ABSTRACT
BACKGROUND: Aedes aegypti rapid larval surveys are mandatory in Brazil. Here, we retrospectively examined whether the house index estimated by larval surveys served as a useful tool in anticipating epidemics within Brazilian municipalities from 2009-2015. METHODS: We used correlation indices and classification analysis stratified by year, region, population size and time after the national larval survey. RESULTS: We found no association between the house index and the proportion of municipalities experiencing an epidemic. The sensitivity of a high score house index in predicting an epidemic was 7.20% (95% CI 6.22 to 8.33%) for all years combined. The positive predictive value of a high score house index to predict a 'true epidemic' was 38.96%, lower than the negative predictive values of a low score house index for predicting 'no epidemic' (56.96%). The highest overall sensitivity was observed in the North region (20.15%; 95% CI 17.14 to 23.53%). The sensitivity of a high score house index demonstrated a monotonic decrease with increasing time from larval collection. CONCLUSIONS: Larval surveys are surveillance tools with the potential to risk-stratify and guide dengue control programs towards judicious resource allocation. However, the national rapid larval survey performed in Brazil, in its present form, consistently underpredicts dengue epidemics.