OBJECTIVE: To analyze the temporal pattern of hospitalization and deaths due to diarrhea among children less than five years old between 1995 and 1998 to support specific prevention actions and disease control strategies. METHODS: Data was collected from the Ministry of Health's Mortality Data System and Hospitalization Data System. Monthly time series of diarrhea hospitalizations and deaths were decomposed into stochastic linear trend, deterministic seasonal and irregular components using structural time series models. RESULTS: The levels of both series showed a decline in time. This pattern being more evident in the hospitalization series. The slope variation was constant in both series; on average less than 5.3 hospitalizations/month (p-value <0.001), and less than 1 death/month (p-value <0.1). The residual analysis of the hospitalization series revealed a positive trend change in January 1996. The seasonal component for both models was statistically significant (p-value <0.0001) with May and June as months of highest hospitalizations and deaths. Normality and time independence assumptions of errors could not be rejected at a 0.05 significance level. CONCLUSIONS: The temporal pattern of moderate and severe diarrhea described and estimated in this study suggests that rotavirus might be a predominant disease agent. In this context, targeted vaccination is recommended for prevention and control. However, further studies on the efficacy of rotavirus candidate vaccines are necessary in order to implement this strategy in Brazil.
Diarrhea; Mortality; Inpatients; Information systems; Time series; Sasonal variations; Models, structural
