Related Experiment Video
Updated: Aug 5, 2026
06:34
A Component-resolved Diagnostic Approach for a Study on Grass Pollen Allergens in Chinese Southerners with Allergic Rhinitis and/or Asthma
Published on: June 4, 2017
Weather-Enhanced Machine Learning for Time-Resolved Risk Stratification of Clinically Managed Hymenoptera-Related
Mohamad Amer Nashtar1, Theodor Baars1, Nicoleta-Alexandra Stille2
1Ruhr University Bochum, Knappschaft Kliniken University Hospital Bochum, Department of Medicine, 44892 Bochum, Germany.
Summary
Predicting insect sting risk using weather data is feasible. Machine learning models show potential for identifying high-risk periods, but require further validation before clinical use.
Area of Science:
- Environmental Health
- Allergy and Immunology
- Data Science and Artificial Intelligence
Background:
- Insect stings from Hymenoptera (wasps, bees) frequently cause severe allergic reactions.
- Predicting short-term sting risk based on environmental conditions is an unmet need.
- Meteorological factors influence insect activity and human exposure, creating a gap in preventive strategies.
Purpose of the Study:
- To explore time-resolved, weather-informed patterns of Hymenoptera sting events.
- To assess the feasibility of using artificial intelligence (AI) for risk stratification.
- To identify periods of increased sting occurrence based on meteorological data.
Main Methods:
- Retrospective analysis of 17 years of clinical sting data (2005-2022) in Bochum, Germany.
- AI-based machine learning (multilayer perceptron) trained on daily weather parameters.
- Expert-informed indicators capturing current and carry-over environmental conditions.
Main Results:
- 346 clinically significant sting events were analyzed.
- Weather variables were associated with intra-seasonal clustering of stings, with earlier onset correlating to higher counts (R = -0.52, p = 0.037).
- AI model achieved moderate recall (0.431) and high specificity (0.86), but did not consistently outperform simpler methods.
Conclusions:
- Combines clinical sting data with meteorological data for weather-informed risk pattern exploration.
- Demonstrates technical feasibility of exposure-based risk stratification for rare events.
- Proof-of-concept requires external validation in multi-center cohorts before clinical implementation.
