Infant mortality and death clustering at the district level in India: A Bayesian approach

This study assessed the clustering of and spatial variations in infant mortality between districts in selected states of India using a Bayesian geoadditive model. The study utilized 10 years of retrospective birth history of women from the fourth round of NFHS-4 (2015-16). Findings suggest, except Kerala, there was a significant amount of clustering of infant deaths in families in the selected Indian states. The maximum impact of clustering was observed in Assam, followed by Madhya Pradesh, Bihar, Uttarakhand, and Uttar Pradesh. The estimated residual spatial effect was statistically significant in all the states, with the maximum effect being in Assam and Chhattisgarh. The risk of infant death in Assam was higher in the north-eastern districts and lower in the southern districts of the state. Mother's age at child birth had a nonlinear effect on infant death in all the states, although significant effects were observed only in Bihar and Assam. In both of these states, mother's age at child birth had a "U-shape," showing that the risk of infant death was higher at both earlier and later ages of mother's reproductive period. With the exception of Kerala, all the other selected states in the study had an "elongated L shaped" pattern, showing that in the early ages of the reproductive period, the risk of infant death was very high and that it gradually decreased with age and remained constant thereafter. In Kerala, mother's age at child birth was a straight line, implying that the risk of infant death was constant across the reproductive age of women. In order to keep infant mortality at a low level and to achieve better maternal and child health outcomes, the government needs to target families experiencing multiple infant deaths. In addition, programs must take into consideration the prevailing state-specific spatial heterogeneity in infant deaths and factors like mother's age at child birth.