Predicting Infectious Disease Incidence After Flooding Using Artificial Intelligence Models: A Retrospective Pre-Post...

Area of Science:

• Epidemiology
• Public Health
• Machine Learning

Background:

• Floods significantly elevate infectious disease risks due to environmental contamination and healthcare system disruptions.
• Existing predictive models for post-disaster disease surveillance are limited.
• Developing accurate predictive models is crucial for effective public health emergency response.

Purpose of the Study:

• To develop and validate machine learning algorithms for predicting infectious disease incidence following flood events.
• To assess the performance of various machine learning classifiers in a post-flood scenario.
• To identify key predictive features for post-flood infectious disease occurrence.

Main Methods:

• A retrospective pre-post cohort study was conducted using electronic health records.
• Data from 30-day pre-flood and 30-day post-flood periods were compared.
• Five machine learning classifiers (Random Forest, Logistic Regression, SVM, Gradient Boosting, ANN) were trained and evaluated using AUC.

Main Results:

• Post-flood infectious disease prevalence increased significantly (39.5% to 47.3%).
• Random Forest achieved the highest predictive performance (AUC=0.76), followed by Gradient Boosting (0.74).
• Age and visit date were the most important predictive features; younger patients had higher post-flood rates.

Conclusions:

• Machine learning models demonstrate moderate predictive capability for post-flood infectious diseases.
• AI-based disaster epidemiology is feasible but requires further development.
• Incorporating environmental and socioeconomic data is essential for actionable prediction systems.