Page 140 - IJPS-2-2

Utilization of institutional delivery services across successive births in India

motivated to deliver in an institution if they experienced no problems in their previous deliveries.
Contrary to this, if women experienced the termination of a previous pregnancy or lost a child, they
would be more conscious about a safe delivery, and therefore, they would tend to deliver in an insti-
tution. However, this is not the case for the third to last birth. This may be because a woman who
had a bad reproductive history has become more conscious for her subsequent births.
In this study the birth order of a child was one important predictor in determining institutional de-
livery in the three-level multivariate analysis. If the birth was the first birth, the mother was more
likely to deliver in an institution compared with their second-order birth. These data are consistent
with studies from developing and low income countries (Feyissa and Genemo, 2014; Kabakyenga,
Östergren, Turyakira et al., 2012; Shimazaki, Honda, Dulnuan et al., 2013); they are also consistent
with studies from different regions of India (Kesterton, Cleland, Slogett et al., 2010; Sugathan, Mi-
shra, and Retherford, 2001). The likely reason is that women with more children perceive delivery as
a normal process and develop the confidence to give birth at home.
Women who had three births in the study period were also more likely to deliver at home com-
pared with women who had two births. Among women who had two births, the likelihood of deli-
vering the first birth in an institution was higher than that of delivering the second birth in an institu-
tion during the observation period. This may be due to concern about the successful delivery of a
healthy first child, and therefore women maybe more likely to use a healthcare facility for delivery
(Gabrysch and Campbell, 2009). However, in the absence of complications, women may forgo an
institutionalized delivery (Stephenson and Tsui, 2002).
An advantage of the current study is the application of a three-level multivariate model. Over the
years, researchers have identified individual and household characteristics that are associated with
the institutional delivery of the most recent birth. However, by modeling the inter-dependence
among successive births, the multivariate model provides additional information. Similar to previous
studies, we tested a three-level model to examine how the effects of women’s characteristics at dif-
ferent levels are associated with delivery in an institution (Yebyo, Alemayehu, and Kahsay, 2015).
We found that the assumption of independent Bernoulli trials is violated in the multilevel model. A
three-level multivariate multilevel model provides a solution to this problem and consists of more
than one response variable. Although multivariate multilevel methods are popular in epidemiological
studies (Oksanen, Kawachi, Subramanian et al., 2013; Subramanian, Kim, and Kawachi, 2005),
prior studies have not addressed the utilization of delivery care. Using this model, we considered the
place of delivery for three births of a mother as the dependent variable(s). Taking advantage of the
multivariate multilevel modeling approach, we were able to investigate the co-variation in socioe-
conomic correlates of place of delivery for all three births for a given mother.
Our models reveal that the variation in the utilization of institutional delivery at the level of the
woman is smaller than that at the PSU level, indicating that even after controlling for house-
hold-level variables, the clustering of use of institutional delivery exists between birth orders for the
same woman. In other words, the utilization of institutional delivery can be considered as a motivat-
ing factor for subsequent deliveries in an institution, and eventually, such repeated behaviors may
elicit habitual behaviors.
Despite the strengths discussed above, the study has the following limitations. First, each woman
did not have the same number of births during the five-year study period, which introduces
some biases in the analyses. Second, the interactions between age, birth cohorts, and place of deli-
very were not included in the analyses because of the complexity of modeling. Therefore, this prohi-
bits us from examining potential cohort and period effects in the use of institutional delivery. Third,
the results could be more informative if data on the quality of health services were collected. Never-
theless, we encourage more research on this topic to further explore the factors associated with the
use of institutional delivery services.
Our findings have important policy implications. Policies aimed at improving the consistent utili-
zation of institutional delivery in India should focus on first-time mothers — particularly targeting

134 International Journal of Population Studies | 2016, Volume 2, Issue 2