Gender role and child health care utilization in Nepal.

Abstract

To study the gender role in child health care utilization in Nepal.
We analysed 8112 individual observations of age </=15 years from 2847 households in 274 communities, obtained from the 1996 Nepal Living Standard Survey. Four steps of a health seeking action, namely illness reporting, choosing an external care, choosing a specific health care provider, and spending money to treat the sick child, were examined using discrete/continuous choice models.
There was no statistically significant difference between boys and girls by demographic, socio-economic and geographical status in the sample. However, gender was associated with all four utilization decision steps. While the net effect of being a boy was modest in illness reporting (p<0.10), it appeared stronger in the choice of external care, in the choice of public provider and in the choice of expenditure with the private provider (p<0.05).
Gender role not only affects illness reporting but also affects the decision to choose a health care provider and how much to spend on the sick child, i.e. it affects the entire steps of a health seeking action.

Full text

Page 1

Health Policy 74 (2005) 100–109
Gender role and child health care utilization in Nepal
Subhash Pokhrela,∗, Rachel Snowb, Hengjin Dongc,
Budi Hidayata, Steffen Flessaa, Rainer Sauerborna
aDepartment of Tropical Hygiene and Public Health, University of Heidelberg, INF 324, D-69120 Heidelberg, Germany
bMichigan Centre on the Demography of Aging, 426 Thompson St., P.O. Box 1248, Ann Arbor, MI 48106, USA
cHealth Economics Research Group, Brunel University, Uxbridge, Middlesex UB8 3PH, UK
Abstract
Objective: To study the gender role in child health care utilization in Nepal.
Methods: Weanalysed8112individualobservationsofage≤15yearsfrom2847householdsin274communities,obtainedfrom
the1996NepalLivingStandardSurvey.Fourstepsofahealthseekingaction,namelyillnessreporting,choosinganexternalcare,
choosing a specific health care provider, and spending money to treat the sick child, were examined using discrete/continuous
choice models.
Results: There was no statistically significant difference between boys and girls by demographic, socio-economic and geo-
graphical status in the sample. However, gender was associated with all four utilization decision steps. While the net effect of
being a boy was modest in illness reporting (p<0.10), it appeared stronger in the choice of external care, in the choice of public
provider and in the choice of expenditure with the private provider (p<0.05).
Conclusion: Gender role not only affects illness reporting but also affects the decision to choose a health care provider and how
much to spend on the sick child, i.e. it affects the entire steps of a health seeking action.
© 2004 Elsevier Ireland Ltd. All rights reserved.
Keywords: Child health; Utilization; Gender bias; Nepal
1. Introduction
Gender refers to “women’s and men’s roles and re-
sponsibilities that are socially determined” [1]. Gen-
der differences in health, therefore, are caused by so-
ciety’s response to secondary sex characteristics that
distinguish males from females. Because a child ap-
∗Corresponding author.
E-mail address: subhash.pokhrel@urz.uni-heidelberg.de
(S. Pokhrel).
pears to be male, society provides opportunities and
constraints that affect his health [2]. Several studies,
in particular, those reported from South Asia, speak
much on gender bias in health and health care [3–8].
Whileeconomicopportunitiesandthreatsarerelatedto
such a difference by some [3], cultural attributes have
been given more importance in influencing the health
caredecisionsbyothers[4,8].Acommonground,how-
ever, is the recognition of the existence of strong son-
preference in South Asian societies including Nepal,
forbotheconomicandnon-economicreasons[3,9–11].
0168-8510/$ – see front matter © 2004 Elsevier Ireland Ltd. All rights reserved.
doi:10.1016/j.healthpol.2004.12.013

Page 2

S. Pokhrel et al. / Health Policy 74 (2005) 100–109
101
InNepal,childmortalityisstillhighanddifferssub-
stantially between boys and girls. The 1–4-year-old fe-
male child deaths per 100 male child deaths in 1970,
1985 and 1996 were, respectively, 108.7, 128.2 and
124.2, showing that the situation for girls relative to
boys increasingly deteriorated during the 1970s and
very little progress could be achieved afterwards [12].
A recent study has opened up a new argument that dif-
ferential child mortality in Nepal may be due to the
differentialaccesstohealthcarearisingprimarilyfrom
illness perception, not necessarily from the decisions
to seek care or spend money on the sick children [13].
The current 5-year plan as well as the long-term
health plan of His Majesty’s Government of Nepal has
targeted at reducing the child mortality rates further
by embarking upon several “first-priority” preventive
and curative care interventions for childhood diseases,
suchasexpandedprogramonimmunization,diarrhoea
and acute respiratory tract infection control, nutrition,
mother and child welfare, etc. [14]. In order to im-
plement these programs more effectively than in the
past, policy makers and program managers need to un-
derstand how different factors, including gender, may
affect the household decisions with regard to the use of
child health care. The purpose of this paper is, there-
fore,toestablishtheextenttowhichgenderroleaffects
child health care utilization in Nepal.
2. Methodology
2.1. Setting
Nepal is situated at the foot-hills of the Himalayas
with two-thirds of its land covered by hills and moun-
tains, the extremely difficult terrain with regard to ac-
cessibility. It has a population of over 23 million, 80%
of which is rural. The per capita income is US$230.
Nepal’s health care system relies heavily on the public
provision of service delivery. The Ministry of Health
provides care through 79 hospitals located at central,
regionalanddistrictlevels,178primaryhealthcenters,
705 health posts and 3132 sub-health posts located at
the community level [14]. Since the new health policy
in 1991, private provision has expanded rapidly. De-
spite the increase in number of service outlets, Nepal’s
health care infrastructure, especially in rural areas, is
characterized by a low level of utilization and poor
quality of care [14,15]. The current health care spend-
ing level in Nepal is about 5% of its gross national
product (GNP), of which 74% is accounted for out-of-
pocket payment [16]. The infant and under-five child
mortalityrateshavedeclinedsteadilyinthepastdecade
but they are still one of the highest in the world and are
characterized by substantial gender differentials [12].
The focus of child health services has been on the ex-
panded program on immunization, nutrition, control
of diarrhoeal and acute respiratory infections, and pro-
gramsthataddressleadingcausesofmortalityandmor-
bidity [14].
2.2. Sample design
We analysed the data obtained from the Nepal Liv-
ing Standards Survey (NLSS) which was administered
byNepal’sCentralBureauofStatistics(CBS),together
with the World Bank, in 1996. This survey included a
nationally representative sample of 3338 households
(18,855 individuals) in 275 communities. A two-stage
stratified sampling procedure was used: first selecting
wards(theprimarysamplingunits(PSU))fromthefour
strata(mountains,urban-hills,rural-hillsandtheTerai,
the plains in the southern part) with probability pro-
portional to sample size, and then choosing a fixed
number of households from each selected PSU with
equal probability. The details of the survey design are
described elsewhere [17]. We analyzed 8112 individ-
ual observations (4162 males and 3950 females) from
2847 households in 274 communities. These observa-
tionsincludeallchildrenaged15yearsorlesscaptured
by the survey.
2.3. Survey instrument
NLSS surveyed on various economic, demographic
and health-related behaviours. Questions related to
health, which we chose for our analysis, included: (a)
whether each household member was perceived to suf-
fer from an illness or injury in the past month, (b)
whethertheindividualsusedhealthcareservicesgiven
this illness/injury, and (c) if so, the type of place and
practitioner that was consulted, and the consultation
and travel costs for the first two visits. If tests were
carried out and drugs were given as the result of the
consultation during the visit, they were also included.
No costs for drug items were recorded separately.

Page 3

102
S. Pokhrel et al. / Health Policy 74 (2005) 100–109
2.4. Statistical methods
The analytical framework employed in this paper is
basedonthefour-stepconstructofhouseholddecision-
making process described by Pokhrel and Sauerborn
[13]. It assumes that the child health care decisions
followapathwaythatinvolvesfourstepsinahierarchy.
The steps are: (a) whether or not a child is reported
ill, (b) whether the sick child is offered an external
care given illness, (c) to which provider the child was
takento,and(d)dependinguponwhattypeofprovider
chosen, how much is spent for the treatment of the sick
child.
Inordertotranslatethisframeworkintoamultivari-
ate analysis, the four steps were considered sequential
decision steps in which each decision is governed by
different stochastic processes [18–20]. This gave rise
to a four-part model in which the decisions to report an
illness, to choose an external care (versus no care), to
chooseatypeofprovider(fromachoicesetcomprising
of public, private, pharmacy, and traditional healers),
andtochoosethelevelofexpendituretotreatthechild,
were modelled separately.
The first two steps (reporting an illness and choos-
ing an external care) were modelled with logit spec-
ifications [21]. The model goodness-of-fit for these
equations was assessed by Hosmer–Lemeshow test
[22,23]. The third step [choosing one provider from
the choice set {no or informal care, public providers,
private providers}] was specified as a multinomial
logit (MNL). Two specification tests were employed
to test the independence of the irrelevant alterna-
tives (IIA) assumption of the model, viz. Hausman
specification test and Small–Hsiao IIA test [21,24].
The final step (the level of expenditure) was esti-
mated by ordinary least squares (OLS). The expen-
diture decisions were modelled for users of private
providers and public providers separately. For nu-
merical significance, pharmacy and traditional healers
were included in the private provisions. This merg-
ing could affect our results because the decision-
making dynamics behind seeking treatment from tra-
ditional healers and from other private providers may
differ significantly, and these differences potentially
have an important gender dimension. Nevertheless,
merging these categories into private provision was
deemed as the most appropriate analytical strategy be-
cause the survey recorded a small number of obser-
vations pertaining to traditional healers and pharmacy
choices.
Since people self-select providers, a correction
might be needed in the OLS estimates of the expen-
diture. A method derived by Lee [25] and applied in a
recent study in Zambia [26] was applied. The selection
terms were found insignificant in both equations, and
hence, the final model did not include selection terms.
The details of the methods and results of these selec-
tionmodelsaregivenin[27].Allregressiondiagnostics
following OLS were carried out.
2.5. Independent variables
Gender was measured by a dummy variable
(0=girls, 1=boys). Other variables that were in-
cluded as the independent variable are: income, age
of the child, mother’s socio-economic status (her edu-
cation and employment as well as whether she heads
the household), household composition (number of
adults),locationvariables(mountain,hillsorTerai),ru-
ral/urban status, types of illness reported, and the time
to reach the nearest public clinic (proxy for health ser-
vice availability). Because information on the severity
of illness was not available, it was not included in the
model. Sanitation variables such as having piped water
was also included as independent variables. Including
types of illness together with age of the child and san-
itation variables helped us separate the gender’s net
effect from biological needs. Income was proxied by
total annual expenditures per capita [26,28,29].
3. Results
Table 1 presents the characteristics of the male and
female children included in the sample. It is important
to note that there was no statistically significant differ-
ence between boys and girls by demographic, socio-
economic and geographical characteristics.
The four-step decision-making process in our sam-
ple is depicted in Fig. 1. About 10% of the children
aged 15 years or less were reported to have suffered
from an acute illness or injury 1 month prior to the
survey date. Over two-thirds of the sick children were
taken to an external help. Of these, 56% were taken to
public providers. The health expenditure ranged from
Rs. 115 (US$2.05) with traditional healers to Rs. 256

Page 4

S. Pokhrel et al. / Health Policy 74 (2005) 100–109
103
Table 1
The characteristics of the sampled population (n=8112a)
CharacteristicsSex of the child
Male Female
p-Value
Mean age of the child (years)
Mean age of the mother (years)
Mothers who attended school (%)
Mothers who were not in self- or
wage-employment (%)
Mothers heading the household (%)
Mean income per capita (Rs.)
Living in rural areas (%)
Mean time taken to reach the nearest
health facility (min)
7.2
33.3
11.5
22.3
7.2
33.4
12.1
22.3
0.91
0.56
0.42
0.64
8.07.4 0.39
0.22
0.79
0.55
6187 6065
93.9
80.7
94.0
82.9
aMale children=4162; female children=3950.
(US$4.57) with the private providers in 1995/1996
value.
3.1. Decision to report an illness
A boy had a 15% more chance to have been re-
ported ill than a girl (the second column of Table 2).
The predicted probabilities of illness reporting were
higher for boys regardless of income (Fig. 2). Con-
sistent with the results from the descriptive analysis
[13], these predictions also show the non-linear rela-
tionship between illness reporting and income-lower
for the poorer who cannot afford to be ill, rising with
wealth, and then falling again as health and wealth are
associated.
3.2. Decision to seek external care given illness
Comparedtogirls,boyswereassociatedwitha42%
more likelihood of seeking external care given illness,
holing constant all other variables including the type
of illness reported (the fourth column of Table 2). This
effect was strong (p<0.05).
3.3. Decision to choose a health care provider
The last four columns of Table 2 report the relative
risk ratio (RRR) given by the multinomial logit model.
The results show that, compared with no or informal
care, being a boy predicted the probability of using the
public provider by 43% more than being a girl. This
effectwasstrong(p<0.05).Genderhadnoeffectinthe
Fig. 1. Child health care utilization decisions in Nepal (data source: NLSS 1996).

Page 5

104
S. Pokhrel et al. / Health Policy 74 (2005) 100–109
Table 2
Estimated odds ratio in logit models for child health care utilization decisions
VariableReport ill Seek care Choose provider
OR
|Z| statOR
|Z| stat Public vs. no/informalPrivate vs. no/informal
RRR
|Z| stat
(2.17)**
(0.04)
(0.45)
RRR
|Z| stat
(0.33)
(0.28)
(1.10)
Male child
Age 1–5 years
Age 6–15 years
Mother heads the household
Mother’s schooling (years)
Mother in wage-employment
Income quartile 2a
Income quartile 3a
Income quartile 4 (highest)a
Natural log of income
Number of adults in the household: 3–5
Number of adults in the household: >5
Piped water source
Mountain region
Hill region
Rural area
Public clinic 1 hour away
1.15
0.75
0.26
1.51
(1.81)*
(2.17)**
(9.89)***
(3.02)***
1.42(2.11)**
1.43
0.99
1.14
1.07
1.11
0.66
0.99
0.70
(0.23)
(1.40)
1.05
1.03
(1.31)
(0.08)
0.99
1.18
1.27
(0.10)
(1.24)
(1.67)*
2.50b
(4.65)***
1.78c
0.88
1.27
(3.68)***
(0.73)
(0.87)
2.63c
1.14
1.24
(4.66)***
(0.55)
(0.60)
0.69
0.63
0.80
(3.83)***
(3.06)***
(2.20)**
0.46
0.51
1.14
0.54
(2.44)**
(3.13)***
(0.37)
(2.77)***
0.93
0.97
1.67
0.57
(0.25)
(0.14)
(1.73)*
(2.93)***
0.03
0.22
0.47
0.92
(3.36)***
(5.72)***
(2.31)**
(0.30)
Reported illness (ref: diarrhoea)
Respiratory problems
Fever
Skin diseases
Injury
Others
0.50
0.57
2.90
0.72
0.57
(1.30)
(2.14)**
(1.38)
(0.77)
(1.96)*
0.67
0.48
0.91
0.49
0.46
(0.92)
(3.19)***
(0.17)
(1.70)*
(3.09)***
1.59
1.59
3.94
1.50
1.21
(0.77)
(1.34)
(1.96)**
(0.66)
(0.50)
Log likelihood
Chi-square (d.f.)
Model fit
Observations
−2452.65
247.43 (10)*
d
8112g
−426.85
64.40 (11)*
e
788h
−730.68
189.69 (34)*
f
788h
OR, odds ratio; RRR, relative risk ratio.
aIncome quartile 1 (lowest) to 4 (highest), quartile 1 being the reference group.
bIncome elasticity at mean=0.30.
cIncome elasticities at mean estimated at 0.21 (public) and 0.60 (private).
dHosmer–Lemeshow Chi-square (8)=5.3 (p=0.73).
eHosmer–Lemeshow Chi-square (8)=4.02 (p=0.85).
fSmall–Hsiao tests of IIA assumption: Chi-square (18)=24.048; prob>Chi-square=0.15 (omitting public); Chi-square (18)=16.04;
prob>Chi-square=0.59 (omitting private). Hausman test of IIA assumption: Chi-square (18)=7.796; prob>Chi-square=0.982 (omitting pub-
lic); did not meet asymptotic assumptions of the test (omitting private).
gMale children=4162; female children=3950.
hMale children=432; female children=356.
*Significant at 10%.
**Significant at 5%.
***Significant at 1%.