Abstract
BACKGROUND: Simple methods have been developed to warn of pre-epidemics and epidemics in small areas using data of infectious diseases surveillance. Epidemic warnings are made if the index of cases per week per sentinel medical institution is greater than a defined value. A pre-epidemic warning means that an epidemic warning will be given in the following four weeks. While the methods are used routinely for surveillance in Japan, they remain to be validated. METHODS: Infectious diseases surveillance data of influenza-like illness and 12 pediatric diseases in the fiscal year between 1999 and 2001 were used in the analysis. We examined the frequency of warnings, temporal changes in the index before and after the onset of a warning, and the sensitivity, specificity, and positive predictive value of pre-epidemic warnings. RESULTS: For the majority of the diseases investigated, the proportion of weeks in which a warning was issued ranged between 0% and 10%. In several diseases including influenza-like illness, we observed a rapid increase and gradual decrease in the index before and after a warning. The sensitivity, specificity, and positive predictive value of a pre-epidemic warning were 90.4%, 93.7% and 23.9% for influenza-like illness, and ranged between 25.1-54.2%, 86.1-99.2%, and 2.5-20.8% for the pediatric diseases (chickenpox, rubella, measles, and mumps), respectively. CONCLUSIONS: The study showed that the methods used for determining whether or not to issue an epidemic warning were satisfactory in some diseases, including influenza-like illness, and may need to be improved in several other diseases.