Predictive modeling and risk factors of under-five child mortality in India using NFHS-5 dataset: a parity-based analysis

Background: Child mortality is one of the important public health issues in developing countries like India. Many previous studies found that there are several risk factors affecting child mortality namely; Socio-economic factors, Bio-demographic factors, etc. Therefore, in the present study, various variables associated with these factors were explored in relation to under-five child mortality. Additionally, three different models were compared to assess their suitability for estimating child mortality across different parity levels of women. This approach was adopted with the objective of framing more effective strategies to reduce child mortality rates in India. Methods: For our purpose, binary logistic regression model was performed to describe the significant risk factors of mortality among under-five children. For finding a best predictive model, three different models are formulated and model fitting was observed by the method of MSE. First model was binomial distribution, second was beta-binomial distribution and third was poisson distribution model. Results: The logistic regression result reveals that the factors like maternal health, education, and socio-economic conditions, rural areas, significantly influence child mortality and the method of comparison on the three different models describes that the beta-binomial distribution model shows the better fit on the data. The estimated values of probability of child deaths at higher parities; parity 3, parity 4, and parity 5, were obtained as 0.153, 0.279, 0.123, respectively. Conclusions: According to this study we found that, child mortality is still a significant issue in India. Therefore, progress on socio economic, bio-demographic and environmental risk factors should be the focus of policymaker’s intervention.