Underlying and proximate determinants of child health: the Cebu Longitudinal Health and Nutrition Study.

Abstract

Between 1983-1985 researchers analyzed socioeconomic data and diarrhea incidence during the 1st year of life from 17 urban areas in metropolitan Cebu the Philippines using a new methodology. This model integrated socioeconomic behavioral and biomedical models to illustrate how policy makers may use it to assess determinants of infant morbidity mortality and growth. They demonstrated that standard statistical procedures (e.g. ordinary least squares or logistic regression) do not correctly estimate the effect of risk factors on health when some variables are endogenous. Further these techniques do not consider the fact that mothers do indeed identify health risks and act to reduce the effects of these risks. In the model which ignores endogeneity (standard analysis) only the variable total calories was statistically significant (alpha=.01) with diarrheal incidence but it was not significant in the model accounting for endogeneity (instrumental variable analysis). Further the standard analysis did not identify certain behaviors (improved excreta disposal preventive health care and exclusive breast feeding) as having an effect on diarrhea incidence yet the instrumental variable analysis confirmed that they did indeed have a significant protective effect (alpha=.01- .05). In addition when the standard analysis indicated the same statistically significant estimates as the instrumental variable analysis (e.g. water quality) the values were substantially biased toward the null. The new analysis also showed that maternal education reduced diarrhea incidence because of improved excreta disposal practices (about 4%) and increased caloric intake (1.23%). It also caused an offsetting increase in diarrhea because of a reduction in the number of mothers who breast fed (exclusive breast feeding 2.08%; any breast feeding 1.66%).