Abstract
We explored the threshold effects of meteorological factors on hand, foot and mouth disease (HFMD) in mainland China to improve the prevention and early warning. Using HFMD surveillance and meteorological data in 2011, we identified the threshold effects of predictors on the monthly incidence of HFMD and predicted the high risk months, with classification and regression tree models (CART). The results of the classification tree showed that there was an 82.35% chance for a high risk of HFMD when the temperature was greater than 24.03 °C and the relative humidity was less than 60.9% during non-autumn seasons. According to the heatmap of high risk prediction, the HFMD incidence in most provinces was beyond the normal level during May to August. The results of regression tree showed that when the temperature was greater than 24.85 °C and the relative humidity was between 80.59% and 82.55%, the relative risk (RR) of HFMD was 3.49 relative to monthly average incidence. This study provided quantitative evidence for the threshold effects of meteorological factors on HFMD in China. The conditions of a temperature greater than 24.85 °C and a relative humidity between 80.59% and 82.55% would lead to a higher risk of HFMD.
Highlights
We explored the threshold effects of meteorological factors on hand, foot and mouth disease (HFMD) in mainland China to improve the prevention and early warning
The results indicated that when the temperature was greater than 24.85 °C and the relative humidity was between 80.59% and 82.55%, the relative risk (RR) of HFMD was 3.49 relative to monthly average incidence during the epidemic period
We found that conditions of a temperature greater than 24.03 °C and a relative humidity under 60.9% (CART 1) or of a temperature greater than 24.85 °C and a relative humidity between 80.59% and 82.55% (CART 2) would lead to a higher risk
Summary
The analysis indicates that there was an 82.35% chance for a high-risk of HFMD when the temperature was greater than 24.03 °C and the relative humidity was less than 60.9% during non-autumn seasons. These results of validation analysis including misclassification analysis and LOOCV reveal that both the two CART models had reasonable accuracy, and its utility in research needs to be further explored
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have