Logistic regression for social-economic and cultural factors affecting diarrhea diseases in children under two years in Egypt.

Abstract

The fieldwork for this research was carried out in Dakahlia governorate in Lower Egypt and Sohag governorate in Upper Egypt. In Dakahlia governorate, 11 clusters (3 urban and 8 rural) and in Sohag governorate, 6 clusters (2 urban and 4 rural) were chosen at random, each with 16 households yielding a sample of 1020 households or mothers to predict diarrhea. Population census data were also obtained from the Central Agency for Public Mobilization and Statistics (CAPMAS) for the two governorates for 1984. Using a multiple regression model a step-up (forward) selection procedure from the set of independent variables involving all 63 variables in the data set was carried out. A linear logistic regression model (LLR) for Dakahlia governorate was based on 660 observations and 6 out of 63 variables. The probability of no diarrhea was low in children whose family disposed of refuse near the house or in surface water, while the probability of no diarrhea increased with owning land, mothers' knowledge of symptoms and causes of diarrhea (watery stools, thirst, and dentition), and mothers' previous use of oral rehydration for treatment. An LLR model for Sohag governorate was based on 360 observations using 10 out of 63 variables. Owning land, mother's knowledge about symptoms of diarrhea (watery stools), increasing numbers of stools, causes of choosing place of treatment, type of residence (urban or rural), eating during diarrhea, and decision about who takes treatment were useful in predicting diarrhea. In Sohag governorate, the logistic regression model with 10 variables achieved a 66.67% agreement between predicting diarrhea and the observed incidence with high sensitivity and moderate specificity compared with the other models.