This paper considers discriminant models for the prediction of diarrhea disease in children aged two or less in two governates in Egypt. Socio-economic and cultural factors are studied as the determinants of child health. Accordingly, a child's health is considered as a product of a whole set of interacting factors, some of which are related to demographic characteristics, environmental conditions, income level and others related to knowledge, attitude, and practices. These interacting factors might have a direct or indirect effect on caring for children having diarrhea disease. Stepwise discriminant analysis is used to choose the best set of predictors from the large number available. These variables were generally categorical and/or nominal originally and so were converted to several binary variables. Models based on the predicted probability that a child has diarrhoea, the linear discriminant score and simplified scores are presented. The performance of each model is measured by calculating their specificity and sensitivity and by carrying out a ROC analysis.