Climate and other environmental factors predict tick abundance and Lyme cases in Minnesota one and two years in...

Environmental factors, like weather and host abundance influence tick populations which in turn affect tickborne disease in endemic regions. It is important to understand how these factors are associated with tick abundance, and whether they can predict disease. We assessed associations between environmental factors, tick abundance, and human Lyme disease cases in the same year and with one- and two-year lags in Minnesota. In parallel, we compare conventional analytical methods with a novel machine learning approach to evaluate their relative strengths and potential for integration.

Environmental and tick abundance relationships were examined using generalized linear mixed-effects models (GLMM), and gradient boosting machine learning incorporating same-year and time lagged effects.

Area under the curve (AUC) indicates higher accuracy in predicting tick abundance in GLMM than gradient boosting. Among GLMM with no time lag, Palmer Drought Severity Index (PDSI), vapor pressure deficit (VPD), precipitation and snow water equivalent (SWE) were significant predictors of tick abundance. Direction of association varies in PDSI and SWE variables with no time lag but show consistency with one- and two-year lags. Among gradient boosted models, days below -18 °C, small mammal count, and mouse-to-small mammal ratio were associated with high tick abundance in the same year, and with one- and two-year lags. AUC are highest with a one-year lag suggesting environmental factors most accurately predict tick abundance one year later. AUC in models with Lyme disease as the outcome are highest with a two-year lag; PDSI, soil moisture, SWE, days below -18 °C, degree days, small mammal count and mouse ratio are all variables of importance in gradient boosted models.

Monitoring environmental factors provides enhanced opportunities for public health interventions through prediction of tick abundance and potential consequences for higher Lyme disease incidence. Incorporating traditional and modern analytic methods offers opportunities for enhanced prediction and early warning.