Falls are a leading cause of injury and death among older adults. There are few studies that use national longitudinal data and multiple weather metrics to explore the complex relationships of environmental risk factors. This thesis compares machine learning (random forests) and classic statistical (mixed-effects Poisson regressions) methods to examine the influence and spatial variation of weather, demographic, and socioeconomic characteristics on county fall-related hospitalization rates among adults 65 years and older in the United States. Three county-level data sources were combined: 1) monthly Medicare hospitalization claims from July 2009 – July 2015 with either primary diagnosis or external cause codes indicating an accidental fall (ICD-9-CM codes E880-E888, excluding E887) 2) weather variables scraped from NOAA’s National Weather Service, and 3) sociodemographic characteristics from the 2010-2014 American Community Survey. National and regional modelling was conducted using both methods and each set of characteristics. Weather, demographic, and socioeconomic variables are all important predictors of fall-related hospitalization rates in older adults. Temperatures were the most influential weather variables at the national level and for all regions except the Northeast. The random forests identified that high fall-related hospitalization rates were associated with temperature variability. The mixed-effects Poisson models demonstrated a U-shaped relationship where both high maximum and low minimum temperatures are associated with greater risk at the national level and in the South and Midwest, controlling for relevant sociodemographics. The two modelling approaches are compared, and the mixed-effects Poisson regressions are consistently more accurate likely due to their adjustment for spatial variation and confounding.