Title : Estimating short term effects of extreme heat on blood pressure using high resolution environmental linkage and lagged multilevel models
Abstract:
Despite well documented cardiovascular effects of extreme heat, its short term impact on blood pressure in routine clinical settings remains poorly characterized. To address this knowledge gap, we examined the short term effects of extreme heat exposure on blood pressure (BP) in real world primary care settings. We first geospatially and temporally linked ambulatory electronic health record data from a national community health center network with a high-resolution metric of heat events (clinic-specific, summer months, 2019-2023). This represents the first large scale assessment of heat-BP relationships in real world primary care settings using geospatially granular heat metrics. Within these data, for each visit, we constructed an 8 day history of Extreme Heat Magnitude Indicator (EHMI), a measure of apparent temperature. We compared BP during visits on days with extreme heat (EHMI > 0 on the visit day) with visits on days without extreme heat (EHMI = 0). Associations were estimated using mixed effects models and distributed lag non linear models with spline based lag functions: under a conditional exchangeability assumption given observed patient demographics, clinical characteristics, calendar time, and clinic, we estimated associations between extreme heat exposure and systolic and diastolic BP using Gaussian mixed effects models with facility level random intercepts to account for clustering and unobserved clinic level heterogeneity. To characterize delayed and cumulative effects, we fit distributed lag non linear models with spline based lag functions across lags 0-7. Extreme heat exposure was associated with small but statistically precise reductions in BP, with the largest effects occurring on the day of exposure and the following day and attenuation thereafter. Cumulative associations across the lag window were larger in magnitude among older adults, while patterns were similar across hypertension and medication status. This work illustrates how linking high resolution environmental exposures with longitudinal clinical data, combined with multilevel modeling and flexible lag structures, can support causal interpretation of short term environmental health effects. These approaches generate policy relevant evidence for identifying heat vulnerable populations in community based care settings.
