Back to browse results
Prediction of child survival in India using developed Cox PH model: a utility for health policy programmers
Authors: Rajvir Singh, Shahina Begum, R.K. Ahuja, Prem Chandra, S. N. Dwivedi1
Source: Statistics in Transition, Vol. 8, No. 1, pp. 97—110
Topic(s): Child health
Childhood mortality
Country: Asia
Published: APR 2007
Abstract: Prediction of any health indicator from a developed model can be achieved accurately only when it comes from a properly validated model. The aim of this study was to develop a model using Cox Proportional Hazard method on child survival data and the prediction of survival probabilities of a child at different points of time. A data set of 2118 children of first birth order was drawn from the National Family Health Survey (NFHS, 1992-93), Uttar Pradesh (UP), India. In multivariate analysis variables such as breastfeeding, immunization, premature birth, antenatal care, type of house and father’s education were found to be significantly associated with child survival. The validation index shrinkage coefficient was 97 percent, indicating only 3 percent lack of fit in the model, and Somer’s D rank correlation (Dxy) was –0.65, indicating good correlation between the log hazard and the observed survival time. Best improvement in child survival was found by 7% in first month, 9% in three months and 11% in twelve months when the variables such as breastfeeding, immunization and the mothers received antenatal care during pregnancy used altogether in developed model for prediction. The above findings may be useful for those who are involved in health policy programs for improving child survival and for better public health management.