Projecting lyme disease risk in the United States: A machine learning approach integrating environmental, socioeconomic and vector factors.

OBJECTIVE

Lyme disease, caused by and transmitted by blacklegged ticks ( species), is the most common vector-borne disease in the United States. Its spatiotemporal dynamics are influenced by environmental and socioeconomic factors, yet the impacts of the COVID-19 pandemic on Lyme disease remain unclear.

METHODS

We analyzed county-level Lyme disease surveillance data (2001-2022) alongside environmental, socioeconomic, and tick vector data. Using machine learning models (Random Forest, Boosted Regression Trees, and XGBoost) and Shapley Additive Explanations (SHAP), we evaluated the influence of key predictors on Lyme disease risk. Predicted cases for 2020-2022 were compared with actual reports to assess the pandemic's effects.

RESULTS

Lyme disease cases rose from 16,862 in 2001 to 61,802 in 2022, with geographic expansion into southeastern regions. Population density, ecological niche of , and maximum temperature were presented as the key predictors of disease risk. The COVID-19 pandemic severely disrupted reporting dynamics, with 2020 and 2021 cases falling 43.9 % (95 % CI: 41.2-46.7 %) and 22.0 % (95 % CI: 19.5-24.5 %) below predictions, respectively-a decline most pronounced in the Northeast and linked to reduced healthcare access and outdoor activity during lockdowns.

CONCLUSION

Our findings highlight the complex interactions of environmental, socioeconomic, and behavioral factors in Lyme disease dynamics, including the significant impact of the COVID-19 pandemic on disease reporting. These insights underscore the need for integrated, data-driven public health strategies to mitigate Lyme disease risk in the United States.