Machine learning identifies straightforward early warning rules for human Puumala hantavirus outbreaks

Area of Science:

• Environmental epidemiology
• Infectious disease modeling
• Mammalian ecology

Background:

• Human Puumala virus (PUUV) infections exhibit multi-annual fluctuations in Germany, correlating with bank vole population dynamics.
• Predicting localized human infection risk is crucial for public health interventions.

Purpose of the Study:

• To develop a robust model for predicting human Puumala virus infection risk at the district level.
• To introduce and apply a PUUV Outbreak Index for quantifying spatial synchrony of outbreaks.

Main Methods:

• A machine-learning classification model was developed using transformed annual incidence values.
• Input features included soil temperature (April, two years prior; September, previous year) and sunshine duration (September, two years prior).
• The PUUV Outbreak Index was introduced to measure the spatial synchrony of local PUUV outbreaks.

Main Results:

• The classification model achieved 85% sensitivity and 71% precision in predicting human infection risk.
• The model effectively utilized only three specific weather parameters from previous years.
• The PUUV Outbreak Index was estimated with a maximum uncertainty of 20% for the period 2006-2021.

Conclusions:

• A straightforward, robust model for predicting human Puumala virus infection risk was established using limited weather data.
• The developed model and PUUV Outbreak Index offer valuable tools for epidemiological surveillance and outbreak management.