IIHMR Working Paper No.

2






A Composite Index-based Approach for
Analysis of the Health System in
the Indian Context









Prashanta Pathak
Kamal K. Gaur

















About the Authors

Prasanta Pathak : After obtaining B. Stat. (Hons.), M. Stat. and Ph.D.
degrees from Indian Statistical Institute, Calcutta, he was on the faculties
of Institute of Business Management, Jadavpur University, and the Xavier
Institute of Management, Bhubaneshwar. Currently, he is Associate
Professor at the Institute of Health Management Research, Jaipur, India.
His areas of interest include Statistics, Operations Research and Regional
Planning. He has published a number of research articles on health and
human resource development in reputed journals, and attended many
national and international conferences and workshops.

Kamal K. Gaur: After obtaining his M.Sc. (Statistics) degree from Indian
Institute of Technology, Kanpur, he served as Research Officer at Institute
of Health Management Research, Jaipur, India. He has more than five
years experience of research in the areas of Health and Family Welfare.







List of Contents


Abstract

1. Introduction 1

2. Methods and Models 3
2.1 Social Environment 4
2.2 Economic Environment 4
2.3 Health Service Facilities 5
2.4 Programme Efforts 5
2.5 Status of Health 7

3. Data and Empirical Analysis 13
3.1 Composite Index for Social Environment (I
S
) 13
3.2 Composite Index for Economic Environment (I
E
) 14
3.3 Composite Index for Service Facilities (I
F
) 15
3.4 Composite Index for Programme Efforts (l
P
) 16
3.5 Composite Index for Health Status (I) 16
3.6 Association of the Composite Index for Health
Status with the Other Composite Indices
17
3.7 A Comparative Analysis of the Various Aspects of the
Health System and the Status of Health in Different States
in India
18

4 Concluding Remarks 21

Acknowledgements 23
References 24















A Composite Index-based Approach for Analysis of the
Health System in the Indian Context


Abstract: Specific features of the Indian socio-economic environment, the health
and family welfare service facilities and the programmatic efforts, which
determine the health status of the Indian community, are studied here from the
health system angle by constructing several composite indices. Each composite
index has been constructed objectively by using Principal Component Analysis
technique and has been meaningfully interpreted. The study establishes the utility
of such a composite index-based statistical approach by associating the
community health status with the availability of service facilities, the programme
efforts and the socio-economic environment by multiple linear regression. It also
suggests a definite approach of identifying the most important factors that are
associated with community health and gives their order of importance. The
methodology has been tested on state-wise Indian data.

1.0 Introduction

The health of an individual has been defined by the World Health Organisation as
a state of complete physical, mental and social well being, and not merely absence
of disease or infirmity. This comprehensive definition has made it necessary to
define the health of a community in a broad perspective, and not merely in terms
of the individual demographic indicators such as Infant Mortality Rate (lMR),
Life Expectancy at Birth (LEB), Mortality Rate by Causes, and so on.

A community could be healthy if almost every individual in it were healthy. The
primary requirements for anyone to be healthy are availability of adequate food,
shelter and clothing; but these necessities could be met, only if there is sufficient
scope of income generation. Also, the working people should have necessary skill
and knowledge, which they might acquire through formal or informal education
and/or experience. Additionally, there have to be certain supportive organisations
or institutions, government/private/international which ensure systematic flow of
goods and services between one group of producers and another group of
producers and/or consumers. These organisations serve areas of common interest
including health, environment, socio/religious/political realities, and so on. All
these together culminate in the overall environment for the people in a
community, and affect the health status of the community also. Thus, the state of
the environment is intimately related to various health indicators of a community
such as IMR, LEB, etc.


The present study tries to look at each aspect of the environment as well as each
aspect of health in its totality and attempts at suggesting a summary or composite
index for each aspect. Conceptualisation of a complex phenomenon, with several
dimensions, through development of a composite index is nothing new.
Construction of a composite health index by Cayolla da Motta [1979] is an
example. However, the procedure of assigning weights to the various constituent
health indicators, described in it was questionable as some degree of subjectivity
was involved in it. Such an index should be constructed objectively, incorporating
the fact that the constituent indicators are all inter-related. The statistical
technique of Principal Component Analysis (PCA) (see Anderson [1984]), could
be useful for such index construction. The technique has been successfully
applied in many studies (e.g., Kendall [1939], Pal [1963, 1971, 1990],
Mukhopadhyaya [1988], Pathak [1991, 1995], Tatlidil [1992]).

However, the technique, in spite of its qualitative superiority, is yet to gain
popularity because the commonly used indices, constructed on the basis of
restrictive assumptions, are easy to estimate and comprehend. Even the Human
Development Index (HDI), proposed in the UNDP reports (see Human
Development Report [1993]) also is too restrictive. Three aspects have been
considered for constructing it, namely, (1) longevity, measured by life
expectancy, (2) knowledge, measured by adult literacy and mean years of
schooling, and 3) well being, measured by adjusted per capita income, and
modified by the Gini coefficient. All three have been assumed to be equally
important, and hence, equal weights have been given to the linear formulation of
the HDI. It is only coincidence that Tatlidil [1992] has found by PCA that the
three indicator variables, representing the three aspects, appear in the HDI with
more or less equal weights. The inter-dependance among the three aspects,
however, has not been taken into consideration in the HDI construction. If this is
taken into consideration in the form of a correlation matrix, with due concern for
the probability distributions of the representing variables for the three aspects, the
construction of the HDI becomes complex and can only be done by applying PCA
technique. An application of it may as well yield unequal levels of importance of
the three aspects (see for example Pal [1963, 1971, 1990], Mukhopadhyay [1988] and
Pathak [1991, 1995]). Here, PCA has been utilised for construction of composite indices,
representing the community health status and various health-related
environmental and programme aspects.

2.0 Methods and Models

The model we follow in this study is represented diagrammatically in Fig.1.
While the health service facilities and the programme efforts are direct intervention
factors for changing the community health status, the factors that indirectly influence the
health status of a community have been categorised here under the broad names, viz.(1)
Social Environment and (2) Economic Environment. These factors have been further
specified by certain population and health programme characteristics (referred to as
variables), that are normally measured in various socio-economic and health surveys. The
variables have been listed under Sub-sections 2.1 to 2.5 with their abbreviations. The
choice of variables may not be the best, but subject to the availability of the published
data on various health and health-related aspects, these were thought to be good ones to
serve the purpose of this study.








Figure 1 : A Model for Health System




2.1 Social Environment

Measuring social environment in absolute terms is very difficult as it is a matter of
individual and/or community perception. The variables that have been chosen to
represent the social environment have all been considered in indicator form, focusing on
literacy, discrimination between males and females, and the constituency of the down-
trodden class.

These are:

a) (i) Sex Ratio of Scheduled Castes [SSRC], (ii) Sex Ratio of Scheduled
Tribes [SSRT] and (iii) Sex Ratio of Non-Scheduled Castes/Tribes
[SSRN]

b) (ii) Literacy Rate among Males [SLRM] and (ii) Literacy Rate among
Females [SLRF]

c) (iii) Percentages of Scheduled Caste/tribe Populations [SSCT] in the total
population.

Discrimination disfavouring females and low literacy are definite signs of social
backwardness. Additionally, a large segment of scheduled castes and tribes in the
population that has remained continuously underprivileged for hundreds of years, is
causing a kind of pull-back effect on the development of the social environment.






2.2 Economic Environment :

This concept also is difficult to measure in absolute terms due to perceptional variation
over time, space and information/technology state. So, this environment is again depicted
by a few variables which are also in indicator form. They summarise the state of poverty,
the state of well-being, and the state of involvement in economic activities.

a) Percentage of Population below Poverty Line [EPL]
b) Per Capita Income [EPI]
c) (i) Proportion of Main Workers [EMW] and (ii) Proportion of Female Main
Workers to Male Main Workers [EFMW].

While the reasons behind choosing the first two variables are self-explanatory, the last
two variables stand for the scope of getting absorbed in various economic activities,
affected by the male-female discrimination aspects.

The variables that have direct influence on the health status of a community have been
categorised under the broad head of Health Service Facility, which includes the
following:

2.3 Health Service Facility

a) (i) Average Coverage of Population by a Sub-centre [HPSC]
(ii) Average Number of Villages Covered by a Sub-centre [HVSC]

b) (i) Average Coverage of Population by a Primary Health Centre
[HPHC]
(ii) Average Number of Villages Covered by a Primary Health Centre
[HVHC]

c) (i) Average Coverage of Population by a Doctor [HPD]
(ii) Average Coverage of Population by a Para-medic [HPP].

The standards set by the Government of India, regarding coverage of population
are 5000 per sub-centre in the plains (3000 for hilly areas), and 25,000 to 30,000
per primary health centre in the plains (20,000 for hilly areas). However, the
actual coverage varies to a great extent all over India. Also, to meet the standards,
the number of villages covered by any of the two health facilities varies very
much, depending on the population sizes of villages. This has a direct logistic
implication as far as the accessibility of the health facilities is considered. So,
both population coverage and village coverage have been considered important
here. The reason behind considering the last two variables is that the doctors
normally provide the curative services and the para-medics provide mostly the
preventive and some support services.

However, health service facilities alone cannot depict the total picture of the kind
of health service environment that has been created to raise the status of
community health. Rather, its level of effective functioning has to be taken into
account to have a more complete picture. This aspect has been considered here
under the broad category, Health Programme Efforts, or simply Programme
Efforts. The term "programme efforts" has been used earlier by Mauldin and Ross
[1991] in a much broader sense, encompassing (1) policy and stage-setting
activities, (2) service and service-related activities, (3) record keeping and
evaluation, and (4) availability and accessibility of fertility-control supplies and
services. Due to limited availability of ready-made data on the first three, only the
information available on the last dimension of programme efforts has been
considered here.

2.4 Programme Efforts

The following variables have been considered to get an idea of programme
efforts.

a) Percentage Achievement of Targets under various components of the
Universal Immunization Programme (UIP) viz.

i) BCG [PBCG] iv) Measles [PMS]
ii) DPT [PDPT] v) Vitamin A [PVA]
iii) OPV [POPVJ vi) DT [PDT]

b) Percentage Achievement of Targets in Giving TT to Pregnant Women
under the Mother and Child Health (MCH) Programme [PTT],

c) Percentage Achievement of Targets through Prophylaxis against
Nutritional Anaemia among Women [PAW],

d) Percentage Achievement of Targets through Prophylaxis against
Nutritional Anaemia among Children [PAC],

e) Couple Protection Rate, achieved through Family Planning (FP)
Programme [PCPR]

f) Reported Rate of Prevalence of the following Diseases in the Population
[PRD],

a) Kala Azar m) Enteric Fever
b) Acute Diarrhoea n) Viral Hepatitis
c) Cholera o) Rabies
d) Tuberculosis p) Syphilis
e) Diptheria q) Gynococcal Infection
f) Poliomyelitis r) Japanese Encephalitis
g) Tetanus (Neo-natal) s) Meningitis
h) Tetanus (Others) t) Positive Malaria
i) Whooping Cough u) Leprosy
j) Measles v) STD
k) ARI w) Mental Disorder
l) Pneumonia

All these variables indicate how effectively the programmes are being
implemented for improving the community health status.


Now, all the variables listed in Sections (2.1), (2.2), (2.3) and (2.4) have been
taken as the deciding variables for the prevalent status of health of a community.
The indicators that have been chosen to define Status of Health include the
following:

2.5 Status of Health

a) (i) Infant Mortality Rate [IMR] in Rural Areas [IMRR] and
IMR in Urban Areas [IMRU],

b) (i) Crude Birth Rate (CBR) in Rural Areas [CBRR] and (ii)
CBR in Urban Areas [CBRU],

c) (i) Life Expectancy at Birth [LEB] (for Males and Females
combined).

None of these variables is used here in a purely demographic sense; rather, they
have been used as the proxy variables, representing sundry physical, mental and
social aspects of health. These inter-related aspects are so complex that gauging
them in absolute terms is very difficult. Perception about physical well being,
mental well being, social well being and so on differs from person to person, from
one community to another, from one geographical region to another, from one
time point to another, and from one information/technology state to another state.
So, there is no absolute concept of health status. It can only have a relative
measure.

The variation in the units of measurement and/or the ways of perceiving all the
above variables needs careful handling because using the variables as such,
without making them comparable on a single scale, may cause confusion in
interpretation. This has been made possible by using the concept of Location
Factor, introduced by Pal [1971]. This is obtained by measuring each variable in
ratio form, making it relative to some acceptable bench-mark or standard value.
Since the present study is only on the state- wise data of India, with the variables
expressed mostly as rates, ratios or proportions/ percentages, the all-India values
of the variables are considered here as the bench-marks for redefining the
variables in the form of location factors. The value of the location factor for a
variable for any state will be obtained by dividing the state value of the variable
by the same at the country level. Thus, the variables in ratio form will normally
have non-negative values and the values will always be unity at the country level.
Table 1 gives the list of composite indices, constructed in the next section, along
with the lists of location factors that have been used for this purpose. The
variables already defined and abbreviated, have been marked with an asterix (*)
to indicate that they have been considered in the location factor form.







Table 1 : Description of Constructed Indices and the
Constituent Variables

Name of the
Composite Index
(Symbol)
Constituent
Indices (Form)
Constituent Variables, used in Location
Factor or other Appropriate Form
I) Social
Environment
Index
[I
S
]
a) Sex
Discrimination
[SSR*]
(Composite)


b) Literacy (SLR*]
(Composite)

c) SC/ST
Concentration
[SSCT*]
(Simple)

i) Sex ratio for Scheduled Castes
(SC) population [SSRC];
ii) Sex ratio for Scheduled Tribes
(ST) population [SSRT];
iii) Sex ratio for non-SC/ST [SSRN].

i) Male literacy rate [SLRM];
ii) Female literacy rate [SLRF].
i) Percentage of SC/ST population
[SSCT].

II) Economic
Environment
Index
[I
E
]
a) Poverty [EPL*]
(Simple)
b) Income [EPI*]
(Simple)
c) Scope of
Employment
[ESW*]
(Composite)

i) Percentage of population below
poverty line [EPL];
i) Per capita income [EPI];
i) Percentage of combined male and
female main workers in the
population [EMW];
ii) Percentage of female main workers'
participation with respect to male
main workers' participation
[EFMW].
III) Health
Service
Facility
Index [I
HI
]
a) Service
Coverage by
Sub-centre (SC)
[HSC*]
(Composite)

b) Service
Coverage by
Primary Health
Centre (PHC)
[HHC*]
(Composite)
c) Service
Coverage by
Doctors and
Para-medicals
[HP*]
(Composite)
i) Average coverage of population by an
SC [HPSC];
ii) Average number of villages covered
by an SC [HVSC].
i) Average coverage of population by a
PHC [HPHC];
ii) Average number of villages covered
by a PHC [HVHC].
i) Average coverage of population by a
doctor [HPD];
ii) Average coverage of population by a
para-medic [HPP].



Name of the
Composite
Index (Symbol)
Constituent Indices
(Form)
Constituent Variables, used in
Location Factor or other
Appropriate Form (Symbols)

IV) Programme
Effort Index
(I
P
)
Achievement in Child
Immunisation [PT}
(Composite)
i) Percentage target achievement
with Index regard to BCG
[PBCG};
ii) Percentage target achievement
with regard to DPT [PDPT];
iii) Percentage target achievement
with regard to OPV [POPV];
iv) Percentage target achievement
with regard to Measles [PMS};
v) Percentage target achievement
with regard to Vitamin A [PVA];
vi) Percentage target achievement
with regard to DT [PDT}.

b) Achievement in
Immunisation of
Pregnant Women
with TT [PTT*]
(Simple)
c) Achievement in
through Prophylaxis
against Nutritional
Anaemia among
Women [PAW*]
(Simple)
d) Achievement
through Prophylaxis
against Nutritional
Anaemia among
Children [PAC*]
(Simple)
e) Achievement in
Family Planning
(FP) Programme
[PCPR*] (Simple)
f) Reported Prevalence
of Diseases of
National Concern
[PRD*] (Simple)
Percentage target achievement with
regard to TT [PTT]


Percentage target achievement with
regard to prophylaxis against
nutritional anaemia among women
[PAW]


Percentage target achievement with
regard to prophylaxis against
nutritional anaemia among children
[PAC]

Achieved couple protection rate
[PCPR].



Percentage of Population having
reported diseases of national concern
[PRD].




E) Status of
Health
Index
(I)
1. Life Expectancy
[LEB*] (Simple)
2. Crude Birth Rate
[CBR*]
(Composite)
i) Life expectancy of males and
females [LEB].

i) CBR in rural area [CBRR];
ii) CBR in urban area [CBRU];

3. Infant Mortality
Rate [IMR*]
(Composite)

i) IMR in rural area [IMRR];
ii) IMR in urban area [IMRU].



The composite index on Social Environment (the first factor) has been constructed in two
stages. At the first stage, two first principal components (fpc) are worked out, one out of a
(i) and a (ii) and another out of b (i) and b(ii), after giving location factor form to all the
variables. The underlying logic here is, when two or more variables are of similar nature
and are highly inter-correlated, a new variable in the form of a linear combination of all
those correlated variables could be constructed by the method of PCA which would
explain most of the total variation of all the constituent variables. This new variable may
be conceptualised either as an overall manifestation or as a generator of all those
correlated variables. In fact, the number of linear combinations of the variables that could
be obtained by PCA is the same as the total number of constituent variables. However, in
a situation where the variables are highly correlated, just one linear combination, called
the first principal component, might be sufficient to explain most of the total variation of all
the variables. Here the situation is similar. Suppose these fpcs are

SSRS* = s
1
SSRC* + s
2
SSRT* (1)
and SLR* = s
3
SLRM* + s
4
LRF* (2)
where s
1
, s
2
, s
3
and s
4
are coefficients, estimated by PCA

The location factor SSRS* is then combined with SSRN* through the following
fpc.

SSR* = s
5
SSRS* + s
6
SSRN* ..(3)

At the second stage, the final PCA yields the following fpc.

I
s
= s
7
SSR* + s
8
SLR* + s
9
SSCT* .. (4)

where s
5
, s
6
, s
7
,s
8
and s
9
are the coefficients, obtained by PCA.

We call this I
5,
Composite Index of Social Environment.

The construction of the composite index of Economic Environment is relatively
simpler, though we have done it in two stages here also. All variables are, first of
all, given location factor form and then the two variables under (c) are combined
in the form of the fpc by PCA to form the composite index ESW*.

This composite index, ESW* may be represented by the following:

ESW* = e
1
EMW* + e
2
EFMW* (5)

where e
1
, and e
2
are fpc coefficients.

The indices corresponding to (a) and (b), symbolised by EPL* and EPI*
respectively, are then combined with ESW* by a second stage PCA. This
Composite Index on Economic Environment is denoted here by I
E
, and it is the
same as





I
E
= e
3
EPL * + e
4
EPI* + e
5
ESW* (6)

The composite index on Health Service Facility (the third factor) has also been
constructed in two stages. At the first stage, the first principal component
corresponding to each of (a), (b) and (c) is worked out after giving location factor
form to the two components of each. Suppose the three fpcs are represented by
the variables HSC*, HHC* and HP* respectively for (a), (b) and (c). These
basically stand for,

HSC* = h
I
HPSC* + h
2
HVSC* .(7)
HHC* = h
3
HPHC* + h
4
HVHC* .(8)
HP* = h
5
HPD* + h
6
HPP* .(9)

where h
1
, h
3
, h
3
, h
4
, h
5
, h
6
are estimated by PCA. At the second stage, these
three fpcs are further combined by PCA to form the fpc, denoted by I
F
, where

I
F
= h
7
HSC* + h
5
HHC* + h
9
HP* .(10)

h
7
, h
8
and h
9
denoting coefficients, estimated by PCA at the second stage.
The index, I
F
, is referred to here as the Composite Health Service Facility. The
composite index on Programme Efforts [i.e. factor (4)] has also been constructed
in two stages. First of all, the location factors of different immunisation
programme efforts, denoted by the variables PBCG, PDPT, POPV, PMS, PVA
and PDT, have been linearly combined through the fpc. This combination,
denoted by PI*, where

PI*= p
1
PBCG*+p
2
PDPT*+ p
3
POPV*+p
4
PMS* +P
5
PVA* +p
6
PDT* (11)

with p
1,
p
2
, p
3,
p
4,
p
5
and
P6
denoting the coefficients, estimated by PCA, may be
interpreted as a composite index for the overall child immunisation programme
effort. This index has subsequently been linearly combined with the location
factors of other programme efforts, denoted by the variables PTT*, PAW*,
PAC*, PCPR* and PRD* by PCA.

Taking only the fpc constructed out of PI*, PTT*, PAW*, PAC*, PCPR* and
PRD*, as the composite index for overall programme efforts, and representing
this index by I
p,
the Composite Index of Programme Efforts may be written as

I
p
= p
7
PI* + p
8
PTT* + p
9
PAW* + p
10
PAC* + P
11
PCPR* + p
I2
PRD* (12)

where p
7,
p
8,
p
9,
p
10,
p
11,
and p
12
are estimated by PCA at the second stage.

Finally, the composite index on overall Health Status is formed by combining
three demographic indicators, namely, Infant Mortality Rate (IMR), Crude Birth
Rate (CBR) and Life Expectancy at Birth (LEB). The first two indicators have
been taken in the location factor form for both rural and urban areas in each state
separately, and then combined by fpc to form two indices, denoted by IMR* and
CBR*. The overall state values of IMR and CBR also could have been taken in
location factor form to replace IMR* and CBR*, but that would not have truly
reflected the disparities in rural and urban conditions. Unfortunately, we do not
have separate urban and rural figures for LEB. So, the index defined,
corresponding to LEB and denoted by LEB*, is computed here as a proportion
with respect to the all-India value of LEB. The Composite Status of Health
Index for overall Health Status, denoted by I, is obtained from the fpc constructed
out of IMR*, CBR* and LEB* in the form

I = i
1
IMR* + i
2
CBR* + i
3
LEB*, . (13)

where i
1
, i
2
and i
3
are the fpc coefficients.

This composite index, which is constructed here to indicate the health status of the
people, is definitely not a comprehensive one, and could be fine-tuned further to
match the reality. For example, one may as well additionally include the index of
Disability Adjusted Life Years (DALY), (for details on DALY see WHO [1994])
or use it as a replacement for LEB in the index construction. Also, one may use
the Total Fertility Rate (TFR) in place of CBR. The primary objective of the
present study is focused on developing a method of constructing meaningful
composite indices.

Thus, we basically construct the following five composite indices:

i) Composite Index for Social Environment (I
s
):

I
s
= s
7
SSR* + s
8
SLR* + s
9
SSCT* {ref to (4)};

ii) Composite Index for Economic Environment (I
E
):

I
E
= e
3
EPL * + e
4
EPI* + es ESW* {ref to (6)};

iii) Composite Index for Health Service Facilities (I
F
):
I
F
= h
7
HSC* + h
8
HHC* + h
9
HP* {ref to (10)};


iv) Composite Index for Programme Efforts (I
p
):

Ip = p
7
PI*+p
8
PTT*+p
9
PAW*+p
10
PAC*+p
11
PCPR*+p
l2
PRD*{ref to
(12)};

v) Composite Status of Health Index (I):

I

= i
1
IMR* + i
2
CBR* + i
3
LEB* {ref to (13)}.







3.0 Data and Empirical Analysis

The results of this study are based on data on 16 major states in India, which have
been compiled by the Central Bureau of Health Intelligence [1991], (Agarwal et
al [1993]) and the Foundation for Research in Health Systems [1993].

For brevity, we discuss only the main findings on the five major composite
indices. Each index expression has been written after normalising the coefficients,
so that the all-India value of each index becomes unity. The detailed index values
corresponding to different fpcs are given in Table 2 of Section 3.7.

3.1 Composite Index for Social Environment (I
s
)

The following gives the estimated coefficients in the expression for I
s
:

I
s
= 1.2407 SSR* + 0.9106 SLR* - 1.1514 SSCT* . (14)

The percentage of total variance, explained by this fpc, is 50.3. The specific
representations of SSR*, SLR* and SSCT* in the composite index, measured by
the constituent variables' correlations with 1, are 0.47, 0.58 and 0.89
respectively. The coefficient values show the great importance of sex ratio and the
SC/ST component of the population (the second and the third variables) in
determining the social environment status. While the increase in sex ratio and
literacy rate improve the environment status, increase in the SC/ST constituent of
the population has a greater pull-back effect on the developmental process of
social environment. The last part of this statement, however, should not be
misinterpreted as something against SC/ST population. The problem basically is
not with the SC/ST population, but with, the social system which has kept it
socio-culturally as well as economically backward over hundreds of years. What
is meant here is, under the existing state of backwardness of SC/ST population,
increase in the proportion of SC/ST population in the total population will have a
pull-back effect on the status of the social environment. The problem is basically
not with the increase in the proportion of the SC/ST population, but with its
associated backwardness, which needs reversal. Thus, here is an index that
measures meaningfully the status of social environment.

3.2 Composite Index for Economic Environment (I
E
)

Here, the expression that represents I
E
is the following.

I
E
= 1.4899 EPL* - 1.3591 EPI* + 0.8692 ESW* . (15)

The percentage of explained variation by this fpc is 59.6. The specific
representations of EPL*, EPI*, ESW* in the composite index, I
E
, are 0.92,-0.88
and 0.37 respectively, measured by the constituent variables' correlation with I
E
.
This index basically measures the extent of economic backwardness. This has a
positive association with the percentage of population below the poverty line and
a negative association with the per-capita income. What needs special attention is
the positive association of this index with the extent of participation of main
workers. Statistically, this has happened due to the multi-collinearity or
interdependence among the three chosen variables.

At first sight, this association looks fallacious because, normally, we think that
with greater work force participation, the extent of economic backwardness
should be on the wane; but, in the context of a developing labour surplus
economy like that of India, this is really not fallacious. The detailed studies by
Todaro [1969, 1976], Saha [1984], Mukhopadhyay and Pathak [1991] are referred
in this context. In a predominantly agricultural economy, with the existing pattern
of use of technological know-how, excessive growth of population in the rural
areas has resulted in extra-ordinary pressure on the land resources. The land
productivity, however, has not increased matching with the population growth. In
fact, the current land and labour input combination gives diseconomic scale of
return. Low wage in the rural areas is its manifestation and this, along with non-
availability of gainful jobs in the rural areas - is the most important cause behind
migration of rural labour to the economically better-of urban areas. On the other
hand, the scope of absorption in the urban areas is also limited and the expansion
of the secondary and tertiary activities in the urban areas is also limited. Even the
expansion of the secondary and tertiary activities in the urban areas has failed to
keep pace with the increasing labour pressure. Except for a few industries, all
others are now in such a state that they can absorb labour only at a decreasing rate
in spite of increasing utilisation of capital and increasing production.

Thus, wherever this labour pressure is more, as indicated by the extent of
participation of main workers, the extent of economic backwardness is also more,
because of greater likelihood of low-wage employment or under-employment.
The weights of the three variables, indicated by the values of the attached
coefficients, are again indicative of the importance of the three variables. The two
most important variables, however, are percentage of population below the
poverty line and per capita income.


3.3 Composite Index for Health Service Facilities (I
F
)

As mentioned already, IF is an index constructed out of three composite indices
viz. (1) composite index, representing coverage by a sub-centre (HSC*), (2)
composite index, representing coverage by a Primary Health Centre (HHC*), and
(3) composite index, representing coverage by health personnel, meaning the
doctors and the para-medics (HP*). The following are the expressions for HSC*,
HHC* and HP*.

HSC* = 0.7602 HPSC* + 0.2398 HVSC* .(16)
HHC* = 0.2913 HPHC* + 0.7087 HVHC* .(17)
HP* = 0.3379 HPD* + 0.6621 HPP* ... .(18)

HSC*, HHC* and HP* may be considered respectively as the composite indices
representing the weakness in service area coverage by the sub-centres, the PHCs
and the manpower component of the health service facilities.



The first two expressions bring out the fact that with increase in average
population coverage, the weakness in the service area coverage increases more in
the case of sub-centres than in the case of PHCs. The variable which is more
important in increasing the weakness in the service area coverage by the PHCs is,
on the other hand, the average number of villages that avail themselves of PHC
facilities. The third expression brings home the fact that the weakness of the
manpower component of the health service facilities increases more with the
increase in the average population coverage by a para-medic than with the
increase in the average population coverage by a doctor.

When these three indices are combined, again by PCA, to form the overall
composite index for health service facilities (I
F
), we obtain the following
expression.

I
F
= 0.3844 HSC* + 0.3118 HHC* + 0.3038 HP* .(19)

This index explains 61.3% of the total variation, and it measures basically the
degree of weakness in the service area coverage by the existing health service
facilities. The specific representations of HSC*, HHC* and HP* in the composite
index, I
F
, are 0.91, 0.63 and 0.78 respectively. The expression gives the
impression that all the three components are more or less equally important in
effecting weakness in the service area coverage.

3.4 Composite Index for Programme Effort (l
p
):

This index also has been constructed in two stages. At the first stage, the
composite index, representing the achievement in the immunisation programme
effort (PI*), has been found out to be the following:

PI* = 0.1918 PBCG* + 0.2041 PDPT* + 0.2066 POPV* + 0.1648 PMS* + 0.1094
PVA* + 0.1233 PDT* .(20)

explaining 63.5% of the total variation. It gives the impression that the efforts put
in on BCG, DPT and OPV vaccinations are the most important ones in
determining the achievement of the current immunisation programme efforts.

The composite index for overall programme efforts (I
p
) found out at the second
stage, is

I
p
= 0.1867 PI* + 0.1724 PTT* + 0.1924 PAW* + 0.1891 PAC*
+ 0.1953 PCPR* + 0.0640 PRD* .(21)


The specific representations of PI*, PTT*, PAW*, PAC*, PCPR* and PRD* in I
p

are 0.93, 0.82, 0.69, 0.61, 0.82 and 0.28 respectively. This summary index
basically measures the overall achievement of the programme efforts. It is clear
that identifying and reporting the cases of the listed diseases seem to be the most
neglected aspect in the overall programme efforts. All other programme efforts
are found to have more or less equal importance.

3.5 Composite Index of Health Status (I)

As mentioned already, this composite index has been constructed by combining in
the first principal component, the IMR and the CBR derived from their location
factor values in urban and rural areas, and also the Life Expectancy at Birth in
index form. The expression for I is found to be

I = 0.8962 IMR* + 0.9571 CBR* - 0.8534 LEB* .(22)

which explains 60.3% of the total variation. The representations of IMR*, CBR*
and LEB* in this composite index, I, are 0.75, 0.78 and -0.68 respectively. It
basically measures how poor the status of health is, or, the degree of
unhealthiness. Thus, with the increase in IMR and CBR, the value of the index
increases, indicating that the situation is becoming more and more unhealthy; but,
the association between I and LEB is negative, meaning thereby that the situation
becomes less and less unhealthy as Life Expectancy at Birth increases.

3.6 Association of the Composite Index for Health Status with
Other Composite Indices

With an intention to explain the variation of I by other composite indices, viz. I
s
,
I
E
, I
F
and I
p
, I is regressed on the remaining indices by the step-wise regression
procedure. It was found that I
s
and I
F
were sufficient to explain the variation of I.
The regression equation is found to be the following:

1= 0.5399 - 0.3795 I
s
+ 0.3901 I
F
.(23)
(0.1720) (0.0826) (0.1425)

with i) multiple correlation (R) = 0.8631
ii) adjusted R
2
(R
2
) = 0.7056

and standard error (s.e.) of the regression parameters, shown within first brackets
under the regression parameter estimates.

The expression (23) in expanded form will be

I = 0.5399 - 0.4708 SSR* - 0.3456 SLR* + 0.4370 SSCT*
+ 0.1500 HSC* + 0.1216 HHC* + 0.1185 HP* ..(24)

by using (14) and (19). The coefficients indicate the relative importances of
various constituent variables.

Both the social environment and the service area coverage by the health service
facilities are found to be the most important factors in determining the status of
health of a community. While better social environment makes the community
more healthy, greater weakness in the service area coverage by the health service
facilities makes the community more unhealthy. Adding I
E
to the above set of
independent variables improves only the R value, but reduces the adjusted R
2

value to 0.7047. Instead of including I
E
, if I
P
is included in the set of independent
variables, the R value increases to 0.8845 and the adjusted R
2
increases
marginally to 0.7279. Then the regression equation turns out to be:

I= 0.9175 - 0.3222 I
S
+ 0.4058 I
F
- 0.2309 I
P
....(25)
(0.3106) (0.0886) (0.1374) (0.1663)

with R = 0.8845 and R
2
= 0.7279 and the s.e. of the regression parameters, shown
within brackets.

The expression (25) in expanded form will be

I = 0.9175 - 0.3998 SSR* - 0.2934 SLR* - 0.3710 SSCT*
+ 0.1560 HSC* + 0.1265 HHC* + 0.1233 HP*- 0.0431 PI*
- 0.0398 PTT* - 0.0444 PAW* - 0.0437 PAC* - 0.0451 PCPR*
- 0.0148 PRD* .(26)

by using (14), (19) and (21). The coefficients are again indicative of the relative
importance of the constituent variables.

Thus, we may at the most accept the regression equation (25) or (26), if the
variation of I needs to be explained a little more than what has been explained
through (23) or (24). In equation (25), I
P
appears with negative regression
coefficient, meaning that the community becomes less and less unhealthy with
more and more achievement of the programme efforts. Now, altering the values of
some of the variables in expressions (24) and (26) may be difficult in short term,
but the values of other variables may be changed through short term interventions.
Depending on the importance of different constituent variables and their short
term or long term changeability, one can decide on the intervention strategies for
lowering I.

3.7 A Comparative Analysis of the Various Aspects of the Health System
and the Status of Health in Different States in India

The values of the indices measuring degree of unhealthiness (I), social
environment status (I
S
), degree of weakness in the service area coverage by the
existing health service facilities (I
F
), achievement out of the programme efforts
(Ip) and economic backwardness (I
E
), are shown in Table 2.
Table 2: State-wise Values of the Composite Indices, Measuring Social
Environment Status, Degree of Weakness in Service Area Coverage, Level of
Achievement through Programme Efforts, Extent of Economic
Backwardness and Degree of Unhealthiness in India*
Composite Indices Measuring
Social
Environ-
ment
Status
Degree of
Weakness in
the Service
Area Coverage
Level of
Achievement
through Programme
Efforts
Extent of
Economic
Backward-
ness
Degree of
Unhealthi-
ness
States

I
S

(Rank)
I
F

(Rank)
I
P

(Rank)
I
E

(Rank)
I
(Rank)
(01)
(02) (03) (04) (05) (06)
Madhya Pradesh

0.2497
(14 )
1.1676
(3)
0.8912
(8)
2.2327
(1)
1.6941
(1)
Uttar Pradesh 0.9161
(8)
1.1086
(4)
0.8257
(11)
1.3799
(5)
1.5128
(2)
Orissa 0.3578
(13)
1.2254
(2)
0.8547
(10)
1.9387
(3)
1.3687
(3)
Rajasthan 0.5219
(12)
0.7725
(11)
0.7068
(13)
0.8638
(8)
1.1077
(4)
Gujarat 1.2662
(5)
0.6733
(13)
1.1751
(3)
-0.0067
(13)
0.9186
(5)
Haryana 1.1829
(6)
0.6764
(12)
1.1172
(4)
-0.8438
(15)
0.9029
(6)
Andhra Pradesh 0.2438
(15)
0.8363
(9)
0.9891
(6)
1.4699
(4)
0.8558
(7)
Assam -0.9878
(16)
0.9024
(5)
0.4587
(16)
0.7867
(10)
0.8324
(8)
Bihar

0.8788
(10)
0.8438
(8)
0.6410
(15)
2.0442
(2)
0.8297
(9)
Karnataka 1.2808
(4)
0.8569
(7)
0.9808
(7)
1.0592
(7)
0.7978
(10)
Punjab 0.8199
(11)
0.5431
(15)
1.3005
(1)
-1.3523
(16)
0.6840
(11)
West Bengal 0.8896
(9)
0.7349
(10)
0.8008
(12)
0.7886
(9)
0.6855
(12)
Himachal
Pradesh
1.0080
(7)
2.2302
(1)
0.6184
(14))
-0.1114
(14)
0.5955
(13)
Maharashtra

1.3906
(3)
0.8924
(6)
1.2589
(2)
0.2731
(12)
0.5832
(14)
Tamil Nadu

1.4475
(2)
0.5832
(14)
1.0883
(5)
1.3691
(6)
0.4554
(15)
Kerala 2.3731
(1)
0.3596
(16)
0.8637
(9)
0.4804
(11)
-0.2886
(16)
INDIA 1,0000 1,0000 1,0000 1,0000 1,0000
* The value predominantly refer to the year 1991.




The states have been listed in decreasing order of I values, i.e. the state which is
most unhealthy has been put at the top and ranked 1, and the state, which is least
unhealthy has been put at the bottom and ranked 16. As per this ordering, the
health status of Madhya Pradesh, Uttar Pradesh, Orissa and Rajasthan cause
maximum concern. The values of II for these states are above the all India value
1. These states are also socio-economically backward with the values of I
S
ranked
as 14, 8, 13 and 12, and the values of I
E
, ranked as 1, 5, 3 and 8 respectively. All
other states have a status of health better than the all-India status.

However, it is worth noting that the achievement through programme efforts in
Rajasthan is rather poor in spite of much less weakness in the service area
coverage by the existing health service facilities. Given that in Rajasthan, the
degree of weakness in the service area coverage is low (with I
F
= 0.7725 and rank
=11) and the extent of economic backwardness is moderate (with I
E
= 0.8638 and
rank =8) in comparison to the all India values, the unhealthy status of the state
(with I = 1.1077 and rank = 4) might have something to do with its poor status of
social environment (with I
S
= 0.5219 and rank = 12) and poor achievement
through various programme efforts (with I
P
= 0.7068 and rank 13). The case of
Bihar is interesting because it is maintaining a moderately unhealthy status in
spite of its socio-economic backwardness and poor achievement through various
programme efforts under moderate weaknesses in the service area coverage. More
in-depth probing is required to understand this phenomenon. Further in-depth
probing is required for Gujarat also which has quite a favourable socio-economic
environment, service facilities and programme efforts, but has an unsatisfactory
health status. The moderately unhealthy status of Andhra Pradesh, on the other
hand, seems to be due particularly to its socio-economic backwardness.

Punjab has much scope to improve its health status by improving the social
environment status. In fact, its health service facilities, programme efforts and
economic status are almost the best among all the states, and yet, it is as unhealthy
as West Bengal, and worse than Himachal Pradesh, Maharashtra, Tamil Nadu and
Kerala. The situation in West Bengal, with respect to service area coverage,
achievement through programme efforts and economic status, is comparable to
that in Rajasthan, and yet, its health status is better because of better social
environment. The importance of social environment in improving health status
gets further established when the situation in Assam is compared with that in
Himachal Pradesh. Achievement through programme efforts is found
unsatisfactory in both these states, may be because of their difficult terrain. The
situation in Maharashtra, in comparison to that in Tamil Nadu and Kerala, makes
one feel that Maharashtra is capable of performing as well as Kerala if it
overcomes its weaknesses in service area coverage.









4.0 Concluding Remarks

The present study attempts to develop a methodology by which the health status
of a community can be quantitatively measured by using some proxy
demographic variables. The WHO definition of health status involving various
physical, psychological and social aspects makes it clear that the health status is
difficult to measure in absolute terms due to perceptional differences over time,
space and information/technology state. Thus, it needs measurement only in
relative terms. Use of the location factor concept has been considered here as an
option for quantifying health status in relative terms.

It takes the health system approach to explain the quantified health status in terms
of various factors in the system, contributing to it. This establishes how important
the factors related to social environment, service area coverage and programme
efforts are in determining a community's health status. Of course, no conclusive
claim can be made in this study on the extent of contribution of various factors
operating in the health system. The reason is that there are a number of data
limitations and the number of observations also is not large. In fact, the indices
that have been developed here could not be made comprehensive and more
logically acceptable due to non-availability of data on a number of other related
variables for the years around 1991. Also, a number of other aspects in the health
system, e.g. the political aspect, the private sector's participation in health service
provision, the service quality aspect, the aspect .of people's knowledge, attitude
and practice, and so on have not been taken into consideration due to lack of
reliable information. The health and the various environment and programme
indices may be improved further by better choice of the constituent variables.

In spite of these limitations, the methodology used here has brought to light in
objective and quantitative terms the utility of the systems approach for finding out
in relative terms the health status of different states in India, and also the
individual importance of various environment and programme factors in
determining the health status. In fact, there is still a lot of scope to rationalize each
of the constructed indices further and also incorporate new indices by taking into
consideration the other aspects which could not be incorporated in this study due
to non-availability of reliable data. There is also some scope for improving the
percentages of explained variations by appropriate mathematical transformation
of some of the variables, and thereby improving the method of index formulation.

The empirical exercise which has been done in this study is basically intended to
demonstrate the appropriateness of both the systems approach and the
methodology. The parameter estimates and the ranking of states require further
refinement by both increasing the number of observations (to increase the degrees
of freedom) and including a larger number or more appropriate variables that are
relevant for specifying the health system in its totality. It is felt that the results
from a more detailed exercise of similar nature covering all the districts in India
would be very useful to the national and state planners.

Acknowledgements

We are extremely thankful to Prof. Rushikesh M. Maru, Director, Institute of
Health Management Research, Jaipur, India, Dr. Barun Kanjilal, Associate
Professor in the same Institute, and the referees for their valuable comments. We
are also thankful to Dr. Ashok Agarwal, Trustee-Secretary of the Institute, for
showing keen interest in and appreciating the findings of this paper. We are
thankful to Prof. G. R. Rao for editing this paper. We greatly appreciate the
patience of Ms. Sudha Behal, who took extreme care in typing the text with
scrupulous attention to the accuracy of the statistical expressions.

References

Agarwal, A.N., H.O. Verma, and R.C. Gupta. [1993]: "India Economic
Information Year Book, 1992-93", National Publishing House, New Delhi, India.

Anderson, T.W. [1984]: "An Introduction to Multivariate Statistical Analysis",
2nd Edition, John Wiley and Sons, USA.

Cayolla da Motta, L.[1979]: "A Summary Indicator of Health Status in the
Municipalities and Districts of Portugal for Health Planning Purposes", in
Measurement of Levels of Health, Ed. Walter W. Holland, Johannes Ipsen and Jan
Kostrzewski, WHO Regional Publications, European Series, No.7, pp.179-193.

Central Bureau of Health Intelligence, Directorate General of Health Services,
Ministry of Health and Family Welfare, Government of India. [1991]: Health
Information of India.

Foundation for Research in Health Systems.[1993]: Health Monitor, Vol.1,No.1.

Kendall, M.G. [1939]: "The Geographical Distribution of the Crop Productivity in
England", Journal of the Royal Statistical Society, Vol.102, pp.21-48.

Mauldin, W. Parker and John A. Ross. [1991] "Family Planning Programs:
Efforts and Results", Studies in Family Planning, Vol.22, No.6, pp.350-367.

Mukhopadhyay, R. [1988]: "An Analysis on Spatial Patterns and Growth of
Indian Agricultural Activities: A Study on the Spatial Variations in Agricultural
Situation by District Group, 1960-61 to 1980-81", Ph.D. Thesis, submitted to
Indian Statistical Institute, Calcutta, India.

Pal, M.N. [1963]: "A Method of Regional Analysis of Economic Development
with Special Reference to South India", Journal of Regional Science, Vol.3, No.1,
pp.1-33. I.

Pal, M.N. [1971]: "Quantitative Techniques for Regional Planning", Indian
Journal of Regional Science, Vol.3, No.1, pp. 1-33.

Pal, M.N.[1990]: "A Generalised Formulation of Multi-collinear Spatial Indices
for Formal Regional Analysis", Indian Journal of Regional Science, Vol.22, No.1,
pp.11 - 33.

Pathak, Prasanta. [1995]: "A Consistent Phase-Order Preserving Equispaced
Optimal Index for Regional Analysis on Non-development of Basic Skill Among
Boys in India" in Essays on Regional Science: Regional Development and
Planning, Vol. (Silver Jubilee Issue), ed. C.R. Pathak, Regional Science
Association, India, I.I.T., Kharagpur, pp.41-56.

Pathak, Prasanta. [1991]: "Diagnostic Regional Analysis on the Shortfalls in
Development and Utilisation of Human Resources in India: A Statistical Study on
the Spatial Variations by Districts of India", Ph.D. Thesis submitted to Indian
Statistical Institute, Calcutta, India.
Saha, Vaskar [1984]: "Regional Analysis of Industrial Productions: A Statistical
Study on the Spatial Variations by Districts of India", Ph.D. Thesis, Jadavpur
University, India.

Tatlidil, Huseyin. [1992]: "A New Approach for Human Development: Human
Development Scores", IDS, Sussex, Mimeo.

Todaro, M.P. [1969]: "A Model of Labour Migration and Urban Unemployment
in Less Developed Countries", American Economic Review, Vol.59, pp.138-148.
Todaro, M.P. [1976]: "Urban Job Expansion, Induced Migration and Rising
Unemployment", Journal of Development Economics, Vol.3, pp. 211-225

United Nations Development Programme (UNDP). [1993]: "Human Development
Report 1993", Oxford University Press, Delhi.

World Health Organisation (WHO) [1994]: "Global Comparative Assessments in
the Health Sector: Disease Burden, Expenditures and Intervention Packages", ed.
C.J.L. Murray and A.D. Lopez.