Page 36 - IJPS-2-2

Infant mortality differentials among the tribal and non-tribal populations of Central and Eastern India

that out of 119,530 total women in the overall sample, around 56% reported that they never attended
a school. So, in the illiterate category we included both the women who never attended school as
well as women who responded that they had zero years of schooling. Mother’s education reduces the
risk of infant death substantially because educated women tend to take care of all the nutritional re-
quirements of the child in the initial years of life and will also understand the importance of exclu-
sive breastfeeding and immunization schedules.
There are many well-established theories showing that higher fertility is associated with higher
risk of infant death (Das, Kapoor and Nikitin, 2010). Thus, we classified birth order into two catego-
ries: 1–2 births versus 3 or more births. The wealth index was created using household assets, amen-
ities, and durable goods. In the DLHS-III, a household was comprised of a person or group of per-
sons who usually live, eat together, and share a common kitchen. Combining household amenities
(such as access to safe drinking water, electricity, improved sanitation, fuel used for cooking, type of
house, per capita space in the house), assets (owns house, owns agricultural land and ownership
of certain animals like cow/buffalo, camels, horses/donkeys, goats, sheep, chicken/ducks, and pigs)
and durables (such as a mattress, pressure cooker, chair, sofa set, cot or bed, table, electric fan, radio
or transistor, black and white television, color television, sewing machine, mobile and other tele-
phone, computer, refrigerator, washing machine, watch or clock, bicycle, motorcycle or scooter, an-
imal-drawn cart, car, tractor, water pump, and thresher), a wealth index is computed at the national
level and divided into quintiles. The factor loading to amenities, assets and durables derived by fac-
tor analysis was used for the computation of the wealth index (factor analyses were not shown).
Households were categorized from the poorest to the richest groups corresponding to the lowest to
the highest quintiles. For analytical purposes, it was re-categorized into three categories: poor, mid-
dle, and rich.
For tribes, religion was classified into three categories as Hindu, Christian, and others, while for
non-tribes it was grouped as Hindu, Muslim, and others. The reason for such a division was because
many of the tribes in this region have converted to Christianity. Furthermore, we considered the
place of a pregnancy registration as a very important program factor because it covers women who
will further receive or use other healthcare services (such as ante-natal care visits, tetanus injection,
and health related information) once she gets registered. The question related to the place of preg-
nancy registration was asked for the last pregnancy and for both live as well as still births. The place
of pregnancy registration with the service provider was categorized into three categories: not regis-
tered, registered in a government facility, and registered in a private facility.

2.3 Analytical Strategy
First, we tried to examine whether STs or non-tribes had a higher risk of infant mortality. Nel-
son-Aalen analysis allows comparison of populations through their hazards curves. The relative ha-
zards curves for tribal and non-tribal populations were drawn in order to get a preliminary idea about
the groups in which infants are at higher risk in the Central and Eastern regions. When studying the
various groups in terms of cumulative hazards functions, the Nelson-Aalen estimator is preferred to
the Kaplan-Meier estimator (Cleves, Gould, Gutierrez et al., 2008). We also plotted the hazard esti-
mates for the two groups over the 12 month analysis period. The main purpose of this graph is to
show the pattern of variation in risk of infant death between tribes and non-tribes.
Second, a log rank test was applied to examine whether there was statistically significant differ-
ence between the survival curve/distributions of two groups. It is a nonparametric test and appropri-
ate to use when the data are right-skewed and censored. The log rank test statistic compared the es-
timates of the hazard functions of the tribal and non-tribal populations at each observed time point of
the event.
Finally, the Cox proportional hazards model (Cox, 1972) was applied to examine the effect of
various socio-economic and demographic covariates on the risk of infant death among both tribes

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