Predicting Lyme Disease: A One Health Approach

Area of Science:

• Epidemiology
• Public Health
• Veterinary Medicine

Background:

• Lyme disease is a prevalent vector-borne illness in North America, necessitating accurate incidence prediction for public health preparedness.
• Previous research established Google Trends search data as a viable tool for predicting monthly human Lyme disease incidence.
• This study aimed to enhance predictive models by incorporating environmental and canine data alongside human data.

Purpose of the Study:

• To evaluate the effectiveness of a One Health approach in predicting state-level human and canine Lyme disease incidence.
• To compare the predictive performance of models integrating human, canine, and environmental data against models using single data types.
• To identify optimal data streams for improving Lyme disease incidence forecasting.

Main Methods:

• Acquired human Lyme disease data from state health departments and canine data from IDEXX Laboratories.
• Developed predictive models incorporating environmental factors, human search trends, canine search trends, and case counts.
• Compared the predictive accuracy (using Mean Absolute Error) of various models, including a comprehensive One Health model.

Main Results:

• The One Health model demonstrated superior performance in predicting human Lyme disease incidence in 6 out of 16 states (MAE = 12.1) compared to other models.
• For canine Lyme disease incidence, the One Health model (MAE = 330.5) performed worse than models utilizing environmental data (MAE = 221.5) or human data (MAE = 248.4).
• Despite improvements, the best-performing models exhibited significant prediction errors, limiting their practical application.

Conclusions:

• Integrating human, animal, and environmental data within a One Health framework shows promise for improving human Lyme disease prediction.
• The predictive utility of different data streams varies between human and canine populations.
• Future research should explore alternative data sources like electronic health records and insurance claims to enhance predictive accuracy for Lyme disease.