of the

PD approach from identifying positive deviants and their behaviors through

implementing a program to spread the behaviors to others. PD inquiries only included the

first stages of identifying positive deviants and their behaviors but did not use this

information to design or implement a program.

Documents were identified by searching a range of academic and gray literature

databases and websites of organizations known to implement and document PD

programs. A predefined search strategy and inclusion criteria were used to identify and

select documents. Documents were included if they describe all or part of a PD program
 10

on any health related topic, or any topic related to known social determinants of health.

A document data collection form (Appendix A) was used to capture the key

information from each document in a standardized format. A single PD program often

had several pieces of documentation associated with it. For example, a training manual,

presentation, evaluation report, and peer reviewed study may all be published about the

same PD program. Therefore, the first step in the analysis involved identifying groups of

documents that related to the same PD program.

Once each unique PD program/PD inquiry and its associated documents had been

identified, within-case analysis (Miles & Huberman, 1994) was used to complete as many

fields as possible in the program / inquiry data collection form (Appendix B). Each

program / inquiry data collection form then represented a single case study of a PD

program. A cross-case analysis (Miles & Huberman, 1994) was conducted in order to

identify common themes and trends across multiple case studies to answer the research

questions.

Research Questions

The following research questions and subquestions guided this study:

1. How is the PD approach currently being used in community health practice?

1.1 What health related topics has the PD approach been used to address?

1.2 What settings has the PD approach been implemented in?

1.3 Which types of organizations are implementing programs that use the PD

approach?

1.4 To what extent do programs based on the PD approach follow the steps

defined by Pascale et al. (2010)?

 11

1.5 To what extent do programs based on the PD approach incorporate

community participation in each step?

1.6 To what extent is the PD approach incorporated into integrated programs

versus stand-alone programs?

2. How has the PD approach diffused through community health practice?

2.1 Has the diffusion followed the principles of theory of diffusion of

innovations?

2.2 How has this diffusion progressed geographically, chronologically, and

between community health issues?

2.3 Can groups of innovators, early adopters, early majority, late majority, and

laggards be identified?

2.4 Are there characteristics of PD that have made it more or less likely to

spread quickly, such as relative advantage, compatibility, complexity,

trialability, and observability?

3. Can the range of PD approaches currently in use be grouped into a typology of

PD approaches?

3.1 Are particular PD approaches more commonly used in specific contexts or

settings?

4. What quantitative and qualitative outcomes have been reported from programs

that use a PD approach, and with what level rigor were the studies conducted?

Theoretical Base

The PD approach is a technique developed by practitioners in the field. It does not

have a specific theoretical underpinning in the literature, and Pascale et al. (2010)
 12

deliberately avoided referring to it as a theory. This study investigated the use of the PD

approach in practice and the evolution of different types of PD approaches. The

conceptual framework used for this analysis was the theory of diffusion of innovations as

proposed by Rogers (1995). The theory of diffusion of innovations describes the spread

of new ideas and innovations throughout a population. It divides the population into

different groups based on how quickly they adopt a new idea; for example, innovators,

early adopters, early majority, late majority, and laggards. It also identifies characteristics

of innovations that make them more likely to spread quickly, such as relative advantage,

compatibility, complexity, trialability, and observability (Rogers, 1995). The spread of

the PD approach throughout community health practice can be seen as the diffusion of an

innovation. By examining the geographic and temporal distribution of PD programs, this

study mapped the diffusion of the PD approach throughout community health practice.

This analysis included identifying innovators and early adopters as well as characteristics

of the PD approach that make it more or less likely to spread further.

Definition of Terms

Positive deviance (PD) approach: An approach to solving community problems

that focuses on positive deviance within the community, rather than focusing on the

community’s needs. The approach uses solutions that already exist in the community to

bring about sustainable behavioral and social change (Pascale et al., 2010).

Positive deviance (PD) concept: The concept that "in every community or

organization, there are a few individuals who have found uncommon practices and

behaviors that enable them to achieve better solutions to problems than their neighbors

who face the same challenges and barriers" (Pascale et al., 2010, p. 206).
 13

Positive deviance Inquiry (PDI): The stage in the PD process where positive

deviants are identified within a community and their successful behaviors and solutions

are documented (Pascale et al., 2010).

Positive deviant: An individual or group who “demonstrates special or uncommon

behaviors and strategies that enable the person or group to overcome a problem without

special resources. However, a person is defined as a PD only in the context of a specific

problem” (Pascale et al., 2010, p. 206).

Positive deviant (PD) behavior: An uncommon behavior practiced by a positive

deviant that allows them to be more successful than their neighbors who have access to

exactly the same resources.

Assumptions

The following assumption was made in this study: There is enough valid and

reliable publicly available documentation on program case studies to provide meaningful

conclusions.

Limitations

This study was limited in two ways. First, the case studies were limited to those

that can be described using publicly available documentation. Second, the case studies

may be subject to publication bias if researchers have not published documents on

programs that failed or had negative results.

Delimitations

Only publicly available documents, including peer reviewed studies and gray

literature documents, were used as sources for analysis. Only programs related to a health

topic or to known social determinants of health were included. Only English language
 14

documents were included.

Significance of the Study

This study addressed a gap in the existing PD literature by helping to better define

the spectrum of PD approaches currently in use, the types of programs currently using the

PD approach, and their effectiveness in a range of contexts. As a result of better defining

the PD approach and potential effectiveness, it is possible to provide evidence for and

against its wider use in mainstream community health practice. This research has direct

implications for practitioners in the field. As the first broad scientific study of the PD

approach itself, it also provides a foundation for future research into the PD approach.

Ultimately this study has the potential to contribute to positive social change by creating

a better understanding of the positive deviance approach, whether or not it should be used

more frequently in community health, and the settings in which it has been shown to be

effective already. This study also contributes to raising awareness of PD in the

community health field, increasing the likelihood that the PD approach will be used to

solve a wider range of community health problems and that it will be incorporated into

community health texts and training programs for the areas where it has been shown to be

successful.

Summary

The PD approach shows promise as a means of solving ingrained community

health problems using existing solutions. However, before PD can become a mainstream

approach in community health, several gaps in our understanding of the approach need to

be addressed. There appears to be wide variation in PD approaches, and the effectiveness

of these approaches in a range of different contexts is unclear. This study used a

 15

qualitative case study and historic approach to help better define the spectrum of PD

approaches currently in use, describe the types of programs that have used the PD

approach, and evaluate their effectiveness in a range of contexts. As a result of better

defining the PD approach and potential outcomes, it is possible to provide evidence for

and against its wider use in mainstream community health practice.

The following chapters describe the study in detail. Chapter 2 gives a full

literature review, Chapter 3 details the methodology, Chapter 4 describes the results,

while Chapter 5 discusses the results and presents recommendations for future action and

research.
 16

Chapter 2: Literature Review

Introduction

This literature review examines the history and past applications of the PD

approach and the need for further research into the approach itself. Following this section

is a brief review of other approaches that share similarities with PD, such as appreciative

inquiry and participatory research. A review of qualitative case study and document

analysis is also included.

Research Strategy

The PD approach has been applied to problems in many different fields.

Therefore, I searched multiple databases in a range of fields simultaneously using the

EBSCO, ProQuest, Ovid, and SAGE interfaces accessed through the Walden University

Library. The final list of databases searched included CINAHL Plus, MEDLINE,

PsycARTICLES, PsycBOOKS, PsycINFO, SocIndex, Mental Measurements Yearbook,

Academic Search Premier, Political Science Complete, Communications & Mass Media

Complete, ProQuest Nursing & Allied Health, ProQuest Dissertations and Theses,

ProQuest Central, ProQuest Health and Medical Complete, ProQuest Psychology

Journals, Science Journals, Social Science Journals, Health Sciences Collection, and

Political Sciences Collection. For the PD component I searched for all articles containing

the term positive deviance. I did not place any restrictions on PD articles in terms of

publication date or location. Articles were restricted to those published in English. I

sourced additional documents on PD from the Positive Deviance Initiative website

(2010e). These included descriptions, evaluation reports, quotes, and interviews from PD
 17
informed programs. I reviewed the references lists for all articles to locate additional

resources.

In addition, I searched the following databases for previous systematic reviews on

PD: Database of Abstracts of Reviews of Effects (DARE), the Cochrane Database of

Systematic Reviews, and the Campbell Library. No existing reviews of PD were

identified.

Positive Deviance

This section describes the history of the PD approach and its benefits and

limitations in practice. The section ends with an overview of the different types of PD

methods found in the literature.

History of Positive Deviance

In popular culture the word deviant usually has negative connotations and is

applied to people, such as criminals and psychopaths, who deviate from the social norms

in a negative way. According to Babalola et al., (2006) the idea that it is possible to have

both positive and negative deviants was first raised by the sociologist Pitirim Sorokin in

the 1950s. The concept of positive deviance has been debated by sociologists ever since,

with camps for (Heckert, 1985, 1998; Spreitzer & Sonenshein, 2004; West, 2004) and

against (Edward, 1985; Goode, 1991) its use as a term.

Although sociologists may debate the use of the term positive deviance, the idea

that individuals can deviate both positively and negatively from the norm makes intuitive

sense. In a normal distribution there will always be a small number of individuals at the

far ends of the spectrum. These individuals deviate from the majority in positive and
 18
negative ways. In other words, positive deviants are those people who experience

“success in spite of hardship” (Hendrickson et al., 2002, p. 84).

The concept of PD first entered the community health arena in the 1970s, initially

in the field of child nutrition. Wishik and Van Der Vynckt (1976) were one of the first

groups to suggest that solutions to childhood malnutrition could be found by looking for

low income families with well-nourished children. They called these families positive

deviants. In their seminal paper, Wishik and Van Der Vynckt (1976) proposed a method

in which a survey of children’s height and weight could be used to identify the small

number of children who were well nourished, despite being born into families with a very

low socioeconomic status. The aim was to find out what the mothers of these children

were doing that allowed their children to be well nourished, while most of the other

children in the community were malnourished. These behaviors could then be

incorporated into a nutrition program, knowing that they were both affordable and

culturally acceptable because they were already practiced by members of the community.

Although they published their proposed methodology, Wishik and Van Der Vynckt

(1976) never published the results of the study.

The term positive deviance was first used to describe this approach by Marian

Zeitlin, as part of her pioneering work in the 1990s documenting positive deviance in

child nutrition (Zeitlin, Ghassemi, & Mansour, 1990). The term positive deviant inquiry

has only come into use only more recently, in papers such as that by Lapping, Schroeder

et al. (2002). Since the concept surfaced in the 1970s, variations of the PD approach

have been used to find indigenous solutions to child malnutrition in a wide range of
 19
countries (Bolles et al., 2002; Dearden, Quan, Do, Marsh, Pachón et al., 2002; Guldan et

al., 1993; Lapping, Schroeder et al., 2002; Levinson et al., 2007; Sethi et al., 2003; Sethi

et al., 2007). The PD approach has also been used to find solutions for nurse-patient

communication in Indonesia (Kim, Heerey, & Kols, 2008), safe sex practices in African

countries (Babalola et al., 2006; Babalola, Awasum, & Quenum-Renaud, 2002), smoking

cessation in Australian prisons (Awofeso et al., 2008), HIV and Hepatitis C prevention

among injection drug users in the U.S. (Friedman, Mateu-Gelabert, Sandoval1, Hagan, &

Des Jarlais, 2008), cancer risk in Guatemala (Vossenaar et al., 2009), healthy eating for

low income pregnant women in the U.S. (Fowles, Hendricks, & Walker, 2005), newborn

care in Pakistan (Marsh, Sternin et al., 2002), pregnancy outcomes in Egypt (Ahrari et al.,

2002), and overall health status in the Netherlands (Mackenbach et al., 1994). The PD

approach has also been used to address nonhealth issues, including public extortion in the

UK, US, India, and China (Horowitz, 2008), gender equity in Brazil (Barker, 2000),

poverty and economic development (Biggs, 2008; Milton & Ochieng, 2007), the

elimination of sweatshops (Arnold & Hartmann, 2005), and increasing sales in for-profit

companies (Pascale et al., 2010).

The PD approach has been incorporated into PD informed programs that aim to

identify and promote PD behaviors. A small number of these programs have been

evaluated, with some showing very positive results. For example, programs using the PD

approach have succeeded in reducing rates of childhood malnutrition (Mustaphi & Dobe,

2005; Trinh Mackintosh et al., 2002), reducing low birth weight (Ahrari et al., 2006),

reducing the incidence of healthcare associated infections in hospitals (Marra et al.,

 20
2010), and improving maternal iron supplementation (Ndiaye, Siekmans, Haddad, &

Receveur, 2009).

The most rigorously studied example of a program using the PD approach is the

work done by Save the Children in Vietnam on child malnutrition (Hendrickson et al.,

2002; Marsh et al., 2002; Marsh et al., 2007; Sripaipan et al., 2002; Trinh Mackintosh et

al., 2002). The initial positive deviance inquiry (PDI) for the program identified a range

of PD behaviors practiced by mothers with well-nourished children. One of the key

behaviors was collecting tiny shrimp from the rice paddies and feeding them to their

children along with the greens from sweet potato tops. Both these foods were freely

available to all members of the community, but most other members of the community

did not believe they were appropriate for children (Sternin, 2002). This finding, along

with others, was incorporated into a nutrition program which resulted in a 74% reduction

in child malnutrition (Marsh et al., 2004) that was maintained 3-4 years after the end of

the program (Trinh Mackintosh et al., 2002). The follow-up evaluation showed that the

PD practices had also been maintained for subsequent children born after the end of the

program. These children were even more likely to be well nourished than their older

siblings who had participated in the program (Trinh Mackintosh et al., 2002). This

program in Vietnam was the first PD program run by Jerry Sternin and Monique Sternin,

the two individuals who appear to have been most active in promoting the PD approach

and who subsequently set up the Positive Deviance Initiative at Tufts University. Their

book, “The Power of Positive Deviance,” describes several of the PD programs they have

worked on and provides guidelines for practitioners wanting to use their PD approach
 21
(Pascale et al., 2010).

Another area where the PD approach has been applied and evaluated is the

prevention of MRSA transmission in U.S. hospitals (Singhal & Greiner, 2007). The use

of PD in this field was also initiated by Jerry Sternin (Pascale et al., 2010) but has since

taken on a life of its own after his death in 2008. In these programs the PD approach has

been used to identify staff members in hospitals who practice behaviors that prevent

MRSA transmission. Gertner (2008) described the example of Jasper Palmer, an orderly

who came up with a new method of removing an MRSA exposed hospital gown and

sealing it inside a glove. This procedure has since been dubbed the Palmer method and

has spread to other staff in the facility. In one hospital MRSA infections declined 55%

when a PD initiative that required no extra resources was implemented; this is compared

to a 35% reduction after using an external quality management system that was

unsustainable in the long term because it required large amounts of funding (Singhal &

Greiner, 2007).

As a result of these studies, the PD approach has now become a relatively well

established technique in nutrition and MRSA prevention. However, despite many calls

for it to be used more often in other areas (Devane, 2009; Lapping, Marsh et al., 2002;

Marsh & Schroeder, 2002; Marsh et al, 2004), and for it to be a standard part of the

community health worker’s toolbox (Lapping, Marsh et al., 2002; Marsh et al, 2004), the

PD approach is still not included as a technique in many health promotion and

community health textbooks, such as those by McKenzie et al (2009), Minkler and

Wallerstein (2008), Issel (2004), and McKenzie, Pinger, and Kotecki (2008). This lack of
 22
attention may in part be due to a lack of scientific research into the approach itself.

Although there are many reports of the PD approach being used in practice, there are no

studies on the wide variation of PD approaches currently in use, the characteristics of

programs using the PD approach, or the effectiveness of the PD approach in a range of

different contexts.

Benefits of the Positive Deviance Approach

The PD approach is an assets based approach, similar to appreciative inquiry and

assets based community development (ABCD; ACCESS, 2009). Assets based approaches

stand in contrast to the “needs based” approach often used in international development

and community health (Sternin, 2002). In a needs based approach, factors in the

community that are causing the problem are identified by bringing in outside solutions,

experts or resources (Sternin, 2002). Even if these outside solutions are culturally

acceptable to the community, experience has shown that they are often not sustainable in

the long term (Lapping, Marsh et al., 2002; Sternin, 2002). A needs assessment will

usually also identify complex social determinants of problems, including poverty,

education, socioeconomic status, discrimination, and lack of infrastructure that can take

many years to address. Conventional approaches to development suggest that all these

underlying determinants must be addressed before a solution can be found (Sternin,

2002).

By comparison, in the PD approach, individuals within the community who have

already found full or partial solutions to the problem using the existing resources are

sought (Positive Deviance Initiative, n.d.; Sternin, 2002). Because these behaviors are
 23
already being practiced by some members of the community they are likely to be

affordable, acceptable and sustainable in the long term (Marsh & Schroeder, 2002;

Sternin, 2002; Trinh Mackintosh et al., 2002). The solutions can also be implemented

immediately, without waiting for all the underlying determinants of the problem to be

addressed (Sternin, 2002).

Many researchers and practitioners have also suggested that the PD approach has

additional intangible benefits for community development – beyond finding practical

solutions to community problems. Anecdotal evidence from programs and studies using

the PD approach has suggested that it can lead to community mobilization and

enthusiasm (Marsh et al, 2004; Save the Children, n.d.; Singhal & Greiner, 2007),

improved social networks (Buscell, 2008; Singhal & Greiner, 2007), reduced aid

dependency (Milton & Ochieng, 2007; Schooley & Morales, 2007), community

empowerment (Hendrickson et al., 2002; Schooley & Morales, 2007), community pride

(Marsh et al, 2004; Marsh, Sternin et al., 2002), group ownership of the problem (Singhal

& Greiner, 2007), enhanced problem solving skills (Marsh et al, 2004), improved

advocacy (Awofeso et al., 2008; Lapping, Marsh et al., 2002; Sternin, 2002), and outside

experts showing increased respect for the community (Sternin, 2002). The following

quotes from PD practitioners and community members illustrate some of the anecdotal

evidence behind these claims:

Quotes from community members involved in programs that use a PD approach:

When people come from outside, it does not feel good. But if we see the things

with our eyes, and try them practically, and see some people practicing them, this
 24
has a good effect on people. (Positive Deviance Initiative, 2010a, para. 3)

PD is like a flashlight. It helps to shine light and illuminate what hides behind the

darkness. It helps us discover what already exists. For example, it helps us

discover our existing strengths we had not realized or utilized. (Positive Deviance

Initiative, 2010b, para. 1)

After a few days of camping in the desert, Abdulkadir the translator, confided to

the PD INGO team what he overheard from a conversation among several of the

elders. They were amazed at the fact that the outsiders had come 3 times to the

water hole to “sit at our feet, and listen to us!” “Not even our own children pay us

that kind of respect,” one man added. (Positive Deviance Initiative, 2010c, para.

3)

Though we did not know, we are very proud of these Afghans who practice good

behaviors most of us are unaware of. (Marsh, Sternin et al., 2002, p. 113).

Quotes from PD practitioners:

Participants have said that they are motivated by learning that they are doing

something right and that a successful solution to their problem already exists

within the community, instead of receiving criticism for their inadequacies.

(Marsh et al, 2004, p. 1177)

…the constant turn outward in search of solutions to national problems has tended

to bury possibilities and to dampen national innovation [the “indifference

trap”]….this paper suggests positive deviance and appreciative enquiry

approaches as organizing frameworks for identifying and amplifying the work of

 25
African innovators, thereby solving the problem of “indifference trap.” (Milton &

Ochieng, 2007, p. 455).

Although anecdotal evidence for these intangible benefits abounds, there has only

been one study that investigated them in detail. Hendrickson et al. (2002) used interviews

to study empowerment among participants in a PD informed nutrition program. Overall

they found a significant improvement in participants’ empowerment compared to the

control group. However, because it was an integrated nutrition program, it was

impossible to separate the effects of the PD component from other parts of the program,

such as health education sessions.

Therefore, due to a lack of research, PD practitioners cannot be certain that the

PD approach does actually lead to the intangible benefits that anecdotal evidence

suggests. It is also unknown which theoretical models of community development might

apply to the PD approach, such as theories of empowerment (Thompson, 2007), capacity

building (Minkler & Wallerstein, 2008), or community organizing (McKenzie, Neiger, &

Thackery, 2009). Because of this, understanding the intangible effects of the PD

approach has been identified as an important area in need of further research (Marsh et al,

2004; Schooley & Morales, 2007).

Limitations of the Positive Deviance Approach

While the PD approach appears to have many benefits, it also has limitations. By

definition, solutions found through the PD approach are those which can be implemented

immediately with existing resources (Sternin, 2002). They do not usually address the

underlying causes of the problem, which require a long term strategy. They are also
 26
highly context specific, and cannot be transferred to other communities, or even between

different seasons of the year (Berggren & Wray, 2002). Each time program planners enter

a new community, the PDI must be repeated. For example, Marsh, Sternin et al. (2002)

showed significant differences in PD behaviors between Afghan refugees and Pakistani

nationals living in the same community. Each group required their own PDI.

As communities change over time, behaviors that were considered positive

initially may eventually be considered negative (The United Nations University, 1990).

For example, not expressing and storing breast milk might be considered a PD behavior

in hot climates where refrigerators are not available, as the chance of contamination is

high. However, if the community develops and most houses acquire a refrigerator and

good hygiene practices then expressing and storing breast milk becomes a positive

behavior because it allows mothers to exclusively breastfeed while still working (The

United Nations University, 1990).

Pascale et al. (2010) acknowledge the PD approach is not panacea solution for all

community health issues. They argue that PD should only be used when no other

solutions have worked, as a last resort. The Positive Deviance Initiative (2009) have

proposed several criteria that can be used to determine whether a PD approach is suitable

for the problem being addressed: (a) The problem requires behavioral or/and social

change rather than a technological solutions; (b) The problem is ongoing and other

potential solutions have failed; (c) Positive deviants exist in the community; and, (d)

community leaders are committed to solving the problem. In addition to these criteria,

Devane (2009) has also identified a range of situations when the PD approach is likely to
 27
be unsuccessful: (a) The problem requires a technical solution and cannot be solved

through behavior change; (b) Complex analysis, data collection, or specialized technical

skills are required to identify a solution; or, (c) There is disagreement about what needs to

be done and/or low certainty that the problem will be solved, even if behavior is changed.

PD is also less likely to be effective in certain cultures, particularly those that are

collective rather than individualistic where celebrating individual achievements is not

considered appropriate (Schooley & Morales, 2007).

Positive Deviance Methodology

Many authors have presented different versions of the steps required to implement

the PD approach (Bradley et al., 2009; Devane, 2009; Marsh et al, 2004; Marsh, Sternin

et al., 2002; Positive Deviance Initiative, 2009; Save the Children, n.d.; Schooley &

Morales, 2007; Walker et al., 2007), although all include variations on the four D’s

presented by Pascale et al. (2010, p.202):

1. Define the problem and desired outcome.

2. Determine common practices and whether positive deviants exist.

3. Discover uncommon but successful behaviors and strategies through

inquiry and observation.

4. Design an initiative based on the findings to spread the positive deviant

behaviors to others in the population.

According to Pascale et al. (2010), the most critical part of the PD approach is

that the community must have full ownership of the process from start to finish. This

means that the “experts” should only participate as facilitators, and should allow the
 28
community to decide on the problem to be solved, and to discover and disseminate the

PD behaviors themselves (Pascale et al., 2010).

The PDI is the second and third step in the development of a PD informed

program (Sternin, 2002). The PDI is a “rapid assessment tool” rather than a survey

(Berggren and Wray, 2002). A rapid assessment tool is a low cost assessment designed to

quickly identify practical solutions that can be implemented immediately. Sternin (2002)

supports this distinction:

The choice of the word ‘inquiry’ in ‘positive deviance inquiry’ is worthy of note.

Rather than call the process a ‘study’, which implies a more rigorous examination

of positive deviant behaviors, including control groups and statistical analysis of

findings, the emphasis is on a quick, practical ‘inquiry’ which enables the

community to act immediately. Hence, scientific rigour is sacrificed in favour of

expeditious programme implementation. (p. 59)

As discussed in the introduction, the literature review identified 24 programs that

used a PD approach, and four articles that gave instructions for conducting a PDI or

program based on the PD approach. There was a wide variation between studies in the

type PD approach used, including qualitative, quantitative, mixed methods exploratory,

and mixed methods explanatory designs. In addition, many of the studies did not follow

the guidelines published by Pascale et al. (2010). For example, many did not have fully

participatory designs, and so the community did not own the process from start to finish.

No review articles were identified which compared or consolidated the different designs.

This supports the need for a meta-analysis of studies that use the PD approach.
 29
Only one study attempted to validate the PDI methodology. Lapping, Schroeder et

al. (2002) compared a quantitative case-control study using a sample of 50 families, with

a small qualitative phenomenological PDI using a sample of 8 families, to see which

approach identified the most PD behaviors associated with child nutrition among Afghan

refugees. They found that the PDI identified many of the same behaviors as the case-

control study. There were also a number of behaviors that were only identified by the

PDI, and were not identified by the case-control study. These were the more complex

behaviors that are difficult to measure on a survey, such as active feeding and maternal

affect. However, the PDI did miss a few behaviors that were identified by the case-

control study, such as immunizations and use of healthcare services. The authors

concluded that the small-scale qualitative PDI is an affordable and valid rapid assessment

tool, although using both methods would be optimal if resources allowed.

In addition to the overall methodology, there are a number of other key areas

where studies differ in their PD approach. The first is the way in which “success” is

defined in order to identify positive deviants. In some studies the outcome used to define

positive deviants is health status, such as lack of infection with HIV or HCV (Friedman et

al., 2008), healthy weight and height (Aruna et al., 2001; Levinson et al., 2007; Wishik &

Van Der Vynckt, 1976), newborn survival (Marsh, Sternin et al., 2002), or the absence of

chronic conditions (Mackenbach et al., 1994). In other cases the outcome is a behavior

that is known to be associated with positive health outcomes, such as delaying first sexual

intercourse (Babalola et al., 2006), practicing safe sex (Babalola et al., 2002), not

smoking (Awofeso et al., 2008), following nutritional guidelines (Vossenaar et al., 2009;
 30
Fowles et al., 2005) and practicing exclusive breast feeding (Dearden, Quan, Do, Marsh,

Pachón et al., 2002).

Studies also differed in what was considered to be a PD behavior. The United

Nations University (1990) defines a PD behavior as something that is “significantly

related to positive-deviant…status; and not yet dictated by rules that are endorsed by

scientists and health professionals” (para 20). However, many other studies included a

behavior as PD if it was not commonly practiced by the community, even if it had

already been endorsed by scientists and health professionals. An example of this is

exclusive breast feeding, which is already known to be beneficial for a child’s health, but

was still identified as a PD behavior by Lapping, Schroeder et al. (2002) and Dearden,

Quan, Do, Marsh, Pachón et al. (2002) because it was not a common practice in the

community.

In some studies, distinctions could be made between PD behaviors that can be

transferred to others, and those that are “True But Useless” (Save the Children, n.d. p.4).

That is, they are true for that particular positive deviant, but cannot be transferred to other

people in the community. For example, Lapping, Schroeder et al. (2002) identified the

following PD behaviors as true but useless for improving child nutrition among Afghan

refugees in Pakistan: The family helping the mother with care giving, the father taking an

active role in family life, and the mother not exhibiting a depressed temperament. These

behaviors cannot be transferred easily because the first two go against deeply ingrained

social norms and the final one is related to individual personality. By comparison, they

considered the feeding of special foods, such as shira (made with sugar, flour, oil and
 31
water) and arkhanak (wild vegetables) to be PD behaviors that were both true and useful

– they could easily be transferred to other members of the community. Marsh, Sternin et

al. (2002) and Save the Children (n.d.) both emphasized the need for the community to be

involved at the end of the PDI to validate the identified PD behaviors, and eliminate those

that are true but useless. However, not all PDI studies make the distinction between

useful PD behaviors and true but useless behaviors. For example, Aruna et al. (2001),

Babalola et al. (2006), Fowles et al. (2005), and Friedman et al. (2008) all included

nontransferable behaviors, sociocultural factors, and personality traits in their identified

list of PD behaviors.

Finally, some programs include “booster PDIs” at key points throughout the

program to identify new adopters of the behavior and determine what has allowed them

to adopt it (Lapping, Marsh et al., 2002; Marsh et al., 2007).

As can be seen from this discussion of PDI design, there are still a lot of gaps in

our understanding of how the PD approach should work, and whether current programs

are using the best approach. More research is needed to determine the range of PD

approaches currently in use, and how the different methodologies relate to the tangible

and intangible benefits it brings to the community (Lapping, Marsh et al., 2002; Marsh et

al, 2004; Marsh & Schroeder, 2002).

Related Approaches and Theories in Community Health

The PD approach shares many similarities with other approaches in community

health, which are described below. A clearly defined relationship between the PD

approach and these other approaches is yet to be developed.

 32
Participatory Approaches

Participatory approaches have a long history. Starting with action research in the

1940s, they have evolved to include a wide range of related definitions and terms,

including participatory research, Participatory Action Research (PAR), Community

Based Participator Research (CBPR), rapid assessment procedures, constructivist inquiry,

emancipatory inquiry, community-partnered participatory research, and participatory

evaluation (Minkler & Wallerstein, 2008).

CBPR - one of the most common terms - is defined by the U.S. National Institutes

of Health (NIH) as:

Scientific inquiry conducted in communities in which community members,

persons affected by condition or issue under study and other key stakeholders in

the community's health have the opportunity to be full participants in each phase

of the work: conception, design, conduct, analysis, interpretation, conclusions,

and communication of results” (NIH, 2008, para. 2).

There is not a single level of community participation within participatory

approaches; rather, there is a spectrum of community participation (Israel et al., 2008).

One of the earliest papers to describe this spectrum was by Arnstein (1969). Arnstein

described the levels of community participation in terms of a ladder. Starting at the top of

the ladder, the three highest levels of participation are citizen control, delegated power,

and partnership. Arnstein described these three rungs as “citizen power”. At these levels

community members hold the majority of decision making power, and can ensure their

wishes are implemented. The next three rungs are placation, consultation, and informing,
 33
which Arnstein refers to as “tokenism”. At these levels community members are able to

have their voice heard by power holders, but are not able to ensure that their wishes are

actually implemented. At the bottom of the ladder are manipulation and therapy, which

are considered “nonparticipation.” At these levels the true objective is allowing those

with power to control the community through supposedly positive initiatives such as

education and care. Historically, participatory research studies have ranged from

tokenism through to true community power and participation (Israel et al., 2008).

The PD approach as defined by Pascale et al. (2010) is intended to be a

participatory approach. One of the guiding PD principles identified by Pascale et al.

(2010) is that “the community must own the entire process” (p.196), with the researchers

and experts taking a back seat. On Arnstein’s scale this would count as the highest rung

on the ladder – citizen control (Arnstein, 1969). However, no review has been conducted

to determine how many of the PD studies and programs actually adhere to this high level

of participation.

In addition to being part of the group of participatory research approaches, the PD

approach shares similarities with some specific participatory approaches. For example,

one of the primary aims of action research is to “produce practical knowledge that is

useful to people in the everyday conduct of their lives.” (Minkler & Wallerstein, 2008, p

226). Clearly then, the PD approach is an example of action research, as its aim is to

discover successful behaviors and solutions from within the community that can be

implemented immediately by the rest of the community as part of their daily lives.

If run according to the guidelines by Pascale et al. (2010), the PD approach should
 34
also directly incorporate participatory evaluation as defined by Minkler and Wallerstein

(2008). The community should decide on how the program is to be monitored, and should

conduct the monitoring and evaluation. Again, there have been no reviews of PD studies

to determine if this is a widely followed practice.

The PD approach should also incorporate the principle of endogenous

development. Endogenous development allows communities to take control of their own

development process, and emphasizes the need for communities to set their own criteria

for development (COMPAS, 2007). The PD approach according to Pascale et al. (2010)

meets this requirement by allowing the community to decide what problem they would

like to address and to opt in or out of the PD process.

In order to enact participatory approaches, a process of community organization is

required. Community organization is defined as a “process through which communities

are helped to identify common problems or goals, mobilize resources, and in other ways

develop and implement strategies for reaching the goals they have collectively set”

(Minkler & Wallerstein, 2005, p. 26). If the PD approach is participatory in the way that

Pascale, Sternin and Sternin (2010) describe, then it will require community organizing in

order for the community to select the problem, identify and implement the solutions, and

collectively evaluate the program.

Assets Based Approaches

When using an assets based approach, a practitioner looks for existing resources

and opportunities within the community, which stands in contrast to the “needs based” or

“problem centered” approach often used in international development and community

 35
health (Sternin, 2002). The PD approach is considered an assets based approach

(ACCESS, 2009; Pascale et al., 2010). Two common assets based approaches are

appreciative inquiry (AI) and assets based community development (ABCD).

Cooperrider, Whitney, and Stavros (2008) defined AI as “a philosophy that

incorporates an approach…for engaging people at any or all levels to produce effective,

positive change” (p. XV). It is based on the assumption that:

Every organization has something that works right – things that give it life when it

is most alive, effective, successful, and connected in healthy ways to its

stakeholders and communities. AI begins by identifying what is positive and

connecting to it in ways that heighten energy, vision, and action for change.

(Cooperrider, Whitney, & Stavros, 2008, p. XV).

At a high level the AI and PD approach look very similar. In both approaches

practitioners look for positive aspects within a community. Both have a process that

involves four D’s. For AI these are discovery, dream, design, and destiny (Cooperrider,

Whitney, & Stavros, 2008), while for PD they are define, determine, discover, and design

(Pascale, Sternin and Sternin, 2010). At present it is hard to make a definitive distinction

between the two, as it is still unclear exactly what is included in the PD approach from

the variety of methodologies published. Based on the definition of the PD approach by

Pascale et al. (2010), one distinction could be made - the AI approach looks for

community success stories first, and then tries to expand them, whereas in the PD

approach, a problem is defined first, and then successful solutions within the community

are explored. However, a better understanding of what constitutes the PD approach is

 36
required before this distinction can be confirmed.

Assets Based Community Development is based on the premise that:

people in communities can organise to drive the development process themselves

by identifying and mobilising existing (but often unrecognised) assets, thereby

responding to and creating local economic opportunity. In particular, ABCD

draws attention to social assets: the particular talents of individuals, as well as the

social capital inherent in the relationships that fuel local associations and informal

networks. (Mathie & Cunningham, 2003, p. 474).

As with AI, ABCD focuses on the assets or positive aspects of a community.

Some of the key differences are that it focuses on community development and

particularly on community economic development. Therefore, the first stage in the

process is for the community to determine how it wants to develop (endogenous

development) and then the ABCD approach helps the community identify the assets it

has that will allow it to develop in the chosen way. ABCD particularly focuses on social

capital and social networks as assets that will allow a community to develop in the

chosen direction. As with AI, it is difficult to clearly distinguish the ABCD approach

from the PD approach without a full understanding of what is included in the PD

approach. However, one difference between the ABCD approach as defined by Mathie

and Cunningham (2003), and the PD approach as defined by Pascale et al. (2010) could

be that the ABCD approach begins with the community deciding how they want to

develop, while in the PD approach the starting point is defining a discrete problem for

which indigenous solutions are found. Again, a better understanding of what constitutes
 37
the PD approach is required before this distinction can be confirmed.

A related concept that also looks for the positive rather than the negative is

“resiliency” (Lapping, Marsh et al., 2002; Zeller, 1991). Models of resiliency have been

used in social work for many years to identify positive factors (assets) that can have a

protective effect on vulnerable youth (Brokenleg, 2010). However, the definition of

resiliency is slightly different to the definition of PD. Resilient individuals are those who

manage to maintain normal outcomes despite adversity, while positive deviants are those

who achieve above average outcomes despite adversity.

Theory of Diffusion of Innovations

The theoretical framework used for this study is the theory of diffusion of

innovations proposed by Rogers (1995). The theory of diffusion of innovations describes

the spread of new ideas and innovations throughout a population. In the theory a

population is divided into different groups based on how quickly they adopt a new idea;

for example, innovators (those who create the new idea), early adopters (those who are

first to adopt it in a population), early majority (the majority of people who adopt it after

the early adopters), late majority (the majority of people who adopt it later, after the early

majority), and laggards (those who resist adopting the new idea). Rogers (1995) also

identifies characteristics of innovations that make them more likely to spread quickly,

such as the relative advantage compared to other solutions, the compatibility with

existing practices and cultures, the complexity, the ease with which the innovation can be

trialed, and how easily the results can be observed.

The spread of new approaches through community health practice can be seen as
 38
the diffusion of an innovation. The innovation is the new approach, while the population

is community health practitioners. Examples of innovations that have spread throughout

community health practice include technological innovations such as new vaccines, as

well as novel approaches to program design and evaluation such as CBPR (Israel et al.

2008). Both of these innovations started as new ideas that were initially only practiced by

a small number of early adopters, and both have now spread extensively to become a

standard part of the community health workers’ toolbox.

Classic health promotion models that have diffused throughout public health

practice include Prochaska and DiClemente’s (2005) stages of change model in which

individuals move through the stages of precontemplation, contemplation, preparation,

action, and maintenance in order to change their behaviors; the health belief model in

which individuals change their behavior based on their perceived susceptibility and

severity of the disease, and the barriers and benefits of adopting the new behavior

(Rosenstock, Strecher & Becker, 1988); the theory of planned behavior in which

individuals change their behavior depending on their behavioral intention, attitude,

subjective norms, and perceived behavioral control (Sheppard, Hartwick, & Warshaw,

1988); and Bandura’s (1986) social cognitive theory in which behavior is determined by

the individual’s self-efficacy, goals, and outcome expectations, all of which can be

influenced by interactions with others, including the modeling of behaviors by others.

The rate and extent to which PD has spread throughout community health practice

is currently unknown. By examining the geographic and temporal distribution of PD

programs, this study will seek to map the diffusion of the PD approach throughout
 39
community health practice. This will include identifying innovators and early adopters, as

well as characteristics of the PD approach that make it more or less likely to spread

further.

Qualitative Case Study Research

This study will use qualitative case study research, with a focus on document

analysis. Case study research has a long history in the field of qualitative research, and

often relies on existing documents as a source of data (Creswell, 2007; Patton, 2002).

According to Creswell (2007), case study research is “a qualitative approach in

which the investigator explores a bounded system (a case) or multiple bounded systems

(cases) over time” (p.73). Therefore, case study research is aligned with the purpose of

this study, which is to identify, describe, and analyze past and present cases of

community health programs that use the PD approach. Information on each case is

usually collected through multiple sources in order to build up a full description of the

case (Creswell, 2007). These sources can include documents, interviews, observations,

and audiovisual material. This is appropriate for the PD approach, as there is a large body

of publicly available program documentation describing the use of the PD approach in

practice. These publicly available documents can be used to develop case studies of

programs that have used the PD approach.

In case study research, a single instrumental case study can be selected to

demonstrate a particular issue or concern, or multiple case studies can be selected

(Creswell, 2007). Because the aim of this study is to describe the variety of PD

approaches and programs in use and the range of contexts in which they are used, a
 40
multiple case study approach is most appropriate. Within-case analysis is used to analyze

data in a single case, while cross-case analysis is used to identify common themes across

the multiple cases (Miles & Huberman, 1994).

There is a long tradition of using case study research in the field of health.

Examples of case study research using documents as a source of information include

cases of patients with rare disorders (Davies, 2011), cases of hospitals undergoing

healthcare reform (Tjerbo, 2009), cases of public health policies being developed

(Daniels & Lewin, 2008), and cases of public health programs being implemented (Khan

et al., 2010).

Two other research traditions were considered for this study, but were deemed

inappropriate due to the nature of available data sources. Systematic reviews (Petticrew

& Roberts, 2006) and qualitative meta-analysis (Timulak, 2009) are two approaches used

to draw conclusions across a large body of published research. The majority of systematic

reviews, such as those published by the Cochrane Collaboration, focus on evaluating

specific health interventions, such as drug treatments, by pooling the results of

randomized controlled trials (RCTs) or other experimental or quasi-experimental studies

(Higgins & Green, 2009). However, Petticrew and Roberts (2006) argued that systematic

reviews can, and should, be used to answer a broader range of research questions when “a

general overall picture of the evidence in a topic area is needed to direct future research

efforts” (Petticrew & Roberts, 2006, p. 21).

Qualitative meta-analysis is used in a similar way to perform secondary

qualitative analysis of published qualitative studies in order to draw more generalizable

 41
conclusions (Timulak, 2009). There is currently no agreed terminology for this method,

and others have referred to similar, if not identical, methods as qualitative meta-synthesis

(Finlayson & Dixon, 2008; Walsh & Downe, 2005), qualitative meta-study (Paterson et

al, 2001), and systematic review of qualitative research (Petticrew & Roberts, 2006).

Both systematic review and qualitative meta-analysis are not appropriate for this

study as the source data is primarily drawn from published peer reviewed studies,

although gray literature reports can sometimes be included. There is relatively little peer

reviewed literature available on the PD approach, and the gray literature available

consists of a wide variety of program documents, including program reports, training

manuals, presentations, workshop notes, web pages, evaluation papers, and online

audiovisual materials. Therefore, multiple case study research is more appropriate in this

instance.

Alternative Research Methods

This dissertation is the first to study the PD approach itself, as opposed to a

specific application of the PD approach in the field. Almost all previous studies published

on PD describe a specific PD inquiry or program that has been implemented in the field.

There has been only one previous study that attempted to investigate the PD approach

itself by comparing a quantitative case-control study using a sample of 50 families with a

small qualitative phenomenological PDI using a sample of eight families to see which

approach identified the most PD behaviors associated with child nutrition among Afghan

refugees (Lapping, Schroeder et al., 2002). However, this method is only appropriate for

comparing the results of the PD approach to the case-control approach for a specific
 42
topic. It is not relevant for investigating the PD approach overall.

Because there has been no previous research into the overall PD approach, no

research methods, measurement instruments, or tools have been developed for this

purpose. Therefore, all the instruments and tool used in this study were developed by the

researcher, or were adapted from other studies.

In conclusion, there are a range of gaps in the PD literature which need to be

attended to before PD can be adopted as a mainstream approach in community health.

This study contributes to addressing some of these gaps. The following chapter describes

the study methodology.

 43
Chapter 3: Research Method

Introduction

The problem being addressed by this study is the lack of research into PD

programs, including the variations of PD approaches currently in use, the characteristics

of programs using the PD approach, the effectiveness of the PD approach in a range of

contexts, and the spread of the PD approach throughout community health practice.

These gaps need to be addressed before decisions can be made about whether or not PD

should become a standard part of the community health worker’s toolbox. Because the

problem being addressed is related to the characteristics of PD programs, it is logical for

this study to use a multiple case study approach in which each case study describes a PD

program. By analyzing a group of case studies describing PD programs, it is possible to

draw conclusions about the characteristics of the programs, the types of PD approaches

used, and the effectiveness of the programs.

Research Design and Approach

The purpose of the study is to identify, describe, and analyze past and present

community health programs that use the PD approach to better define the spectrum of PD

approaches currently in use and their effectiveness. Because the study involves the

analysis of a series of individual programs, it is appropriate to use a multiple case study

approach, with each case study representing one program (Creswell, 2007). The study is

not seeking to develop a theory, understand the essence of experience, describe a culture

sharing group, or document the life experiences of an individual. Therefore, a grounded

theory, phenomenological, ethnographic, or narrative approach was not appropriate

 44
(Creswell, 2007). As described in Chapter 2, systematic review and qualitative meta-

analysis were both considered as alternative research traditions for this study. They were

deemed inappropriate as they both use published peer reviewed literature as the primary

source of data, and there is relatively little of this available for the PD approach.

Publicly available program documentation was used as the source of case study

data, including program reports, training manuals, presentations, workshop notes, web

pages, evaluation papers, online audiovisual materials, and peer reviewed studies. This

documentation is appropriate as a source of data because it is readily available and a large

amount exists in the public domain. Because existing documentation was used as the

source of data, the case study research had a focus on document analysis.

The conceptual framework used for the analysis was the theory of diffusion of

innovations as proposed by Rogers (1995). The theory of diffusion of innovations

describes the spread of new ideas and innovations throughout a population. In the theory

a population is divided into different groups based on how quickly they adopt a new idea.

For example, groups include innovators (those who create the new idea), early adopters

(those who are first to adopt it in a population), early majority (the majority of people

who adopt it after the early adopters), late majority (the majority of people who adopt it

later, after the early majority), and laggards (those who resist adopting the new idea).

Rogers (1995) also identified characteristics of innovations that make them more likely to

spread quickly, such as relative advantage, compatibility, complexity, trialability, and

observability. The spread of the PD approach throughout community health practice can

be seen as the diffusion of an innovation. By examining the geographic and temporal

 45
distribution of PD programs, this study maps the diffusion of the PD approach throughout

community health practice. The study also identifies innovators and early adopters, as

well as characteristics of the PD approach that make it more or less likely to spread

further.

Role of the Researcher

The role of the researcher in this study was as the primary individual collecting

available documentation, using it to describe the program case studies, and analyzing the

case studies to answer the research questions. As the researcher I worked closely with the

peer debriefer. Our working relationship is described in more detail in later sections.

Research Questions

1. How is the PD approach currently being used in community health

practice?

1.1 What health related topics has the PD approach been used to

address?

1.2 What settings has the PD approach been implemented in?

1.3 Which types of organizations are implementing programs that use

the PD approach?

1.4 To what extent do programs based on the PD approach follow the

steps defined by Pascale et al. (2010)?

1.5 To what extent do programs based on the PD approach incorporate

community participation in each step?

1.6 To what extent is the PD approach incorporated into integrated

 46
programs versus stand-alone programs?

2. How has the PD approach diffused through community health practice?

2.1 Has the diffusion followed the principles of the theory of diffusion

of innovations?

2.2 How has this diffusion progressed geographically, chronologically,

and between community health issues?

2.3 Can groups of innovators, early adopters, early majority, late

majority, and laggards be identified?

2.4 Are there characteristics of PD that have made it more or less

likely to spread quickly, such as relative advantage, compatibility,

complexity, trialability, and observability?

3. Can the range of PD approaches currently in use be grouped into a typology

of PD approaches?

3.1 Are particular PD approaches more commonly used in specific

contexts or settings?

4. What quantitative and qualitative outcomes have been reported from

programs that use a PD approach, and with what level rigor were the studies

conducted?

Setting and Sample

Sampling for this study consisted of identifying and selecting publicly available

program documents. To ensure consistency, sampling was done using a predefined search

strategy and inclusion criteria. Because this is the first such study of the PD approach, an
 47
exhaustive sampling strategy was used in which all documents that meet the inclusion

criteria are sampled.

Inclusion Criteria

The criteria for including a document in the study were as follows. The document

must describe all or part of a past, present, or future program/study that uses the PD

approach. This is based on the author identifying it as using the PD approach. It need not

adhere to the core principles of the PD approach as defined by Pascale et al. (2010). The

document must be on a health topic or a topic related to known social determinants of

health including, but not limited to social class, gender, ethnicity, education, occupation,

income, economic development, housing, working conditions, food availability, social

exclusion, and access to infrastructure such as health services and sanitation (WHO,

2010). The document must be published in English. The document must be available in

full text online.

No restrictions were placed on the date of publication or type of document.

Program reports, training manuals, presentations, workshop notes, web pages, evaluation

papers, online audiovisual materials, and peer reviewed studies were all included.

Documents containing a general discussion of the PD approach, without any reference to

a specific program, were excluded.

Sampling Methods

The protocol used to identify and sample the documents varied based on the type

of document being sampled. Sampling of published peer reviewed studies was done

according to the following protocol:

 48
1. All databases listed in Appendix C were searched for any articles that

contain the term positive deviance. Because the inclusion criteria for

includes studies on social determinants of health, a wide range of health

and social science databases, and a number of multidisciplinary databases,

were included in Appendix C.

2. The abstracts of all articles were retrieved.

3. I reviewed all the abstracts and exclude studies that did not meet the

inclusion criteria.

4. The full text of the remaining articles was retrieved.

5. In addition to searching databases, the Food and Nutrition Bulletin was

hand searched to identify all PD studies. The Food and Nutrition Bulletin

published the largest number of PD studies identified for this proposal,

and so it was possible that there could be additional articles published in

the journal that would not be identified by a database search. Hand

searching of key journals such as this has been shown to improve the

accuracy of systematic reviews, which have a similar approach to this

study (Petticrew & Roberts, 2006).

Sampling of program documentation in gray literature was done according to the

following protocol:

1. All sources of gray literature listed in Appendix D were searched for any

articles or reports that contain the term positive deviance. Items were

reviewed and included if they meet the inclusion criteria defined above.
 49
2. All documents available on the Positive Deviance Initiative website

(2010e) were reviewed and included if they met the inclusion criteria.

3. All documents referenced in the book “The Power of Positive Deviance”

by Pascale et al. (2010) were reviewed and included if they met the

inclusion criteria.

4. An Internet search was conducted using the search engine Google and the

search term positive deviance. Results were reviewed and included if they

meet the inclusion criteria.

The reference lists for all documents was reviewed to locate additional

documents. These were reviewed and included they meet the inclusion criteria.

Data Collection and Analysis

This section describes how the data were collected and analyzed, as well as the

steps taken to ensure validity and reliability.

Data Collection

I read through all the documents that meet the inclusion criteria. Initial data

collection involved the extraction of key data from the documents using the document

data collection form in Appendix A. A single PD program often had several pieces of

documentation associated with it. For example, there may be a training manual,

presentation, evaluation report and peer reviewed study all published on the same PD

program. Therefore, the first step in the analysis involved identifying groups of

documents that relate to the same PD program. During this process it was found that

many of the PD programs were actually only PD inquiries. A PD inquiry is a study in

 50
which positive deviants and their behaviors are identified, but the findings are not used to

design and implement a program. Given the substantial number of them identified I made

the decision to include PD inquiries in the data set, but to analyze them separately from

the full PD programs. A program or inquiry was included as long as there was at least one

substantial document describing it. A substantial document was one that allowed at least

80% of the program / inquiry data collection form (Appendix B) to be completed.

Once each unique PD program and its associated documents were identified,

within-case analysis was used to complete as many fields as possible in the program /

inquiry data collection form (Appendix B). Each program / inquiry data collection form

then represented a single case study of a PD program or PD inquiry. A deeper qualitative

analysis was conducted of these case studies in order to answer the research questions.

Once each case study was developed using the program / inquiry data collection

form (Appendix B), those cases which had data available on the results of an evaluation

had the level of rigor of the evaluation assessed using quality appraisal forms. For case

studies with a quantitative evaluation design, the quality assessment tool for quantitative

studies was used (Appendix E). For case studies with a qualitative evaluation design, the

quality assessment tool for qualitative studies was used (Appendix F). In case studies in

which a mixed-methods evaluation design was utilized, both assessment forms will be

completed. If the case study was missing information on critical items, such as data

collection, no conclusions were drawn on its effectiveness, and the quality assessment

tools were given a rating of unknown.

All documents and completed forms were loaded into a Microsoft database for
 51
storage. A backup copy of the database was kept on an external hard drive. The database

will be kept for a minimum of five years after the end of the project. An auditable log

was kept in a Word document and Excel spreadsheets with details of the data sampling,

collection and analysis processes. This included the location and date when each

document was identified, the date of data extraction, any iterative modifications to data

collection forms, minutes of peer debriefing sessions, and notes on the evolution of my

ideas during the analysis process.

Validity and Reliability

The initial data extraction process into the document data collection form required

minor interpretive steps, but was basically a process of data transfer and did not raise

significant questions of validity and reliability. However, the creation of case studies

using the program / inquiry data collection form, the in-depth analysis of patterns and

themes, and the development of a typology of PD approaches raised issues of validity and

reliability common to most qualitative research. The primary techniques used to ensure

validity and reliability across the entire study was by external audits and peer debriefing

with an independent researcher (Creswell, 2007). The external auditor and peer debriefer

were the same person for practical reasons. This person has the equivalent of a Masters

qualification in Social Science, and over 20 years’ experience in planning and evaluation

of programs, qualitative and quantitative research, data collection and analysis,

facilitation, and policy development. This includes experience using participatory

research and community development approaches similar to Positive Deviance, as well as

qualitative case study research. He has been published in peer reviewed journals, such as
 52
the Journal of Applied Behavioural Science (Onyx & Bullen, 2000) and Developing

Practice (Bullen, 2004), has written book chapters (Onyx & Bullen, 2001), and presented

at conferences such as the Australasian Evaluation Society conference. I previously

worked with the peer debriefer from 2000 to 2005 as a consulting analyst at the

management consulting firm for which he is the director.

An audit log was kept with details of the data sampling, collection, and analysis

tasks performed. The external auditor reviewed the first five case studies developed in the

program data collection form to ensure that the data extraction process was accurate. This

involved the auditor reviewing all source documents for the case studies and comparing

them to the data contained in the program / inquiry data collection form to ensure that the

case study accurately reflected the information in the program documents. Based on this

initial audit, the forms and data collection protocols were refined and enhanced. The

auditor then reviewed a further 10% of all program / inquiry data collection forms during

the course of the study.

The completion of the quality assessment tools for qualitative and quantitative

evaluations (Appendices E and F) also raised issues of validity and reliability. To address

this, the external auditor independently completed the quality assessment tools for all

case studies that had sufficient information. These results were then compared with my

results, and any discrepancies were resolved through discussion to arrive at a final rating

for each evaluation study. The external auditor did not complete independent quality

assessments of case studies which did not contain sufficient information to complete the

quality assessment tool adequately. Criteria for how this is decided can be found in the
 53
quality assessment tools for qualitative and quantitative evaluations (Appendices E and

F). This was to avoid wasting the external auditor’s time by having him review cases

which I had already identified as having insufficient information.

Peer debriefing was used during the analysis process. I identified patterns and

themes in the initial data, and discussed and validate these with the second researcher.

Together we identified areas for deeper analysis. Written minutes were kept of all peer

debriefing sessions (see Appendix G). In addition to peer debriefing, the development of

the typology of PD approaches used negative case analysis as a technique to enhance

validity (Creswell, 2007). An initial typology was developed, and it was further refined

and enhanced until all cases were able to fit into the final typology.

Data Analysis

Data analysis consisted of an open-ended iterative approach based on the theory

of diffusion of innovations, with the purpose of answering the four primary research

questions. The first step was the analysis of data in the document data collection forms to

generate an overall picture of the documentation available. This used tables and charts in

Microsoft Excel to examine the number and type of documents, the range of topics, the

types of organizations involved, and the dates and locations of publication. Using this

information documents were grouped into those that describe the same PD program.

Once each unique PD program case study was developed in the program / inquiry data

collection form using with-case analysis, a cross-case analysis (Miles & Huberman,

1994) was conducted to identify common themes and trends across multiple case studies

in order to answer the research questions.

 54
The first and second research questions were answered using a series of tables and

charts in Microsoft Excel to examine the following characteristics of programs: health

topics, geographic distribution, chronological distribution, settings, organizations

involved, PD steps, community participation, and integration. This analysis used the

theory of diffusion of innovations (Rogers, 995) as a framework to chart the spread of the

PD approach throughout the global practitioner community.

The third research question involved developing a typology of PD approaches by

reviewing each case study in turn, and building up a set of categories until all case studies

were able to fit into the classification system. During peer debriefing sessions, the

categories were reviewed and refined until a final set was developed. This involved

removing extraneous categories and dividing larger categories into smaller categories.

Once the final typology was developed, it was applied to all the case studies. A

series of tables and charts in Microsoft Excel were then used to examine relationships

between the type of PD approach and other contextual factors, such as the health topic,

geographic location, and timeframe of the program.

The final research question was related to the quantitative and qualitative

outcomes from programs that use a PD approach. This analysis used the methods and

outcome data in the program / inquiry data collection form, the assessment of the

evaluation report completeness, and the completed quality assessment tools for

quantitative and qualitative studies (Appendices E and F). Each case study was reviewed

to identify any positive and negative outcomes, and to assess these in relation to the rigor

of the evaluation and completeness of the report. Due to the wide range of topics and
 55
methods, it was not be possible to perform a quantitative meta-analysis of program

outcomes. Therefore, qualitative analysis was used to generate an overall picture of

program effectiveness in different contexts (Petticrew & Roberts, 2006). When

developing the qualitative analysis, greater emphasis was placed on the outcomes of

programs that use a more rigorous evaluation design, such as randomized controlled trials

and quasi-experimental designs.

Instrumentation and Materials

The study used two data collection forms: the document data collection form

(Appendix A) and the program / inquiry data collection form (Appendix B). I developed

both forms for the purposes of this study and they have been extensively tested by both

myself and the expert reviewer on a range of PD documents identified during the

sampling strategy test. Some adjustments were made during the research process. The

most significant adjustment was including PD inquiries into what was originally only the

program data collection form. This involved identifying the specific fields in the program

form that were also relevant for inquiries. In addition to this change, some minor

modifications were also made to accommodate new variables and categories that were

developed during the analysis of the data.

The document data collection form has two sections – document details and

program details – and is mainly comprised of open ended fields. Although different types

of documents (e.g. reports, presentations, audiovisual materials, etc.) had slightly

different information available, I decided to keep the document data collection form the

same for all types of documents. This approach dramatically simplified the database
 56
design and data entry process, which was very important given the large volume of

documents that needed to be processed. I made some minor modifications to the original

document data collection form during data collection. For example, I split the source field

into separate fields for source, volume / issue number, URL, doi, etc. so that it was easier

to sort and search the database.

The program / inquiry data collection form included open ended and multiple

choice fields for document sources, program details, steps used in the PD approach,

community participation during each step, practical implementation, case study

completeness, qualitative results, and quantitative results. The PD steps and community

participation sections were based on the PD approach described by Pascale et al. (2010).

In addition to the data collection forms there were also two quality assessment

tools, one for quantitative studies (Appendix E) and one for qualitative studies (Appendix

F). The tool for quantitative studies is based on the Effective Public Health Practice

Project (EPHPP) quality assessment tool for quantitative studies (EPHPP, 2010). This

tool was selected because it is relatively short compared to other assessment tools, which

is more practical given the very large number of case studies in this analysis. It is also

applicable to all types of quantitative evaluation designs, and has been used widely in

systematic reviews of quantitative literature (Petticrew & Roberts, 2006). The EPHPP

tool was used in conjunction with the EPHPP Dictionary (EPHPP, 2009) which provides

definitions and guidance for completing each item. A large number of the gray literature

studies contained insufficient information to complete many sections of the tool.

Therefore, some minor adjustments were made to allow for the identification of missing
 57
data. This included the addition of unknown as a rating in addition to strong, moderate

and weak. An evaluation study was classified as having an overall unknown level of rigor

if there was insufficient information to complete three or more sections of the tool.

The quality assessment tool for qualitative studies is based on the National Health

Service (NHS) Critical Appraisal Skills Programme (CASP) quality assessment questions

for qualitative studies (NHS CASP, 2006). This tool was selected because it is limited to

the 10 most important assessment criteria, making it more concise than many of the other

tools available, such as those by Long and Godfrey (2004) which has more than 30 items

and Spencer et al (2003) which has 18 items. This was important given the large number

of case studies it was be applied to, and the fact that each item requires a yes / no

response as well as more extended comments. As with the EPHPP tool, the NHS CASP

tool was modified to allow missing data to be identified. A global rating of strong,

moderate or weak, was also added using the same approach as the EPHPP tool (EPHPP,

2010) to allow for consistency in the analysis of qualitative and quantitative studies.

Protection of Human Participants

The proposal was approved by the Walden Institutional Review Board (IRB;

approval number 10-26-11-0124829). Only publicly available documents were included

in the study. No confidential documents, individual patient records or private health

information, were assessed as part of the review.

Dissemination of Findings

The primary audiences for this study are community health practitioners working

in the field and academics, particularly those who set the content for community health
 58
textbooks, and training programs. Therefore the dissemination strategy will include

disseminating the results to practitioners by sending the study to the Positive Deviance

Institute for inclusion in their online collection of positive deviance resources. The study

will also be submitted for publication in a peer reviewed journal that has an audience of

both academics and practitioners.

In conclusion, this study used a multiple case study analysis based on publicly

available documentation to help better define the spectrum of PD approaches currently in

use, the characteristics of PD programs, and their effectiveness in a range of contexts.

Chapter 4 describes the results of the study.

 59
Chapter 4: Results

Sampling and Data Collection

The initial search for the term positive deviance using the sources listed in

Appendices A and B returned 274 potentially relevant peer reviewed studies and 615

potentially relevant gray literature documents (see Appendix H for search results by

database). Of these, 158 peer reviewed studies and 341 gray literature documents were

excluded based on their title and abstract. The main reasons for exclusion were that the

document only mentioned PD briefly (46 peer reviewed journal articles and 260 gray

literature documents), and that the document was not on a health-related topic (104 peer

reviewed articles and 53 gray literature documents). Examples of non-health topics

included climate change, sociology, and sports psychology. The full text of the remaining

116 peer reviewed studies and 274 gray literature studies were retrieved for further

assessment. The largest number of documents were retrieved from the Positive Deviance

Initiative website which returned 276 results, 147 of which were retrieved for further

analysis, and the USAID website which returned 157 results, 45 of which were retrieved.

Figure 1 summarizes the sampling process. Of the 116 peer reviewed studies and

274 gray literature documents retrieved for further assessment, 47 peer reviewed studies

and 107 gray literature documents were excluded based on the full text. The most

common reasons for exclusion were that the document discussed the PD approach in

general without referring to a specific program (36 peer reviewed articles and 78 gray

literature documents) and that the full text could not be located (four peer reviewed

journal articles and 21 gray literature documents). The details of the remaining 69 peer
 60
reviewed studies and 167 gray literature documents were entered into the document data

collection form.

The first round of within-case analysis involved collecting the documents into

groups that described the same programs. This was done by comparing the country, topic,

and dates on the documents, and then comparing the title of the program they described

and the name of the organization implementing the program. Using this method the first

round of within-case analysis identified 91 individual PD programs described using 236

documents. At this point an additional 19 gray literature documents were identified

describing the PD programs. These were found by searching Google for the names of the

programs.

A more detailed analysis of each program was conducted when I entered the data

from the relevant documents into the program data collection form. During this process I

excluded 17 programs as there were not enough data available to complete 80% of the

form. In addition, 26 of the individual programs were actually part of the same larger

program and so they were combined together. Programs were considered to be part of the

same larger program if they were run in the same location, on the same topic, and by the

same organization or donor.

 61
Peer Reviewed
Gray Literature
Journals

274 potentially 615 potentially

relevant on the basis relevant on the basis
of including the term of including the term
158 excluded on the basis of “positive deviance” “positive deviance” 341 excluded on the basis of
title and abstract: title and brief assessment:
· 46 only mentioned positive · 260 only mentioned
deviance briefly positive deviance briefly
· 104 on non-health topics · 53 on non-health topics
· 2 not available in English · 8 not available in English
· 6 for other reasons · 20 for other reasons
116 full text retrieved 274 full text retrieved
for further for further
assessment assessment
47 excluded on the basis of full 107 excluded on the basis of
text: full text:
· 36 discussed PD in general · 78 discussed PD in general
not a specific PD program not a specific PD program
· 5 duplicates · 21 could not locate full text
· 4 could not locate full text · 3 did not contain any
· 2 did not contain any references to PD in full text
69 documents 167 documents · 3 duplicates
references to PD in full text
continued to within- continued to within- · 2 very short without detail
case analysis case analysis

Total of 236 documents. Preliminary within-

case analysis identified 91 individual PD
programs.
19 additional grey literature
documents identified for program
17 programs (23 documents) excluded due to case studies.
insufficient data.

26 programs combined with other programs

that were part of the same larger program.

39 documents (8 programs) that only included a

PD Inquiry study, not a full PD program,
removed for separate analysis.

6 PD Inquiry
studies excluded: Total of 193 documents describing 40
· 4 insufficient individual PD programs.
data.
· 3 duplicates.

Total of 32 documents describing

32 PD Inquiry Studies.

Figure 1. Sampling process.

 62
There were four large programs which contained a significant number of

subprograms. The first was the child nutrition program implemented by Save the

Children in Vietnam. Save the Children implemented various versions of their program in

Vietnam from 1990 to 2007. Rather than treating each individual variation as a separate

program, they were all treated as one ongoing program. A similar program was

implemented on child nutrition in Indonesia. In this program multiple different NGOs

received funding from USAID to implement the PD approach to reduce child

malnutrition. All the NGOs followed a similar approach, and all were included in the

final evaluation report for USAID. Therefore, all of them were considered to be part of

one larger program. The third large program was on child nutrition in India. A large

number of PD programs were identified in India, being run by a range of different NGOs.

However, all programs were being implemented as part of the government run Integrated

Child Development Services (ICDS), which was considered to be one overarching

program. Finally, six hospital based PD programs for preventing hospital acquired

infections were all part of the same pilot program implemented using a grant from the

Robert Wood Johnson Foundation. Because all the hospitals were included in the final

evaluation report, all were considered to be part of one larger program.

Combining multiple subprograms into one larger program was necessary for

practical reasons, as counting each individual subprogram separately would have made

the data more difficult to analyze and interpret. However, combing the subprograms does

have some limitations. Individual subprograms, particularly in India, had a wide range of

variations in context, setting, and approach, which cannot be fully explored when they are
 63
grouped together as a single program.

While completing the program data collection forms, I also found that a

significant number of the programs were not actually programs, but studies. They only

included the results of a PD inquiry to identify positive deviant behaviors but did not

spread these behaviors to other people as part of a program. I decided to extract these PD

inquiries from the set of programs so they could be analyzed separately. Hereafter these

case studies are referred to as PD inquiries, as distinct from PD programs. Only some of

the fields in the original program data collection form were relevant for these studies, and

so I modified the form slightly to identify those fields that were relevant for PD inquiries.

Of the PD inquiries, four had to be excluded due to insufficient data, and three were

excluded because they reported on the results of the same study, just in different formats

(e.g., a poster and peer reviewed journal article).

The final result of the sampling process was 40 PD program case studies

described by 193 documents, and 32 PD inquiries described by 32 documents. Table 1

summarizes the types of documents used to describe them. The most common types of

documents describing PD programs were program reports (54 documents) and peer

reviewed journal articles (39 documents). However, there was a wide range of other

document types, including videos, webpages, training manuals, and presentations. PD

inquiries were more likely to be described in a peer reviewed journal, with 25 (78%) of

all PD inquiries being research studies that were published as a journal article. Appendix

I provides details of all PD programs and PD inquiries included in the final sample,

including the number and types of documents related to each case study.
 64
Table 1

Type of Documents Describing PD Programs and PD Inquiries

No. documents describing

Document type
PD programs PD inquiries
Peer Reviewed Journal Article 39 25
Program Report 54 3
Video 25
Periodical Article 16
Webpage 17 1
Manual/Training Materials 13
Presentation 14
Book 10
News Article 3
Poster 1 1
Thesis/Dissertation 1 2
Total 194 32

Data Management

All document data collection forms and program / inquiry data collection forms

were entered into a Microsoft Access Database with a unique identification number. The

database was relational, with each program / inquiry data collection form being linked to

multiple document data collection forms through the unique identifying numbers.

Examples of completed forms are shown in Appendix J. A backup of the database was

kept at all times.

Findings

Research Question 1

Topics. Table 2 shows the range of topics addressed by the PD programs and PD

inquiries identified. The most common area in which the PD approach has been used is
 65
child nutrition, with a total of 20 PD programs (50%) and 18 PD inquiries (56%). All

child nutrition programs were related to malnutrition rather than overnutrition,

overweight, or obesity. The next most common topic was hospital acquired infections

with a total of six PD programs (15%). There were several topics where only one or two

programs and/or inquiries have been identified. These included topics such as smoking

and chronic diseases.

Table 2

PD Programs and PD Inquiries by Topic

No. PD No. PD
Topic
Programs Inquiries
Child nutrition 20 18
Hospital acquired infections 6
Pregnancy outcomes 4 3
Child trafficking and soldiers 2
Health services 2 2
School education 2
Female Genital Mutilation 1
HIV/AIDS and reproductive health 1 4
Prisoner wellbeing 1
Smoking 1
Diet and weight control 3
Gender Equity 1
Chronic diseases 1

Settings. There are two parts to the issue of setting. The first is the type of

country in which PD programs and inquiries are being conducted, including the

geographic location and socioeconomic status. The other is the specific setting within the

country, such as urban communities, rural communities, institutional settings, and so

 66
forth. Table 3 shows the number of PD programs and inquiries conducted in each

geographic region. The largest numbers of PD programs have been conducted in Sub-

Saharan Africa (a total of 18 PD programs and PD inquiries) with the next largest number

in South Asia (15), North America (12), East Asia and Pacific (11), and Latin America

and the Caribbean (10). Very few PD programs or PD inquiries have been conducted in

Western Europe, Eastern Europe and Central Asia, or the Middle East and North Africa.

Table 3

Number of PD Programs and PD Inquiries by Geographic Region

# PD # PD
Region Total
programs Inquiries
East Asia & Pacific 5 6 11
Eastern Europe & Central Asia 1 1
Latin America & Caribbean 5 5 10
Middle East & North Africa 2 1 3
North America 4 8 12
South Asia 6 9 15
Sub-Saharan Africa 15 3 18
Western Europe 2 2
Figure 2 shows the percentage of PD programs and inquiries by the income level

of the country according to the World Bank (2011). Low income countries included

Afghanistan, Bangladesh, Ethiopia, and Haiti. Middle income countries included China,

Egypt, India, and Brazil. High income countries included the United States, United

Kingdom, Denmark, and Australia. The majority of PD programs and PD inquiries have

been conducted in low income countries (45% of PD programs and 13% of PD inquiries)

and middle income countries (38% of PD programs and 61% of PD inquiries). Relatively

few have been conducted in high income countries.

 67

Figure 2. Percentage of PD programs and PD inquiries conducted in low, middle and

high income countries

Figure 3 shows the percentage of PD programs and inquiries conducted in various

settings. The largest number of PD programs and inquiries were conducted in rural

communities (60% of PD programs and 46% of PD inquiries) followed by urban

communities (11% of PD programs and 41% of PD inquiries). This was followed by

hospitals, which were the setting for 18% of PD programs and 8% of PD inquiries. A

small number of PD programs were set in schools and prisons.

 68

Figure 3. Percentage of PD programs and inquiries conducted in various settings

Scale. The majority of PD programs and inquiries were relatively small. Of the 40

programs, fifteen had an unknown number of participants, four had less than 100

participants, 13 had hundreds of participants, five had thousands of participants, one had

tens of thousands of participants, and only two had more than 100,000 participants. The

largest single program was a child nutrition program in India that was comprised of many

loosely connected subprograms. All the subprograms were linked with the government

run Integrated Child Development Services program. Although it is difficult to calculate

the total participation, it is likely that together these many subprograms covered millions

of children in India, making it the largest implementation of the PD approach.

Types of organizations. Data on the type of organization was only available for

PD programs, as the majority of PD inquiries were published as peer reviewed journal

articles which did not clearly identify the organizations involved.

Of the 40 PD programs, 32 (80%) had an international organization involved in

 69
implementation. The most common international organizations involved in PD programs

were USAID, which was listed as a donor for 16 (40%) of the PD programs, and Save the

Children which was listed as an implementing agency for 10 (25%) of the PD programs.

The only programs that did not involve an international organization were those

implemented in high income countries (e.g. United States, Australia), with the exception

of one program in Pakistan on medical education. Out of the 40 PD programs, 18 (45%)

involved a national organization, such as a local NGO or university in the

implementation, and 21 (53%) involved the government at either national or subnational

level. However, it is important to note that national or government partners may have

been involved in the implementation of some programs without being mentioned in the

program documentation or reports.

PD steps. Thirty nine (98%) of the 40 PD programs implemented all PD steps

defined by Pascale et al. (2010) in some way, including defining the problem,

determining whether positive deviants existed, discovering PD behaviors, designing the

program, implementing the program, and monitoring the program. However, not all of

them implemented the steps exactly as Pascale et al. describe. For example, Pascale et al.

stated that PD programs should use action learning to spread the PD behaviors to others.

Some PD programs used other methods to spread the behaviors, such as mass media

campaigns, which did not specifically include action learning. This is discussed further in

the following section on the classification of PD approaches. There was one PD program

which did not include any description of the steps followed for implementation.

By definition, a PD inquiry included the first three steps only: defining the
 70
problem, determining whether positive deviants existing, and discovering their unique

practices. Of the 32 PD inquiries, all 32 (100%) defined the problem, 31 (97%) identified

positive deviants, and 29 (91%) attempted to discover positive deviant behaviors.

However, not all of them implemented the steps exactly as Pascale et al. describe. For

example, Pascale et al. state that the discovery of PD behaviors should happen through

open-ended inquiry and observation that allows for the discovery of new behaviors that

the researcher may not have previously considered. However, some PD inquires used

only quantitative data analysis to identify the PD behaviors. This means that all the data

was collected using a predefined survey instrument, and there was no opportunity to

identify new behaviors that the researcher may not have considered when they were

designing the survey. This issue is also discussed further in the following section on the

classification of PD approaches. Only one PD inquiry reported eliminating true but

useless behaviors and only five PD programs reported doing this.

Positive deviant behaviors identified. Each PD program and inquiry identified

different behaviors practiced by positive deviants, although they were not always

described in detail, particularly in the reports of PD programs. Table 4 shows examples of

positive deviant behaviors identified by a range of programs and inquiries.

In some cases the PD behaviors identified were new to both community members

and program staff. For example, collecting shrimp from the rice paddies and combining

with the green tops of sweet potatoes to make nutritious baby food, was not something

that the program staff working in Vietnam had considered before, nor was it a common

practice in the community. It was only the positive deviants who were practicing this
 71
behavior and so it could be considered a new behavior that had been discovered through

the PD inquiry.

In other cases the PD behaviors identified were new to the community members,

but were not new to program staff. For example, in Pakistan community members were

surprised to discover that positive deviants with good pregnancy outcomes also had

regular antenatal visits, and increased their food intake during pregnancy. This

“discovery” was not new to the program staff, who already knew from their public health

training that regular antenatal visits and nutrition are linked to good pregnancy outcomes.

In other cases the PD behaviors were well known by community members, but

were new to the program staff. For example, hiding condoms in the storage space of

motorcycles so their wives could not find them was a common strategy used by moto-taxi

drivers in Indonesia, but it was new for the researchers who “discovered” it during a PD

inquiry.

Finally, there were some cases where the PD behaviors were not new to

community members or program staff. For example, PD child soldiers in Uganda were

considered to be girls who worked harder and longer than others, and who attended

school. These behaviors were already known to be positive by both community members

and program staff before the PD inquiry was conducted, and so they could not be

considered new behaviors that had been discovered.

Unfortunately due to incomplete reporting of PD program and inquiry results it

was not possible to calculate the total number and type of PD behaviors identified across

all the case studies. The majority of documents and journal articles only gave examples
 72
of the PD behaviors identified without categorically listing and classifying them.

Table 4

Examples of Positive Deviant Behaviors Identified

Topic Country Examples of PD behaviors identified Behaviors were new for

Child Vietnam Collecting shrimp from the rice paddies and Community members and
nutrition combining with the green tops of sweet program staff
potatoes to make nutritious baby food.
Pregnancy Pakistan Going for antenatal consultation and tetanus Community members
outcomes toxoid injection; increasing food intake
during pregnancy; giving the birth attendant
a clean blade; insisting that nothing be
applied on the umbilical cord after it was cut
and tied.
HIV/AIDS Indonesia Hiding condoms in the storage space of the Program staff
and motorcycle; hiding condom in one’s
reproductive underwear; if caught by his wife with
health condoms, he says that he was approached in
the street by population promoters who gave
them to him, and he didn’t have time to
refuse.
Child Uganda Working longer and harder than others; No one
soldiers respecting aunts, parents, and community
elders; attending school and performing
well.
Health United Cross-referencing records within the Health service staff and
services Kingdom database to speed up data entry time. management
Prisons Denmark Offering inmates a tour of the prison instead Prison staff and management
of a formal interview; inviting inmates on
parole for supper in the prison and eating
together with the guards.

Community Participation. According to Pascale et al. (2010), community

participation should be incorporated into every step of the PD process. During the

analysis I attempted to rate the PD programs and inquiries using a scale of community

participation similar to that developed by Arnstein (1969). However, this proved to be

extremely difficult, as the majority of PD program and PD inquiry reports did not include

a description of the community participation process – only whether or not community

 73
participation had occurred. Therefore, I had to limit the analysis to identifying whether or

not any type of community participation had occurred at all. A case study was marked as

including community participation if it described any type of participation process,

including community consultation meetings, community volunteers, community

members helping to implement activities, etc. In order to be considered participatory, the

community members had to be involved in the implementation of the program or inquiry,

not just be participants in it. For example, a PD inquiry was not considered participatory

if the researchers simply interviewed community members as subjects in the study, but it

was considered participatory if community volunteers worked with the researchers to

conduct the interviews.

Even with this very broad definition of community participation, only four (10%)

PD programs included community participation during all PD steps. This included one

program related to child nutrition, one related to health services, one related to school

education, and one related to child sex trafficking. Three PD programs did not include

participation during any step. This included one program related to child nutrition, one

related to hospital acquired infections, and one related to reproductive health. Figure 4

shows the percentage of PD programs that included community participation in each step

of the process. The lowest level of community participation was found during the first

step to define the problem, with only five PD programs (13%) describing community

participation during this step. The remainder of the programs either did not describe any

community participation, and so the level of participation is unknown, or stated that

people other than community members defined the problem.

 74
Community members were involved in determining the existence of positive

deviants, discovering their behaviors and designing the program in 23 (58%), 25 (63%)

and 26 (65%) PD programs respectively. The step with the highest level of participation

was implementation of the program, with 34 PD programs (85%) including the

community in this step. In most cases participation it this step involved the positive

deviants or other community volunteers spreading the positive deviant behaviors to

people in their community. However, only 21 (53%) of PD programs involved the same

community members in monitoring the success of the program. In most cases community

involvement in the monitoring step involved community members or volunteers

measuring the direct outcome of the program, such weighing children to determine if they

were gaining weight from a PD child nutrition intervention.

Figure 4. Percentage of PD programs that included community participation in each PD

step
 75
Table 5 gives examples of three PD programs with high, medium, and low levels of

participation. High participation was defined as involving the community in all steps,

moderate was defined as involving the community in one to five steps, while low was

defined as not involving the community in any step. The first program on child nutrition

in Guinea included community members in all PD steps. Child nutrition was identified as

a problem by community members, PD mothers were identified by community volunteers

who worked together with program staff to discover PD behaviors, design, implement,

and monitor the program. The second program on pregnancy outcomes in Egypt only

included community members during the implementation of the program. Program staff

identified pregnancy outcomes as a problem in the community, they conducted the PD

inquiry to identify PD mothers, discover their behaviors, and design education sessions.

They invited the PD women and community volunteers to help implement the education

sessions, but did not involve them in the ongoing monitoring of the program. The third

program on reproductive health in Cote d’Ivoire did not include community members at

any stage of the process. The program staff identified reproductive health as an issue,

conducted their own PD inquiry, and designed a mass media campaign, which they then

implemented and monitored.

 76
Table 5

Comparison of Three PD Programs with Different Levels of Community Participation

PD Step Overall level of community participation

High Moderate Low / None
(Child Nutrition, Guinea) (Pregnancy Outcomes, (Reproductive Health, Cote
Egypt) d’Ivoire)
Define Child nutrition was Pregnancy outcomes were Reproductive health for
identified as a high priority identified as a high priority young people was identified
issue by community issue by program staff as a high priority issue by
members program staff
Determine PD mothers were PD women were identified PD young people were
identified by community by program staff based on identified by program staff
volunteers measurements taken at the with assistance from local
health centre NGOs
Discover PD mothers and Program staff discovered Program staff discovered
community volunteers PD behaviors by PD behaviors by designing
discovered PD behaviors interviewing the PD a survey based on
with assistance from women they had identified interviews and
program staff administering the survey to
a large number of young
people
Design PD mothers and Program staff designed Program staff designed a
community volunteers education sessions based mass media campaign based
designed educational on the PD behaviors they on the PD behaviors they
sessions with assistance had identified had identified
from program staff
Implementation PD mothers and PD women and Program staff implemented
community volunteers community health workers the mass media campaign
implemented educational implemented educational
sessions to teach the PD sessions to teach the PD
behaviors to other mothers behaviors to other
with malnourished pregnant women in the
children. Both mothers and community
fathers contributed food
and resources to the
educational sessions
Monitor PD mothers and Program staff and health Program staff conducted
community volunteers centre staff conducted monitoring surveys to
regularly weighed the monitoring to assess the assess the effectiveness of
children to assess the effectiveness of the the program
effectiveness of the program
program
 77
While the majority of PD programs included community participation in at least

some parts of the PD process, the same is not true for PD inquiries. Thirty (94%) of PD

inquiries did not describe community participation at any stage of the process (see Figure

5).

Figure 5. Percentage of PD inquiries that included community participation in each PD

step

Integration. Figure 6 shows 32 PD programs (80%) were part of a larger

integrated program that included a range of different activities, of which PD was only one

part. For example, the PD child nutrition program in India is only one part of the

government run Integrated Child Development Services which includes a wide range of

activities to support early childhood development, including immunization, health

checks, referral services, preschools, health education, and nutrition. Another example of

PD being incorporated into a larger integrated program is a hospital acquired infection

prevention program implemented at the Veterans Affairs Medical Center in the United

States. The PD approach was part of a bundle of interventions implemented at the

 78
hospital, which also included nasal screening of patients, isolation of positive patients,

and standardization of hand hygiene.

3%

17%
PD is part of an
integrated program
PD is a stand-alone
program
Uknown

80%

Figure 6. Percentage of PD programs that were integrated with other program activities

or run as stand-alone programs

The question of integration is not relevant for PD inquiries, as they do not involve

implementing any type of program. Some integrated programs included specific actions

aimed at institutionalizing PD to increase the chance that it would continue to be

implemented after the program finished. In community settings this included training

local government and community workers on the approach, so they could continue to

implement it. For organizationally based programs at hospitals, schools, and prisons this

included securing management support and including PD in the routine policies and

practices of the organization.

 79
Research Question 2

Chronological diffusion. Figure 7 is a timeline of the PD programs and PD

inquiries, while Figure 8 shows the number of new programs and inquiries that started

each year. These two figures illustrate how the PD approach spread to different countries

and topics. From Figure 7 it is clear that the PD approach was first proposed in the 1970s

for child nutrition, but was not implemented for another 1.5 decades at the start of the

1990s. During the 1990s one to two PD programs and inquiries were implemented each

year. These were mainly on child nutrition, although there was one PD inquiry related to

health services in 1995, and one PD program related to female genital mutilation that

started in 1998. The early 2000s were similar, with most programs on child nutrition.

There was one PD inquiry on gender equity in 2000 and one HIV/AIDS program that

started in 2001.

The year 2002 appears to mark a change in this trend. More programs and

inquiries were published on PD in 2002 than in any previous year. This is due to the

publication of a special supplement in the Food & Nutrition Journal focusing only on PD.

After the 2002 publications a larger number of PD programs and inquiries have been run

each year on an increasingly broad range of topics. The first programs on pregnancy

outcomes started in 2002, child trafficking and schools started in 2003, hospital acquired

infections in 2005, and smoking in 2006. The last three years, from 2009 to 2011 saw the

expansion of PD into even more new areas, including diet and weight control, prisoner

wellbeing, and chronic diseases.

 80
Year
Type Country 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11
Inquiry Unspecified Carribean
Program Vietnam
Inquiry India
Inquiry Morocco
Inquiry India
Inquiry China
Program Haiti
Inquiry United States
Program Bangladesh
Inquiry India
Inquiry Bangladesh
Program Bangladesh
Program Egypt
Program Mali
Inquiry Brazil
Program India
Program Guinea
Program Haiti
Inquiry India
Program Cote d'Ivoire
Inquiry Pakistan
Inquiry Vietnam
Inquiry Vietnam
Inquiry Vietnam
Inquiry Rwanda
Inquiry Vietnam
Program Guinea
Program Pakistan
Program Philippines
Program Indonesia
Program Indonesia
Program Senegal
Program Argentina
Program Egypt
Inquiry India
Program Ethiopia
Program Kenya
Program Tajikistan
Program Rwanda
Program Madagascar
Inquiry Mexico
Inquiry Mozambique
Inquiry United States
Program Malawi
Program United States
Program Mali
Program Colombia
Program United States
Program Afghanistan
Program United States
Program Australia
Inquiry Ghana
Inquiry India
Inquiry India
Program Uganda
Program Nigeria
Inquiry United States
Inquiry Indonesia
Inquiry United States
Program Rwanda
Program Brazil
Program United Kingdom
Program Zambia
Inquiry Guatemala
Program Canada
Program Denmark
Inquiry United States
Inquiry Guatemala
Inquiry United States
Inquiry United States
Program Pakistan
Inquiry United States

Key
Child nutrition Female Genital Mutlilation HIV/AIDS & reproductive health Prisoner wellbeing
Child trafficking and soldiers Gender Equity Hospital acquired infections School education
Chronic diseases Health services Pregnancy outcomes Smoking
Diet and weight control

Figure 7. Timeline of PD programs and PD inquiries

 10

9
Number of PD programs /inquiries starting in the year

8 Smoking
School education
7
Prisoner wellbeing
Pregnancy outcomes
6
Hospital acquired infections
5 HIV/AIDS and reproductive health
Health services
4 Gender Equity
Female Genital Mutlilation
3
Diet and weight control
Chronic diseases
2
Child trafficking and soldiers
1 Child nutrition

Figure 8. Number and topic of PD programs/inquiries starting each year

81
 82
Geographic diffusion. In terms of geography, the PD approach started almost

exclusively in low and middle income countries, on a range of different continents.

Vietnam appears to be a key country, as it was the first major implementation of the

approach in the field of child nutrition, and has the longest running program. One study

was conducted in the U.S. in 1995 on nutrition and hemodialysis in a healthcare setting.

However, the PD approach did not really diffuse to the U.S. or other developed countries

until it was first used for hospital acquired infections in 2005. After 2005 the PD

approach was used multiple times in the U.S., and at least once each in Canada, Australia,

and Denmark.

Topic diffusion. From a topic perspective, the PD approach started in child

nutrition, with a significant number of projects being implemented in this area from the

1990s until 2011. Despite its wide application to child malnutrition in low and middle

income countries, no case studies were identified which applied the PD approach to child

overnutrition / obesity in high or middle income countries.

Other topics outside nutrition have been applied a few times at most, with the

exception of hospital acquired infections which has a larger number of programs. There

have been no further applications of PD to female genital mutilation, despite an extensive

project being implemented in Egypt from 1998 – 2006. There have also been no further

applications to child trafficking and soldiers despite two programs being implemented in

the area, starting in the early 2000s.

Diffusion actors. The theory of diffusion of innovations identifies different

groups based on how quickly they adopt a new idea. Innovators are those who create the
 83
new idea, early adopters are those who are first to adopt it in a population, early majority

are the large group of people who adopt it after the early adopters, late majority are the

group who adopt it after the early majority, and laggards are those who resist adopting the

new idea (Rogers, 1995).

For the PD approach Jerry and Monique Sternin could be considered the primary

innovators. While they did not coin the term positive deviance, they did implement the

first large scale PD program and are directly or indirectly (through Save the Children)

linked to at least 15 PD programs and five PD inquiries. This includes at least eight

occasions when PD was applied to a new topic for the first time, such as the first large

scale implementation for child nutrition (Vietnam, 1990-2007), the first applications to

female genital mutilation (Egypt, 1998-2006), pregnancy outcomes (Pakistan, 2002-

2004), HIV/AIDS (Indonesia, 2002), school education (Argentina, 2003-2004),

trafficking (Indonesia, 2003-2008), hospital acquired infections (United States, 2005-

2007) and prisoner and guard wellbeing (Denmark, 2009-2011).

The individuals and organizations who adopted the PD approach after seeing one

of Jerry and Monique Sternin’s early programs could be considered early adopters, and

those that followed them early majority. For child nutrition, the PD approach has spread

from the innovators (the Sternins) to other early adopters in USAID and Save the

Children, and now appears to be spreading through other international development

organizations to become a relatively well established approach. PD in hospital acquired

infections was also started by Jerry Sternin, and quickly spread to other early adopter

hospitals via the Plexus Institute. If this diffusion continues the PD approach for hospital
 84
acquired infections may spread to the early majority, bringing it closer to becoming a

mainstream approach.

By comparison, the application of PD to other topics in most cases has only

spread from the innovators to a small number of early adopters (e.g. pregnancy outcomes,

school education, HIV/AIDS, and reproductive health). In some cases the approach has

not spread at all past the first implementation (e.g. female genital mutilation).

One factor which may have affected the speed of diffusion is the locations where

the PD programs and inquiries were published. The fact that the majority of PD programs

were only described in internal program documents may have slowed the diffusion of the

PD approach significantly, as the only people who normally read internal program reports

are the program staff and donors. Once a large number of programs were published in the

Food and Nutrition Bulletin in 2002 the rate of diffusion increased. The rate of diffusion

may also have been increased by the Positive Deviance Initiative (2010e) website which

aggregates a range of peer reviewed and gray literature on PD, and promotes the

approach for a range of topics.

Diffusion criteria. Theory of diffusion of innovations describes characteristics of

innovations that make them more likely to spread quickly. This includes the relative

advantage compared to other solutions, the compatibility with existing practices and

cultures, the complexity, the ease with which the innovation can be trialed, and how

easily the results can be observed (Rogers, 1995).

The two areas where PD appears to have spread more rapidly - child nutrition and

hospital acquired infections - are both areas where applying the PD approach meets
 85
several of the criteria required for diffusion. The reports from PD programs in child

nutrition and hospital acquired infections show that the results are immediate and easily

observed. For child nutrition programs children put on weight, and in hospital programs

the rate of hospital acquired infections goes down. In both these cases, according to

reports from PD programs, the PD approach is easy to trial and does not require too many

resources. It is simple to implement and it is compatible with the desires and culture of

the majority of people. Because the results of the programs can be easily observed and

compared to other programs the relative advantage of the PD approach in these areas is

easy to see.

Some areas where the PD approach appears to have been used once or twice and

never again (e.g. female genital mutilation, trafficking, child soldiers) do not meet the

same criteria for diffusion. For example, the results are not able to be immediately

observed and the issues are complex so the PD approach may not be easy to implement.

In the case of female genital mutilation, the PD approach is also not compatible with the

local culture. For example, the program evaluation reported many “dissident voices who

consider an intervention against FGM/C as a ‘Western conspiracy’ against Egyptian

traditions and values” (Population Council, 2008, p.6). As a result, positive deviants were

not celebrated by the community in the same way as they were in other programs.

Because of this, the relative advantage of the PD approach over other approaches may not

have been as apparent compared to child nutrition and hospital acquired infections and so

this may have slowed or stopped the diffusion of the PD approach in these areas.
 86
Research Question 3

The peer debriefer and I used qualitative analysis and discussion to develop a

typology of PD approaches in order to address Research Question 3. This process

involved reviewing the case studies of PD programs and inquiries to look at the different

ways in which they could be grouped. Initially we decided to group them by the

methodology used to identify positive deviants and their behaviors (qualitative,

quantitative, mixed-methods, etc). However, after trying to apply these categories to the

case studies we found that it was necessary to separate the methodologies for three

different steps - identifying the positive deviants, discovering their behaviors, and

spreading those behaviors to others. In parallel to this we tried grouping the PD programs

and inquiries by topic. The first grouping resulted in the classifications of child nutrition,

adult health, organizational, and socio cultural. Eventually, through several rounds of

discussion, we decided to modify these groups so they aligned with the levels of

intervention seen in many public health frameworks - that is, individual, organizational /

institutional and social / cultural. See Appendix G for the minutes of these discussions.

The final conclusion after extensive discussion and analysis was that there are

four factors to consider when categorizing the exiting PD approaches: the level of

intervention, the method used to identify positive deviants, the method used to determine

their positive deviant behaviors, and the method used to spread the behaviors to others.

Each case study uses a different combination of the four factors, although some

combinations are more common than others.

Level of intervention. The case studies of PD programs and inquiries were

 87
grouped into those that apply the PD approach to individual behavior change, those that

apply it to organizational or institutional change, and those that apply it to social or

cultural change. These levels are common in public health, and are often portrayed as

concentric circles or layers similar to Figure 9.

Society / culture

Organization / institution

Individual
behavior

Figure 9. Levels of intervention

Out of the 40 PD programs in the sample, 26 (65%) were related to individual

behavior change, and out of the 32 PD inquiries, 29 (91%) were related to individual

behavior change. This includes 20 programs and 18 inquiries related to child nutrition,

and six programs and 11 inquiries related to other health behaviors, including chronic

diseases, diet and weight control, HIV/AIDS and reproductive health, pregnancy

outcomes, and smoking.

There were 11 PD programs (28%) and two PD inquiries (6%) related to

organizational or institutional change. This included programs and inquiries related to

school retention, prisoner and guard wellbeing, health services, and hospital acquired
 88
infections. All these programs and inquiries were in organizational / institutional settings

(schools, hospitals, and prisons), and the focus was to achieve changes at the

organizational level, even if this was related to the behavior of individuals.

There were three PD programs (8%) and one PD inquiry (3%) related to social or

cultural change. This included those related to child trafficking, child soldiers, female

genital mutilation, and gender equity. Although these programs may have included

changing the behavior of particular individuals (e.g. returned child soldiers, parents of

girls at risk of circumcision or trafficking) the overall aim was to change the sociocultural

practices at the community level. Table 6 shows the number of PD programs and

inquiries for each topic, and at each level of intervention.

 89
Table 6

Number of PD Programs and Inquiries for Each Topic and Level of Intervention

Level of # PD # PD
intervention Topic programs inquiries Total
Child nutrition 20 18 38
Chronic diseases 1 1
Diet and weight control 3 3
Individual
HIV/AIDS and reproductive health 1 4 5
behavior
Pregnancy outcomes 4 3 7
Smoking 1 1
Total 20 18 38
Health services 2 2 4
Hospital acquired infections 6 6
Organizational /
Prisoner wellbeing 1 1
institutional
School education 2 2
Total 11 2 13
Child trafficking and soldiers 2 2
Female Genital Mutilation 1 1
Social / cultural
Gender Equity 1 1
Total 3 1 4

Figure 10 shows the timeline of PD programs and inquiries grouped by the level

of intervention. It is clear that PD first started at the individual behavior change level in

child nutrition, and has only spread to other individual behavior change topics since

2000. With the exception of one early study on health services in 1995, PD has only been

applied to organizational / institutional level change since 2003, when it was first applied

in a school setting. PD was first applied to social / cultural change in 1998 but has not

spread to a large number of programs in that area.

 90
Individual behavior change
Year
Type Country 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11
Inquiry Unspecified Carribean
Program Vietnam
Inquiry India
Inquiry Morocco
Inquiry India
Inquiry China
Program Haiti
Program Bangladesh
Inquiry India
Inquiry Bangladesh
Program Bangladesh
Program Mali
Program India
Program Guinea
Program Haiti
Inquiry India
Program Cote d'Ivoire
Inquiry Pakistan
Inquiry Vietnam
Inquiry Vietnam
Inquiry Vietnam
Inquiry Rwanda
Inquiry Vietnam
Program Guinea
Program Pakistan
Program Philippines
Program Indonesia
Program Senegal
Program Egypt
Inquiry India
Program Kenya
Program Tajikistan
Program Rwanda
Program Madagascar
Inquiry Mexico
Inquiry Mozambique
Inquiry United States
Program Malawi
Program Mali
Program Afghanistan
Program Australia
Inquiry Ghana
Inquiry India
Inquiry India
Program Nigeria
Inquiry United States
Inquiry Indonesia
Inquiry United States
Program Rwanda
Program Zambia
Inquiry Guatemala
Inquiry Guatemala
Inquiry United States
Inquiry United States
Inquiry United States Year
Type Country 76 77 78 79 80 81 82 83 84 85 86 87 Organizational
88 / institutional
89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11
Inquiry United States
Program Argentina
Program Ethiopia
Program Colombia
Program United States
Program United States
Program United States
Program United Kingdom
Program Brazil
Program Canada
Program Denmark
Program Pakistan
Inquiry United States Year
Type Country 76 77 78 79 80 81 82 83 84 85 86 87 88 89 Social
90 9192 93/ cultural
94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11
Program Egypt
Inquiry Brazil
Program Indonesia
Program Uganda

Individual behavior change Organizational / institutional Social / cultural

Child nutrition HIV/AIDS & reproductive health Hospital acquired infections Female Genital Mutlilation
Chronic diseases Pregnancy outcomes Health services Child trafficking and soldiers
Diet and weight control Smoking Prisoner wellbeing Gender Equity
School education

Figure 10. Timeline of PD programs and PD inquiries grouped by level of intervention

 91
Method used to identify positive deviants. After the level of intervention, the

next factor used to classify PD approach was the method used to identify positive

deviants. Four different methods have been used to identify positive deviants in the PD

case studies. The most common method was quantitative screening in which participants

are assessed against specific quantitative criteria to determine whether they are positive

deviants. For example, in the case of child nutrition programs, children are weighed and

measured, and the socioeconomic status of their family is assessed. Those who are

growing normally despite having low socioeconomic status are considered positive

deviants. This was the most common approach overall, particularly for programs and

inquiries addressing issues of individual behavior change. Twenty five (66%) of child

nutrition programs and inquiries used this approach, and 34 (67%) of all individual

behavior change programs and inquiries used it.

The next most common method used to identify positive deviants is individual

choice or convenience sampling. Using this approach, a key informant such as the village

chief or health worker is asked to identify individuals or families who they believe are

positive deviants. Two (50%) programs and inquiries addressing social / cultural issues

used this approach, as did 3 (30%) of the programs and inquiries addressing

organizational / institutional issues.

Some programs use a more rigorous qualitative screening to identify positive

deviants. This is more systematic than individual choice or convenience sampling. For

example, in some case studies large numbers of interviews were conducted with members

from the target group and systematically analyzed to identify positive deviants. This is a
 92
relatively uncommon method, only used by five (7%) of all programs.

Finally, six (50%) of the programs that addressed organizational / institutional

level changes used staff meetings as the method for identifying positive deviants. Staff

would meet to discuss who among the staff could be considered a positive deviant based

on their performance (e.g. having a high student retention rate).

Method used to discover positive deviant behaviors. Four different methods

have been used to discover the unique behaviors of positive deviants once they have been

identified. The most common method, used in 50 (75%) of PD programs and inquiries,

was a qualitative inquiry using interviews, focus groups, and/or observations of the

positive deviants. This method was used for 41 (80%) of all programs and inquiries

addressing individual behavior change, five (42%) of those addressing organizational /

institutional issues and four (100%) of those addressing social / cultural issues.

The next most common method was a quantitative survey. This involves

conducting a survey of both positive deviants and normal individuals, and using

statistical analysis to determine which behaviors are related to positive deviant status.

Using this method, the range of possible positive deviant behaviors is predefined by the

researcher or program staff when they design the survey, and so there is no opportunity

for identifying new behaviors that the researcher or program staff may not have

considered. This method was not used for any PD programs, but was used for 10 (31%)

of all PD inquiries. For many PD inquiries, the quantitative screening and analysis was

used on secondary data to identify retrospectively which behaviors were associated with a

positive outcome.
 93
One case study used a combination of qualitative and quantitative methods to

determine positive deviant behaviors. This involved interviewing a small number of

positive deviants. The behaviors that were identified were turned into a quantitative

survey, which was then distributed to the population to validate whether the behaviors

identified were actually related to positive deviant status.

Finally, six (50%) of the programs that addressed organizational / institutional

level changes used staff meetings as the method for identifying PD behaviors. Once staff

had identified who among the staff was a positive deviant based on their performance

(e.g. having a high student retention rate), they would then discuss what their positive

deviant behaviors were (e.g. specific teaching methods).

Table 7 and Table 8 show the number of PD programs and inquiries respectively

that used each combination of methods to identify positive deviants and discover their

behaviors.
 94
Table 7

Methods Used to Identify Positive Deviants and Their Behaviors by Level of Intervention

for PD Programs

Method used to Level of intervention

Identify positive Discover behaviors Individual Organizational Social /
deviants behavior / institutional cultural Total
Individual choice /
Qualitative inquiry 6 2 2 10
convenience
Qualitative screening Qual.& Quant. 1 1
Quantitative screening Qualitative inquiry 15 2 1 18
Staff meetings Staff meetings 6 6
Unknown Unknown 4 1 5

Table 8

Methods Used to Identify Positive Deviants and Their Behaviors by Level of Intervention

for PD Inquiries

Method used to Level of intervention

Identify positive Discover behaviors Individual Organizational Social /
Total
deviants behavior / institutional cultural
Individual choice /
Qualitative inquiry 7 1 8
convenience
Qualitative screening Qualitative inquiry 3 1 4
Qualitative inquiry 10 10
Quantitative screening
Quantitative survey 9 1 10

A qualitative analysis of the methodology sections of PD programs and PD

inquiries found that the majority failed to describe the relationship between program staff

/ researchers and community members. In particular, they did not describe this
 95
relationship during the process of identifying PD behaviors. For example, in one case

study, community members found that positive deviant mothers with well-nourished

children also cut their children’s hair regularly. The program staff probably knew that

cutting the children’s hair was not likely to be related to their nutrition, but it is unclear

whether they used this knowledge to guide community members away from hair cutting

as a PD behavior, or whether they allowed hair cutting to be considered a PD behavior.

Although according to Pascale et al. (2010) community members should identify the PD

behaviors themselves, it seems likely that some filtering may occur by program staff to

guide participants to select those behaviors which fit with a western theory of medicine.

However, this filtering process, or its absence, was not described in the majority of

program reports and studies.

Method used to spread PD behaviors to others. Five different methods were

identified for spreading positive deviant behaviors to others in the community (see Table

9). The most common method, used by 19 (95%) child nutrition programs, two other

programs targeting individual behaviors, and one organizational level program, is action

learning sessions (usually referred to as “Hearth” sessions in child nutrition – the hearth

is a fireplace/oven where women meet together to prepare food). Action learning sessions

involve the positive deviants practicing their behaviors with other members of the

community so that the other members can learn them. For example, in most of the child

nutrition programs, Hearth sessions typically run for a period of two weeks. During the

sessions, positive deviant mothers teach the other mothers how to cook nutritious meals

using locally available ingredients they have discovered (for example, in Pakistan
 96
cooking porridge with Shea butter was a PD behavior, while in Vietnam it was shrimp

and sweet potato tops). They also deliver health education sessions on key topics such as

breastfeeding and hand washing. During each session mothers must contribute food and

cook a nutritious meal and snacks for their children, which they eat at the session in

addition to their regular meals. The mothers learn the positive deviant behaviors from the

PD mother by practicing them together with their children.

Six programs, including some addressing individual behavior change and social /

cultural issues, used peer educators or community volunteers to spread the PD behaviors.

The peer educators and / or community volunteers would learn the positive deviant

behaviors and then visit other members of the target group, often in their homes or in

groups, to teach them the behaviors.

Two programs used mass media campaigns (e.g. TV spots, radio advertisements,

events, etc) to spread the behaviors, while one program on female genital mutilation used

advocacy activities targeting community leaders and decision makers. The majority of

programs for organizational / institutional change used staff meetings or on-the-job

training, mentoring or demonstrations to spread the behaviors to others.

 97
Table 9

Method Used to Spread PD Behaviors by Level of Intervention

Level of intervention
Individual Organizational Social /
Method used to spread PD behaviors behavior / institutional cultural Total
Advocacy campaign 1 1
Action learning sessions (incl. Hearth) 21 1 22
Mass media campaign 1 1 2
Peer educators or community
volunteers 4 2 6
Staff meetings or on-the-job training 8 8
Unknown 1 1

Combinations. Table 10 shows the combinations that have been used for

identifying, discovering, and spreading PD behaviors in each of the intervention levels.

The most common combination, used by 13 (33%) of the programs (11 for child

nutrition), was to identify positive deviants using quantitative screening against specific

criteria, to identify their behaviors using qualitative interviews and focus groups, and to

spread their behaviors to others using action learning (Hearth) sessions. This approach

matches the steps proposed by Pascale et al. (2010) exactly.

The next most common combination, which was only used in organizational

settings, was to use staff meetings to identify positive deviants and determine their

behaviors, and staff meetings or on-the-job training or demonstrations to spread the

behaviors to others.
Table 10

Method Used to Identify, Discover, and Spread PD Behaviors by Level of Intervention

Method used to Level of intervention

Individual Organizational Social /
Identify positive deviants Discover behaviors Spread behaviors Total
behavior / institutional cultural
Advocacy campaign 1 1
Action learning sessions (incl. Hearth) 5 5
Individual choice / convenience Qualitative inquiry Mass media campaign 1 1
Peer educators or community volunteers 1 1 2
Staff meetings or on-the-job training 1 1
Qualitative screening Qual. & Quant. Mass media campaign 1 1
Action learning sessions (incl. Hearth) 12 1 13
Quantitative screening Qualitative inquiry Peer educators or community volunteers 3 1 4
Staff meetings or on-the-job training 1 1
Staff meetings Staff meetings Staff meetings or on-the-job training 6 6
Unknown Unknown Action learning sessions (incl. Hearth) 4 4
Unknown 1 1

98
 99
Research Question 4

Number and type of evaluations. Thirty PD programs (75%) included some type

of quantitative evaluation and 15 programs (38%) included some type of qualitative

evaluation. Of these, 12 (30%) included both quantitative and qualitative evaluation in a

mixed methods approach. Table 11 shows the number and type of quantitative and

qualitative PD program evaluations that were identified by topic, and by level of

intervention. Child nutrition programs have the largest number of quantitative and

qualitative evaluations, including the only randomized controlled trial. For some topics,

such as diet and weight control, gender equity, and chronic disease, no PD program

evaluations were identified. Table 12 shows the number of quantitative and qualitative

PD program evaluations by the method used to identify, discover and spread positive

deviant behaviors. The largest number of evaluations (14, 31%) have been conducted on

programs that used quantitative screening to identify positive deviants, qualitative inquiry

to identify their behaviors, and action learning (e.g. Hearth) sessions to spread the

behaviors to others. The next largest number (9, 20%) were on programs which had the

same approach, but which used individual choice / convenience sampling instead of

quantitative screening to identify positive deviants. All other methodological

combinations have five or fewer evaluations.

Table 11

Number and Type of PD Program Evaluations by Topic and Level of Intervention

Quantitative evaluations
Pre and
Pre and posttest,
Randomized posttest, Pre and
Level of unmatched Posttest Qualitative
Topic Controlled matched posttest, Total
intervention comparison only evaluations
Trial comparison no control
group
group
Child nutrition 1 3 2 9 1 10 26
Chronic diseases
Diet and weight control
Individual HIV/AIDS and reproductive
behavior health 1 1
Pregnancy outcomes 2 1 1 4
Smoking 1 1
Total 1 6 3 11 1 10 32
Health services 1 1 1 3
Organization Hospital acquired infections 1 3 4
al / School education 1 1
institutional
Prisoner wellbeing 1 1
Total 1 1 5 2 9
Child trafficking and soldiers 1 2 3
Social / Female genital mutilation 1 1
cultural Gender Equity
Total 1 3 4
Total 1 7 4 17 1 15 45

100
Table 12

Number and Type of PD Program Evaluations by Method Used to Identify, Discover, and Spread Positive Deviant Behaviors

Method used to: Quantitative evaluations

Pre and Pre and
posttest, posttest,
Identify Randomized matched unmatched Pre and
positive Discover Controlled compariso comparison posttest, Posttest Qualitative
deviants behaviors Spread behaviors Trial n group group no control only evaluations Total
Advocacy campaign 1 1
Action learning sessions 1 1 3 4 9
Individual
Qualitative Mass media campaign
choice /
inquiry
convenience Peer educators or
community volunteers 2 1 3
Staff meetings 1 1
Qualitative Qual. &
Mass media campaign 1 1
screening Quant.
Action learning sessions 3 3 3 5 14
Quantitative Qualitative Peer educators or
screening inquiry community volunteers 1 2 2 5
Staff meetings 1 1
Staff Staff
meetings meetings Staff meetings 3 1 4
Action learning sessions 3 1 1 5
Unknown Unknown
Unknown 1 1
Total 1 7 4 17 1 15 45

101
 102
Of the 30 quantitative evaluations identified, a large number use relatively weak

study designs, including 17 evaluations that used a pre and posttest without any type of

comparison or control group, and one that used posttest only. Out of the 15 qualitative

evaluations identified, 10 were for child nutrition, with only a small number on other

topics.

Only four PD programs included any type of long term evaluation to assess the

sustainability of the impact after the PD program had finished being implemented. Of

these, three were for child nutrition, and one was related to child sex trafficking.

Evaluation outcomes. All of the quantitative evaluations except one showed at

least some positive impacts from the PD program, even if they were not as extensive as

hoped, or did not cover all target groups. The only exception was one study on a child

nutrition program in Rwanda which showed no statistically significant difference

between the intervention group and unmatched comparison group. Examples of positive

outcomes reported from quantitative evaluations are shown in Table 13. Many of the

quantitative evaluations reported improvements that were not only statistically

significant, but also large enough to have a substantial impact on program beneficiaries.
 103
Table 13

Examples of Positive Outcomes Reported from Quantitative Evaluations of PD Programs

Topic Country Evaluation Results

design
Child nutrition Vietnam Pre and posttest, The prevalence of severe malnutrition decreased from
matched 23% to 6% (p<0.001); the prevalence of normal
comparison group weight-for-age increased from 42% to 56% (p<0.001).
Younger siblings of participant children had better
weight for age scores compared to the comparison
group (p < .021).
Hospital Brazil Pre and posttest, There was more than a 2-fold difference in the number
acquired matched of alcohol gel aliquots dispensed per month from
infections comparison group before to after the PD intervention. The rate of hospital
acquired infections per 1,000 device days decreased
from baseline to the end of the PD program in the
intervention unit (p<.05).
Pregnancy Egypt Pre and posttest, The mean birth-weight was significantly higher at
outcomes matched endline compared to baseline in both target and
comparison group comparison groups. However, the increase in birth-
weight from endline to baseline was significantly
higher (p<0.01) in the target areas.
Child Uganda Pre and posttest, 6.4 % of girls said they regained self-respect, self-
trafficking and no comparison esteem, and confidence by participating in the PD
soldiers group program. 92.7% said they have been identified as ‘role
models’ by community members.
Health United Pre and posttest, Time savings of between 1 and 2.5 hours per staff
services Kingdom no comparison member per week were made after implementing the
group PD program. Time savings of up to 30 per cent of a
day for social workers were also made by transferring
call handling to administration, which was part of the
PD program.
HIV/AIDS Cote Pre and posttest, Perceived peer support increased from 47% to 55%.
and d'Ivoire no comparison Percent that discussed sexual abstinence with either
reproductive group parent increased from 14% to 23%. Percent that
health reported consistent condom use remained the same
overall, but increased relative to campaign exposure
(17% low exposure, 22% moderate exposure, 26%
high exposure)
School Argentina Pre and posttest, The school where the PD program was implemented
education no comparison reported 45%-50% increase in retention rates within
group one school year.
Smoking Australia Pre and posttest, Eleven of 14 long-sentence prisoners who participated
matched in the program remained abstinent for at least 12
comparison group months, compared with 16 out of 41 patients similarly
followed-up at other sites. The number of smokers
making efforts to quit smoking on their own was
significantly higher at the intervention site.
 104
All qualitative evaluations which investigated the impact of the PD programs

reported that participants and other stakeholders felt the program had a positive impact.

Some qualitative evaluations did not focus on the programs’ impact or effectiveness, but

instead focused on identifying successes and challenges associated with the

implementation of the PD program. Example of findings from child nutrition programs

include that when mothers are required to contribute their own food for the communal

meal cooked during Hearth education sessions this can be a barrier to participating in the

program, while the immediate changes they observe in their child (including looking

visibly healthier and more playful) during the program are a motivator to continue

participating. Several qualitative evaluations found that the PD behaviors identified were

often similar every time that the PD inquiry was implemented. However, a number also

recommended continuing to implement the PD inquiry every time, as the real benefit of

the PD inquiry is the process of implementing it where community members identify

their own positive deviants, rather than the specific positive deviant behaviors identified.

Three of the four long term evaluations to assess the sustainability of the PD

program also reported at least some positive outcomes – both in terms of impact on

health, and the continuation of PD behaviors. One long term evaluation assessed the

effectiveness of a child nutrition program in Indonesia four to five years after it had been

run. The evaluation included a follow-up study of 103 children who had previously

participated in the program, as well as a sibling study of 448 former participants and 62

younger siblings from five intervention communities and three control communities. The

purpose of studying younger siblings of program participants is to assess whether their

 105
parents continued to implement the positive deviant behaviors, such as feeding children

extra snacks during the day, they had learned with their next children who were born after

the program. The results of the study showed no significant differences between children

who had participated in the program and comparison communities. However, there were

statistically significant differences for their younger siblings (p<.05). In the intervention

communities the rate of moderate and severe malnutrition among younger siblings was

22% and 9.8% respectively, while among comparison communities it was 43% and 9.5%.

A similar study conducted in Rwanda two to three years after program implementation

found no significant differences between the intervention and comparison communities

for either participants or younger siblings. A third study using the same method on a child

nutrition program in Vietnam also showed no statistically significant differences for

participants, but did show that the weight for age Z-scores of younger siblings in the

intervention community were better than for the comparison community (age-adjusted

mean WAZ –1.82 versus –2.47, respectively, p < .021). A range of behaviors such as

breastfeeding and hand washing were more common in the intervention communities

than in the comparison communities.

The final long term evaluation was a qualitative evaluation of a child trafficking

program in Indonesia, using interviews, focus groups, and participatory sketching in

which participants in the program drew pictures to represent the changes they had

experienced. The qualitative evaluation was conducted on a village that had implemented

the PD program five years previously. The village was specifically selected because it

had shown the most significant improvement of the three pilot areas. The results of the
 106
evaluation showed that some substantial changes had occurred since the PD program was

implemented. This included that no new girls had been trafficked out of the village in the

five year period, there had been twenty averted cases of trafficking, the use of travel

papers was now strictly enforced, a community watch group had been set up, and a girls’

club had been established. Participatory sketches by community members emphasized the

ability for girls to work outside the village safely, and greater community interaction.

According to the interviews the community had started using PD for other development

issues, in addition to trafficking.

No case studies discussed the environmental sustainability of the PD behaviors. In

some cases it is possible that when the PD behaviors are spread to large numbers of other

people they may not be sustainable in the long term. For example, in Vietnam one PD

behavior was collecting shrimp from the rice paddies to feed to children. This is

sustainable if only a small number of mothers are collecting these shrimp. However, it is

possible that when large numbers of mothers learn this behavior they will all collect

shrimp, and so the numbers of shrimp may be reduced to a point where they are no longer

available.

Quality assessment. Apart from looking at the overall study design, assessing the

rigor and quality of the evaluations was extremely difficult, as only seven quantitative

evaluations out of 30 (23%) had enough information available to complete the quality

assessment tool, and only four qualitative evaluations out of 15 (27%) had enough

information.

Table 14 shows that the majority of evaluation reports had incomplete

 107
descriptions of the setting and population, and many had incomplete descriptions of the

PD approach and evaluation methodology used, which meant that the quality assessment

tools could not be completed. Only nine evaluation reports had enough information

available in all sections to complete the quality assessment tools.

Table 14

Number of Evaluations with Complete, Incomplete, and Missing Report Sections

Report section Complete Incomplete Missing

Setting and population 8 23 2
Description of PD approach 12 18 3
Description of evaluation methodology 12 13 8
Description of evaluation results 14 19

Even when the evaluations did have enough information available to complete the

quality assessment tools, it was difficult to complete the tools since they were designed to

assess scientific research studies. The program evaluations used a range of methods that

would be considered acceptable and appropriate for a program evaluation in the field, but

which do not meet the criteria for scientific research. For example, a program evaluation

conducted solely by program staff typically does not require the same ethical

considerations as a research study. However, this is a criterion in the quality assessment

tool for qualitative studies.

Therefore, a decision was made to use a slightly flexible interpretation of the

quality assessment tools so that each evaluation could be assessed based on what it

claimed to be, rather than assessing it strictly against research criteria (see the minutes in

Appendix G for details of this discussion). The peer debriefer and I first decided to
 108
identify what the evaluation was intended to be used for. For example, evaluations

published in peer reviewed journals were intended to be used as research, while

evaluations published in program reports were intended to be used by field staff to

improve the program implementation on the ground, but were not intended as research.

Once we had agreed what the purpose of the evaluation was we then allowed some

flexibility in the quality assessment criteria based on that purpose. For example, one of

the quantitative quality assessment criteria was “Was (were) the outcome assessor(s)

aware of the intervention or exposure status of participants?” For evaluations done for the

purposes of research the answer to this question is very important for assessing the

overall quality of the research. However, evaluations done for the purposes of field work

almost always involve program staffs who are naturally aware of the intervention status

of program beneficiaries. Therefore, we decided not to mark down the quality of field

evaluations too much if they did not meet this criterion, as long as they acknowledged it

as a weakness and included reflection on how the involvement of the program staff may

have influenced the results.

Each evaluation was given an overall rating based on a combination of the

individual ratings for subsections in the quality assessment tools (see Appendices E and F

for details of the subsections). Evaluations received an overall rating of strong if none of

the subsections contained a weak of unknown rating. Evaluations received overall rating

of moderate if the subsections contained one weak or unknown rating, while evaluations

with two or more weak or unknown ratings were given an overall rating of weak. Using

this approach, three of the seven quantitative evaluations with enough information
 109
available to complete the tools were rated strong, while four were rated moderate. One of

the four qualitative studies with enough information available to complete the tool was

rated as strong, one was rated as moderate, and two were rated as weak (see Table 15).

Table 15

Quantitative and Qualitative Assessment Ratings for Evaluations with Enough

Information to Complete the Quality Assessment Tools

Evaluation Quality Assessment Rating

Topic Country
Quantitative Qualitative
Child nutrition Bangladesh Moderate Moderate
Child nutrition Bangladesh Moderate Not applicable
Child nutrition Rwanda Moderate Not applicable
Child nutrition Vietnam Strong Weak
Female Genital Mutilation Egypt Not applicable Strong
Pregnancy outcomes Senegal Strong Not applicable
Pregnancy outcomes Egypt Strong Not applicable
Child trafficking Indonesia Not applicable Weak
Hospital acquired infections Brazil Moderate Not applicable

A number of methodological flaws were identified in multiple studies. None of

the quantitative evaluations clearly stated to which populations the results of the study

could be generalized. It was unclear whether they were generalizing to other people in the

same community who met those specific criteria, or to other communities or settings.

This issue has a significant impact on how the quality of the statistical analysis is

assessed.

For quantitative evaluations the unit of allocation and unit of analysis were often

unclear. Most evaluations (even those related to individual behavior change) assigned
 110
entire villages to the intervention and control groups, rather than assigning individuals.

However, the same evaluations used the individual as the unit of allocation and analysis.

Since all members of a village do not act independently, the unit of allocation and

analysis should technically have been the village. This would dramatically reduce the

sample size and power of the analysis in these studies.

The methods used in the qualitative evaluations were relatively poor, as only one

out of four received a rating of strong. Examples of methodological flaws included

inappropriate sampling – such as only selecting interviewees from areas where the

program was known to be successful – as well as failing to provide a description of the

context, role of the researcher, or detailed interview results. In particular, one evaluation

which claimed to be qualitative actually used quantitative analysis of very small sample

sizes, which is a significant methodological flaw.

Overall, the majority of program outcome evaluations are relatively weak. Most

used a weak study design and had incomplete reports which made it difficult to assess

their quality. Of those that did have complete reports the quality ranged from weak to

strong on an established quality assessment tool. Based on these results, there is

substantial evidence to support the effectiveness of the PD approach in child nutrition,

but very little high quality evidence to support its effectiveness in other areas.

Evidence of Quality

Quality was assured using a peer debriefing and auditing process. The peer

debriefer conducted an audit of the first five program case studies, followed by three

(10%) of all remaining program case studies, and three (10%) of all inquiry case studies,
 111
which were randomly selected using a random number generator. A total of 11 records

were reviewed by the peer debriefer. The peer debriefer provided a total 25 comments on

nine of the records reviewed. Two records had no comments. I discussed each comment

with the peer debriefer. In 13 of the cases we decided not to make any changes to the

record as the comment was simply to clarify or confirm something in the text. In the

other 12 cases corrections were made to the record. In most of these cases the corrections

were minor (e.g. slight change to the start date or program title) as a result of different

interpretations of the documents – particularly for documents where the language was

unclear or details were missing. In some cases the audit comments resulted in a change to

the process of data entry. For example, in cases where multiple subprograms were

included in one program we decided to use only the most comprehensive subprogram

documents to complete the case study, which was noted in the comments for the case

study. The detailed results of the audit and corrective actions are shown in the audit log in

Appendix K.

Throughout the entire research process the peer debriefer and I had regular phone

calls to discuss emerging patterns, trends, and ideas. This included discussing the

categorization of PD approaches, the pattern of diffusion of the PD approach, and

philosophical / conceptual issues raised by the case studies. This process was essential for

ensuring the quality of the study, as it allowed me to discuss my ideas with an

independent person who could then provide alternative viewpoints and suggestions,

which I then incorporated into my analysis. For example, to answer Research Question 3

I initially categorized the PD programs and inquiries by the methodology used to identify
 112
the positive deviants and their behaviors. When faced with the same problem of

classification the peer debriefer decided to classify them by topic and level of

intervention. Both classifications systems proved useful for understanding the data, and

so ultimately we decided to combine both into the final classification system presented in

the results of this study. This process was particularly fruitful because my background is

in international development in low and middle income countries, while the peer

debriefer’s background is in NGO management in high income countries. Between us we

were able to bring two very different perspectives to the analysis. Minutes of all peer

debriefing meetings can be found in Appendix G.

The quantitative quality assessment tool and qualitative quality assessment tool

were completed for all program evaluations by the researcher. The peer debriefer

independently completed the quality assessment tools for all program evaluations which

had enough information to fully complete the tool. The peer debriefer and I then

compared our overall quality ratings of each study and resolved any discrepancy in the

ratings through discussion. Of the seven quantitative evaluations with enough

information available to complete the tool, four received the same rating from the peer

reviewer and I, while three received different ratings. Of the four qualitative evaluations

with enough information available to complete the tool, three received the same rating,

while one received a different rating. All of the discrepancies except for one were related

to how strictly the criteria in the quality assessment tools should be applied, particularly

for field evaluations that were not intended to be research. This issue is discussed in

detail in the previous section on quality assessment. For example, in one case I applied
 113
the criteria for Q6, Q7, and Q8 in the qualitative quality assessment tool very strictly

since they were not explicitly address in the evaluation report, while the peer reviewer

assumed that they were implied by the study design and did not need to be explicitly

stated as it was a field evaluation rather than a research study. After discussion it was

agreed to give an overall rating of strong, given that the evaluation was a very high

quality field evaluation even though it did not strictly meet all the criteria for scientific

research. The only case where the discrepancy was not related to a difference in the

strictness of the criteria was one on hospital infections where the peer debriefer identified

a flaw in the study design which I had overlooked. A log of the quality assessment tool

ratings and discussion can be found in Appendix L. Overall the majority of quality

assessment ratings were aligned between the peer debriefer and I. Rather than indicating

a lack of quality in this study’s methodology, these discrepancies show the difficulty of

applying scientific research quality assessment tools to field evaluations, and the need for

more research grade evaluations of PD programs.

In conclusion, the results show that most PD programs and evaluations were

conducted for child nutrition in middle and low income countries. The approach has only

recently diffused to other topics and high income countries, and there were few

evaluations available for these programs. Four factors were identified which differentiate

between PD approaches: the level of intervention; the method used to identify positive

deviants; the method used to discover their behaviors; and the method used to spread the

behaviors to others. The following section discusses the implications of these findings.
 114
Chapter 5: Discussion, Conclusions, and Recommendations

Overview

This study used multiple case study analysis based on publicly available

documents to identify, describe, and analyze 40 case studies of past and present

community health programs that used the PD approach and 32 cases studies of PD

inquiries. Analysis of these case studies provides an overall picture of how, where, and

when the PD approach has been used and its diffusion through community health

practice. The analysis has also been used to develop a categorization of PD approaches

currently in use and to assess the qualitative and quantitative evidence for their

effectiveness. The ultimate aim of the study was to provide evidence for or against the

wider use of the PD approach in mainstream community health practice.

The results of the study showed that the majority of PD programs and evaluations

have been conducted in the area of child nutrition in middle and low income countries. It

is only relatively recently that the PD approach has diffused to other topics and to high

income countries. In some cases the rate of diffusion can be linked to the presence of

specific elements from the theory of diffusion of innovations, such as the Sternins who

have acted as innovators by pushing the PD approach into new areas. Four factors were

identified that can be used to classify positive deviance approaches: (a) the level of

intervention; (b) the method used to identify positive deviants; (c) the method used to

discover their behaviors; and (d) the method used to spread those behaviors to others.

Many different combinations of these factors have been used, although some are more

common than others. Finally, with the exception of child nutrition, relatively few
 115
evaluations of the PD approach were identified, and even fewer could be classified as

high quality research. The rest of this section discusses the implications of the findings

and recommendations for future actions.

Current Use of PD Approach in Community Health Practice

The idea that individuals may hold the solutions to long standing problems has a

natural appeal to it which motivates and inspires people. Successful applications of the

PD approach make for good stories, in which the heroes are local people and the

solutions were “right under their nose.” This natural appeal of PD was described by

Lewis (2009) when she first heard about the approach:

The concept of ‘positive deviance’ (PD) captured my imagination and the

hearts of many other alumni of the Oxford University/HEC Paris

Consulting… Jerry and Monique’s passion, humour and humility engaged

us, as did their powerful stories of what PD had achieved. (p. 282)

The powerful stories of what PD has achieved can be seen in the current and past

applications of PD to community health practice, in particular, the first applications of the

PD approach to child nutrition in low and middle income countries, which were highly

successful. Even today, the majority of PD programs and inquiries identified as part of

this study were for child nutrition programs implemented in low and middle income

countries by international development organizations such as USAID and Save the

Children. This is in line with previous literature describing the history of the PD approach

(Pascale et al., 2010; Positive Deviance Initiative, 2010e.; Sternin, 2002). Given that most

PD programs and inquiries were for child malnutrition, it is surprising that no PD

 116
programs or inquiries were found for child overnutrition or obesity. The potential

application of the PD approach to this area is also not discussed in any of the previous

literature.

Apart from child nutrition, other individual behavior change issues to which the

PD approach has already been applied are smoking, pregnancy outcomes, HIV/AIDS and

reproductive health, diet and weight control, and chronic diseases. While most of these

topics have been mentioned in previous PD literature (Pascale et al., 2010; Positive

Deviance Initiative, 2010e), they had far fewer PD programs and inquiries than child

nutrition, and so it is reasonable to say that the PD approach has not become firmly

established in these areas. It is interesting that some of the major health topics in

developed countries, such as diet and weight control, chronic diseases, and smoking, had

very few PD programs and inquiries, even though they would be obvious targets, because

(like child nutrition) they are related to individual behaviors and have easily measurable

outcomes. The fact that the PD approach has not been used widely for many issues

related to high income countries may be related to its origins with international

development practitioners such as Wishik and Van Der Vynckt (1976) and Jerry and

Monique Sternin (Pascale et al., 2010). International development practitioners mainly

work in low and middle income countries, and so it will take time for frameworks used in

this field to diffuse into developed countries.

Another group of topics to which PD has been applied are those in organizations

or institutional settings. This group included programs run in schools, hospitals, and

prisons, such as improving school retention rates, reducing hospital acquired infections,
 117
and improving prisoner wellbeing. While all of these topics are mentioned in previous PD

literature (Pascale et al., 2010; Positive Deviance Initiative, 2010e) there are questions

about whether the PD approach is truly a unique approach in these situations or whether

it is really another name for existing organizational change frameworks. This will be

discussed further in later sections.

The third group of topics to which PD has already been applied is those involving

more complex sociocultural issues or social determinants of health. These included child

trafficking and child soldiers, female genital mutilation, and gender equity. Relatively

few PD programs and inquiries had been conducted on these topics, although some were

been quite extensive. As with organizational change, it is possible that the PD approach

for social/cultural change may actually be another name for existing frameworks, even

though they appear in previous PD literature (Pascale et al., 2010; Positive Deviance

Initiative, 2010e). This issue will be discussed further in later sections.

Local Involvement and Community Participation

It is interesting to see that the PD approach has mainly been implemented in low

and middle income countries, and only recently was implemented in a meaningful way in

high income countries. This diffusion process runs counter to many other approaches

which start in developed countries and are applied to developing countries. However,

despite the fact that it started in low and middle income countries, the majority of PD

programs are still implemented by international organizations, such as USAID, rather

than local organizations or government.

Given that the majority of PD programs are implemented by international

 118
organizations, involving the local community in the development, implementation and

monitoring of the program is particularly important. According to most descriptions of

PD found in the literature (Bradley et al., 2009; Devane, 2009; Marsh et al, 2004; Marsh,

Sternin et al., 2002; Pascale et al., 2010; Positive Deviance Initiative, 2009; Save the

Children, n.d.; Schooley & Morales, 2007; Walker et al., 2007) the PD approach is a

form of participatory action research that should fully involve the community and be

consistent with the principles of endogenous development and participatory evaluation

described by Minkler and Wallerstein (2008). However, the results of this study show

that the actual level of community participation in PD programs and inquiries is far lower

than this, even when a very broad definition of community participation is used. Almost

no programs or inquiries involved the community in the first step of the PD approach:

defining the problem. Involving the community in this step is critical for achieving

citizen control, which is the highest rung on Arnstein’s (1969) ladder of community

participation. Without participation in the first step, PD programs can only reach lower

rungs on the ladder, such as delegated power, partnership, placation, consultation, and

informing. A probable reason for the lack of community participation in many programs

is the time and resource pressures faced by staff. Involving the community is a time

intensive process that can often lead to delays in a project or changes to the program

design and goals. This unpredictability does not fit well with many donors who require

projects to be implemented within a fixed timeframe and to deliver predefined outputs.

Community participation in PD programs was still higher than in PD inquiries.

The majority of PD programs included community participation in at least one step

 119
(usually implementation), while almost no PD inquiries included community

participation in any step. Therefore, at best, the PD inquiries only reached the level of

“tokenism” on Arnstein’s (1969) ladder. In addition to time and resource pressures,

another possible reason for the low level of community participation in PD inquiries is

that the majority of them were conducted by researchers rather than practitioners.

Researchers may have less access to community members as they do not work in the

community every day.

In addition to the lack of community participation, the methodologies of almost

all PD programs and PD inquiries did not describe the relationship between program staff

/ researchers and community members throughout the process, and particularly when

identifying PD behaviors. Although according to previously published best practice

guidelines PD (Pascale et al., 2010), community members are supposed to identify the

PD behaviors themselves, it seems likely that some filtering may occur by program staff

or researchers to guide participants to selecting those behaviors which fit with a western

theory of medicine.

While previous literature describes the PD approach as fully participatory, the

many case studies of successful PD programs and inquiries without any community

participation show that it is technically possible to identify positive deviants and spread

their behaviors in a nonparticipatory way. Therefore, it could make sense to separate the

concept of the PD approach from the concept of community participation. While they are

separate concepts, using both together is likely to be more effective in achieving the types

of intangible community benefits described in previous literature, such as community

 120
mobilization (Marsh et al, 2004; Save the Children, n.d.; Singhal & Greiner, 2007),

empowerment (Hendrickson et al., 2002; Schooley & Morales, 2007), and increased

respect for community members (Sternin, 2002). Some would argue that community

participation and endogenous development is intrinsically beneficial (Arnstein, 1969;

Minkler & Wallerstein, 2008), particularly in low and middle income countries where

outside experts can easily ignore the opinions of local people. Therefore, although the

results of this study show that participation is not technically necessary to implement the

PD approach, there is still reason to believe that the meaningful involvement of

community members in the implementation of the PD approach will result in greater

positive social change.

On a more promising note, 80% of PD programs were part of a larger integrated

program, rather than being a stand-alone program. While this does make it difficult to

distinguish the effects of the PD approach from the other activities, it also increases the

chances of having a long term, sustainable, impact. The fact that PD is usually part of an

integrated program also addresses the criticism raised by Schulte (1993) that “the positive

deviance approach may imply that we are continuing to simply do nothing about the

major underlying socio-economic constraints” (p. 3). Most PD programs are one part of a

broader integrated program that addresses the underlying social, economic and resource

issues, rather than stand-alone programs which assume that the full solution can be found

without any additional resources at all.

PD Steps

The majority of PD programs and PD inquiries identified follow the steps defined
 121
by Pascale et al. (2010) in some way, including defining the problem, determining

whether positive deviants exist, discovering their behaviors, and designing, implementing

and monitoring the program. Overall, the steps defined by Pascale et al. appear to be a

useful basis for defining what the PD approach should include, and are followed by many

programs. However, as discussed later, not all of them implemented the steps exactly as

Pascale et al. intended, and some applications may better be described using other

frameworks. Very few PD programs or inquiries eliminated “true but useless” behaviors;

that is, behaviors which the positive deviants practice, but which cannot be spread to

other members of the community for practical reasons. This is not part of the Pascale et

al. (2010) steps, but is recommended by Save the Children (n.d.).

The types of PD behaviors identified varied between the different programs and

inquiries. In some cases the behaviors of positive deviants were new to both community

members and program staff. These behaviors were truly novel, and came as a surprise to

everyone involved. In other cases the behaviors were new for community members, but

were well known as good practices by program staff – for example, attending antenatal

visits, breastfeeding, and good nutrition. These types of behaviors are not recognized as

positive deviance by some scholars, who believe that PD behaviors should be things not

already known to scientists (The United Nations University, 1990). In a few cases the PD

inquiry identified behaviors that were well known by community members, but were new

for program staff, such as the methods used by moto-taxi drivers to hide condoms.

There were also some programs and inquiries in which the PD behaviors

identified were already well known by both community members and program staff
 122
before the program began, such as working hard or attending school. Given that most

previous literature defines the PD approach as a process to discover the uncommon but

successful practices of positive deviants (Bradley et al., 2009; Devane, 2009; Marsh et al,

2004; Marsh, Sternin et al., 2002; Pascale et al., 2010; Positive Deviance Initiative, 2009;

Save the Children, n.d.; Schooley & Morales, 2007; Walker et al., 2007), it is difficult to

see how behaviors which are already known to both community members and program

staff match the definition of PD.

Diffusion of PD Approach Through Community Health Practice

Almost one and a half decades passed between the first suggestion to use the PD

approach in child nutrition, and the first large scale application of the approach to child

nutrition in Vietnam. According to Fitzpatrick (2008) it is not uncommon for there to be a

long delay, even up to 30 years, between the publication of research results and its

implementation in practice. The first suggestion of the positive deviance approach was

published in a single peer reviewed journal article (Wishik & Van Der Vynckt, 1976).

Therefore, it is not surprising that it took such a long time for practitioners to pick up on,

and implement, the idea. The Sternins, who first implemented the PD approach on a large

scale, described PD at that time as an “obscure research construct” (p.23, Pascale et al.,

2010) which was the domain of researchers in peer reviewed journals, rather than

practitioners on the ground.

Since the first large scale application of the PD approach by the Sternins in the

1990s the number of PD programs per year has steadily increased. It seems that a series

of key publications in 2002 instigated by the Sternins (Ahrari et al., 2002; Berggren &
 123
Wray, 2002; Bolles et al., 2002; Dearden, Quan, Do, Marsh, Pachón et al., 2002;

Dearden, Quan, Do, Marsh, Schroeder et al., 2002; Hendrickson et al., 2002; Lapping,

Marsh et al., 2002; Lapping, Schroeder et al., 2002; Sripaipan et al., 2002; Trinh

Mackintosh et al., 2002) immediately preceded the spread of the PD to a range of new

topics and areas.

The theory of diffusion of innovations is a useful way to think about the spread of

PD through community health practice. Within the theory of diffusion of innovations,

innovators are the individuals who develop a new idea or product, and then begin the

process of spreading it to others – particularly early adopters (Rogers, 1995). The role of

the Sternins as innovators cannot be overstated. The results from this study show that, not

only did they take the idea proposed by Wishik et al. (1976) and apply it to the first large

scale study, but they were also instrumental in spreading it to other early adopters

(Pascale et al., 2010).

However, even though the PD approach was introduced to a range of new topics

(often by the Sternins), the rate of spread seems to vary between topics. For child

nutrition, the PD approach has spread from the innovators (the Sternins) to other early

adopters, and now appears to be spreading through the early majority to become

relatively well established. PD in hospital acquired infections also appears to be on the

cusp of spreading to the early majority. By comparison, the application of PD to other

topics in most cases has only spread from the innovators to a small number of early

adopters (e.g. pregnancy outcomes, school education, HIV/AIDS and reproductive

health). In some cases the approach has not spread at all past the first implementation
 124
(e.g. female genital mutilation).

Based on the results of this study it seems that some of these differences can also

be explained by the theory of diffusion of innovations. According to the theory,

innovations are more likely to spread if they have certain characteristics, including

relative advantage compared to other solutions, compatibility with existing practices and

cultures, a low level of complexity, easy ability to trial, and an easy method for observing

results (Rogers, 1995). The PD approach in child nutrition and hospital acquired

infections meets all these criteria, which may explain why the approach has diffused so

rapidly in these areas. The PD approach in other topics meets fewer of the criteria, which

may also explain why the approach has not diffused, or is diffusing at a slower rate. This

alignment between the theory of diffusion of innovations criteria and the diffusion of the

PD approach gives further support to the theory of diffusion of innovations. It gives

another example of how the theory can apply to the adoption of new approaches by

researchers and practitioners, not just the adoption of new technologies by the general

public.

Possibility of Grouping PD Approaches in Typology

Four factors were identified which can be used to distinguish between the PD

approaches used in the case studies: the level of intervention, the method used to identify

positive deviants, the method used to discover their behaviors, and the method used to

spread their behaviors to others. These factors were identified through a qualitative

analysis of the case studies, and they have not previously been used to group or classify

PD approaches. A fundamental question is whether all combinations of these factors

 125
should be considered the PD approach, as distinct from other approaches, or whether

some of them should not be called the PD approach.

The first factor is the level of intervention. This study has identified three

different levels of intervention where PD programs and inquiries have been implemented:

individual behavior change, organizational / institutional change, and social / cultural

change. Categorizing case studies using these three levels is useful because most other

frameworks and theories are related to a specific level (NCI, 2005). By grouping them

this way it is easier to compare the PD approach to other frameworks and theories at the

same level.

At the level of individual behavior change, the PD approach appears to be quite

distinct when compared to other approaches based on theories, such as stages of change

(Prochaska & DiClemente, 2005), the health belief model (Rosenstock, Strecher &

Becker, 1988), and the theory of planned behavior (Sheppard, Hartwick, & Warshaw,

1988). No other individual behavior change frameworks, approaches, or theories use the

same method of identifying positive deviants, discovering their uncommon behaviors,

and spreading those behaviors to others. The most similar approach is the concept of

modeling good behaviors according to social cognitive theory (Bandura, 1986). However,

the PD approach goes beyond simply modeling good behaviors, to identify new

behaviors and strategies that people may not have considered before. Therefore, at the

individual behavior change level I believe that the PD approach should be considered a

unique and separate approach.

However, at the organizational / institutional level the PD approach is very similar

 126
to a combination of organizational change approaches that already exist. PD at this level

appears to combine elements of appreciative inquiry, best practices, quality improvement,

and learning organizations approaches. Appreciative inquiry involves all employees

looking within the organization for successful practices which can then be expanded and

enhanced (Cooperrider, Whitney, and Stavros, 2008). Best practices involves looking for

an innovative “technique, method, process, activity or incentive which has proven to be

most effective” (The Best Practice Network, 2009, para.5) and applying it in an

organization. Quality improvement involves continuously monitoring the results of a

process in order to find ways to improve it (Riley & Moran, 2010), while the learning

organizations framework is based on full employee participation in which “people at all

levels, individuals and collectively, are continually increasing their capacity to produce

results they really care about" (Richard Karash, n.d., para.2).

Most of these organizational change frameworks include the idea of finding the

unique and successful practices of others and spreading them throughout the organization

in a participatory way. Given this, the PD approach could be described as a repackaging

of existing frameworks under a new name. This conclusion is in line with Lewis (2009)

who used the PD approach for health services improvement in the United Kingdom, and

concluded that “PD is a highly practical way of delivering change and cuts through the

psychobabble of some change management approaches, while being entirely consistent

with current thinking about the human aspects of change, employee engagement, and

how to overcome resistance” (Lewis, 2009, p.287). As Lewis states, PD is “entirely

consistent” with existing thinking on organizational change. Its main benefit then is the
 127
way in which it repackages these existing frameworks under the new name of PD. As

stated earlier, PD is a naturally inspiring concept, and is supported by powerful stories

from the field of international development. Even if the approach is not fundamentally

different from existing frameworks, the PD name and inspiring stories from the field may

help convince staff to participate during organizational change initiatives. That said, the

PD approach is still relatively rare in the organizational literature. A search of

organizational and management literature databases revealed only 25 results for PD,

while there were 70 for appreciative inquiry, 803 for learning organizations, 10,423 for

quality improvement and 9,571 for best practice. Therefore, even though the PD

approach may have some benefits in organizational change, its relative advantage

according to the theory of diffusion of innovations may not be great enough for it to

replace the existing organizational change frameworks which use essentially the same

approach.

At the social / cultural level it also appears that the PD approach may not actually

be a new approach, but simply another name for role models, which have been used in

social change programs for many years and are a core element of social cognitive theory

(Bandura, 1986). While the current uses of the PD approach at the individual behavior

change level go beyond the concept of modeling by identifying unique behaviors and

spreading them, the small number of PD programs and inquiries implemented for social /

cultural change focus mainly on positive deviants as role models for good behaviors that

are already known. That is, the PD behaviors identified by these programs were already

known to be good behaviors by both community members and program staff before the
 128
program started. As discussed previously, I do not believe these behaviors should be

considered PD because they are not new behaviors.

For example, in the case study of a PD program to help reintegrate returned child

soldiers, girls were selected as “positive deviants” if they practiced predefined positive

deviant behaviors such as attending school, running a small business, and respecting

adults. These behaviors were already considered good behaviors by parents, program

staff, community leaders, and the girls themselves. They were not new behaviors that had

been discovered as part of the PD inquiry. Therefore, I believe that it is better to say the

girls in the program were role models for good behaviors, rather than saying they were

positive deviants practicing new or unusual behaviors.

In the evaluation, participants also referred to the girls as role models: “The PD

girls are role models. They are respectful and resourceful. They wear clean clothes and

have food on the table. They work hard and have taken control of their lives” (Singhal &

Dura, 2008, p. 49). In addition to role models, the girls in the child soldiers’ program

could also be considered examples of resiliency (Lapping, Marsh et al., 2002; Zeller,

1991).

The PD program on female genital mutilation is similar. The positive deviants in

this program were leaders or family members who opposed female genital mutilation.

While this is a good behavior to model, it is not a new or unique behavior that has been

identified through a positive deviance inquiry, and due to cultural reasons some

community members considered it to be a bad behavior rather than a good behavior.

Several other programs targeting female genital mutilation used exactly the same
 129
approach, but referred to these people simply as “role models” rather than positive

deviants (Oloo, Wanjiru, & Newell-Jones, 2011; World Bank & UNFPA, 2004).

In addition to the concept of role models, the PD programs targeting social /

cultural change also draw on elements of community development frameworks. For

example, the trafficking case study in Indonesia included obvious community organizing

elements described by Minkler and Wallerstein (2005), such as mobilizing local support

for the initiative. It also included elements of assets based community development as

described by Mathie and Cunningham (2003), such as identifying existing local resources

which could be used to bring about positive change. Given this, and the relatively small

number of case studies in the social / cultural change area, I believe that the PD approach

in this area does not offer anything new. It is easier and clearer to refer to the positive

deviants in these programs simply as role models.

The fact that the PD approach at the organizational / institutional and social /

cultural levels overlaps with existing frameworks is not surprising, since it was first

introduced to both of these areas by the Sternins. By his own admission, Jerry Sternin

was not a specialist in theoretical frameworks:

Jerry Sternin, despite having been Counseling Dean of Harvard

University, never considered himself an academic or theorist. ‘I’m sorry,

but I’m not a methodologist. I don’t know what’s out there. I only know

what I know,’ he said to CCC alumnus Roberto Saco (Lewis, 2009,

p.286).

Therefore, it is entirely possible that when the Sternins moved from their original
 130
programs in child nutrition to new areas in organizational and sociocultural change, they

may not have been familiar with all the existing theories and approaches in those areas.

Instead, they applied the PD approach which they had found very successful in their early

work on child nutrition.

After the intervention level (individual behavior, organizational / institutional, or

social / cultural), the next factors which can be used to classify PD approaches are the

method used to identify the positive deviants and the method used to discover their

behaviors. Looking at these methods for programs targeting individual behavior change,

several combinations of methods seem to be effective (see Figure 11). The most common

combination, which could also be considered best practice, is to use quantitative

screening against specific criteria to identify positive deviants and then use qualitative

interviews, focus groups, and/or observations to identify their behaviors. This method

could be considered best practice as all positive deviants will be systematically identified,

and the inquiry to determine their behaviors is open-ended and flexible so that behaviors

which the program staff may not have considered before will be identified. A variation on

this approach is to use systematic qualitative screening against specific qualitative criteria

to identify positive deviants, and then additional qualitative interviews and focus groups

to identify their behaviors. This method could be useful when addressing outcomes which

cannot be measured quantitatively, although it has only been applied in a small number of

cases. A limitation of this approach is that it is extremely time consuming, as it requires

a large number of interviews or focus groups to be analyzed.

A more cost-effective approach is to use individual choice or convenience

 131
sampling to identify positive deviants, and then qualitative interviews or focus groups to

determine their behaviors. While this approach was very common in the PD program and

inquiry case studies, it was normally only used as a substitute for quantitative or

qualitative screening when there was not enough time or resources to systematically

identify positive deviants. A key limitation of this approach is that some behaviors may

not be identified because the positive deviants have not been systematically identified.

The results may also be biased by the person selecting the positive deviants. For example,

if the village chief is asked to identify women with good child care practices they may

choose their own relatives or friends, so the results of this type of PD inquiry may not be

as convincing for community members as systematically collecting the data. Despite

these limitations, the paper by Lapping, Schroeder et al. (2002) which compared this

approach to a case-control study shows that it can still be effective when there is limited

time and resources.

One study on HIV (Babalola et al., 2006) used qualitative screening to identify

positive deviants, and then used qualitative interviews to identify their behaviors. These

behaviors were then included in a quantitative survey to see if they were actually related

to positive deviant status in the general population. This approach is more likely to

identify real positive deviant behaviors since it validates the results of the qualitative

inquiry with a survey. The downside is the additional time and resources required, and

the specialist expertise required for developing a valid and reliable survey instrument

based on the positive deviant behaviors identified.

Many of the PD inquiries used only quantitative methods to identify positive

 132
deviants and their behaviors. This typically involved administering a predefined survey or

using a secondary data set. The quantitative data were then analyzed to identify any

statistical correlations between behaviors and positive deviant status. I do not believe this

should be considered the PD approach, as all questions on the survey were predefined by

the researcher or program staff. This means that the PD inquiry will not identify any new

or uncommon behaviors that researchers or program staff may not have considered

before.

The final factor which can be used to categorize PD approaches is the method

used to spread the behaviors to others. If organizational / institutional and social / cultural

programs are ruled out, only three methods of spreading behaviors to other people

remain: action learning sessions (Hearths), peer educators or community volunteers, and

mass media campaigns. The current best practice for spreading PD behaviors to others is

action learning sessions, as there were several high quality evaluations conducted on

programs which have used this method to successfully spread behaviors in both the short

and long term. Action learning sessions are also recommended as best practice by

previous literature on PD (Pascale et al., 2010; Positive Deviance Initiative, n.d.) as they

are highly participatory. An alternative to action learning sessions is to use peer educators

or community volunteers to spread the behaviors to others. Although there were only

three evaluations available for this approach, they included the only randomized

controlled trial available, as well as two quasi-experimental evaluations with matched

comparison groups, all of which reported at least some positive results. Therefore, even

though this is not considered a best practice by previous PD literature it could be

 133
considered a best practice based on the results of this study, as it has a higher level of

evidence than other methods. Finally, one individual behavior change program

successfully used mass media to spread the behaviors to others. Mass media is not a

participatory approach, and with only one existing evaluation, it cannot yet be considered

a best practice. Figure 11 summarizes the different combinations of methods found in the

case studies, and identifies which could be considered a unique PD approach, which are

already covered by existing frameworks, which should not be considered PD, and which

could be considered best practice given the available evidence. These are new

classifications based on the results of this study, and have not previously been proposed

in the PD literature. The combinations of methods found in the case studies do not

represent all possible combinations that could theoretically be included in the PD

approach. For example, using quantitative screening followed by a qualitative inquiry and

then spreading the behaviors via mass media is a combination that is theoretically

possible, but was not included in any of the case studies. As the body of PD literature

grows more combinations may be added to the list.

 134

Method used to identify Method used to Method used to spread

Level of Intervention positive deviants discover behaviors behaviors

Action learning sessions (Hearth)

Individual choice / convenience Qualitative inquiry
Peer educators or community volunteers

Qual. inquiry and quant. survey Mass media campaign

Individual behavior Qualitative screening
change Qualitative inquiry

Action learning sessions (Hearth)

Qualitative inquiry
Peer educators or community volunteers
Quantitative screening
Quantitative survey

Mass media campaign

Individual choice / convenience Qualitative inquiry
Staff meetings and on-the-job training

Positive Deviance Organizational / Action learning sessions

Quantitative screening Qualitative inquiry
approach in practice institutional change
Staff meetings and on-the-job training

Staff meetings Staff meetings Staff meetings and on-the-job training

Advocacy campaign
Individual choice / convenience Qualitative inquiry
Peer educators or community volunteers
Social / cultural
change Qualitative screening Qualitative inquiry

Quantitative screening Qualitative inquiry Peer educators or community volunteers

PD is a unique approach and has Best practice PD approach based on Never been implemented as a PD is the same as existing
Key been implemented successfully current evidence program, only in a PD inquiry frameworks
Should not be considered PD

Figure 11. PD approaches used in practice

134
 135

Outcomes and Rigor of PD Program StudiesWhile previous literature on the PD

approach has described many examples of individual PD programs (Ahrari et al, 2002;

Aruna, Vazir, & Vidyasagar, 2001; Awofeso et al., 2008; Babalola et al., 2006; Bradley

et al., 2009; Dearden, Quan, Do, Marsh, Pachón et al., 2002; Friedman et al., 2008; Kim

et al., 2008; The United Nations University, 1990; Vossenaar et al., 2009; Walker et al.,

2007; Wishik & Van Der Vynckt, 1976), no study has attempted to review all the

available evidence for the effectiveness of the PD approach in general.

The results of this study show that the only topic with a substantial body of

literature on the effectiveness of the PD approach was child nutrition, with a total of 26

evaluations. However, even in this area the study designs used to evaluate the programs

were relatively weak and in most cases substantial sections of the report were incomplete

or missing. There was only one randomized controlled trial, three quasi-experimental

designs with a matched comparison group, and one design with an unmatched

comparison group. All of the child nutrition evaluations except one showed at least some

positive impacts from the PD program, even if they were not as extensive as hoped, or

did not cover all target groups. These results provide more concrete support for previous

recommendations by individual such as Berggren and Wray (2002) to use the PD

approach in child nutrition. Therefore, although there is room to expand the number of

evaluations, particularly those using more rigorous study designs, the current level of

evidence does suggest that the PD approach in child nutrition can have positive impacts

that may be sustained in the long term.

135
 136

No other topic had enough published evaluations to draw any meaningful

conclusions from this study, although all the quantitative and qualitative evaluations

reported some positive results. Both hospital acquired infections and pregnancy outcomes

had a small number of quantitative studies of a reasonable level of rigor that showed

positive results, although more studies are needed to confirm their results. These results

do not confirm or contradict previous literature, as there have not been any overall

reviews of the PD approach in topics outside child nutrition. However, organizations such

as the Positive Deviance Initiative (2010e) have been promoting the use of PD in a wide

range of areas. The results from this study suggest that there may not be strong evidence

for its use in all areas, although there is definitely room for more pilot programs and

evaluation studies to determine its effectiveness. It is surprising that all published

evaluations except one showed at least some positive results, and it is possible that

publication bias means that PD programs with negative results or no effects have

deliberately not been published.

Assessing the rigor of the evaluations using the quality assessment tools proved

challenging. Very few evaluation reports had enough information available to complete

the tools. Even when the information was available, it was difficult to complete the tools

as they were designed to assess scientific research studies rather than program

evaluations. Therefore, some flexibility was required in the interpretation of the tools.

This highlights both the need for separate tools to assess program evaluations, and the

need for more actual research (rather than program evaluations) into the PD approach. If

new tools were developed in the future it would be ideal to have separate tools for
 137

research studies and program evaluations. The tools for research studies would be the

same as the quality assessment tools used in this study, while the tools for program

evaluations would include evaluation criteria that are matched to the standards published

by organizations such as the American Evaluation Association (AEA, n.d.). Both

scientific research and field evaluations have the potential to contribute to our

understanding of PD. Scientific studies provide a higher level of rigor for assessing the

efficacy of the PD approach in ideal conditions, while field evaluations are able to

provide a broader picture of its overall effectiveness when implemented in the field.

Limitations

There are four main limitations to this study as a result of the methodology used.

First, and most importantly, the case studies were limited to those that could be described

using publicly available documentation online. Given that the majority of PD program

case studies were based on internal program reports that had been loaded onto websites, it

is entirely possible that large numbers of other PD programs have been implemented, but

their programs reports are not publicly available. It is also possible that some older

programs or inquiries might not have been available online.

Secondly, the potential for publication bias means that some programs which

showed no, or negative, results may have been deliberately not published. This would

affect the conclusions about the effectiveness of the PD approach.

Thirdly, this study only included PD programs and inquiries related to health or

social determinants of health. Based on the search results during sampling it is clear that

PD is also being used in a range of nonhealth areas, such as climate change and economic
 138

development. Future studies could investigate the use of PD in these areas.

Finally, the aggregation of subprograms into larger programs may have affected

the interpretation of some results. Aggregating subprograms into their larger programs

resulted in an overall lower number of programs, even though the number of individual

implementations was very high. However, the aggregation was justified given that many

of the large programs had combined evaluations covering all subprograms, and counting

each individual subprogram separately would have made the data more difficult to

analyze and interpret.

Implications for Positive Social Change

This study provides compelling evidence for the wide use of PD in child nutrition.

Making the PD approach a standard approach for child nutrition could have significant

positive social change implications. Early childhood nutrition is related to long term

outcomes in the areas of health, education and social development, and so reducing rates

of malnutrition will have a corresponding positive impact in these areas. Universities,

trainers, writers, and community health professionals should incorporate PD for child

malnutrition into their training materials, textbooks, and programs.

The results of this study also show the potential for PD to contribute to positive

social change for other individual behavior change issues, including smoking, weight

loss, HIV prevention, and childhood obesity, although researchers and practitioners

should conduct more trials and high quality outcome evaluations to determine exactly

which topics the PD approach is useful for. This study can provide a basis for further

research in these areas, which will ultimately contribute to positive social change by
 139

identifying effective programs based on the PD approach which can be implemented to

improve community health.

Finally, this study has drawn attention to the significant discrepancy between the

ideal of PD as a highly participatory approach (Pascale et al., 2010) and the reality of PD

programs and inquiries on the ground that rarely involve the community in a meaningful

way. While the results show that community participation is not technically necessary for

implementing the PD approach, it is still important for creating positive social change at

the community level that is aligned with the real needs and opinions of local people

(Minkler & Wallerstein, 2005). Therefore, this study will hopefully motivate practitioners

to involve community members in a meaningfully way during all steps of the PD process.

Recommendations for Action

The results of this study show there is compelling evidence for including the PD

approach as a standard approach to child malnutrition, and implementing it widely in low

and middle income countries. However, in other areas of individual behavior change,

such as pregnancy outcomes, smoking, HIV/AIDS, weight loss, and reproductive health,

the PD approach is still in its infancy. PD practitioners should consider piloting PD

programs in these areas, and using rigorous outcome evaluations to assess their results,

before implementing them widely.

PD should not be considered a separate approach for issues related to

organizational / institutional change and social / cultural change, as there are existing

frameworks in these areas which cover essentially the same approach and are already

more widely used. The term positive deviance should also be reserved for behaviors that
 140

are new to community members and/or program staff. Behaviors that were already

considered positive by community members and program staff should not be called

positive deviant behaviors.

When implementing PD programs practitioners should consider using the best

practices identified by this study. This includes using quantitative screening to identify

positive deviants, qualitative inquiry to identify positive deviant behaviors, and action

learning or peer educators / community volunteers to spread the behavior to others.

Although it is not technically necessary, practitioners should also consider the potential

benefits of implementing the PD approach in a fully participatory way.

There is a strong argument for testing the application of PD to child overnutrition

in middle and high income countries given how successful it has been at addressing child

malnutrition in middle and low income countries. Child overnutrition, and its relationship

to overweight, obesity and chronic disease, is a growing concern, with relatively few

effective programs available to combat it. If positive deviance was as successful with

overnutrition as it is with malnutrition, this could make a very substantial contribution to

positive social change.

The PD approach could easily be applied to this problem by having a group of

community members measure and weigh children to identify those who are a normal

weight for their age despite having many of the social risk factors associated with

childhood obesity, such as poverty, lower parental education, and a parent who is

overweight or obese. This would prove to community members, particularly the parents

of children who are overweight or obese, that it is possible to have a normal weight child
 141

even in difficult circumstances, and that positive deviant parents have clearly found

unique solutions for the problems they face. Community members could then interview

and observe the positive deviant parents to see what unique behaviors they use, including

particular types of affordable foods, recipes, and types of exercise. The positive deviant

parents could teach these behaviors to other parents using action learning sessions similar

to the Hearth sessions for malnutrition. Participatory monitoring could be done by

community members by weighing and measuring the children throughout the program to

see if it is successful.

One challenge associated with applying the PD approach to childhood obesity

compared to child malnutrition is that a parent’s motivation to help a child lose weight

may not be as strong as their motivation to help a malnourished child gain weight.

Malnourished children are at immediate risk of acute health problems, and they gain

weight very quickly once the Hearth sessions start, which motivates the parents to keep

going. The health risks of childhood obesity are longer term and so parents may not be as

motivated. In addition, losing weight takes longer than gaining weight, which does not

provide immediate reinforcement to parents. Practitioners involved in child nutrition in

middle and high income countries should trial PD as a high priority.

Recommendations for Further Study

Since the results of this study show that the majority PD program evaluations

were weak methodologically, researchers should conduct more trials and high quality

outcome evaluations to determine exactly which individual behavior change topics the

PD approach is effective for (e.g. smoking, weight loss, HIV/AIDS, etc). Implementers of
 142

future PD programs should be strongly encouraged to evaluate their programs using a

rigorous study design. This can often be difficult for program staff that are busy

implementing the program and may not have the research skills or time required to

conduct a rigorous evaluation. Therefore, it will be important for researchers at academic

institutions to partner with implementing agencies for this purpose, in addition to

conducting their own high quality research into PD. These evaluations should be

disseminated through peer reviewed publications rather than languishing in incomplete

and sometimes incoherent internal reports, as was the case for many PD programs in this

study.

Almost all existing evaluations compare the PD approach with no intervention. It

would be interesting to see whether a comparison between action learning sessions based

on PD behaviors and action learning sessions based on regular good practice showed any

differences. It is possible that the effectiveness of the PD approach in child nutrition is

not actually related to the PD approach itself, but to factors such as community

participation, action learning, and regular feedback to parents.

Researchers should consider restricting their use of the term PD to cases where it

clearly meets the definitions provided by Pascale et al. (2010). Using the term PD too

freely when other terms already exist may cause confusion and dilute the meaning of PD.

Specifically, researchers should not use the term PD when they are simply looking for

correlations between variables and a positive outcome using a predefined survey or

secondary data. In these types of studies there is no possibility of identifying new

behaviors that have not previously been considered. PD should be reserved for studies
 143

where new behaviors are being identified using primary data collection. Researchers

should also carefully consider the use of term PD in relation to organizational /

institutional change and social / cultural change when frameworks already exist to

describe the same approach.

As part of this study a large number of stand-alone PD inquiries were found in

which PD behaviors were identified but were not incorporated into a program or

disseminated to the community. By definition, PD behaviors are those which could easily

be incorporated into programs because they are already being practiced by some

community members and do not require additional resources to implement. Therefore,

researchers who discover new PD behaviors should be strongly encouraged to

disseminate their findings directly to practitioners in the field so they can be incorporated

into existing programs.

When applying the PD approach to a new area, or trying to promote its wider use

in an existing area, it might also be useful to consider the theory of diffusion of

innovations. Research findings which meet the requirements for quick diffusion are more

likely to be taken up by practitioners than those which do not. Researchers should also

consider where they publish the results of their programs and studies. While the Positive

Deviance Initiative website (2010e) is probably contributing to the spread of the approach

into other areas, a greater number of complete reports published other locations targeting

practitioners would also help this.

 144

Reflection on the Researcher’s Experience

This study used a qualitative analysis of multiple case studies, and so my own

preconceived biases, ideas, and values, as well as those of the peer debriefer, may have

influenced our interpretation and analysis of the data. There are several factors and

experiences which may have affected my interpretation of the data. During this study I

was working in Cambodia in an international development setting. My experiences

working in this setting made me more aware of issues related to international

development, such as the involvement of local governments, local NGOs, sustainability,

and the severe lack of trained staff and resources faced by many programs. By

comparison, the peer debriefer’s experience is in management consultancy for NGOs in

developed countries. Therefore, he approached the topic from a very different perspective

which complemented my approach. He asked more questions related to the application of

PD in developed countries, and particularly to organizations. He was very instrumental in

highlighting the overlap between the PD approach and other organizational frameworks

since he has extensive experience in that area. I believe our two contrasting approaches

worked well together to give a balanced analysis of the data from both a developing and

developed country perspective.

This research has also influenced my own perspectives on the PD approach. I first

heard about PD when reading the book “Influencer” (Patterson, Grenny, Maxfield,

McMillan & Switzler, 2007) as part of the required reading for one of my university

courses. At the time I was impressed by the simplicity of the idea and its potential to

discover solutions from within the community itself. As I discussed it with other people I
 145

found they were also inspired by it, and particularly by the success stories of positive

deviants discovered by early programs. My original PhD proposal was an application of

the PD approach to hygiene and sanitation in rural Tanzania, in partnership with a local

community organization. Unfortunately, or fortunately, I decided not to pursue this

proposal because of the time required to obtain ethical approval from the Tanzanian

government to conduct research.

While conducting the literature review for my first proposal I discovered that the

PD approach was not a tried and tested method as I had expected from the success

stories, but a relatively new approach which had only been implemented in a small

number of areas. This made it very difficult for me to design the Tanzanian proposal, and

it is what motivated me to conduct this current study.

Through completing this study I have found that, while the success stories of

positive deviants are very inspiring, they can also give the impression that the PD

approach is more participatory than it really is, and has a lot more evidence than it

actually does. In reality the PD approach has only been applied to a few areas and, with

the exception of child nutrition, it does not yet have a full body of evidence to support it.

Although it is described as a fully participatory approach, the reality is that most

implementations fall far short of this ideal. For me, this is a personal lesson in being

critical of new approaches and “buzzwords” which may be very appealing and inspiring,

but might not be supported by a lot of evidence.

 146

Conclusion

Despite the powerful stories of what PD has achieved and its natural appeal, this

study has shown that the PD approach has only been extensively applied to child

nutrition. While it has been applied in other areas in the last decade - often as the result of

innovations by the Sternins – PD is still at the very early stages of the diffusion process.

Based on the results of this study I believe there is sufficient evidence to conclude that

the PD approach in individual behavior change is something unique, while in

organizational / institutional change and social / cultural change, PD is already covered

by existing frameworks and theories. When applying PD to individual behavior change

there are clearly some methods of identifying positive deviants, discovering the

behaviors, and spreading those behaviors to others which are aligned with past PD

literature and are also supported by more high quality outcome evaluations. These could

be considered current best practices in the field.

There is significant evidence for the effectiveness of PD in child malnutrition.

However, in other areas of individual behavior change, such as pregnancy outcomes,

smoking, HIV/AIDS, weight loss, and reproductive health, the PD approach is still in its

infancy. More research is needed to determine whether the PD approach is effective in

these areas before it can be considered a standard approach.

In conclusion, PD is an innovative approach which is naturally appealing and

inspiring. It is clearly effective in the area of child malnutrition, and has the potential to

contribute to many other areas of individual behavior change in low, middle and high

income countries.
 147

References

Ahrari, M., Houser, R. F., Yassin, S., Mogheez, M., Hussaini, Y., Crump, P.,…Levinson,

F. J. (2006). A positive deviance-based antenatal nutrition project improves birth-

weight in Upper Egypt. Journal of Health, Population & Nutrition, 24(4), 498-

507. Retrieved from http://www.jhpn.net

Ahrari, M., Kuttab, A., Khamis, S., Farahat, A. A., Darmstadt, G. L., Marsh, D. L., &

Levinson, F. J. (2002). Factors associated with successful pregnancy outcomes in

Upper Egypt: A positive deviance inquiry. Food and Nutrition Bulletin, 23(1), 83-

88. Retrieved from http://www.foodandnutritionbulletin.org

American Evaluation Association. (n.d.). The program evaluation standards. Retrieved

from http://www.eval.org/evaluationdocuments/progeval.html

Arnold, D. G., & Hartmann, L. P. (2005). Beyond sweatshops: positive deviancy and

global labour practices. Business Ethics: A European Review, 14(3), 206-222.

Retrieved from http://www.wiley.com/bw/journal.asp?ref=0962-8770

Arnstein, S. R. (1969). A ladder of citizen participation. Journal of the American Institute

of Planners, 35(4), 216-224. Retrieved from http://lithgow-schmidt.dk/sherry-

arnstein/ladder-of-citizen-participation.pdf

Aruna, M., Vazir, S., & Vidyasagar, P. (2001). Child rearing and positive deviance in the

development of preschoolers: A microanalysis. Indian Pediatrics, 38, 332-339.

Retrieved from http://www.indianpediatrics.net

Australian Community Development and Civil Society Strengthening Scheme

[ACCESS]. (2009). ACCESS Phase II: A Toolkit for Understanding and Using
 148

Strength Based Approaches in Development. Retrieved from http://www.access-

indo.or.id

Awofeso, N., Irwin, T., & Forrest, G. (2008). Using positive deviance techniques to

improve smoking cessation outcomes in New South Wales prison settings. Health

Promotion Journal of Australia, 19(1), 72-73. Retrieved from

http://www.healthpromotion.org.au/journal

Babalola, S., Awasum, D., & Quenum-Renaud, B. (2002). The correlates of safe sex

practices among Rwandan youth: a positive deviance approach. African Journal

of AIDS Research, 1, 11–21. Retrieved from http://www.ajol.info/index.php/ajar

Babalola, S., Ouedraogo, D., & Vondrasek, C. (2006). Motivation for late sexual debut in

Côte d’Ivoire and Burkina Faso: A positive deviance inquiry. Journal of

HIV/AIDS Prevention in Children & Youth, 7(2), 65-87.

doi:10.1300/J499v07n02_05

Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory.

Englewood Cliffs, NJ: Prentice Hall.

Barker, G. (2000). Gender equitable boys in a gender inequitable world: reflections from

qualitative research and programme development in Rio de Janeiro. Sexual and

Relationship Therapy, 15(3), 206-282. Retrieved from

http://www.tandf.co.uk/journals/SRT

Berggren, W. L., & Wray, J. D. (2002). Positive deviant behavior and nutrition education.

Food and Nutrition Bulletin, 23(4) (Suppl.), 7-8. Retrieved from

http://www.foodandnutritionbulletin.org
 149

Best Practice Network, The. (2009). The best practice network – definition of a best

practice. Retrieved from http://www.best-practice.com/

Biggs, S. (2008). Learning from the positive to reduce rural poverty and increase social

justice: Institutional innovations in agricultural and natural resources research and

development. Experimental Agriculture, 44, 37–60.

doi:10.1017/S001447970700595

Bolles, K., Speraw, C., Berggren, G., & Lafontant, J. G. (2002). Ti Foyer (hearth)

community-based nutrition activities informed by the positive deviance approach

in Leogane, Haiti: A programmatic description. Food and Nutrition Bulletin,

23(4) (Suppl.), 9-15. Retrieved from http://www.foodandnutritionbulletin.org

Bradley, E. H., Curry, L. A., Ramanadhan, S., Rowe, L., Nembhard, I. M., & Krumholz,

H. M. (2009). Research in action: using positive deviance to improve quality of

health care. Implementation Science, 4(25). Online publication. doi:10.1186/1748-

5908-4-25

Brokenleg, M. (2010). The resilience revolution: Our original collaboration. Reclaiming

Children and Youth, 18(4), 8-11.

Bullen, P. (2004). Some practice questions in evaluating services. Developing Practice:

The Child, Youth and Family Work Journal, 11, 38. Retrieved from

http://reclaimingjournal.com

Buscell, P. (2008). Pathways to prevention: Mapping the positive deviance/MRSA

prevention networks at Pennsylvania and Montana healthcare facilities shows

promise. Prevention strategist, Autumn. Retrieved from the Positive Deviance

 150

Initiative website:

http://www.positivedeviance.org/pdf/evaluation/Pathways%20to%20Prevention.p

df

Catalani, C., & Minkler, M. (2010). Photovoice: A review of the literature in health and

public health. Health Education & Behavior, 37, 424-451. doi:

10.1177/1090198109342084

Creswell, J. (2009). Research design: Qualitative, quantitative, and mixed methods

approaches (3rd ed). Thousand Oaks, CA: Sage Publications.

Creswell, J. W. (2007). Qualitative inquiry & research design: Choosing among five

approaches (2nd ed). Thousand Oaks, CA: Sage Publications.

Comparing and Supporting Endogenous Development [COMPAS]. (2007). What is

endogenous development? Retrieved from: http://www.compasnet.org/ed_1.html

Cooperrider, D., Whitney, D., & Stavros, J. M. (2008). Appreciative inquiry handbook:

For Leaders of change (2nd ed). Brunswick, Ohio: Crown Custom Publishing, Inc.

Daniels, K., & Lewin, S. (2008). Translating research into maternal health care policy: a

qualitative case study of the use of evidence in policies for the treatment of

eclampsia and pre-eclampsia in South Africa. Health Research Policy And

Systems, 6, 12. Retrieved from: http://www.health-policy-systems.com

Davies, S. R. (2011). A rare complication of a unilateral vertebral artery occlusion, which

resulted in a basilar emboli after a C5-C6 bifacet dislocation in a professional

rugby player: case study. The Spine Journal, 11(3), p.e1-4.

doi:10.1016/j.spinee.2011.01.030
 151

Dearden, K. A., Quan, L. N, Do, M., Marsh, D. R., Pachón, H., Schroeder, D. G., &

Lang, T. T. (2002). Work outside the home is the primary barrier to exclusive

breastfeeding in rural Viet Nam: insights from mothers who exclusively breastfed

and worked. Food and Nutrition Bulletin, 23(4) (Suppl.), 99-106. Retrieved from

http://www.foodandnutritionbulletin.org

Dearden, K. A., Quan, L. N., Do, M., Marsh, D. R., Schroeder, D. G., Pachón, H., &

Lang, T. T. (2002). What influences health behavior? Learning from caregivers of

young children in Viet Nam. Food and Nutrition Bulletin, 23(4) (Suppl.), 117-

127. Retrieved from http://www.foodandnutritionbulletin.org

Devane, T. (2009). The positive deviance approach - A briefing. Retrieved from Positive

Deviance Initiative website:

http://www.positivedeviance.org/pdf/research/The_Positive_Deviance_Briefing_

T_Devane.pdf

Edward, S. (1985). Positive deviance: an oxymoron. Deviant Behavior, 6(2), 169-181.

Effective Public Health Practice Project [EPHPP] (2009). Quality assessment tool for

quantitative studies dictionary. Retrieved from

http://www.ephpp.ca/PDF/QADictionary_dec2009.pdf

Effective Public Health Practice Project [EPHPP] (2010). Quality assessment tool for

quantitative studies. Retrieved from

http://www.ephpp.ca/PDF/Quality%20Assessment%20Tool_2010_2.pdf

Fitzpatrick, J. J. (2008). Lag time in research to practice: are we reducing or increasing

the gap? Applied Nursing Research, 21(1), 1. Retrieved from

 152

http://www.appliednursingresearch.org

Fowles, E. R., Hendricks, J. A., & Walker, L. O. (2005). Identifying healthy eating

strategies in low-income pregnant women: Applying a positive deviance model.

Health Care for Women International, 26, 807–820. Retrieved from

http://www.tandf.co.uk/journals/UHCW

Friedman, S. R., Mateu-Gelabert, P., Sandoval, M., Hagan, H., & Des Jarlais, D. C.

(2008). Positive deviance control-case life history: a method to develop grounded

hypotheses about successful long-term avoidance of infection. BMC Public

Health, 8(94). Online publication. doi:10.1186/1471-2458-8-94

Finlayson, K. W., & Dixon, A. (2008). Qualitative meta-synthesis: a guide for the novice.

Nurse researcher, 15(2), 59-71. Retrieved from

http://nurseresearcher.rcnpublishing.co.uk

Gertner, J. (2008, December 14). Positive deviance. The New York Times. Retrieved from

http://www.nytimes.com/2008/12/14/magazine/14ideas-section3-t-005.html

Goode, E. (1991). Positive deviance: a viable concept? Deviant Behavior, 12(3), 289-

309. Retrieved from http://www.tandf.co.uk/journals/tf/01639625.html

Guldan, G. S., Zhang, M. Y., Zhang, Y. P., Hong, J. R., Zhang, H. X., Fu, S. Y., & Fu, N.

S. (1993). Weaning practices and growth in rural Sichuan infants: a positive

deviance study. Journal of Tropical Pediatrics, 39(3), 168-175.

doi:10.1093/tropej/39.3.168

Heckert, D. M. (1985). Toward a theory of positive deviance (Doctoral dissertation).

University of New Hampshire, US.

 153

Heckert, D. M. (1998). Positive deviance: A classificatory model. Free Inquiry in

Creative Sociology; 26(1), 23-30. Retrieved from Sage Journals.

Hendrickson, J. L., Dearden, K., Pachón, H., Hoi An, N., Schroeder, D. G., & Marsh, D.

R. (2002). Empowerment in Rural Viet Nam: Exploring changes in mothers and

health volunteers in the context of an integrated nutrition project. Food and

Nutrition Bulletin, 23(4) (Suppl.), 83-91. Retrieved from

http://www.foodandnutritionbulletin.org

Higgins J. P. T., & Green, S. (Eds.). (2009). Cochrane handbook for systematic reviews

of interventions version 5.0.2. Retrieved from The Cochrane Collaboration

website: http://www.cochrane-handbook.org.

Horowitz, B. (2008). Using positive deviance methods to help legal service providers to

ethically and successfully handle public extortion situations. Retrieved from the

Positive Deviance Initiative website:

http://www.positivedeviance.org/projects/Anti-

extortion%20concurrent%20BBL%20Results%20FINAL.pdf

Israel, A. B., Schultz, A. J., Parker, E. A., Becker, A. B., Allen, A. J., & Guzman, R.

(2008). Critical issues in developing and following CBPR principles. In Minkler,

M. & Wallerstein, N. (Eds.), Community-based participatory research for health:

From process to outcomes (2nd Ed.) (pp. 47-66). San Francisco, CA: John Whiley

& Sons, Publishers.

Issel, L. M. (2004). Health program planning and evaluation: A practical, systematic

approach for community health. Sudbury, MA: Jones and Bartlett Publishers, Inc.
 154

Karash, R. (n.d.). Learning organizations: Where the learning begins. Retrieved from

http://www.see.ed.ac.uk/~gerard/MENG/MEAB/lo_index.html

Khan, A. A., Khan, N. U., Bile, K. M., & Awan, H. (2010). Creating synergies for health

systems strengthening through partnerships in Pakistan -- a case study of the

national eye health programme. Eastern Mediterranean Health Journal, 16, S61-

8. Retrieved from http://www.emro.who.int/emhj.htm

Kim, Y. M, Heerey, M., & Kols, A. (2008). Factors that enable nurse–patient

communication in a family planning context: A positive deviance study.

International Journal of Nursing Studies, 45, 1411–1421.

doi:10.1016/j.ijnurstu.2008.01.002

Lapping, K., Marsh, D. R., Rosenbaum, J., Swedberg, E., Sternin, J., Sternin, A., &

Schroeder, D. G. (2002). The positive deviance approach: Challenges and

opportunities for the future. Food and Nutrition Bulletin, 23(4) (Suppl.), 128-135.

Retrieved from http://www.foodandnutritionbulletin.org

Lapping, K., Schroeder, D., Marsh, D., Albalak, R., & Zahir Jabarkhil, M. (2002).

Comparison of a positive deviant inquiry with a case-control study to identify

factors associated with nutritional status among Afghan refugee children in

Pakistan. Food and Nutrition Bulletin, 23(4) (Suppl.), 26-33. Retrieved from

http://www.foodandnutritionbulletin.org

Levinson, F. J., Barney, J., Bassett, L., & Schultink, W. (2007). Utilization of positive

deviance analysis in evaluating community-based nutrition programs: An

application to the Dular program in Bihar, India. Food and Nutrition Bulletin,
 155

28(3), 259-265. Retrieved from http://www.foodandnutritionbulletin.org

Long, A. F., & Godfrey, M. (2004) An evaluation tool to assess the quality of qualitative

research studies. International Journal of Social Research Methodology Theory

and Practice, 7 (2): 181-196. Retrieved from

http://www.tandf.co.uk/journals/tsrm

Mackenbach, J. P., Van Den Bos, J., Joung, Van De Mheen, H., & Stronks, K. (1994).

The determinants of excellent health: Different from the determinants of iII

health? International Journal of Epidemiology, 23(6), 1273-1281. Retrieved from

http://ije.oxfordjournals.org

Marsh, D. R, Schroeder, D. G., Dearden, K. A., Sternin, J., & Sternin, M. (2004). The

power of positive deviance. BMJ, 329, 1177-1179.

doi:10.1136/bmj.329.7475.1177

Marsh, D. R., & Schroeder, D. G. (2002). The positive deviance approach to improve

health outcomes: experience and evidence from the field—preface. Food and

Nutrition Bulletin, 23(4) (Suppl.), 3-5. Retrieved from

http://www.foodandnutritionbulletin.org

Marsh, D. R., Khanh, V. N., Ha, P. B., Kiem, T. T., Vu, N. A., & Jones, E. (2007). Save

the Children working paper no. 2: Acceptability, feasibility, quality, effect, and

sustainability of a “PD-Plus” approach for improving newborn, child, and

maternal care in Quang Tri province, Vietnam. Retrieved from the Positive

Deviance Initiative website:

http://www.positivedeviance.org/pdf/research/CS%2018%20OR%20Report-1.pdf
 156

Marsh, D. R., Pachón, H., Schroeder, D. G., Ha, T. T., Dearden, K., Lang, T.

T.,…Claussenius, D. (2002). Design of a prospective, randomized evaluation of

an integrated nutrition program in rural Viet Nam. Food and Nutrition Bulletin,

23(4) (Suppl.), 34-44. Retrieved from http://www.foodandnutritionbulletin.org

Marsh, D. R., Sternin, M., Khadduri, R., Ihsan, T., Nazir, T., Bari, A., & Lapping, K.

(2002). Identification of model newborn care practices through a positive

deviance inquiry to guide behavior-change interventions in Haripur, Pakistan.

Food and Nutrition Bulletin, 23(4) (Suppl.), 107-116. Retrieved from

http://www.foodandnutritionbulletin.org

Mathie, A., & Cunningham, G. (2003). From clients to citizens: Asset-based community

development as a strategy for community-driven development. Development in

Practice, 13(5), 474-486. Retrieved from http://www.developmentinpractice.org

McKenzie, J. F., Neiger, B. L., & Thackery, R. (2009) Planning, implementing &

evaluating health promotion programs: A primer (5th Ed.). San Francisco, CA:

Benjamin Cummings.

McKenzie, J. F., Pinger, R. R., & Kotecki, J. E. (2008). An introduction to community

health (6th Ed.). Sudbury, MA: Jones and Bartlett Publishers, Inc.

Milton, C., & Ochieng, O. (2007). Development through positive deviance and its

implications for economic policy making and public administration in Africa: The

case of Kenyan agricultural development, 1930–2005. World Development, 35(3),

454–479. 2007doi:10.1016/j.worlddev.2006.04.003

Minkler, M., & Wallerstein, N. (2005). Improving health through community

 157

organization and community building: A health education perspective. In

Minkler, ed. Community organizing and community building for health (2nd ed).

New Brunswick, NJ: Rutgers University Press.

Minkler, M., & Wallerstein, N. (Eds.) (2008). Community-based participatory research

for health: From process to outcomes (2nd Ed.). San Francisco, CA: John Whiley

& Sons, Publishers.

Miles, M. B., & Huberman, M. (1994). Qualitative data analysis: An expanded

sourcebook (2nd Ed.). Sage Publications: London & Thousand Oaks, California

Mustaphi, P., & Dobe, M. (2005). Positive deviance - the West Bengal experience. Indian

Journal of Public Health, 49(4), 207-13. Retrieved from http://www.ijph.in

National Institutes of Health [NIH]. (2008). Community-based participatory research

scientific interest group (CBPR-SIG). Retrieved from

http://grants.nih.gov/grants/training/esaig/cbpr_sig.htm

National Health Service [NHS] Critical Appraisal Skills Programme [CASP] (2006). 10

questions to help you make sense of qualitative research. Retrieved from

http://www.sph.nhs.uk/sph-files/casp-appraisal-

tools/Qualitative%20Appraisal%20Tool.pdf

Ndiaye, M., Siekmans, K., Haddad, S., & Receveur, O. (2009). Impact of a positive

deviance approach to improve the effectiveness of an iron-supplementation

program to control nutritional anemia among rural Senegalese pregnant women.

Food and Nutrition Bulletin, 30(2), 28-36. Retrieved from

http://www.foodandnutritionbulletin.org
 158

Oloo, H., Wanjiru, M., & Newell-Jones, M. (2011). Female genital mutilation practices

in Kenya. Retrieved from

http://www.feedtheminds.org/downloads/FGM%20Report_March2011.pdf

Onxy, J., & Bullen, P. (2000). Measuring social capital in five communities. Journal of

Applied Behavioural Science, 36(1), 23-42. Retrieved from

http://jab.sagepub.com/

Onxy, J., & Bullen, P. (2001). Chapter 4, The different faces of social capital in NSW. In

Dekker, P. & Uslaner, E.M. Social capital and participation in everyday life.

London, UK: Routledge.

Pascale, R., Sternin, J., & Sternin, M. (2010). The power of positive deviance. Boston,

MA: Harvard Business Press.

Paterson, B. L., Thorne, S. E., Canam, C., & Jillings, C. (2001). Meta-study of qualitative

health research: A practical guide to meta-analysis and meta-synthesis. Thousand

Oaks, CA: Sage.

Patterson, K., Grenny, J., Maxfield, D., McMillan R., & Switzler, A. (2007). Influencer:

The power to change anything. New York: McGraw Hill.

Patton, M. Q. (2002). Qualitative research & evaluation methods (3rd Ed). Thousand

Oaks, CA: Sage.

Petticrew, M., & Roberts, H. (2006). Systematic reviews in the social sciences: A

practical guide. Malden, MA: Blackwell Publishing.

Population Council (2008). Female genital mutilation abandonment program: Evaluation

summary report. Retrieved from

 159

http://www.unicef.org/evaldatabase/files/EGY_FGM_AP_report.pdf

Positive Deviance Initiative (2009). Basic guide to the positive deviance approach.

Retrieved from

http://www.positivedeviance.org/about_pd/Final%20revised%2006-03-

09Basic_PD_Steps_2.pdf

Positive Deviance Initiative (2010a). Voices from the field: Public health. Retrieved from

http://www.positivedeviance.org/from_the_field/voices-public-health.html

Positive Deviance Initiative (2010b). Voices from the field: Vulnerable populations.

Retrieved from http://www.positivedeviance.org/from_the_field/voices-

vulnerable-populations.html

Positive Deviance Initiative (2010c). Voices from the field: Other. Retrieved from

http://www.positivedeviance.org/from_the_field/voices-other.html

Positive Deviance Initiative (2010d). Stories: Public health. Retrieved from

http://www.positivedeviance.org/from_the_field/stories-public-health.html

Positive Deviance Initiative (2010e). Positive Deviance Initiative website. Retrieved from

http://www.positivedeviance.org/

Positive Deviance Initiative (n.d). Positive deviance approach for behavior & social

change. Retrieved from

http://www.positivedeviance.org/pdf/generic_for_website.pdf

Prochaska, J. O., & DiClemente, C. C. (2005). The transtheoretical approach. In

Norcross, J. C. & Goldfried, M. R. (Eds.)., Handbook of psychotherapy

integration (2nd ed). (pp. 47–171). New York: Oxford University Press.
 160

Riley, W. J., & Moran, J. W. (2010). Defining quality improvement in public health.

Journal of Public Health Management & Practice, 16(1), p 5–7. doi:

10.1097/PHH.0b013e3181bedb49

Rogers, E. M. (1995). Diffusion of innovations (4th ed). New York: The Free Press.

Rosenstock, I. M., Strecher V. J., & Becker, M. H. (1988). Social learning theory and the

health belief model. Health Education Quarterly, 15(2), 175-83. Retrieved from

http://hej.sagepub.com

Save the Children (n.d.). Positive deviance: A strength based approach to behavior

change and social mobilization. Retrieved from the Positive Deviance Initiative

website http://www.positivedeviance.org/projects/Brief_PD_Pakistan.pdf

Schooley, J., & Morales, L. (2007). Learning from the community to improve maternal–

child health and nutrition: The positive deviance/hearth approach. Journal of

Midwifery & Women’s Health, 52(4), 376-383. doi:10.1016/j.jmwh.2007.03.001

Schuftan, C. (1993). Positive deviance in child nutrition: a critique. Ecol. of Food and

Nutrition, 30(2). Retrieved from http://humaninfo.org/aviva/ch25.htm

Sethi, V., Kashyap, S., Aggarwal, S., Pandey, R.M., & Kondal, D. (2007). Positive

deviance determinants in young infants in rural Uttar Pradesh. Indian Journal of

Pediatrics, 74(6), 594-595. Retrieved from http://www.ijppediatricsindia.in

Sethi, V., Kashyap, S., Seth, V., & Agarwal, S. (2003). Encouraging appropriate infant

feeding practices in slums: A positive deviance approach. Pakistan Journal of

Nutrition, 2(3), 164-166. Retrieved from http://www.pjbs.org/pjnonline

Sheppard, B. H., Hartwick, J., & Warshaw, P. R. (1988). The theory of reasoned action:
 161

A meta-analysis of past research with recommendations for modifications and

future research. Journal of Consumer Research, 15, 325–343. Retrieved from

http://ejcr.org

Singhal, A., & Dura L. (2008). A positive deviance approach to child protection: A

comparative assessment of two projects. Retrieved from

http://www.positivedeviance.org/projects/Singhal-Dura-Draft-composite-9-29-

08.pdf

Singhal, A., & Greiner, K. (2007). “Do what you can, with what you have, where you

are” A quest to eliminate MRSA at the VA Pittsburgh healthcare system. Plexus

Institute, Deeper Learning, 1(4) Retrieved from the Positive Deviance Initiative

website: http://www.positivedeviance.org/pdf/publications/DoWhatYouCan.pdf

Spencer, L., Ritchie, J., Lewis, J., & Dillon, L. (2003). Quality in qualitative evaluation:

a framework for assessing research evidence. (Report ISBN: 07715 04465 8.)

Retrieved from

http://www.civilservice.gov.uk/Assets/a_quality_framework_tcm6-7314.pdf

Spreitzer, G.M., & Sonenshein, S. (2004). Toward the construct definition of positive

deviance. The American Behavioral Scientist, 47(6), 828-847. Retrieved from

http://abs.sagepub.com

Sripaipan, T., Schroeder, D. G., Marsh, D. R., Pachón, H., Dearden, K. A., Ha, T. T., &

Lang, T. T. (2002). Effect of an integrated nutrition program on child morbidity

due to respiratory infection and diarrhea in northern Viet Nam. Food and

Nutrition Bulletin, 23(4) (Suppl.), 67-74. Retrieved from

 162

http://www.foodandnutritionbulletin.org

Sternin, J. (2002). Positive deviance: A new paradigm for addressing today’s problems

today. The Journal of Corporate Citizenship, 5, 57-62. Retrieved from

http://www.greenleaf-publishing.com/default.asp?ContentID=7

Timulak, L. (2009). Meta-analysis of qualitative studies: A tool for reviewing qualitative

research findings in psychotherapy. Psychotherapy Research, 19(4-5), 591-600.

doi 10.1080/10503300802477989

Tjerbo, T. (2009). The politics of local hospital reform: a case study of hospital

reorganization following the 2002 Norwegian hospital reform. BMC Health

Services Research, 9, 212. doi:10.1186/1472-6963-9-212

The United Nations University (1990). Positive deviance in child nutrition - with

emphasis on psychosocial and behavioural aspects and implications for

development. Retrieved from

http://www.unu.edu/unupress/unupbooks/80697e/80697e00.htm

Trinh Mackintosh, U. A., Marsh, D. R., & Schroeder, D. G. (2002). Sustained positive

deviant child care practices and their effects on child growth in Viet Nam. Food

and Nutrition Bulletin, 23(4) (Suppl.), 16-25. Retrieved from

http://www.foodandnutritionbulletin.org

Vossenaar, M., Mayorga, E., Jose Soto-Mendez, M., Medina-Monchez, S. B., Campos,

R., Anderson, A. S., & Solomons, N. W. (2009). The positive deviance approach

can be used to create culturally appropriate eating guides compatible with reduced

cancer risk. The Journal of Nutrition, 139, 755–762. doi:10.3945/jn.108.100362

 163

Walker, L. O., Sterling, B. S., Hoke, M. M., & Dearden, K. A. (2007). Applying the

concept of positive deviance to public health data: A tool for reducing health

disparities. Public Health Nursing, 24(6), 571–576. Retrieved from

http://www.wiley.com/bw/journal.asp?ref=0737-1209

Walsh, D., & Downe, S. (2005). Meta-synthesis method for qualitative research: A

literature review. Journal of Advanced Nursing, 50(2), 204–211. Retrieved from

http://onlinelibrary.wiley.com/journal/10.1111/(ISSN)1365-2648

Wang, C., & Burris, M. A. (1994). Empowerment through photo novella: Portraits of

participation. Health Education & Behavior, 21, 171-186.

doi:10.1177/109019819402100204

West, B. (2004). Synergies in deviance: Revisiting the positive deviance debate.

Electronic Journal of Sociology, p0-0. Retrieved from http://www.sociology.org

Wishik, S. M., & Van Der Vynckt, S. (1976). The use of nutritional 'positive deviants' to

identify approaches for modification of dietary practices. AJPH, 66(1), 38-42.

Retrieved from http://ajph.aphapublications.org/

World Health Organization [WHO]. (2010). Equity, social determinants and public

health programmes. Retrieved from

http://whqlibdoc.who.int/publications/2010/9789241563970_eng.pdf

World Bank. (2011). Country and lending groups. Retrieved from

http://data.worldbank.org/about/country-classifications/country-and-lending-

groups#Low_income

World Bank & United Nations Population Fund [UNFPA] (2004). Female genital
 164

mutilation / cutting in Somalia. Retrieved from

http://siteresources.worldbank.org/INTSOMALIA/Data%20and%20Reference/20

316684/FGM_Final_Report.pdf

Zeitlin, M., Ghassemi, H., & Mansour, M. (1990). Positive deviance in child nutrition:

With emphasis on psychosocial and behavioral aspects and implications for

development. The United Nations University. Retrieved from

http://unu.edu/unupress/unupbooks/80697e/80697E00.htm

Zeller, K. R. (1991) Nutritional resilience in a hostile environment: positive deviance in

child nutrition. Nutrition Reviews, 49(9), 259-268. Retrieved from

http://onlinelibrary.wiley.com/journal/10.1111/(ISSN)1753-4887
 165

Appendix A: Document Data Collection Form

1 Document ID

2 Title

3 Author(s)

4 Year of Publication

5 Source

6 Volume / Issue

7 Page numbers

8 DOI

9 Date retrieved

10 URL

11 Date retrieved

12 Source type Peer reviewed journal

Gray literature source

13 Type of document

14 Topic

15 Country

16 Comments
 166

Appendix B: Program / Inquiry Data Collection Form

All fields are relevant for PD programs. Field numbers marked with an (I) are

relevant for PD inquiries.

1 Sources

1.1 (I) Type Program / Inquiry

1.2 (I) ID

1.3 (I) Related document IDs

1.4 (I) Document types

2 Program Details

2.1 (I) Country

2.2 (I) Province / area

2.3 (I) Region

2.4 (I) World Bank Income

Level

2.5 (I) Topic

2.6 (I) Objective

2.7 Target beneficiaries

2.8 (I) Setting

2.9 Program name

2.10 Program type Stand-alone

Integrated

2.11 Funding organization

 167

2.12 Implementing
organization

2.13 (I) Linked to the Sternins Yes / No

2.14 International Yes / Not reported or unknown

organization involved

2.15 National organization Yes / Not reported or unknown

involved (NGO,
university etc)

2.16 National government Yes / Not reported or unknown

involved (includes any
level of govt)

2.17 (I) Program start year

2.18 Program end year

2.19 Program length (years)

3 Positive Deviance Steps Implemented

3.1 (I) Define Yes/No/Unknown

3.2 (I) Determine Yes/No/Unknown

3.3 (I) Discover Yes/No/Unknown

3.4 Design Yes/No/Unknown

3.5 Implementation Yes/No/Unknown

3.6 Monitor Yes/No/Unknown

4 Any Community Participation in Positive Deviance Steps

4.1 (I) Define Yes/No/Unknown

4.2 (I) Determine Yes/No/Unknown

4.3 (I) Discover Yes/No/Unknown

 168

4.4 Design Yes/No/Unknown

4.5 Implementation Yes/No/Unknown

4.6 Monitor Yes/No/Unknown

5 Practical Implementation

5.1 (I) Method used to identify

PDs and their behaviors
– description

5.2 (I) Method used to identify

PDs – category

5.3 (I) Method used to discover

PD behaviors – category

5.4 (I) Elimination of “True But Yes / No

Useless” PD behaviors

5.5 (I) PD behaviors identified

– description

5.6 Method used to spread

PD behaviors to other
member of the
community – description

5.7 Method used to spread

PD behaviors – category

5.8 Community level

facilitation done by

5.9 Institutionalization
activities

6 Type of Evaluation

6.1 Includes quantitative Yes / No

evaluation
 169

6.2 Includes qualitative Yes / No

evaluation

6.3 Has a long term Yes / No

evaluation after the
program was completed

7 Evaluation Report Completeness

7.1 Description of setting / Complete/Incomplete/Missing

population

7.2 Description of PD Complete/Incomplete/ Missing

approach

7.3 Description of study Complete/Incomplete/ Missing

methodology

7.4 Description of results Complete/Incomplete/ Missing

8 Quantitative Evaluation

8.1 Testing Pre and posttest

Posttest only

8.2 Control group Yes/No

8.3 Randomization Randomized

Matched
Neither randomized nor matched
N/A – no control group

8.4 Sample size

8.5 Outcome variables

8.6 Results

8.7 Results category Positive impact

No impact
Negative impact
 170

8.8 Global quality rating Strong

(from quality assessment Moderate
tools)
Weak
Unknown

9 Qualitative Evaluation

9.1 Methods

9.2 Results

9.3 Results category Positive impact

No impact
Negative impact

9.4 Global quality rating Strong

(from quality assessment Moderate
tools)
Weak
Unknown

10 Comments
 171

Appendix C: Databases

The following databases were used to identify documents describing PD programs:

Databases accessed through the EBSCO multiple database search interface:

· CINAHL Plus with Full Text

· MEDLINE
· PsycARTICLES
· PsycINFO
· SocIndex with Full Text
· Political Science Complete
· Communications & Mass Media Complete
Databases accessed through the ProQuest multiple database search interface:

· ProQuest Nursing & Allied Health

· ProQuest Central
· ProQuest Health and Medical Complete
· ProQuest Psychology Journals
· Science Journals
· Social Science Journals
· Research Library
Databases accessed through the SAGE multiple database search interface:
· Health Sciences
· Social Science & Humanities
Multidisciplinary Databases:

· Academic Search Complete

· Web of Science
· ResearchNow
· Science Direct
 172

Appendix D: Gray Literature Sources

The following gray literature sources were used to identify documents describing PD
programs:
Gray literature databases:
· The British Library Integrated Catalogue, including conference proceedings index
· Walden University Library Catalogue
· Copac National, Academic, and Specialist Library Catalogue
· ProQuest Dissertations and Theses
· Networked Digital Library of Theses and Dissertations
· Open System for Information on Gray Literature (SIGL)
Organization websites (selected based on the findings of the proposal literature search):
· The Plexus Institute
· Positive Deviance Project Canada
· CARE
· Caritas International
· Family Health International
· OXFAM
· PLAN International
· Positive Deviance Resource Center
· Save the Children
· The Australian Government's overseas aid program (AusAID)
· UK Department for International Development (DFID)
· UNICEF
· United Nations University
· United States Agency for International Development (USAID)
· World Health Organization
· World Vision
 173

Appendix E: Quality Assessment Tool For Quantitative Studies

The following tool is based on the EPHPP Quality Assessment Tool for Quantitative
Studies (EPHPP, 2010). Some minor adjustments have been made to allow for a larger
amount of missing data in gray literature studies. The tool is to be used in conjunction
with the EPHPP Quality Assessment Tool for Quantitative Studies Dictionary (EPHPP,
2009).

COMPONENT RATINGS

A) SELECTION BIAS

(Q1) Are the individuals selected to participate in the study likely to be

representative of the target population?
Very likely
Somewhat likely
Not likely
Can’t tell

(Q2) What percentage of selected individuals agreed to participate?

80 - 100% agreement
60 – 79% agreement
less than 60% agreement
Not applicable
Can’t tell

RATE THIS SECTION

STRONG MODERATE WEAK UKNOWN
See dictionary
 174

B) STUDY DESIGN

(Q1) Indicate the study design

Randomized controlled trial
Controlled clinical trial
Cohort analytic (two group pre + post)
Case-control
Cohort (one group pre + post (before and after))
Interrupted time series
Other specify ____________________________
Can’t tell

(Q2) Was the study described as randomized? If NO, go to Component C.

No Yes

(Q3) If Yes, was the method of randomization described? (See dictionary)

No Yes

(Q4) If Yes, was the method appropriate? (See dictionary)

No Yes

RATE THIS SECTION

STRONG MODERATE WEAK UKNOWN
See dictionary

C) CONFOUNDERS

(Q1) Were there important differences between groups prior to the intervention?
Yes
No
Can’t tell
 175

The following are examples of confounders:

Race
Sex
Marital status/family
Age
SES (income or class)
Education
Health status
Pre-intervention score on outcome measure

(Q2) If yes, indicate the percentage of relevant confounders that were controlled
(either in the design (e.g. stratification, matching) or analysis)?
80 – 100% (most)
60 – 79% (some)
Less than 60% (few or none)
Can’t Tell

RATE THIS SECTION

STRONG MODERATE WEAK UKNOWN
See dictionary

D) BLINDING

(Q1) Was (were) the outcome assessor(s) aware of the intervention or exposure
status of participants?
Yes
No
Can’t tell
 176

(Q2) Were the study participants aware of the research question?

Yes
No
Can’t tell

RATE THIS SECTION

STRONG MODERATE WEAK UKNOWN
See dictionary

E) DATA COLLECTION METHODS

(Q1) Were data collection tools shown to be valid?

Yes
No
Can’t tell

(Q2) Were data collection tools shown to be reliable?

Yes
No
Can’t tell

RATE THIS SECTION

STRONG MODERATE WEAK UKNOWN
See dictionary

F) WITHDRAWALS AND DROP-OUTS

(Q1) Were withdrawals and drop-outs reported in terms of numbers and/or reasons
per group?
Yes
No
Can’t tell
 177

Not Applicable (i.e. one time surveys or interviews)

(Q2) Indicate the percentage of participants completing the study. (If the percentage
differs by groups, record the lowest).
80 -100%
60 - 79%
less than 60%
Can’t tell
Not Applicable (i.e. Retrospective case-control)

RATE THIS SECTION

STRONG MODERATE WEAK UKNOWN
See dictionary

G) INTERVENTION INTEGRITY

(Q1) What percentage of participants received the allocated intervention or

exposure of interest?
80 -100%
60 - 79%
less than 60%
Can’t tell

(Q2) Was the consistency of the intervention measured?

Yes
No
Can’t tell
 178

(Q3) Is it likely that subjects received an unintended intervention (contamination or

co-intervention) that may influence the results?
Yes
No
Can’t tell

RATE THIS SECTION

STRONG MODERATE WEAK UKNOWN
See dictionary

H) ANALYSES

(Q1) Indicate the unit of allocation (circle one)

community organization/institution practice/office individual

(Q2) Indicate the unit of analysis (circle one)

community organization/institution practice/office individual

(Q3) Are the statistical methods appropriate for the study design?
Yes
No
Can’t tell

(Q4) Is the analysis performed by intervention allocation status (i.e. intention to

treat) rather than the actual intervention received?
Yes
No
Can’t tell

RATE THIS SECTION

STRONG MODERATE WEAK UKNOWN
See dictionary
 179

GLOBAL RATING

COMPONENT RATINGS
Transcribe the information from the previous rating boxes. See dictionary on how to rate
this section

A SELECTION BIAS STRONG MODERATE WEAK UKNOWN

B STUDY DESIGN STRONG MODERATE WEAK UKNOWN
C CONFOUNDERS STRONG MODERATE WEAK UKNOWN
D BLINDING STRONG MODERATE WEAK UKNOWN
E DATA STRONG MODERATE WEAK UKNOWN
COLLECTION
METHOD
F WITHDRAWALS STRONG MODERATE WEAK UKNOWN
AND DROPOUTS
G INTERVENTION STRONG MODERATE WEAK UKNOWN
INTEGRITY
H ANALYSES STRONG MODERATE WEAK UKNOWN

GLOBAL RATING FOR THIS PAPER (circle one):

STRONG (no WEAK or UNKNOWN ratings)
MODERATE (one WEAK or UNKNOWN rating)
WEAK (two or more WEAK or UNKNOWN ratings)
UKNOWN (three or more UNKNOWN ratings)

IF the global rating for the first reviewer is UNKNWON the paper is not reviewed by a
second reviewer.

IF the global rating for the first reviewer is STRONG, MODERATE or WEAK the paper
is reviewed by a second reviewer.
 180

With both reviewers discussing the ratings: Is there a discrepancy between the two
reviewers with respect to the component (A-F) ratings? Yes / No

If yes, indicate the reason for the discrepancy

1 Oversight
2 Differences in interpretation of criteria
3 Differences in interpretation of study

Final decision of both reviewers (circle one):

STRONG
MODERATE
WEAK
UNKNOWN
 181

Appendix F: Quality Assessment Tool For Qualitative Studies

The following tool is based on the NHS CASP quality assessment questions for
qualitative studies (NHS CASP, 2006). Some minor adjustments have been made to
allow for a larger amount of missing data in gray literature studies. A global rating has
been added using the same approach as the EPHPP Quality Assessment Tool for
Quantitative Studies (EPHPP, 2010) to allow for consistency in the analysis of qualitative
and quantitative studies.

1. Was there a clear statement of the aims of the Yes / No

research? Comments:
Consider:
what the goal of the research was
why it is important
its relevance

2. Is a qualitative methodology appropriate? Yes / No / Unknown

Consider: Comments:
if the research seeks to interpret or illuminate the
actions and/or subjective experiences of research
participants

3. Was the research design appropriate to Yes / No / Unknown

address the aims of the research? Comments:
 182

4. Was the recruitment strategy appropriate to Yes / No / Unknown

the aims of the research? Comments:
Consider:

if the researcher has explained how the

participants were selected

if they explained why the participants they selected

were the most appropriate to provide access to the
type of knowledge sought by the study

if there are any discussions around recruitment (e.g.

why some people chose not to take part)

5. Were the data collected in a way that Yes / No / Unknown

addressed the research issue? Comments:
Consider:

if the setting for data collection was justified

if it is clear how data were collected (e.g. focus

group, semi-structured interview etc)

if the researcher has justified the methods chosen

if the researcher has made the methods explicit

(e.g. for interview method, is there an indication of
how interviews were conducted, did they used a
topic guide?)

if methods were modified during the study. If so,

has the researcher explained how and why?

if the form of data is clear (e.g. tape recordings,

 183

video material, notes etc)

if the researcher has discussed saturation of data

6. Has the relationship between researcher and Yes / No / Unknown

participants been adequately considered? Comments:
Consider whether it is clear if the researcher
critically examined their own role, potential bias
and influence during:

formulation of research questions

data collection, including sample recruitment and

choice of location

how the researcher responded to events during the

study and whether they considered the

implications of any changes in the research

design

7. Have ethical issues been taken into Yes / No / Unknown

consideration? Comments:
Consider:

if there are sufficient details of how the research

was explained to participants for the reader to
assess whether ethical standards were maintained

if the researcher has discussed issues raised by the

study (e. g. issues around informed consent or
confidentiality or how they have handled the effects
of the study on the participants during and after the
 184

study)

if approval has been sought from the ethics

committee

8. Was the data analysis sufficiently rigorous? Yes / No / Unknown

Consider: Comments:

if there is an in-depth description of the analysis

process

if thematic analysis is used. If so, is it clear how the

categories/themes were derived from the data?

whether the researcher explains how the data

presented were selected from the original sample to
demonstrate the analysis process

if sufficient data are presented to support the

findings to what extent contradictory data are taken
into account

whether the researcher critically examined their

own role, potential bias and influence during
analysis and selection of data for presentation

9. Is there a clear statement of findings? Yes / No / Unknown

Consider: Comments:

if the findings are explicit

if there is adequate discussion of the evidence both

for and against the researcher’s arguments

if the researcher has discussed the credibility of

 185

their findings (e.g. triangulation, respondent

validation, more than one analyst.)

if the findings are discussed in relation to the

original research questions

10. How valuable is the research? Valuable / Not valuable /

Unknown
Consider:
Comments:
if the researcher discusses the contribution the
study makes to existing knowledge or understanding
(e.g. do they consider the findings in relation to
current practice or policy, or relevant research-
based literature?)

if they identify new areas where research is

necessary

if the researchers have discussed whether or how

the findings can be transferred to other populations
or considered other ways the research may be used

GLOBAL RATING FOR THIS PAPER (circle one):

STRONG (no NO answers)
MODERATE (one NO answer)
WEAK (two or more NO answers)
UKNOWN (three or more UNKNOWN answers)

IF there are more than three questions where the answer is UNKNOWN the paper is not
reviewed by a second reviewer.
 186

IF the global rating for the first reviewer is STRONG, MODERATE or WEAK the paper
is reviewed by a second reviewer.

With both reviewers discussing the ratings: Is there a discrepancy between the two
reviewers with respect to the component (A-F) ratings? Yes / No

If yes, indicate the reason for the discrepancy

1 Oversight
2 Differences in interpretation of criteria
3 Differences in interpretation of study

Final decision of both reviewers (circle one):

STRONG
MODERATE
WEAK
 187

Appendix G: Peer Debriefing Meeting Miuntes

Meeting Minutes # 1

Date: 5 Nov 2011

Time: 4:30pm – 5:30pm Phnom Penh time (GMT+7)
Participants: Piroska Bisits Bullen, PhD student
Paul Bullen, Peer debriefer
Location: Phone call
Prepared by: Piroska Bisits Bullen
Minutes:
- I walked Paul through my oral presentation slides to update him on the final
problem statement, research questions, theoretical base, and methodology for the
proposal, and his role as auditor and peer debriefer.
- Paul and I discussed the tasks that he needs to complete as part of the peer
debriefer role / responsibilities. This includes:
o Reviewing the database structure, and auditing:
 The first 5 records in the Programs table, followed by 4 other
randomly selected records (10%) (total of 9 records).
 3 randomly selected records (10%) in the Studies table.
o Reviewing the timeline of PD programs and studies to identify patters,
particularly any related to Diffusion of Innovations theory. We will
compare the patterns that Paul identifies to the ones that I identify, and
work together to decide on the final set of trends / patterns that will be
included in the results of the dissertation.
o Reviewing the classification system I will develop for PD approaches and
identifying any improvements or changes that need to be made.
o Completing the Quantitative Quality Assessment Tool for the 7 programs
that have a complete quantitative evaluation, and completing the
Qualitative Quality Assessment Tool for the 4 programs that have a
complete qualitative evaluation. These will then be compared to the tools
that I have completed, and we will use discussion to arrive at an overall
quality rating for each study.
- Paul clarified some questions about the database structure, such as the relationship
between the Documents table, Program table and Studies table.
 188

- Paul and I agreed on a schedule of calls over the following week to complete the
required tasks. The next calls will be:
o 9 Nov 2011, 9am Phnom Penh time (GMT+7)
o 11 Nov 2011, 9am Phnom Penh time (GMT+7)
o 12 or 13 Nov 2011, time to be confirmed

Meeting Minutes # 2

Date: 9 Nov 2011

Time: 9:15am – 10:15am Phnom Penh time (GMT+7)
Participants: Piroska Bisits Bullen, PhD student
Paul Bullen, Peer debriefer
Location: Phone call
Prepared by: Piroska Bisits Bullen
Minutes:
- Paul gave his initial thoughts after reviewing the data in the timeline and doing
some rough analysis of patterns:
o The distinction between programs and studies is very important.
o Topic and location are particularly important.
o For location it would be useful to systematically group countries into
continents and socioeconomic levels.
o For topic it may be useful to group topics into three categories: child
health, adult health and other. When this is done there are lots of patterns.
- Paul and I discussed the issue of PD programs implemented in organizational
settings (schools, hospitals, prisons):
o Are they really an issue of individual behavior change like the other PD
programs, or are they related to organizational issues?
o What is the difference between PD and other organizational improvement
frameworks, such as:
 Quality improvement: Analyses a process and uses the data to
improve the process using a continuous improvement cycle.
Mainly used in manufacturing. Distinct from PD because it focuses
on the process not the individual.
 189

 Best practice: Looks at which organizations or programs are the

best and how they do it. Then tries to replicate that inside the
organization or program. Focuses on the organization or program,
rather than the individual. Meet PD criteria because the
organizations / programs investigated are those that are positively
deviated. But its focus is organizations / programs, rather than
individuals.
 Appreciative inquiry: Organizationally focused. Gets people
talking about what works well and is successful in the
organization. By telling more of those stories it hopes to get more
of it happening.
 Resilience: Looks for resilient individuals or cases and tries to
learn what made them resilient.
o There are difference conceptual models related to individuals, families,
communities and organizations (of which one sub-set are programs). It
would be useful to split programs into those that are individual focused,
and those that are organizationally focused.
o Is there a group of people using the term PD when they should be using a
conceptual framework from organizational development?
- Paul and I discussed the importance of where the programs and studies are
published:
o It is important to look at where the documents are published, since this
may affect how PD spreads.
o Some may be published in academic journals about nutrition, some in
adult health journals, some in general development publications, and some
as internal documents. In theory, the organizational programs / studies
would be published in organizational journals.
o Need to add two field to the data set: One long one to indicate all the
sources, and a short one to indicate the type of primary source.
- I described my initial set of categories for different types of PD approaches:
o The method used to identify the positive deviants and their behaviors:
 Quantitative > Qualitative: Quantitative screening followed by
qualitative inquiry to identify behaviors
 Qualitative > Quantitative: Qualitative inquiry to identify
behaviors followed by quantitative survey to see if they are related
to the outcome of interest
 190

 Individual opinion > Qualitative > Positive deviants are identified

by an individual’s opinion or convenience sample and then
qualitative inquiry is used to identify their behaviors
 Qualitative: A more rigorous qualitative approach is used to
identify positive deviants and then qualitative inquiry to identify
their behaviors.
 Quantitative: A predefined survey is administered, and then
statistical analysis is used to see which variables are related to a
positive outcome. Often uses secondary data (not sure if this really
counts as PD since the behaviors are predefined).
 Staff meetings: Staff in an organizational setting have a meeting
and decide who the positive deviants are, and what they are doing
(not sure if this really counts as PD since it is very similar to
organizational best practice / appreciative inquiry).
o The method used to spread the PD behaviors to others:
 Hearth sessions
 Mass media (e.g. TV, radio, newspaper, mass events etc)
 Peer educators / community volunteers
 Advocacy campaign (target to decision makers)
 Staff meetings
- Paul and I agreed that I would add the following fields to the data and we would
re-analyze the patterns on the next call:
o Topic group: Child nutrition, adult health, organizational, other.
o Publication sources.
o Primary publication source.
o Length of program.
o Continent.
o Country income level (using World Bank High, Middle and Low income
countries).
o Category for the method used to identify PD behaviors
o Category for the method used to spread PD behaviors to others
- The next call will be on the 10 Nov at 9am Phnom Penh time (GMT+7).
 191

Meeting Minutes #3

Date: 10 Nov 2011

Time: 9:00am – 10:15am Phnom Penh time (GMT+7)
Participants: Piroska Bisits Bullen, PhD student
Paul Bullen, Peer debriefer
Location: Phone call
Prepared by: Piroska Bisits Bullen
Minutes:
- Paul gave his thoughts after contemplating the role of PD in organizational /
institutional change:
o Paul’s search of an online management literature database for Positive
Deviance only returns approximately 20 results, compared to hundreds or
thousands of results for other similar organizational management terms
(e.g quality improvement, best practice).
o Paul’s search of an online evaluation literature database for Positive
Deviance returned no results.
o This suggests that PD is a very uncommon term in organizational
management. Perhaps it was brought into the field by people from the
international development field (e.g. the Sternins) who were not familiar
with the existing organizational management frameworks that essentially
cover the same approach as PD.
- Paul gave his thoughts after further reviewing the full data set and patterns:
o There is no doubt that the programs and inquiries are very different from
each other on a range of variables (topics, countries, methods etc).
o There are clearly four different stories related to the topic categories
discussed on the previous call:
 Child nutrition
 Adult health (e.g. weight, chronic disease, pregnancy outcomes,
HIV/AIDS, reproductive health – this may be a single group or
multiple groups)
 Organizational (e.g. school education, hospital infections, health
services prisoner wellbeing)
 192

 Sociocultural (this is a new category that I add for the programs /

inquiries that were primarily related to social or cultural issues
rather than individual behavior change e.g. child soldiers, child
trafficking, female genital mutilation, gender equity)
o If you chart the topic category, date, program / inquiry and involvement of
the Sternins as a 3D scatter plot it is very clear that the Sternins were the
first people involved at the program level of all four topic categories. It is
also clear that all the stories are interrelated.
o The other variable that is important for diffusion is where the articles were
published. Reducing the primary publication source to Nutrition Journal,
Other Journal, Internal Report and Other also shows patterns related to
topic category and program / inquiry. It seems that the majority of
inquiries are published in journals, while the majority of programs are
only reported in internal documents with some in journals (with the
exception of the organizational programs.
- During the phone call I discovered an error in the data set for PD inquiries where
the ID numbers are not aligned with the other fields. The error only affects the
data set extracted from the database, but does not affect the database itself. The
error was probably caused when the data were sorted in Excel, but Excel did not
sort all the columns of the inquiry data because a large number of columns were
blank. I will re-export the data from the database and check that it does not
contain the error so that the data can be re-analyzed.
- The “Sternins” field is missing from the extracted data set. I will add it again and
re-send the full data set.
- We decided it would be useful to group the topic categories into three levels so it
can be compared to the other types of other frameworks that are available:
o Individual behavior change: This includes child nutrition and all the other
adult health topics. It seems that at this level PD deviance may be a new
approach, not covered by existing theories and frameworks.
o Organizational / institutional change: This includes the organizational
topics (schools, hospitals, prisons). At this level PD may already be
covered by existing frameworks such as appreciative inquiry, quality
improvement, and best practice. It may not be adding anything new, and
may be confusing practitioners by being applied here.
o Social / cultural change: This includes topics focused on social or cultural
change within the community rather than individual behaviors or
organizations (child soldiers, child trafficking, female genital mutilation,
gender equity). The examples of PD used at this level are very similar to
the role model frameworks used as part of social work, and/or Social
 193

Cognitive Theory which includes the importance of modeling the desired

change. So PD may not be adding anything new, and may be confusing
practitioners by being applied here.
- We discussed the methods used to identify the positive deviants and their
behaviors, and which ones could be considered “good practice” in PD:
o Quantitative > Qualitative: Quantitative screening followed by qualitative
inquiry to identify behaviors. This could be considered good practice in
PD since it systematically identifies PDs using objective quantitative data,
and then identifies their unique behaviors using qualitative inquiry.
o Qualitative > Quantitative: Qualitative inquiry to identify behaviors
followed by quantitative survey to see if they are related to the outcome of
interest. This is more systematic than using individual opinion or
convenience sampling only, as the quantitative survey is used to validate
the findings of the interviews / focus groups. However, conceptually it
may not be as “clean” or easy to explain to community members and field
staff as a quantitative > qualitative approach.
o Individual opinion > Qualitative: Positive deviants are identified by an
individual’s opinion or convenience sample and then qualitative inquiry is
used to identify their behaviors. This is a commonly used approach,
possibly due to limited time and resources. It is not as systematic as using
a quantitative approach to identify the PDs and could potentially miss
some important behaviors or be biased by the individuals who identified
the PDs. It is a second best option if time and resources are limited, and it
does meet the Sternins’ criteria.
o Qualitative Only: A more rigorous qualitative approach is used to identify
positive deviants and then qualitative inquiry to identify their behaviors.
This could be considered good practice in PD since it systematically
identifies PD using qualitative inquiry, and then identifies their unique
behaviors also using qualitative inquiry.
o Quantitative Only: A predefined survey is administered, and then
statistical analysis is used to see which variables are related to a positive
outcome. Often uses secondary data. We agreed that this should not count
as PD since the behaviors are predefined defined by the researcher when
they write the survey instrument, while the purpose of PD is to identify
new behaviors that researchers and program staff may not have
considered.
o Staff meetings: Staff in an organizational setting have a meeting and
decide who the positive deviants are, and what they are doing. PD appears
to be covered by other organizational management frameworks, and so we
 194

agreed that this method of identifying PDs in an organizational setting

should not actually be considered PD.
- We discussed the methods used to spread the behaviors to other people, and
which ones could be considered “good practice” in PD. It seems that the Sternins
use a particular adult learning theory in their programs that must involve action
leaning. However, in reality a range of approaches seem to have been used:
o Hearth sessions: This is the most commonly used approach for child
nutrition programs. The evaluations have shown it to be successful, and it
meets the Sternins’ requirement for an “action learning” initiative.
Therefore, it could be considered good practice.
o Mass media (e.g. TV, radio, newspaper, mass events etc): This does not
meet the Sternins’ requirement for an “action learning” initiative.
However, there is at least one case study that used mass media to
successfully spread positive deviant behaviors related to reproductive
health. If future evaluations show it to be a successful method of spreading
the behaviors then it could be considered an option, particularly for
behaviors that cannot be taught through action learning.
o Peer educators / community volunteers: This does not necessarily meet the
Sternins’ requirement for an “action learning” initiative. However, there
are a few case studies of programs that successfully spread positive
deviant behaviors using peer educators. If future evaluations show it to be
a successful method of spreading the behaviors then it could be considered
an option.
o Advocacy campaign (target to decision makers): The only case study that
used this method was for female genital mutilation. Since PD programs at
the sociocultural level may be better covered by existing frameworks
rather than PD, we agreed that advocacy should not actually be considered
PD.
o Staff meetings: PD appears to be covered by other organizational
management frameworks, and so we agreed that this method of identifying
PDs in an organizational setting should not actually be considered PD.
- We discussed one article which I identified that was particularly critical of PD
because they felt that focusing on solutions which do not require additional
resources may be used as an excuse for the community not demanding the
resources that it needs:
o Paul felt that this could be a possibility in theory, but you would need to
look at the program case studies to see how often it happened in practice.
o I noted that in the majority of program case studies PD was being
implemented as part of an integrated program that normally included
 195

investments in infrastructure, training, etc and it did not seem to be a

“second best” option that was only chosen when no resources were
available.
o We agreed that it probably happens sometimes but does not appear to be
common in the PD case studies.
- Paul agreed to do the following by the next call:
o Search the organizational management databases and social work
literature he has access to see how many references there are to existing
frameworks that might already cover PD.
o Start the process of record auditing and completing the Quality
Assessment Tools.
- I agreed to fix the errors in the data set and re-send the data set, along with a
video I have of a PD sessions at a hospital which may help identify which other
organizational management frameworks may apply.
- The next call will be on the 12 Nov at 9am Phnom Penh time (GMT+7).
 196

Meeting Minutes #4

Date: 12 Nov 2011

Time: 9:15am – 10:30am Phnom Penh time (GMT+7)
Participants: Piroska Bisits Bullen, PhD student
Paul Bullen, Peer debriefer
Location: Phone call
Prepared by: Piroska Bisits Bullen
Minutes:
- Paul sent a list of ID numbers for programs and studies which had been selected
for auditing. The randomly selected ID numbers were chosen using a random
number generator. Paul did not have all the documents required to audit, so I will
need to send them to him.
- The list of program ID numbers for the Quality Assessment Tools needs to be
updated to match the list that I sent previously so that it only includes those
program evaluations for which enough information is available to complete the
tools.
- Paul reviewed the structure of all the database tables and concluded that it made
sense and was logical.
- Paul was not able to open the Program table in Version 8 of the database. I will
save it again as Microsoft Access 2007 and resend it.
- Paul raised some concerns about the completion of the Quality Assessment Tools:
o The case studies are cases of PD programs that have been evaluated by
staff in the field. Assessing the quality of an evaluation is fundamentally
different to assessing the quality of a study.
o Both tools seem to be based on certain assumptions which are more
related to the quality of studies than evaluations.
o The Qualitative Quality Assessment Tool includes Yes / No / Unknown as
answer options, when really these should be a scale. However, the
comments sections of the tool are helpful in getting people to think about a
qualitative study.
o The response options for Q10 for the Qualitative Quality Assessment Tool
in the database have an error. They should be Valuable / Not valuable
rather than Yes / No. I will fix this a re-send it.
 197

o The Quantitative Quality Assessment Tool appears to be based on the

assumption that the evaluation is an experimental study (e.g. randomized
controlled trial, cohort, etc). Some types of study designs will not fit well
into the tool.
o I confirmed that my focus was on the global rating, rather than the
individual questions. When a section of the quantitative tool was not
applicable I rated it “can’t tell” and gave an overall rating for the section
as “moderate” so that it did not affect the global rating of the study.
o Paul will complete the quality assessment tools in separate Word
documents rather than in the database so that he is doing it independently
and cannot see my ratings.
o We agreed that the quality assessment tools are not well suited to deciding
whether the program evaluations are useful or not. This is related to their
design, and also the fact that the majority of the case studies do not
include enough information to complete them.
o Therefore, I will do a separate assessment of the usefulness of the program
evaluations without the tools. It is possible that my conclusions may differ
from the conclusions reached by using the tools.
- Paul and I discussed a quote I had found from a PD practitioner in health services
who felt that the PD approach was the same as existing frameworks, but re-
packages them in an easy to use way:
o Paul and I agreed with this, and both felt that PD could be considered a
way to sell or market concepts that already exist. In particular, the stories
that go with PD (e.g. from child nutrition) are very powerful, particularly
in an organizational context.
o Paul noted that in both the title of the dissertation, and the quoted
definition of PD the reference is to communities, not to organizations. Is it
reasonable to talk about PD in an organizational setting? Or is it really
about individuals in communities?
- Paul and I agreed that the variable for method used to identify PD behaviors
should be split into two separate variables:
o Method to identify positive deviants
o Method to discover their behaviors
In theory, any combination of these is possible, even if it hasn’t been used in any
of the case studies.
- Paul and I discussed the section of the dissertation regarding the role of the
researcher:
 198

o Paul agreed that I approach the analysis from an international development

perspective, while he approaches the analysis from an organizational
management in developed countries perspective.
o Paul recommended that I include in the section a narrative about why this
dissertation exists, including how I first heard about PD, my initial trip to
Africa and challenges, insights I have found along the way etc. I agreed
that this was a good idea.
o A similar story could be told about why any of the case studies exist
and/or the story of the Sternins.
- Paul and I discussed the possibility that PD is just a marketing package that sells
existing mechanisms / motivators for individual behavior change (similar to
organizational change):
o For example, all the successful programs include a strong motivation
mechanism (e.g. parents are motivated to feed malnourished children), a
peer pressure mechanism and a feedback mechanism. All these are
important in non-PD programs too. So if the PD was removed would the
program still work?
o Paul and I agreed that applying PD to child obesity is likely to be less
effective than applying it to child malnutrition. This is because the
motivation for the parent is stronger when their child is malnourished than
when they are overweight. Also, the feedback from seeing the child
improve is faster and more dramatic for malnutrition than obesity.
- I agreed to:
o Send the full set of documents for the programs and studies that need to be
audited and quality assessed.
o Re-save the database in 2007 and re-send it.
- Paul agreed to:
o Update me on progress by the end of 13th Nov.
o Try to complete the auditing and quality assessment by 20th Nov.
- The next call will be after the auditing and quality assessment tools are complete,
or before then if Paul has any questions or concerns regarding this.
 199

Meeting Minutes #5
Date: 19 Nov 2011
Time: 4:30pm – 5:30pm Phnom Penh time (GMT+7)
Participants: Piroska Bisits Bullen, PhD student
Paul Bullen, Peer debriefer
Location: Phone call
Prepared by: Piroska Bisits Bullen
Minutes:
- Paul sent the results of the audit, which we reviewed together. This included
general questions / issues, specific comments on the selected programs and
studies that were audited, and broader philosophical issues that he identified while
reviewing case studies during the audit.
- The general questions / issues raised were:
o How were the country income levels determined? These were determined
using the World Bank income categories. The World Bank classified
countries into High, Middle and Low income. Within these classifications
there are sub-classifications (e.g. High income OECD and non-OECD,
High-middle income, Low-middle income etc). Only the three broad
classifications (High, Middle, and Low) were used for simplicity. Paul
will not check the World Bank classifications during the audit as these
were assigned using an automated lookup of the country name.
o How was the link to the Sternins determined? The link to the Sternins was
determined in one of three ways: 1) The Sternins were listed as authors on
the document or their role in the program was described in the text of the
document; 2) The program was implemented by Save the Children, which
is the international NGO where the Sternins worked for many years and
first implemented PD; 3) Another document not included in the case study
(such as the Sternins own book) mentions their role in the program.
o How is the involvement of the national government determined? What
level of involvement do they need to have for them to be considered
“involved” (e.g. organizing organization, complementing organization)?
The national government was considered “involved” if there was any
involvement mentioned at all in the documents. This includes the
government organizing the program, implementing the program, training
staff, linking to the program, complementing the program, or being
involved as a stakeholder or in M&E.
 200

o What is the meaning of the stages Define, Determine, Discover etc? These
stages are taken from the document by Pascal, Sternin & Sternin (2010):
 1) Define the problem and desired outcome.
 2) Determine common practices.
 3) Discover uncommon but successful behaviors and strategies
through inquiry and observation.
 4) Design an action learning initiative based on the findings.
The stages of Implementation and Monitoring were added based on the
testing at the beginning of this year. I have included all interpretations of
the stages, even when they use different techniques to those underlined
above (e.g. using mass media rather than action learning).
o When is a program considered to have qualitative evaluation? I have
completed the qualitative evaluation section for any programs which
explicitly include interviews, focus groups, participatory drawing etc for
the purpose of evaluating or monitoring the program in any way. This
includes process, outcome and impact monitoring. General comments or
observations by the document author are not included as qualitative
evaluation.
- Paul and I reviewed the specific comments for each program and study audited.
See the Audit Log for a detailed list of these comments and the corrective actions
taken. Program 6 was particularly challenging as it had many sub-programs. Paul
and I agreed that the most complete sub-program (Document 135) would be used
as an example to complete the program field, and the most complete evaluation
would be used to complete the evaluation fields (Documents 15 and 134). I will
add a special commentary on Program 6 in the discussion to ensure that it receives
enough weight – otherwise it is easy to overlook this very large and complex
program because all the sub-programs have been combined into only one case
study.
- The main philosophical issue that Paul raised after reviewing the audit case
studies in detail is that none of the case studies he reviewed were
methodologically sound in how they identified the positive deviant behaviors, and
what prior knowledge is required to do this. After discussion our conclusions
were:
o Many programs state that they are fully participatory and community
members identified the positive deviant behaviors that the positive
deviants were practicing.
o However, it is unclear what frame of reference was used to identify these
behaviors (e.g. traditional practice, western medicine etc).
 201

o It is also unclear what role program staff played in filtering out suggested
behaviors that did not have scientific evidence according to the western
model of medicine, or adding suggested behaviors that did have scientific
evidence. Based on both of our experience it is highly likely that some
level of filtering by program staff occurred, but this was not described in
any of the methodologies.
o There were a two examples of this issue being raised in programs related
to hospital acquired infections. One in which kitchen staff suggested
cleaning kitchen utensils more thoroughly, and the program staff did not
tell them that it was unlikely for MRSA to be transmitted this way. In
another example hospital staff suggested that all doctors should have
stethoscope covers. Rather than telling them this was not evidence based,
the program staff suggested they research it more thoroughly, and the
result was that they found that stethoscope covers actually transmitted
MRSA.
- In addition to the issue of filtering, it was also clear that the majority of programs
and studies did not include community members in the first step of defining the
problem. They were either clearly not involved, or the method to define the
problem was not described, in which case it is assumed they were probably not
involved. This is a very critical issue.
- It appears that PD may not be what it seems. It claims to be fully participatory,
placing it at the top of the participation ladder with full community ownership.
However, most programs do not involve community members at the most critical
first step, and there is likely to be filtering of community members suggestions
occurring which is not being described in the methodologies. This suggests that
PD may not be at the top of the participation ladder, but rather, several rungs
down.
- We agreed to add an additional field to the database to show the scale of the
program using the total number of participants by order of magnitude (e.g. 10s,
100s, 1000s…100,000s etc).
 202

Meeting Minutes #6
Date: 28 Nov 2011
Time: 9:00pm – 11:00pm Phnom Penh time (GMT+7)
Participants: Piroska Bisits Bullen, PhD student
Paul Bullen, Peer debriefer
Location: Phnom Penh, face-to-face meeting
Prepared by: Piroska Bisits Bullen
Minutes:
- Paul presented his results from the independently completed quality assessment
tools. We compared his ratings to my ratings and discussed the discrepancies to
arrive at a final rating. See the Quality Assessment Log for details of this process.
- Paul raise several important issues based on his experience completing the quality
assessment tools:
o All the evaluations are different, and have a different intended purpose.
For example, there are impact studies, evaluations, case studies and
program descriptions. All are described as “evaluations” but are actually
doing different things.
o There are no actual research studies, conducted solely for the purpose of
research. This means that PD has a lack of research base, and so the
program evaluations are trying to fill the research gap.
o The questions in the two quality assessment tools have been designed to
assess the quality of research, not program evaluations. This makes it
particularly difficult / inappropriate to apply them to some of the program
evaluations.
 An evaluation tries to determine the value of something.
 Research tries to build knowledge.
 While they can be interconnected they are quite distinct.
o There are some cases where a strict application of the quality assessment
tools would result in a weak rating, but in fact it is a strong evaluation.
Should we treat is as a bad piece of research or a good evaluation? We
have decided to treat the document as what it claims to be, using a slightly
flexible interpretation of the quality assessment tool criteria. For example,
if it claims to be a program evaluation rather than a research study then it
will be assessed as such, and some of the criteria may be considered not
applicable.
 203

o The evaluations are not typically explicit about what population they are
generalizing to (which is probably connected to the fact they are program
evaluations as distinct to research studies). For example, are they
generalizing to other people in the program, other villages, etc. That
question has a big impact on judgment you make about the analysis
strategy.
o Some evaluations have multiple parts with different study designs. We
agreed to use the study design which receives the highest quality score as
the one to complete the quality assessment tool.
o The section on blinding in the quantitative quality assessment tool is not
applicable to many of the studies, such as those on child nutrition, where it
is impossible to conduct a double blind trial. The section on ethical
approval is also not relevant for program evaluations conducted by
program staffing in the same was as ethical approval for research studies
by researchers independent of the program. So should these criteria be
used as something which lowers the quality of the research, or should it be
marked not applicable? We agreed to mark it not applicable.
o The unit of allocation and unit of analysis are often not clear. Most studies
claim the unit of allocation and analysis is the individual, but in reality it is
probably the community since all members of a village do not act
independently of each other. The ambiguities here probably also relate to
the issued noted above about the difference between research studies and
program evaluations. A change in perspective could change the sample
size and power of the analysis in these studies.
 204

Appendix H: Search Results

The search results for different databases often contained the same studies, so the full text
of the relevant studies was only retrieved once – the first time that it was identified.
Total # of
Total # of
full text
results
Search Type of documents
Database (including
Order Database retrieved for
duplicate
further
results)
assessment
1 CINAHL Plus with Full Text Peer-reviewed 36 31
2 MEDLINE Peer-reviewed 57 29
3 PsycARTICLES Peer-reviewed 0 0
4 PsycINFO Peer-reviewed 37 2
5 SocIndex with Full Text Peer-reviewed 28 1
6 Political Science Complete Peer-reviewed 2 0
7 Communications & Mass Media Complete Peer-reviewed 3 1
8 ProQuest Nursing & Allied Health Peer-reviewed 20 0
9 ProQuest Central Peer-reviewed 84 14
10 ProQuest Health and Medical Complete Peer-reviewed 27 0
11 ProQuest Psychology Journals Peer-reviewed 6 0
12 ProQuest Science Journals Peer-reviewed 5 0
13 ProQuest Social Science Journals Peer-reviewed 5 0
14 ProQuest Research Library Peer-reviewed 33 0
15 Health Sciences Peer-reviewed 13 3
16 Social Science & Humanities Peer-reviewed 85 2
17 Academic Search Complete Peer-reviewed 42 2
19 ResearchNow Peer-reviewed 0 0
20 Science Direct Peer-reviewed 14 6
21 Food and Nutrition Bulletin Peer-reviewed 17 9
22 Web of Science - Regular search Peer-reviewed 45 15
23 Web of Science - Cited reference search Peer-reviewed 28 1
24 Positive Deviance Initiative website Gray literature 276 147
25 The British Library Integrated Catalogue Gray literature 3 0
26 Walden University Library Catalogue Gray literature 0 0
Copac National, Academic, and Specialist
27 Gray literature 27 2
Library Catalogue
28 ProQuest Dissertations and Theses Gray literature 14 8
Networked Digital Library of Theses and
29 Gray literature 1 0
Dissertations
 205

Total # of
Total # of
full text
results
Search Type of documents
Database (including
Order Database retrieved for
duplicate
further
results)
assessment
Open System for Information on Gray
30 Gray literature 0 0
Literature (SIGL)
31 The Plexus Institute website Gray literature 24 7
32 Positive Deviance Project Canada website Gray literature 1 1
33 CARE International websites Gray literature 4 0
34 Caritas International website Gray literature 0 0
35 Family Health International website Gray literature 1 1
36 OXFAM website Gray literature 0 0
37 PLAN International website Gray literature 3 0
38 Positive Deviance Resource Center website Gray literature 1 1
39 AusAID website Gray literature 0 0
40 DFID website Gray literature 3 1
41 USAID website Gray literature 157 45
42 BASICS website Gray literature 10 4
43 CORE Group website Gray literature 23 12
42 World Health Organization website Gray literature 40 6
43 UNICEF website Gray literature 35 12
44 United Nations University website Gray literature 3 1
45 World Vision website Gray literature 6 3
46 Save the Children website Gray literature 6 1
47 The Power of Positive Deviance Book Gray literature 22 10
48 Google Search Engine Gray literature 4 4
49 Reference lists of all identified articles Gray literature 1 1
50 Other sources Gray literature 7 7
 Appendix I: Final Sample Of PD Programs And Inquiries

PD programs included in the final sample

No. documents

Periodical Article
Program Report
Journal Article
Peer Reviewed

News Article
Presentation

Dissertation
Program Program

Materials
Webpage

Manual /
Training

Thesis /
N Topic Country Setting

Poster
Video

Total
Book
start year end year

1 Child nutrition Afghanistan Rural community 2005 2007 3 3

2 Child nutrition Bangladesh Rural community 1995 1995 1 1
3 Child nutrition Bangladesh Rural community 1997 1998 1 1
4 Child nutrition Guinea Rural community 2000 2007 1 1 2
5 Child nutrition Haiti Rural community 2000 2002 1 1
6 Child nutrition Haiti Rural community 1993 1997 1 1
7 Child nutrition India Rural and Urban 2000 2011 7 2 4 1 1 1 16
8 Child nutrition Indonesia Rural and Urban 2003 2008 8 3 1 2 14
9 Child nutrition Kenya Rural community 2004 2009 2 2
10 Child nutrition Madagascar Rural community 2004 2004 1 1
11 Child nutrition Malawi Rural community 2005 2009 1 1
12 Child nutrition Mali Rural community 1999 2000 2 1 1 4
13 Child nutrition Mali Rural and Urban 2005 2008 1 1 2
14 Child nutrition Nigeria Rural community 2007 2007 1 1
15 Child nutrition Philippines Rural community 2002 2003 1 1
16 Child nutrition Rwanda Rural community 2004 2006 1 1 2
17 Child nutrition Rwanda Rural community 2008 2010 1 1
18 Child nutrition Tajikistan Rural community 2004 2007 1 1

206
 No. documents

Periodical Article
Program Report
Journal Article
Peer Reviewed

News Article
Presentation

Dissertation
Program Program

Materials
Webpage

Manual /
Training

Thesis /
N Topic Country Setting

Poster
Video

Total
Book
start year end year

19 Child nutrition Vietnam Rural community 1990 2007 7 3 1 1 1 1 2 16

20 Child nutrition Zambia Rural community 2008 2008 2 2
Child trafficking and
21 Uganda Rural community 2007 2008 2 1 1 1 1 6
soldiers
Child trafficking and
22 Indonesia Rural community 2003 2008 1 2 1 1 1 6
soldiers
Female genital
23 Egypt Rural and Urban 1998 2006 1 6 2 1 1 1 1 1 14
mutilation
24 Health services UK Hospital/clinic 2008 2008 1 1
25 Health services Pakistan Hospital, School 2010 2010 1 1
HIV/AIDS &
26 Cote d'Ivoire Urban community 2001 2003 1 1 1 3
reproductive health
Hospital acquired
27 Brazil Hospital/clinic 2008 2009 3 1 4
infections
Hospital acquired
28 Canada Hospital/clinic 2009 2011 1 2 2 1 2 1 9
infections
Hospital acquired
29 Colombia Hospital/clinic 2005 2007 2 2 1 5
infections
Hospital acquired
30 United States Hospital/clinic 2005 2007 8 4 12 9 1 2 2 38
infections
Hospital acquired
31 United States Hospital/clinic 2006 2008 1 1 1 3
infections
Hospital acquired
32 United States Hospital/clinic 2005 2008 1 1 2
infections
33 Pregnancy outcomes Senegal Rural community 2003 2005 1 1
34 Pregnancy outcomes Egypt Rural community 2003 2004 2 1 3

207
 No. documents

Periodical Article
Program Report
Journal Article
Peer Reviewed

News Article
Presentation

Dissertation
Program Program

Materials
Webpage

Manual /
Training

Thesis /
N Topic Country Setting

Poster
Video

Total
Book
start year end year

35 Pregnancy outcomes Guinea Rural community 2002 2004 1 2 1 4

36 Pregnancy outcomes Pakistan Rural community 2002 2004 1 1 1 1 3 1 1 9
37 Prisoner wellbeing Denmark Prison 2009 2011 1 2 1 1 5
38 School education Argentina School 2003 2004 1 1 1 3

39 School education Ethiopia School 2004 2009 2 2

40 Smoking Australia Prison 2006 2008 1 1

PD inquiries included in the final sample

No. documents

Periodical Article
Program Report
Journal Article
Peer Reviewed

News Article
Presentation

Dissertation
Materials
Webpage

Manual /
Training

Thesis /
N Topic Country Setting Year

Poster
Video

Total
Book
1 Child nutrition Bangladesh Rural community 1997 1 1
2 Child nutrition China Rural community 1993 1 1
3 Child nutrition Ghana Rural community 2007 1 1
4 Child nutrition India Rural community 1991 1 1
5 Child nutrition India Rural community 1992 1 1
6 Child nutrition India Rural community 1996 1 1

208
 No. documents

Periodical Article
Program Report
Journal Article
Peer Reviewed

News Article
Presentation

Dissertation
Materials
Webpage

Manual /
Training

Thesis /
N Topic Country Setting Year

Poster
Video

Total
Book
7 Child nutrition India Urban community 2004 1 1
8 Child nutrition India Rural and urban 2007 1 1
9 Child nutrition India Urban community 2007 1 1
10 Child nutrition Mexico Rural and urban 2005 1 1
11 Child nutrition Morocco Urban community 1991 1 1
12 Child nutrition Mozambique Rural community 2005 1 1
13 Child nutrition Pakistan Rural community 2002 1 1
a
14 Child nutrition Caribbean country Unknown 1976 1 1
15 Child nutrition Vietnam Rural community 2002 1 1
16 Child nutrition Vietnam Rural community 2002 1 1
17 Child nutrition Vietnam Rural community 2002 1 1
18 Child nutrition b
India Rural community 2001 1 1
19 Chronic diseases United States Unknown 2011 1 1
Diet and weight
20 Guatemala Rural and urban 2009 1 1
control
Diet and weight
21 Guatemala Rural and urban 2010 1 1
control
Diet and weight
22 United States Urban community 2010 1 1
control
23 Gender equity Brazil Urban community 2000 1 1
24 Health services United States Hospital/clinic 2010 1 1
25 Health services United States Hospital/clinic 1995 1 1

209
 No. documents

Periodical Article
Program Report
Journal Article
Peer Reviewed

News Article
Presentation

Dissertation
Materials
Webpage

Manual /
Training

Thesis /
N Topic Country Setting Year

Poster
Video

Total
Book
HIV/AIDS &
26 Rwanda Rural, Urban 2002 1 1
reproductive health
HIV/AIDS &
27 United States Urban community 2008 1 1
reproductive health
HIV/AIDS &
28 United States Urban community 2010 1 1
reproductive health
HIV/AIDS &
29 Vietnam Urban community 2002 1 1
reproductive health
HIV/AIDS &
30 Indonesia Hospital/clinic 2008 1 1
reproductive health
31 Pregnancy outcomes United States Urban community 2005 1 1
32 Pregnancy outcomes United States Urban community 2008 1 1

a
Specific Caribbean country was not identified in the study.
b
Study included a wider range of child development issues in addition to child nutrition, but has been classified as child nutrition for
practical reasons.

210
 140
Appendix J: Examples Of Completed Data Collection Forms

Document Data Collection Form

1 Document ID 126

2 Title Child Survival (CS19) Afghanistan Midterm Evaluation

Report. Excerpted sections related to Positive Deviance
Approach to Child Nutrition.

3 Author(s) Ihsan, T.

4 Year of Publication 2006

5 Source Positive Deviance Initiative website

6 Volume / Issue n/a

7 Page numbers n/a

8 DOI n/a

9 Date retrieved 4 Nov 2010

10 URL http://www.positivedeviance.org/projects/CS19_MTE_PD_Hearth_excerpt.pdf

12 Source type Gray literature source

13 Type of document Program Report

14 Topic Child nutrition

15 Country Afghanistan

16 Comments

n/a
 141
Program / Inquiry Data Collection Form (for PD Program)

1 Sources

11.1 (I) Type Program

1.1 (I) ID 32

1.2 (I) Related document IDs 132, 262, 395, 396

1.3 (I) Document types Internal report, video, website

2 Program Details

2.1 (I) Country Mali

2.2 (I) Province / area Sogola and Fulabula

2.3 (I) Region Sub-Saharan Africa

2.4 (I) World Bank Income Low income

Level

2.5 (I) Topic Child nutrition

2.6 (I) Objective To reduce the prevalence of child malnutrition

2.7 Target beneficiaries Mothers and carers in the target areas with
malnourished children

2.8 (I) Setting Rural community

2.9 Program name Save the Children PD/Hearth Program Mali

2.10 Program type Stand-alone

Integrated

2.11 Funding organization USAID

2.12 Implementing Save the Children

organization

2.13 (I) Linked to the Sternins Yes

2.14 International Yes

organization involved
 142

2.15 National organization Not reported or unknown

involved (NGO,
university etc)

2.16 National government Not reported or unknown

involved (includes any
level of govt)

2.17 (I) Program start year 1999

2.18 Program end year 2000

2.19 Program length (years) 1

3 Positive Deviance Steps Implemented

3.1 (I) Define Yes

3.2 (I) Determine Yes

3.2 (I) Discover Yes

3.3 Design Yes

3.4 Implementation Yes

3.5 Monitor Yes

4 Any Community Participation in Positive Deviance Steps

4.1 (I) Define Not reported / Unknown

4.2 (I) Determine Yes

4.2 (I) Discover Yes

4.3 Design Yes

4.4 Implementation Yes

4.5 Monitor Yes

5 Practical Implementation
 143

5.1 (I) Method used to identify 1. All children in the target group were weighed
PDs and their behaviors 2. A list of well-nourished children was made
– description
3. Wealth/poverty indicators were measured
4. Those children who were well nourished and
who came from relatively poor families were
identified
5. Home visits were conducted to identify the
strategies that parents of PD children were using.

5.2 (I) Method used to identify Quantitative screening

PDs – category

5.3 (I) Method used to discover Qualitative inquiry

PD behaviors – category

5.4 (I) Elimination of “True But Not reported / Unknown

Useless” PD behaviors

5.5 (I) PD behaviors identified - Breastfeeding until 24 months

– description - Introduction of porridge with shea butter
- Giving enriched porridge when the child starts to
crawl
- Child eats 6 times per day
- Child does not eat food that has called on the
ground
- Child eats until he/she is full
- No nutritional taboos
- Father plays an active role in care giving
- Father and grandparents supervision
- Supervision of child at all times by an
experienced person
- Active feeding
- Child lives in a stable environment
- Child's hands are washed before and after a meal
- Child is fully vaccinated
- Gives liquids when the child has diarrhea
 144
- Any adult can get medical care for the child
- Regular follow-up of child by father
- Membership in CSCOM

5.6 Method used to spread Hearth sessions in which mothers practice the PD
PD behaviors to other behaviors together by cooking nutritious meals and
member of the feeding their children together.
community – description

5.7 Method used to spread Action learning (Hearth)

PD behaviors – category

5.8 Community level Save the Children staff and Health Committee
facilitation done by comprised of community members

5.9 Institutionalization Not described in detail. State that it is "integrated"

activities into the village activities, but it is unclear exactly
how.

6 Type of Evaluation

6.1 Includes quantitative Yes

evaluation

6.2 Includes qualitative No

evaluation

6.3 Has a long term No

evaluation after the
program was completed

7 Evaluation Report Completeness

7.1 Description of setting / Missing

population

7.2 Description of PD Incomplete

approach

7.3 Description of study Missing

methodology

7.4 Description of results Incomplete

8 Quantitative Evaluation
 145

8.1 Testing Pre and posttest

8.2 Control group No

8.3 Randomization N/A – no control group

8.4 Sample size Two villages (number of children not specified)

8.5 Outcome variables Nutrition status

8.6 Results Results from programs in both villages revealed a

decrease in moderate malnutrition from 76% to
19% and a decrease in severe malnutrition from
24% to 7% , between October 1999 to October
2000.

8.7 Results category Positive impact

8.8 Global quality rating Unknown

(from quality assessment
tools)

9 Qualitative Evaluation

9.1 Methods n/a

9.2 Results n/a

9.3 Results category n/a

9.4 Global quality rating n/a

(from quality assessment
tools)

10 Comments

Initial workshop was conducted by Monique Sternin. The PD methodology is based on

workshop notes where Monique Sternin presented the PD approach, but may not actually
be what was implemented.
 146
Program / Inquiry Data Collection Form (for PD Inquiry)

1 Sources

1.1 (I) Type Inquiry

1.2 (I) ID 8

1.3 (I) Related document IDs 8

1.4 (I) Document types Peer reviewed journal article

2 Program Details

2.1 (I) Country United States

2.2 (I) Province / area n/a

2.3 (I) Region North America

2.4 (I) World Bank Income High income

Level

2.5 (I) Topic Pregnancy nutrition

2.6 (I) Objective Identify strategies that enabled some low-income

pregnant women to eat healthy meals while others
do not.

2.7 Target beneficiaries n/a

2.8 (I) Setting Urban community

2.9 Program name n/a

2.10 Program type n/a

2.11 Funding organization n/a

2.12 Implementing n/a

organization

2.13 (I) Linked to the Sternins No

2.14 International n/a

organization involved
 147

2.15 National organization n/a

involved (NGO,
university etc)

2.16 National government n/a

involved (includes any
level of govt)

2.17 (I) Program start year 2005

2.18 Program end year n/a

2.19 Program length (years) n/a

3 Positive Deviance Steps Implemented

3.1 (I) Define Yes

3.2 (I) Determine Yes

3.2 (I) Discover Yes

3.3 Design n/a

3.4 Implementation n/a

3.5 Monitor n/a

4 Any Community Participation in Positive Deviance Steps

4.1 (I) Define No

4.2 (I) Determine No

4.2 (I) Discover No

4.3 Design n/a

4.4 Implementation n/a

4.5 Monitor n/a

5 Practical Implementation
 148

5.1 (I) Method used to identify Low-income pregnant women were recruited to the
PDs and their behaviors study using a convenience sampling process
– description through the federal WIC program. Women were
interviewed and completed demographic and
psychosocial instruments. Information was on the
women’s age, educational, level, marital status,
income, ethnicity, and current weight, number of
previous pregnancies, number of live children, and
smoking patterns. Dietary intake was based on a
self-reported 24-hour dietary recall and was used
to calculate a dietary score. 6 of 18 women were
identified as PD. The transcripts and field notes
from the interview sessions were analyzed to
identify nutritional behaviors and practices of PD
women compared to women with inadequate
nutritional scores.

5.2 (I) Method used to identify Individual choice / convenience

PDs – category

5.3 (I) Method used to discover Qualitative inquiry

PD behaviors – category

5.4 (I) Elimination of “True But No

Useless” PD behaviors

5.5 (I) PD behaviors identified Women with healthy diets knew to eat balanced
– description meals, had family support, were willing to prepare
foods that were different than other family
members, and ate at home more frequently than
women with unhealthy diets.

5.6 Method used to spread n/a

PD behaviors to other
member of the
community – description

5.7 Method used to spread n/a

PD behaviors – category

5.8 Community level n/a

facilitation done by

5.9 Institutionalization n/a

activities
 149

6 Type of Evaluation

6.1 Includes quantitative n/a

evaluation

6.2 Includes qualitative n/a

evaluation

6.3 Has a long term n/a

evaluation after the
program was completed

7 Evaluation Report Completeness

7.1 Description of setting / n/a

population

7.2 Description of PD n/a

approach

7.3 Description of study n/a

methodology

7.4 Description of results n/a

8 Quantitative Evaluation

8.1 Testing n/a

8.2 Control group n/a

8.3 Randomization n/a

8.4 Sample size n/a

8.5 Outcome variables n/a

8.6 Results n/a

8.7 Results category n/a

8.8 Global quality rating n/a

(from quality assessment
tools)

9 Qualitative Evaluation
 150

9.1 Methods n/a

9.2 Results n/a

9.3 Results category n/a

9.4 Global quality rating n/a

(from quality assessment
tools)

10 Comments

n/a
 Appendix K: Audit Log

Date : 21 Nov 2011

Auditor: Paul Bullen, Peer debriefer
Audit Log prepared by: Piroska Bisits Bullen, PhD student

Note: N is the case study reference number used in this document. Database ID is the
unique identifier used in the Microsoft Access Database.

Programs – First five programs

N Database Field Auditor Corrective Action Taken

ID Comment
4 4 2.16 Is the government No action taken. After discussion
National actually involved it was agreed that any mention of
government in implementation government, even if was only
involved or is the program very brief, would be counted as
only government involvement.
complementing
government
programs.
4 4 2.17 Start date should Start date changed to 2000. The
Program be 2000 rather original date of 2001 was based
start year than 2001 on the first set of data discussed
in the body of the text, but it
appears from the table on Page 5
that program activities actually
started in 2000.
4 4 8.4 Sample Sample size – it is The sample size description was
size unclear whether changed to between 65 and 400
there are actually per year.
more than 400
families involved,
or whether it is
the same children
participating each
year.
 152

N Database Field Auditor Corrective Action Taken

ID Comment
3 5 No comments
3 6 All Multiple studies After discussion we agreed that
with multiple document # 135 would be used
levels of to complete the program record
sophistication. as the most complete example of
Which part is the many implementations
recorded in the included in this program. The
Program? most rigorous study design
available (pre and posttest with
control group - docs 15 and 134)
will be used to complete the
evaluation section.
3 6 2.17 / 2.18 Is the timeframe No changes made. After
Program of 2000-2011 discussion we agreed to use the
start /end correct? start date of the earliest
year implementation and the end date
of the latest implementation,
which is 2000 – 2011.
3 6 2.2 Dakshin Dinajpur Added
Province / should be
area included
3 6 2.9 Program May be better Changed to "Banchete Shekha
name "Banchete Shekha Process (ICDS)"
Process" as ICDS
is also an
Organization
3 6 2.11 Cannot see a Left as is. The reference to
Funding reference to USAID is from another sub-
organization USAID. UNICEF program but included to allow
is included. the total number of programs
with funding by USAID to be
calculated.
3 6 Evaluation Is this correct for No changes made. Agreed that
section Doc 135 or is it the evaluation is completed using
from another document 15 and 134 as they
study? have a more rigorous study
design than 135. A comment will
be included to reflect this.
 153

N Database Field Auditor Corrective Action Taken

ID Comment
14 7 2.8 Setting Setting marked as Correction due to typo. Change
Rural/Urban, but made.
should be rural
only.
14 7 2.16 National Correction made.
National government
government involved should
involved be ticked.
14 7 2.17 Did the program Agreed to leave as is, since the
Program start in 2003 or body of the text suggests that
start year 2007 while the overall program ran
from 2003, the Positive Deviance
part only ran from 2007. A
comment will be added to the
program record to clarify this.
14 7 8.4 Sample The sample size A comment has been added to
size lists a first and clarify this.
second round of
beneficiaries. It is
unclear whether
the second round
was conducted in
the same location,
with the same
beneficiaries, or
in a different
location.
18 8 2.2 Need to add Aini Added
Province / District
area
18 8 2.16 It appears that the Correction made. Agreed to spot
National Ministry of check some more of the national
government Health was government involved fields to
involved involved, and so ensure they are consistent.
this field should
be ticked.
 154

N Database Field Auditor Corrective Action Taken

ID Comment
18 8 2.17 Did the program No change made. The overall
Program start in 2004 or program started in 2002, but the
start year 2002? Positive Deviance part only
started in 2004. A comment will
be added to clarify this.
18 8 8.4 Sample Was the sample No change made. The target
size size 131 group for the overall program
communities or was 131 communities, but
338 children? Positive Deviance was only used
for 338 children.

Programs – 10% of programs randomly selected

N Database Field Auditor Corrective Action Taken

ID Comment
12 32 2.2 Need to add Added
Province / Bougouni
area
25 63 2.8 Setting Is University or No change made. We agreed it is
School the correct only a language issue between
term to use? Australian English and American
English (where “School” can be
used to refer to university).
25 63 2.17 Is the 2010 start No change made, but a comment
Program date implied from has been added to clarify that the
start year the date of start date is assumed from the
publication? date of publication of the
document.
34 66 All This program Agreed that Document 28 will be
appears to have used to complete the program
two linked sub- form, as it is the latest document
programs. Which and includes the implementation
one is used to part of the program, while the
complete the earlier document only includes
program form? the PD inquiry.
 155

N Database Field Auditor Corrective Action Taken

ID Comment
34 66 2.9 Is IMPRESS the No change made. Agreed that
Program program name, or IMPRESS is one of the
name one of the interventions, but since there is
interventions? no program name in the
document it will be used as the
program name since it is the most
memorable acronym from the
program.

PD Inquiries – 10% of inquiries randomly selected

N Database Field Auditor Corrective Action Taken

ID Comment
30 7 2.8 Setting It is based in a No change made. It was agreed
clinic, so is that all healthcare settings will be
“Hospital/Clinic” grouped together (including
the best setting hospitals and clinics) since
category. different types of healthcare
settings are not necessarily
comparable between developing
and developed countries (e.g. a
hospital in a developing country
can be more similar to a clinic in
a developed country).
30 7 4.1, 4.2, 4.3 Should this study No change made. We agreed that
Participation be marked as a study will only be marked as
including including participation if
participation community members are
since PD nurses involved in implementing the
were interviewed study, not if they are only
to discover their participants in the study (i.e. the
behaviors? study is being done to them).
 156

N Database Field Auditor Corrective Action Taken

ID Comment
18 30 2.5 Topic The topic is child No change made. Agreed to
development not include a small number of child
child nutrition. development programs / studies
in the child nutrition category for
clarity. A note will be made of
this in the results section of the
dissertation.
15 93* No comments

* The actual number randomly selected was 32. Since there was no document # 32,
document #31 was selected as the closest number to the one randomly chosen. Document
#31 contains multiple studies, so one study (#93) was randomly selected for audit.
 157
Appendix L: Quality Assessment Log

Date : 29 Nov 2011

Researcher: Piroska Bisits Bullen, PhD student
Peer Reviewer: Paul Bullen, Peer debriefer
Log prepared by: Piroska Bisits Bullen, PhD student

This log compares the overall results of the quality assessment tools completed by the
researcher and peer reviewer, and explains the final decision on the overall quality rating
for quantitative and qualitative evaluations.

Discrepancies in the answers to individual questions nearly always related to the issues
discussed in the meeting minutes from 29 Nov 2011 in Appendix K.

Note: N is the case study reference number used in this document. Database ID is the
unique identifier used in the Microsoft Access Database.

Quantitative Evaluations

N Database Global Quality Rating Comments

ID
Researcher Peer Debriefer Final
3 5 Moderate Moderate Moderate No discrepancy between
the ratings.
2 18 Moderate Moderate Moderate No discrepancy between
the ratings. If it is assessed
as a descriptive study then
a moderate rating is
appropriate, but if it is
assessed as an outcome
study then it would be
considered weak.
16 21 Moderate Moderate Moderate No discrepancy between
the ratings.
 158

N Database Global Quality Rating Comments

ID
Researcher Peer Debriefer Final
19 24 Moderate Strong Strong The researcher and peer
reviewer differed on the
interpretation of Section C
(confounders) in the
assessment tool. The study
addressed most of the
essential confounders, but
did not address every
possible confounder. After
discussion it was agreed to
assign an overall strong
rating since the study did
address most of the
important confounders.
33 65 Moderate Strong Strong The researcher assessed
Section D on Blinding as
unknown, but the peer
reviewer assessed it as not
applicable due to the fact
that it would not be
possible to conduct a
double blind trial on this
type of program. After
discussion a decision was
made to give a final rating
of strong.
34 66 Strong Strong Strong No discrepancy between
the ratings.
 159

N Database Global Quality Rating Comments

ID
Researcher Peer Debriefer Final
27 80 Strong Moderate Moderate The peer reviewer raised
an additional point on the
research design, that
although the total number
of hand sanitizer uses was
recorded to increase, there
was no observation of
hospital staff to see what
they actually did with the
hand sanitizers. This is a
flaw in the design, and so it
was agreed to give an
overall rating of moderate.

Qualitative Evaluations

N Database Global Quality Rating Comments

ID
Researcher Peer Debriefer Final
3 5 Moderate Moderate Moderate No discrepancy between
the ratings. If it is assessed
as an evaluation then a
moderate rating is
appropriate, but if it is
assessed as research then it
would be considered weak.
19 24 Weak Weak Weak No discrepancy between
the ratings. Although this
study was described as
qualitative, it actually
attempted quantitative
analysis on a very small
sample size, which was a
serious flaw in the
methodology.
 160

N Database Global Quality Rating Comments

ID
Researcher Peer Debriefer Final
23 52 Weak Strong Strong The researcher applied the
criteria for Q6, Q7, and Q8
in the quality assessment
tool very strictly since they
were not explicitly address
in the report, while the peer
reviewer assumed that they
were implied by the study
design. After discussion it
was agreed to give an
overall rating of Strong,
given that the report is an
evaluation rather than
research, and as an
evaluation it is strong.
22 72 Weak Weak Weak No discrepancy between
the ratings. This was a
program description rather
than an evaluation or
outcome study.
 161
Curriculum Vitae

Piroska Bisits Bullen

ACADEMIC EXPERIENCE

2007-2012 Doctor of Philosophy – Public Health (Community Health),

Walden University, Minneapolis, Minnesota, US

2005 Honors Year – Physiology, University of New South Wales,

Sydney, Australia

2001 –2004 Bachelor of Science – Life Science (Neuroscience), University of

New South Wales, Sydney, Australia

PROFESSIONAL EXPERIENCE

Cambodia Health Education Media Service (CHEMS)

Technical Management Advisor (Cambodia, June 2011 – Present)

CHEMS uses creative media (TV, radio, print, online, social media) to address health and
development issues. My responsibilities include:
· Developing innovative ideas for media projects, including the use of new
technologies and social marketing approaches
· Proposal and report writing
· Managing relationships with donors and government agencies
· Monitoring and evaluation

Cambodian HIV/AIDS Education and Care (CHEC)

Technical Management Advisor (Cambodia, June 2011 – Present)

CHEC provides HIV/AIDS prevention, care and training services. My responsibilities

include:
· Technical oversight of the Strengthening Economic Livelihood Opportunities for
Low Income and HIV Positive Women (SECLO) project funded by UN Women
· Monitoring, evaluation and report writing for the SECLO project
· Proposal writing for new projects
· Training and staff development

International SOS
International SOS is the world’s largest medical assistance provider, operating in over 70
countries. My responsibilities included:
 162
Group Medical Implementation Manager, Consulting (UK, Nov 2009 – May 2011)
· Standardization and quality assurance for all consulting services globally,
including public health, occupational health and corporate health
· Training of regional and country teams
· Proposal and report writing
· Managing client relationships
· Strategic planning
· Extensive travel in Asia, Europe, the Middle East and U.S. to complete projects
· Coordination of all public health projects globally while the Public Health
Program Director was on maternity leave

Medical Projects Manager, R&D (Australia & UK, Oct 2006 – Nov 2009)
· Creating, piloting and implementing new global health programs
· Integrating new technologies into programs
· Research into new strategies
· Project management

Public Health Officer (Australia, Oct 2005 – Jan 2006)

· Technical research on avian and pandemic influenza
· Development of communication materials and planning tools

Management Alternatives
Consulting Analyst – Part Time (Australia, Nov 2000 – Oct 2005)

Management Alternatives is a boutique consulting agency specializing in human services

and social capital research. My responsibilities included:
· Data collection and data entry
· Analysis of qualitative and quantitative data
· Preparation of charts, reports, and communication materials

COMMUNITY SERVICE

Aug 2009 – Present CHINACO

Tanzania.

CHINACO operates the Chilunga Cultural Tourism program in Morogoro, Tanzania. I

supported the program by setting up and maintaining the website and online booking
system for cultural tours.

2006 - 2010 Tabitha Foundation

Cambodia.

The Tabitha Foundation provides a range of services to the Cambodian community,

 163
including micro-savings, well building, house building, and cottage industry. I supported
the foundation by preparing email newsletters and conducting research on arsenic
contamination of wells.

PROFESSIONAL ORGANIZATIONS

American Public Health Association

International Union for Health Promotion & Education
The Royal Society for Public Health

PROFESSIONAL PRESENTATIONS AND PAPERS

Bisits Bullen, P. & Cox, J. (2010). Designing workplace health promotion for a global
workforce: A case study of one multinational company operating in 25
countries. Annual conference of the International Union for Health Promotion
& Education (IUHPE).

Bisits Bullen, P. et al. (2008). Executive Report: International SOS Case Trends in the
Energy Mining and Infrastructure sector. International SOS: London.

Participant in the 2008 World Health Organization and World Economic Forum joint
working group on “Employee Health as a Strategic Priority in India”. New
Delhi, India.

Guibert, P. & Bisits Bullen, P. (2007). Online Learning: A complementary approach to

employee health education. Annual conference of the International Society of
Travel Medicine (ISTM).

REFERENCES

Available on request.