Professional Documents
Culture Documents
542
MATERNAL HEALTHCARE UTILIZATION IN INDIA 543
KEY MESSAGES
This study examined factors associated with maternity healthcare utilization in nine high focus states (which share the
utmost burden of maternal mortality in India) accounting for individual-, household-, community- and district-level
factors.
Along with individual-/household-level factors, communities with high concentration of poor and illiterate women are
disadvantageous in utilizing three recommended maternal healthcare services.
District-level factors such as average population coverage per PHC, availability of labour room in PHC and the percentage
of registered pregnancies in the district significantly influence the utilization of maternal healthcare services.
The findings of this study suggest evidence based and targeted interventions that should adopt a multilevel approach to
address barriers to maternity healthcare services utilization beyond individual and household factors.
Figure 1 MMR for India and major states, Sample Registration System, 2007–09. Notes: Abbreviations shown for the states are as follows: AP,
care. However, it is important to recognize that the particular of community position and the health system attributes at
model is probably an oversimplification and that, at each level district level.
of care, there may be a series of complex factors affecting care
(Bhutta 2011). This study adopts the health behaviour model
Data
and its subsequent modifications (Aday and Andersen 1974;
This study analysed data from the third round of the District
Andersen 1995) along with the Social Determinants of Health
Level Household and Facility Survey (DLHS-3) conducted
(SDH) framework proposed by the World Health Organisation
during 2007–08 (IIPS 2010). The DLHS is a nationally repre-
(WHO) Commission on SDH (CSDH 2007). The health behav-
sentative cross-sectional survey carried out in 34 states covering
ioural model proposed that health seeking behaviour is a
601 districts in India. DLHS-3 questionnaires were canvassed in
function of three sets of individual characteristics: (1) predis-
720 320 households and from 643 944 ever-married women
posing characteristics, e.g. age, household size, education,
aged 15–49 years. The facility survey covered 18 068 sub-centres,
number of previous pregnancies, etc., (2) enabling characteris-
8619 primary health centres (PHCs) and 4162 community
tics, i.e. income, contact to health workers, etc. and (3) need
health centres. The broad objective of DLHS-3 was to provide
characteristics, i.e. perceived health status, and expected benefit
reproductive and child health outcome indicators at the district
from treatments. Moreover, the SDH framework extends the set
level to monitor and provide corrective measures. The financial
of healthcare factors to include the importance of governance,
support for DLHS-3 was provided by the Ministry of Health and
and government policies. The SDH framework underscores
Family Welfare (MOHFW), Government of India, United
mainly two sets of factors, i.e. (1) structural and (2) inter-
Nations Population Fund (UNFPA) and United Nations
mediary determinants. The structural determinants include the
Children’s Fund (UNICEF).
socio-economic and political context, as well as indicators of
social position such as education, income, occupation, social
class, gender and ethnicity. The weight and relevance of the Sampling techniques and study population
assigned social position is influenced by the socio-economic and A multi-stage stratified systematic sampling design was em-
political context, including governmental policies, cultural ployed in the DLHS-3 survey. In each district, 50 primary
values and the macroeconomic conditions. Furthermore, the sampling units (PSUs) were selected, which were villages in
impact of these structural determinants on equity in health and rural areas and census enumeration blocks (CEBs) in urban
wellbeing is mediated by different intermediary determinants areas. In rural areas, first, villages were selected by probability
such as living conditions and exposure that directly influence proportional to size (PPS) systematic sampling, and in the
health, as well as access to and quality of care received when second stage, households were selected by systematic sampling.
encountering the health system (CSDH 2007). This study For urban areas, first, wards were selected by PPS systematic
considers these healthcare frameworks, and applies a multilevel sampling, and in the second stage, CEBs were selected by PPS
approach in explaining the utilization of maternal healthcare sampling, and then households were selected by systematic
services, where along with an individual’s social position and sampling in the third stage.
other intermediary factors at one level, indicators of governance The term ‘community’ used throughout this article refers to
and government policies at other levels are examined in tandem the clustering within the same geographical living environment.
(considering the structure of potential indicators). In particular, These communities were measured based on sharing a common
along with the background or predisposing characteristics of PSU in the DLHS-3 data. PSUs were census villages in rural
the population at risk, the analysis explicitly delves into the role areas and CEBs in urban areas based on the 2001 census.
MATERNAL HEALTHCARE UTILIZATION IN INDIA 545
Details of sampling techniques and sampling weights are and Fatusi 2009). The three community-level indicators such as
available in the DLHS-3 report (IIPS 2010). The study sample type of residence (urban/rural), proportion of illiterate women
includes 125 721 ever-married women aged 15–49 years in nine in PSU and proportion of women belonging to the poorest
high focus states, who had at least one child during the 3 years wealth quintile in PSU are considered. The urban health
preceding the survey. These sample women hail from 14 385 advantage has often been attributed to the improved modern
PSUs (communities) across 292 districts in the study area. healthcare system that facilitates public health interventions
(Fotso 2006). The high concentration of people belonging to the
Outcome variables poor (Gage and Calixte 2006) and illiterate (Kravdal 2004) in
The study measures three outcome variables such as 4 ANC the community could have an impact on health and mortality
visits, skilled birth attendance (SBA) and PNC within 2 days outcomes. The theories of social capital (Coleman 1990)
after delivery as indicators of maternal healthcare utilization. including social networks and relationships (Gage 2007),
These three selected indicators of maternal healthcare utiliza- community awareness and diffusion (McNay et al. 2003; Gage
tion and their components are based on the guidelines and Calixte 2006) provide a framework for understanding the
developed by the MOHFW, Government of India and the mechanism through which communities may influence the
World Health Organization (MOHFW 2012). Delivery conducted utilization of healthcare services.
either in a medical institution or a home delivery assisted by
doctor/nurse/Lady Health Visitor/Auxiliary Nurse Midwife District-level factors
(ANM)/other health professionals are termed ‘SBA’ (WHO In the majority of the developing countries like India, the
Table 1 Exposure variables used in modelling the utilization of maternity healthcare services in nine high focus states in India, DLHS-3 (2007–08)
Note: Analytical sample size and percentage distribution for all the exposure variables are presented in Appendix 1.
The dependent variable in our analysis is whether a woman proportions. However, women residing in rural areas and in
had at least 4 ANC visits (no/yes), had SBA (no/yes) and communities with high concentration of illiterate and poor
received post-natal check-ups (no/yes). Due to the binary women utilized lower proportions of all three maternity care
nature of the dependent variable, the multilevel model with services. The three maternity care services also varied with
logit link function can be described as follows: district-level indicators including population coverage per PHC,
functionality of PHC, availability of labour room in PHC and
picd
ln ¼ þ xicd þ wcd þ zd þ ucd þ vd , proportion of registered pregnancies in the district.
1 picd
where ln(picd/(1 picd)) is the logit in which picd is the probability
Factors associated with the utilization of maternal
of woman ‘i’ in community (PSU) ‘c’ in district ‘d’ using maternity
healthcare services
healthcare services; xicd , wcd and zd are vectors of individual-/
household-, community- and district-level characteristics; a is a In the multilevel models, community (PSU) and district of
constant, while , and are vectors of estimated parameter residence were modelled to be random. The results of the
coefficients; and ucd and vd are unexplained residual terms at the random intercept only model (empty model) are shown in
community and district level, respectively. Thus, a multilevel Table 3. There were considerable variations in the utilization
model with three levels was fitted to assess the influences of of the three maternal healthcare services across communities
measured individual-household-, community- and district-level (14–16%) and districts (6–7%) in nine high focus states. Table 4
factors as fixed effects, and community (ucd) and district (vd) as presents the results of unadjusted (univariate) and adjusted
random effects on the utilization of maternity healthcare services. (multivariate) models when individual-household-, commu-
The odds of SBA seemed to decline significantly with the Post-natal care
reduction in the level of women’s education. Women who had In the full adjusted multilevel model, individual-/household-
not seen or heard about the safe delivery services were less level variables such as birth order, women’s education, aware-
likely to have SBA [odds ratios (OR) ¼ 0.75; 95% confidence ness of information/messages related to ANC and safe delivery
interval (CI) ¼ 0.73–0.77], compared with women who were services, and household economic status (wealth quintile)
aware about the delivery services. The likelihood of SBA was transpired as significant predictors of PNC services utilization.
43% lower among the poorest women compared with the The odds of PNC were lower among women with 3 order birth
richest. Rural women were 22% less likely to avail themselves (OR ¼ 0.58; 95% CI ¼ 0.56–0.59) compared to women with first
of SBA compared with their urban counterparts. The results order birth. PNC seemed to decline with the decrease in the
suggest that the increasing number of illiterate women and of level of women’s education, and household economic status.
women belonging to the poorest wealth quintile at the Women who were not exposed to messages related to delivery
community level were negatively associated with SBA. Except services (OR ¼ 0.90; 95% CI ¼ 0.88–0.93) and ANC (OR ¼ 0.91;
Uttarakhand, the odds of SBA were higher in all states 95% CI ¼ 0.88–0.94) were less likely, respectively, to utilize PNC
compared with Uttar Pradesh. SBA declined with the decreas- services compared with women who had heard/seen messages
ing proportion of PHC with labour room availability and related to delivery and ANC services. At the community and
registered pregnancies in the district. Women who did not district levels, type of residence, average population covered per
take 4 ANC visits were less likely to have SBA compared with PHC, per cent PHC with labour room facility and percentage of
women who had 4 ANC visits. registered pregnancies were significantly associated with the
MATERNAL HEALTHCARE UTILIZATION IN INDIA 549
utilization of PNC services. Previous experiences with the use of services. The individual/household characteristics found to be
maternal care services appeared to have strong influence on the particularly important are birth order, women’s education,
utilization of PNC services. knowledge about maternity care services, social and religious
affiliations and household economic conditions. Studies from
developing countries have shown an inverse relationship
between birth order and maternal healthcare utilization (Bell
Discussion et al. 2003). Since the lower parity women are inexperienced
Despite the progress that has been made towards improving the with pregnancy-related activities and perceive greater health
utilization of maternal healthcare services during the last two risks, they tend to heed skilled delivery (Raj et al. 2009).
decades, MMR remains unacceptably high in these nine high However, time and resource constraints faced by those with
focus states. Although widely published evidence from India larger families among higher parity women could restrict their
has highlighted the factors that contribute to the utilization of use of maternity healthcare services (Singh et al. 2012b).
maternity healthcare services, few studies have exemplified it Many hypotheses have been proposed for the mechanism
by using a comprehensive framework using potential factors at through which increased education could lead to reduction in
different levels. This study investigates factors associated with maternal and child mortality rates, through timely use of
three maternal healthcare outcomes (namely, 4 ANC visits, healthcare services, economic advantages, high autonomy and
SBA and PNC within 2 days after delivery) in nine high focus improved status in family and society (Levine and Rowe 2009).
states by applying a multilevel approach considering a range of This finding is particularly imperative for these nine high focus
potential individual-household-, community- and district-level states in the wake of the fact that nearly three in five women in
covariates. these states had no or below primary level education. According
The results show a number of individual-, household-, to the 2011 Census, of 292 districts in the study area, 72
community- and district-level factors that significantly influ- districts recorded female literacy of <50%, while only 1 district
ence the decision to utilize the three maternal healthcare had male literacy of below 50% (RGI and CC 2011). Lack of
550 HEALTH POLICY AND PLANNING
Table 2 Percentage of women who had at least one live birth during Table 2 Continued
the last 3 years preceding the survey by usage pattern of maternity
Background characteristics 4 ANC SBA PNC
healthcare services and by background characteristics, high focus states visits
in India, DLHS-3 (2007–08)
Community-level variables
Background characteristics 4 ANC SBA PNC Type of residence
visits
Urban 30.6 63.0 52.6
Individual-household-level variables
Rural 11.3 32.6 28.1
Mother’s age at child birth
Proportion of illiterate women in PSU
35–49 6.8 23.5 22.5
0–25% 32.5 51.6 35.2
25–34 13.3 33.9 29.9
26–50% 16.5 43.4 34.8
13–24 15.4 40.9 33.9
>50% 7.7 33.1 29.9
Birth order
Proportion of women in PSU from lowest wealth quintile
1 21.9 52.8 40.8
0–25% 18.3 42.7 37.7
2 17.3 40.9 34.4
26–50% 9.5 32.4 26.4
3 7.3 25.4 24.6
>50% 11.9 26.2 22.2
Mother’s childcare burden
District-level variables
No burden 15.4 37.8 31.9
Average population covered per PHC in district
Table 3 Parameter coefficients for the multilevel model (random intercept only model, without covariates) for various
indicators of the use of maternal healthcare services, high focus states in India, DLHS-3 (2007–08)
Notes: PSU ¼ Primary Sampling Unit; SE ¼ Standard Error; VPC ¼ Variance Partition Coefficient.
knowledge about the messages related to the safe delivery and 1400 in rural areas after delivering in a government or
ANC have shown a negative effect on the utilization of accredited private health facility. However, a few recent
maternal healthcare services, which highlights the importance evaluation studies show that the poorest and illiterate women
of disseminating specific healthcare knowledge that brings have not always been benefited by the JSY scheme (Lim et al.
about changes in people’s attitude towards timely utilization of 2010). For instance, a study in Rajasthan has documented that
skilled healthcare services (Ghosh 2006). more than 50% of the eligible women were not benefited from
The significant impact of religion (Hazarika 2011) and castes the JSY, and that the reach of JSY remained inequitable for
3 0.30 (0.29–0.31)*** 0.68 (0.66–0.70)*** 0.30 (0.29–0.31)*** 0.58 (0.56–0.59)*** 0.47 (0.46–0.49)*** 0.90 (0.87–0.92)***
Mother’s childcare burden
No burden 1.00 1.00 1.00 1.00 1.00 1.00
At least one burden 0.65 (0.63–0.67)*** 1.01 (0.98–1.04)ns 0.95 (0.88–0.99)** 0.94 (0.85–1.01)** 0.92 (0.90–0.95)*** 1.01 (0.99–1.03)ns
Women’s education
Higher secondary and above (Ref.) 1.00 1.00 1.00 1.00 1.00 1.00
Middle 0.40 (0.38–0.42)*** 0.77 (0.74–0.79)*** 0.41 (0.39–0.42)*** 0.77 (0.74–0.79)*** 0.53 (0.51–0.55)*** 0.90 (0.87–0.94)***
Primary 0.26 (0.25–0.28)*** 0.70 (0.67–0.73)*** 0.27 (0.26–0.28)*** 0.71 (0.69–0.73)*** 0.39 (0.38–0.41)*** 0.89 (0.86–0.92)***
Illiterate 0.11 (0.11–0.12)*** 0.62 (0.60–0.64)*** 0.15 (0.14–0.15)*** 0.67 (0.65–0.70)*** 0.25 (0.24–0.26)*** 0.87 (0.84–0.90)***
Husband’s education
Higher secondary and above (Ref.) 1.00 1.00 1.00 1.00 1.00 1.00
Middle 0.48 (0.46–0.50)*** 0.90 (0.88–0.93)*** 0.54 (0.52–0.56)*** 0.93 (0.91–0.96)*** 0.63 (0.61–0.65)*** 0.98 (0.95–1.01)ns
Primary 0.34 (0.33–0.36)*** 0.84 (0.81–0.87)*** 0.38 (0.37–0.40)*** 0.90 (0.87–0.92)*** 0.50 (0.48–0.52)*** 0.99 (0.96–1.03)ns
Illiterate 0.20 (0.19–0.21)*** 0.82 (0.79–0.85)*** 0.25 (0.24–0.26)*** 0.85 (0.83–0.88)*** 0.35 (0.34–0.36)*** 0.93 (0.90–0.96)***
Heard/seen message on delivery services
Yes (Ref.) 1.00 1.00 1.00 1.00 1.00 1.00
No 0.35 (0.33–0.36)*** 0.84 (0.81–0.87)*** 0.37 (0.36–0.38)*** 0.75 (0.73–0.77)*** 0.44 (0.43–0.46)*** 0.90 (0.88–0.93)***
Heard/seen message on ANC
Yes (Ref.) 1.00 1.00 1.00 1.00 1.00 1.00
No 0.21 (0.20–0.22)*** 0.69 (0.65–0.72)*** 0.34 (0.33–0.35)*** 0.85 (0.81–0.89)ns 0.40 (0.38–0.42)*** 0.91 (0.88–0.94)***
Social groups
Others (Non-SC/ST) (Ref.) 1.00 1.00 1.00 1.00 1.00 1.00
Scheduled Tribes (ST) 0.63 (0.60–0.66)*** 0.83 (0.80–0.87)*** 0.54 (0.52–0.56)*** 0.83 (0.80–0.86)*** 0.57 (0.55–0.59)*** 0.91 (0.88–0.95)***
Scheduled Castes (SC) 0.58 (0.55–0.60)*** 0.90 (0.87–0.92)*** 0.67 (0.65–0.69)*** 0.91 (0.89–0.94)*** 0.69 (0.67–0.72)*** 0.98 (0.96–1.01)ns
(continued)
(continued)
Level of significance ***P < 0.001; **P < 0.05; *P < 0.10; (Ref.) ¼ reference categories are indicated in parentheses after the names of the characteristics being considered. na, not applicable; ns, not significant.
Likelihood ratio (LR) tests for multilevel logistic vs ordinary logistic regression model: for 4 ANC (2 ¼ 4315.44, P < 0.001); for SBA (2 ¼ 6715.68, P < 0.001); for PNC (2 ¼ 14267.03, P < 0.001). These LR tests are
conservative and provided only for reference; calculations are based on the Laplacian approximation. Final models are based on penalized quasi likelihood estimation that does not provide LR test.
Sample size: at level 1 (individual) ¼ 125 721; level 2 (community) ¼ 14 385; level 3 (district) ¼ 292.
This study also found the average population covered per the interview. However, many studies have examined maternal
PHC, proportion of PHC with labour room availability and healthcare use retrospectively even up to the last 5 years period;
percentage of registered pregnancies at the district level as the present study minimized these errors while considering
significant influencing factors. According to the guidelines of births up to 3 years period. Not all predictors of maternal
the Indian Public Health Standards a PHC covers 20 000 to healthcare services use were included in the study due to
30 000 population, but due to a huge shortage of PHCs in the limitations of the data and scope of the study. For instance,
study area, it has adverse implications for quality assurance and healthcare use could be determined by certain community
maintenance (Rai et al. 2011). Since the launch of the JSY norms including cultural practices related to pregnancy care
scheme, the government health facilities have become over- etc., that were difficult to measure from the available dataset.
crowded. A study conducted in Jharkhand, India found that the This study also acknowledges the limitation in considering
majority of the women were discharged within 24 h, and 50% measures of quality of healthcare services such as waiting time,
of them within 3 h due to lack of resources in hospitals (Rai staff attitudes and behaviour, which may influence women’s
et al. 2011). In addition, SBA and PNC were also determined by decisions on whether or not to make use of given facilities.
the percentage of PHCs equipped with a labour room in the However, the district-level variables included in this study
district. A recent evaluation has registered lower availability of provide some scope to assess the structural features of health
labour rooms in PHCs in the nine high focus states ranging facilities available in the study area. This study defined
from the highest of 28% in Rajasthan to the lowest in community based on PSUs and CEBs in rural and urban
Jharkhand (4%), compared with the national average of 40% areas, respectively, and these boundaries may not conform to
The findings also highlight the need to look beyond individ- References
ual- and household-level factors while assessing the determin-
Aday LA, Andersen RM. 1974. A framework for the study of access to
ants of maternal healthcare services use. To increase the
medical care. Health Services Research 9: 208–20.
utilization of maternal healthcare services, focus on commu- Ahmed S, Creanga AA, Gillespie DG, Tsui AO. 2010. Economic
nity- and district-level interventions is critical. There is an status, education and empowerment: implications for maternal
immense need to engage directly with communities, including health service utilization in developing countries. PLoS One 5:
women, men, community leaders and other stakeholders e11190.
towards raising awareness about the adverse consequences of Ahmed S, Li Q, Liu L, Tsui AO. 2012. Maternal deaths averted by
lower or indeed untimely utilization of maternal healthcare for contraceptive use: an analysis of 172 countries. Lancet 380: 111–25.
women and their newborns. A clear evidence of benefit from Andersen RM. 1995. Revisiting the behavioural model and access to
the greater focus on infrastructural development at district level medical care: does it matter? Journal of Health and Social Behaviour
has appeared from the present analysis. Expanding basic health 36: 1–10.
facilities such as increasing number of PHCs to reduce the Babalola S, Fatusi A. 2009. Determinants of use of maternal health
burden from existing facilities, and ensuring labour room services in Nigeria—looking beyond individual and household
availability at PHCs are some of the key steps to strengthen factors. BMC Pregnancy Childbirth 9: 43.
district level health infrastructure. Exploring strategies to Barros AJ, Ronsmans C, Axelson H et al. 2012. Equity in maternal,
enhance the district level pregnancy registration is another newborn, and child health interventions in Countdown to 2015: a
essential aspect that significantly affects current and future retrospective review of survey data from 54 countries. Lancet 379:
1225–33.
Hazarika I. 2011. Factors that determine the use of skilled care during Rai SK, Dasgupta R, Das MK, Singh S, Devi R, Arora NK. 2011.
delivery in India: implications for achievement of MDG-5 targets. Determinants of utilization of services under MMJSSA scheme
Maternal and Child Health Journal 15: 1381–8. in Jharkhand ‘Client Perspective’: a qualitative study in a
International Institute for Population Sciences (IIPS). 2010. District Level low performing state of India. Indian Journal of Public Health 55:
Household and Facility Survey, 2007–08: India. Mumbai: IIPS. 252–9.
International Institute for Population Sciences (IIPS). 2012. Raj A, Saggurti N, Balaiah D, Silverman JG. 2009. Prevalence of child
Concurrent Evaluation of National Rural Health Mission, 2009-10: India. marriage and its effect on fertility and fertility-control outcomes of
Mumbai: IIPS. young women in India: a cross-sectional, observational study.
Lancet 373: 1883–9.
Jat TR, Ng N, San SM. 2011. Factors affecting the use of maternal
health services in Madhya Pradesh state of India: a multilevel Raj AU, Vaidya A, Shakya-Vaidya S, Petzold M, Krettek A. 2012.
analysis. International Journal of Equity in Health 10: 59. Establishing a health demographic surveillance site in Bhaktapur
district, Nepal: initial experiences and findings. BMC Research Notes
Kesterton AJ, Cleland J, Slogett A, Ronsmans C. 2010. Institutional
5: 489.
delivery in rural India: the relative importance of accessibility and
economic status. BMC Pregnancy and Childbirth 10: 30. Rasbash J, Charlton C, Browne WJ, Healy M, Cameron B. 2009.
MLwiN Version 2.1. Centre for Multilevel Modelling, University of
Kiros GE, White MJ. 2004. Migration, community context, and child
Bristol.
immunization in Ethiopia. Social Science and Medicine 59: 2603–16.
Registrar General of India (RGI). 2006. Maternal Mortality in India: 1997-
Koblinsky M, Matthews Z, Hussein J et al. 2006. Going to scale with
2003, Trend Causes and Risk Factors. New Delhi, India: Ministry of
professional skilled care. Lancet 368: 1377–86.
Home Affairs, Government of India.
Kravdal Ø. 2004. Child mortality in India: the community-level effect of
Continued
Background characteristics n %
Per cent PHC with labour room available in district
>50% 81 818 65.1
25–50% 30 958 24.6
<25% 12 945 10.3
Percentage of pregnancies registered at district level
>50% 75 915 60.4
25–50% 32 082 25.5
<25% 17 724 14.1
High focus states
Uttar Pradesh 37 260 29.6
Uttarakhand 4113 3.3
Rajasthan 12 428 9.9
Bihar 21 138 16.8
Assam 9882 7.9