Research paper

Cognitive work analysis to evaluate the problem of

patient falls in an inpatient setting
Karen Dunn Lopez,1 Gregory J Gerling,2 Michael P Cary,3 Mary F Kanak4

< Supplementary material is ABSTRACT systems engineering tools to transform a complex

published online only. To view Objective To identify factors in the nursing work domain and often error-laden health system into a safer
these files please visit the that contribute to the problem of inpatient falls, aside healthcare system.10e12 Although systems engi-
journal online http://jamia.bmj.
com/content/vol17/issue3 from patient risk, using cognitive work analysis. neering approaches have addressed other adverse
1 Design A mix of qualitative and quantitative methods events in healthcare,12e14 research is needed that
Department of Health Systems
Science, College of Nursing, The were used to identify work constraints imposed on explores the problem of patient falls.
University of Illinois Chicago, nurses, which may underlie patient falls.
Chicago, Illinois, USA Measurements Data collection was done on

Downloaded from https://academic.oup.com/jamia/article/17/3/313/741198 by guest on 09 April 2024

2
Department of Systems and
BACKGROUND
a neurology unit staffed by 27 registered nurses and The true impact of patient falls in the inpatient
Information Engineering,
University of Virginia,
utilized field observations, focus groups, timeemotion setting is unknown as current reporting mecha-
Charlottesville, Virginia, USA studies and written surveys (AHRQ Hospital Survey on nisms significantly underestimate the incidence of
3
School of Nursing, University of Patient Culture, NASA-TLX, and custom Nursing patient falls.15 However, research over the last
Virginia, Charlottesville, Virginia, Knowledge of Fall Prevention Subscale).
USA decade underscores the significance of the problem.
4 Results Four major constraints were identified that Reports estimate between 4 and 12 falls per 1000
Veterans Affairs Midwest
Health Care Network, inhibit nurses’ ability to prevent patient falls. All patient days in US hospitals.6 15e17 It is estimated
Minneapolis, Minnesota, USA constraints relate to work processes and the physical that 15e30% of patient falls cause fractures
work environment, opposed to safety culture or nursing requiring a cast, traction or surgery, but these esti-
Correspondence to knowledge, as currently emphasized. The constraints
Dr Gregory J Gerling, mates do not include other serious injuries and, in
were: cognitive ‘head data’, temporal workload, some cases, death related to falls.1 2 16e20
Department of Systems and
Information Engineering, inconsistencies in written and verbal transfer of patient Most patient falls are predictable and simple
University of Virginia, PO Box data, and limitations in the physical environment. To deal patient risk assessment tools can predict over 70%.6 8
400747, 151 Engineer’s Way, with these constraints, the nurses tend to employ four Identified risk factors include poor health status,
Charlottesville, VA 22904, USA; workarounds: written and mental chunking schemas,
[email protected] gait instability, balance problems, age, agitation,
bed alarms, informal querying of the previous care nurse, confusion, urinary frequency and incontinence,
Received 16 June 2009 and informal video and audio surveillance. These patient history of falls, and certain sedative and
Accepted 3 February 2010 workarounds reflect systemic design flaws and may only hypnotic drugs.8 21 22 While many of these risk
be minimally effective in decreasing risk to patients. factors are well known and documented by nursing
Conclusion Cognitive engineering techniques helped staff, the assessment tools and prevention strategies
identify seemingly hidden constraints in the work domain have made little headway in reducing the incidence
that impact the problem of patient falls. System redesign of falls.6 7 23
strategies aimed at improving work processes and In addition to patient-related factors, three
environmental limitations hold promise for decreasing the elements of the systemdnurse staffing, nurse
incidence of falls in inpatient nursing units. knowledge, and safety culturedare emphasized in
the adverse event literature. Studies of nurse staffing
(nurse-to-patient ratios) and skill mix (higher per-
INTRODUCTION centage of Bachelor of Science prepared and regis-
The greatest number of non-fatal injuries in the tered nurses than nursing assistants and licensed
acute care setting are caused by patient falls.1 2 In practical nurses),16 17 24 and the expansion of
addition to causing injuries, patient falls can lead to nursing knowledge in risk assessment for falls25e27
increased healthcare expenditures, and are a major have found mixed results on fall rates. Although the
motive behind lawsuits.3 The direct financial burden relationship between fall prevention and safety
to hospitals increased in October 2008 when the US culture has not been rigorously studied, safety
Center of Medicare and Medicaid Services began to culture is often cited in the safety literature as being
disallow reimbursement charges associated with an important component of error prevention.9 28 29
falls that incurred injuries during a hospitalization.4 A culture of safety acknowledges the error-prone
Others speculate that private insurers will institute nature of hospital work, establishes a blame-free
similar ‘no pay for bad performance’ policies in the environment, and dedicates resources to safety
future.5 promotion.30e32 In recent studies, staff perception
Patient falls are adverse events that are largely of a blame-free, safety culture and the emphasis on
preventable.6e8 The magnitude of adverse events in safety promotion has been shown to be associated
US hospitals was described in the Institute of with improved falls reporting.22 29 33 Increases in fall
Medicine’s (IOM) report of healthcare errors, reporting offer opportunities for system improve-
which estimated that 48 000 to 98 000 patient ments that may prevent falls which would not exist
deaths occur yearly due to preventable medical if the falls went unreported. Despite considerable
errors.9 Since the release of this landmark report, research, patients continue to fall at unacceptable
many experts have advocated the application of rates. Therefore, research is needed to uncover other

J Am Med Inform Assoc 2010;17:313e321. doi:10.1136/jamia.2009.000422 313

 Research paper

factors that could help prevent falls within the acute care envi- neurological unit that treated stroke and epileptic patients, with
ronment, but which are poorly understood at present. an average daily census of 23 patients. The unit nursing staff
consisted of 27 full-time equivalent (FTE) registered nurses
SPECIFIC AIMS (RNs), 15 FTE nursing assistants (NA), and 4 FTE nursing unit
The specific aims of this research were to identify constraints in clerks. The nurse to patient ratio was 1:4 for the stroke popu-
work processes and the work environment (physical, organiza- lation and 1:5 for other patients. Figure 1 shows the unit layout.
tional systems and culture, individual, and technical) imposed
on acute care nurses that may increase the likelihood of patient Data sources
falls, and to identify workarounds employed by nurses to deal 1. Structured field observations (two total, 4.5 h each) of RNs,
with these constraints. NAs, and unit clerks in their normal routines were conducted
separately by nursing and cognitive engineering investigators.
These observations explored workflow, communication, the
OVERVIEW OF COGNITIVE WORK ANALYSIS physical environment, equipment, and nursing documenta-
Cognitive work analysis (CWA) is a formal approach in systems tion. Greater focus was placed on RNs.
and cognitive engineering for examining the functional structure 2. Focus groups (n¼3) were conducted by investigators with
of the work domain, including tasks performed, processes fol-

Downloaded from https://academic.oup.com/jamia/article/17/3/313/741198 by guest on 09 April 2024

a mix of RNs, NAs, and unit clerks in a semi-structured
lowed, environmental constraints, and information followed discussion format.
available.11 34 Such analysis helps system designers, managers, re- 3. Written surveys (n¼3) were administered to RNs.
designers, and accident investigators understand how workers a. The Hospital Survey on Patient Culture35 is a 44-item Likert
interact with their environment and how modifications might scale survey designed to measure staff perception about patient
create more productive work or reduce errors. In this study we safety, medical error, and event reporting for inpatient hospital
used CWA as an umbrella approach for data collection and anal- settings. Items are grouped into 14 dimensions: frequency of
ysis. Data were collected both by observation of workers and time event reporting, overall safety perception, safety grade, number
motion studies to identify constraints placed on the workers, and of events reported, organizational learning, teamwork within-
by surveys and focus groups to attain input from workers to unit, teamwork across-units, openness of communication,
identify workarounds. error communication, non-punitive response to error, staffing,
manager actions, management support, and hand-offs. The
METHODS instrument has undergone psychometric testing with accept-
Setting able to high reliability per dimension (Cronbach’s alphas range
The study took place at a 572-bed academic medical center on from 0.63 to 0.84).36 The Agency for Healthcare Research and
the east coast. The center is recognized by the American Nurses Quality (AHRQ) maintains a comparative database to allow
Credentialing Center as a magnet organization for excellence in for national benchmarking of survey results. In addition to its
nursing service. The specific unit studied was a 26-bed, general favorable psychometrics and ability to benchmark results, this

Figure 1 Hospital unit floor plan. The

three rooms with a direct viewing angle
are marked by the gray, dashed
rectangles.

314 J Am Med Inform Assoc 2010;17:313e321. doi:10.1136/jamia.2009.000422

 Research paper

survey was chosen because of its inclusion of key areas within Table 1 Demographics of focus groups and surveys
the concept safety culture possibly related to patient falls, Focus group
including reporting, hand-offs, staffing, non-punitive response, (%) Surveys (%)
and management response. Age 18e30 25
b. Nursing Knowledge of Fall Prevention Subscale (see supple- 31e40 25
mentary material online) is a subset of knowledge-related 41e50 44
questions from the Fall Prevention Survey currently in >50 6
development that seeks to identify factors extrinsic to
patient risk that may increase the likelihood of patient falls. Years of experience <1 13 <1 13
In this analysis, we include the 13-item subscale that assesses 1e3 6 1e5 27
nurse perception of knowledge of fall prevention and 4e7 13 6e10 7
protocols, as this is an important initial step in determining 8e12 19 11e15 19
if worker knowledge is a constraint. While existing surveys >12 44 16e20 27
assess the overall safety of the internal culture and peer N/A 0 >21 7
relationships, there is a void of psychometrically validated
Years of experience on <1 13 <1 27

Downloaded from https://academic.oup.com/jamia/article/17/3/313/741198 by guest on 09 April 2024

surveys to assess fall prevention knowledge. The subscale
nursing unit 1e3 19 1e5 33
contains eleven 4-point Likert item questions ranging from
4e7 19 6e10 13
strongly agree to strongly disagree; two items were
8e12 19 11e15 20
dichotomous (yes/no) questions. Content validity of the
> 12 30 16e20 7
complete survey was assessed by advanced practice nurses
N/A 0 > 21 0
(N¼3) and cognitive engineering experts (N¼2). While
additional psychometric testing is in progress, full reliability
and validity testing of this survey has not been done and is informed via written consent that the study sought to identify
a limitation addressed in the study limitations section. and analyze factors related to the risk of patient falls. There
c. The NASA-TLX Workload Instrument37 38 is a measure of were no direct benefits to individual subjects. The primary risk
worker perception of workload. It was administered to the of participation was the potential loss of confidentiality;
RNs to shed light on specific areas (assessing, determining, however, this risk was considered minimal and unlikely. Partic-
performing, documenting) of work constraint in their fall ipants were informed that all data collected would be published
prevention activities. The survey ranks six areas (1e100 only in aggregate form. Completed surveys were returned
visual analog scale).37 The survey has been shown to be anonymously to a box near the nursing station.
sensitive in situations, like fall prevention, that require
vigilance39; it performs well psychometrically38 and has Analysis methods
been used extensively to study workload in the aviation A different method of analysis was used for each data source.
and other complex industries,40e42 but has been used to Focus group discussion notes were summarized during the
a lesser extent in the healthcare domain.43 44 session by investigators. Included in the analysis were major
4. Four timeemotion studies measured the duration that nurses points confirmed by participant consensus. Single or minority
spent performing various tasks over a 4 h period. Four opinion points and points with minor relation to patient falls
randomly selected RNs were ‘shadowed’ by one of two were not included. The Hospital Survey on Patient Culture was
investigators (from advanced care nursing and cognitive analyzed by computing composite frequencies of the positive
engineering). Tasks in progress were reported every 5 min. It responses for items within the 14 safety culture dimensions
was possible to code in smaller blocks of time; however, this using Likert response categories (‘neutral’ responses were
practice rarely was necessary. During these studies, each omitted). The number of annual events reported was averaged
nurse cared for 4e5 ‘moderate to acutely ill’ patients. by individual respondent. NASA-TLX survey scores were derived
5. Three key informant interviews (each 30 min) with the nurse from ‘X’ marks on a visual analog scale (0e100) and individual
manager were used to acquire background information on scores were averaged. Likert items on the Nursing Knowledge
expected work processes and tools, as pertains fall prevention. of Falls Prevention Subscale were also averaged per item
This input helped to familiarize the researchers with the unit (1¼strongly disagree, 2¼disagree, 3¼agree, 4¼strongly agree).
but did not generate results. Field observations findings were discussed between the two
observers, following their separate observations. A comparison
Demographics of notes was used to reach a consensus regarding relevant data to
Table 1 presents focus group and survey demographic data. Data include. Timeemotion data were added by category, across the
were collected in categorical form to protect the identity of each four data collection periods.
participant. The majority of respondents (74%) in the Hospital All data analysis was conducted iteratively with the data
Survey on Patient Culture reported working 40e59 h per week. collection process. Preliminary conclusions from the data sources
Although the survey did not include questions on shift length, were synthesized into the abstractionedecomposition (AeD)
the nurse manager reported that the majority of nurses work hierarchy (figure 2), to help uncover relationships between
a combination of 12 and 8 h shifts (full-time nurses are sched- disparate components and sub-components of the work domain.
uled for two 12 h shifts and two 8 h shifts each week). The AeD hierarchy is a cognitive engineering model that
describes the work domain in a single representation and serves
Human subjects protection to focus on identifying constraints and workarounds.
The study was approved by the local Institutional Review
Board. Participants had the option of participating in all of the RESULTS
data collection methods detailed above or could choose to only Overall, 33 of the 46 (71.7%) members of the nursing unit
to participate in some of the methods. Participants were personnel, including registered nurses (N¼27), nurse assistants

J Am Med Inform Assoc 2010;17:313e321. doi:10.1136/jamia.2009.000422 315