Page 137 - IJPS-2-2

Priyanka Dixit and Laxmi Kant Dwivedi

their births in an institution than those in the central region. For the third pregnancy, except for
women in the northeastern region (OR northeastern = 1.00; 95% CI: 0.60, 1.66) women in the other four
regions had a significantly higher chance of delivering in a hospital compared with those in the cen-
tral region. Living in urban areas was associated with a higher likelihood of delivering all three suc-
cessive births in a hospital. Muslim women were 17% less likely (OR Muslim = 0.83; 95% CI: 0.74,
0.93) to deliver their last birth in an institution compared with Hindu women. However, there was no
association between religion and an institutional delivery for the second to last and the third to
last births in this study.
Mothers from scheduled castes/tribes were less like to use institutional delivery for all three suc-
cessive births. However, we found no significant association between scheduled castes/tribes and the
use of institutional delivery for their recent second and third births. Mothers with higher education
were more likely to deliver all three births (OR first birth = 3.44; 95% CI: 3.06, 3.87), (OR second birth =
3.81; 95% CI: 3.19, 4.56) (OR third birth = 2.61; 95% CI: 1.65, 4.14) in an institution compared with
illiterate women. Additionally, the likelihood of using institutional delivery increased with increasing
levels of the respondent’s education; with the exception of the third birth where there was no differ-
ence for illiterate women. A positive association between the use of institutional delivery and educa-
tion among the partners of women was also found for the last and the second to last births. For ex-
ample, women with partners with middle-school or above educational attainment had 26–31%
greater odds of deliveries at a medical institution versus home for the last and the second to
last children compared to women with illiterate partners. However, no educational difference was
found for the third to last birth.
The difference in the use of institutional delivery by wealth quintile was more pronounced. Moth-
ers in a higher wealth quintile were associated with significantly higher odds of delivering all three
successive births in an institution. For example, mothers in the second (poor), third (middle), fourth
(rich), and fifth quintiles (richest) of the wealth index were all more likely to use institutional deli-
very than the poorest. A possible reason why women from poor families had higher odds of using
institutional delivery than those from the poorest families could be the introduction of different “free
maternity benefit schemes,” which may attract women to utilize institutional delivery even if they
were in the second quintile group (poor households).
Women who experienced the loss of a child during their reproductive period (not necessarily
within the 5-year study period)were more likely to have all three successive births in an institution.
Women who experienced a terminated pregnancy had higher odds of delivering their most recent and
second to last birth in an institution; however, this variable was not significant for the third to last
birth. Compared with second-order births, first-order births were more likely to be delivered in an
institution, but third- and higher-order births had less of a chance to be delivered in an institution.
Table 3 shows the random parameter estimates for the three-level multivariate regression models
for the utilization of institutional delivery in India from the 2005–2006 NFHS-III using the
second-order approximation penalized quasi-likelihood. The results indicate that the covariance be-
tween place of birth of the most recent (last) and second to last birth was higher (0.46) than between
the last and third to last births (0.28). These results indicate that utilization of institutional delivery
for the two subsequent births was significantly related to each other. Using the second-order ap-
proximation and maximize quasi-likelihood method, we obtained random parameter estimates for
the three-level multivariate regression models for institutional delivery in India (see Table 4).The
unconstrained binomial variance parameter for the three pregnancy outcome was estimated to be
σ e 2 11 = 1.00, σ e 2 22 = 0.97, and σ e 2 33 = 0.95, which was very close to one. This indicates that the data
in the multivariate multilevel model supported the binomial assumption. The estimated covariates
derived from the multivariate multilevel model were positive and highly significant. The interesting
finding is that the correlation between successive pregnancies was high compared with the estimates
from the non-successive pregnancies, suggesting that once the decision has been made for the place

International Journal of Population Studies | 2016, Volume 2, Issue 2 131