Early infant morbidity and infant mortality in the City of São Paulo, Brazil: A probabilistic record linkage approach

Despite declines in under five mortality rates in the developing world in recent decades, infant mortality has declined less and neonatal mortality rates have not changed much. Determinants of infant health status are difficult to study at individual level because suitable data on birth weight, for example, a widely used indicator of infant morbidity, are difficult to obtain in developing countries. The aim of this dissertation is to provide a comprehensive picture of early infant morbidity and infant mortality in a developing country. In order to accomplish that, we studied 209,628 births and 3,842 infant deaths from the 1998-birth cohort of the City of Sao Paulo, Brazil. The indicators of early infant morbidity chosen were low Apgar scores at one minute and low Apgar scores at five minutes (seven or less). Neonatal mortality and post-neonatal mortality were also used. We examine more proximate, less proximate and distal determinants of such indicators, following a slight modified version of the Mosley-Chen framework for child survival in the developing world. Our data came from two separate files from vital statistics data for the City of Sao Paulo, the birth and death files, to take advantage of variables only available in the birth record such as birth weight. To link births and deaths we used a probabilistic record linkage procedure, since the two files did not contain unique identifiers. To analyze the matched data, we used weighted logistic regression, weighting observations according to how certain we were that each matched pair corresponded to the same infant. To account for similarities between mothers in a same community we used robust standard errors. Main results from multivariate analysis are as follows: Infants of older mothers were more likely to have low Apgar scores at one minute, but were less likely to die in the neonatal period. Infants of adolescent mothers were more likely to die in the neonatal and post-neonatal period. Second, low Apgar scores at one and five minutes were highly associated with low levels of maternal schooling. For mortality, after controlling for proximate determinants, only college-degree remained as a significant predictor of infant survival. Third, parity one or higher decreased the likelihood of low Apgar scores; parity four or higher significantly increased the likelihood of post-neonatal mortality. Fourth, lower numbers of prenatal care visits remained significant predictors of both low Apgar scores and mortality. Fifth, cesarean delivery was associated with increased likelihood of low scores, but not the likelihood of mortality. Finally, even controlling for birth weight and gestational age, low Apgar scores at one and five minutes increased the likelihood of both neonatal and post-neonatal mortality.