|
Abstract:
|
Active Bacterial Core Surveillance (ABCs) conducts active, laboratory- and population-based surveillance at 10 U.S. sites, covering about 12% of U.S. population. A common practice is using ABCs to predict infections in the United States (US), assuming that incidence rate in each demographic strata (i.e., age, gender, race) is the same for the entire US. This approach does not account for differences in incidence rates by health conditions such as diabetes and obesity (two important risk factors for Invasive group A Streptococcus [IGAS] infection). The objective of this study is to develop a new method to predict IGAS cases in non-ABCs areas accounting for not only demographics but also health conditions. We constructed a multilevel logistic regression model for IGAS, using individual-level demographics, health conditions, and county-level random effects. We then predicted IGAS cases for non-ABCs areas using parameters estimated from ABCs. The estimated IGAS cases in the US were compared with that obtained by the traditional synthetic method. The new method can be applied to predict other infectious diseases in non-ABCs or other non-surveillance areas.
|
