Background: Several longitudinal studies have investigated factors associated with childhood diarrhea in developing countries. However, most studies have neglected important dynamic features of the longitudinal design and hierarchical interrelationships among the potential risk factors.
Methods: We conducted a longitudinal study of 902 children, age 0 to 36 months at baseline, in a large urban center in northeastern Brazil. Diarrhea data were collected by following children from October 2000 until January 2002 with biweekly home visits. We used a dynamic time-to-event analysis to account for several longitudinal features. We applied an effect-decomposition strategy to quantify direct and indirect effects of risk factors grouped in different blocks.
Results: Child's age and an autoregressive effect of past diarrhea episodes explained some of the decline of diarrhea incidence observed throughout the study (from more than 14 episodes to 2 episodes per child-year), a phenomenon already observed but not explained in other longitudinal diarrhea studies. We identified the following major diarrhea determinants: low socioeconomic status, poor sanitation conditions, presence of intestinal parasites, and absence of prenatal examination. The effect of socioeconomic status was mediated mostly by living and sanitation conditions.
Conclusion: Our study shows important advantages of applying a dynamic analysis approach to longitudinal observational studies of diarrhea or other acute diseases and highlights the complex interrelationships of diarrhea determinants. Our results confirm the importance of sanitation as a major determinant of child health in urban settings of developing countries.