Linking Heat Conditions to Mental Health Hospitalizations: A Data-Driven Analysis for Tyrol, Austria

Heat conditions pose a substantial threat to population health, with certain groups being particularly vulnerable. People with mental health disorders may be especially at risk due to structural and social stressors (e.g., living environment, limited access to cooling), physiological factors (e.g., medication effects on thermoregulation), and psychological factors (e.g., reduced self-care). This study is part of the Austrian climate–health project Parahsohl, in which we develop a data-driven workflow to assess health-relevant heat indicators, focusing on individuals with mental health disorders in the federal state of Tyrol. The objective is to develop regression models linking hospitalizations to weather conditions while accounting for relevant confounders, enabling interpretation in an impact-based weather-warning context and providing a basis for subsequent climate risk assessments.

The analytical workflow comprises three stages: (i) data preparation, (ii) model development, and (iii) model evaluation. Daily hospital admissions (n = 83,673) between May and September from 2007–2023 were used as the response variable for the nine administrative districts of Tyrol, focusing on mental and behavioural disorders (ICD-10 diagnosis codes: F00–F99). Weather predictors were derived from high-resolution (1×1 km) gridded observation data (SPARTACUS) for the same time period and aggregated using a population-weighted approach. Thus, we could account for differences in exposure between densely and sparsely populated areas. Lag variables over multiple temporal windows were generated for key meteorological metrics to capture delayed health effects. Hospitalization counts were modelled using generalized additive mixed models (GAMMs) with a negative binomial distribution. Weather variables were included as fixed effects, while day-of-week, year, and district were treated as random effects. Population offsets allowed incidence-based interpretation. Model performance was evaluated using standard statistical criteria for fit and predictive accuracy, and predictive skill was further assessed through temporal cross-validation across years and months. Initial results indicate that higher daily mean temperatures are significantly associated with increased hospitalization counts, with lagged temperature effects further enhancing model performance. Partial-effect plots and relative risk estimates provide interpretable quantitative measures of heat-related impacts on mental health outcomes. For instance, the hottest day in the study period was associated with an estimated increase in hospitalization risk exceeding 10% compared with average summer conditions.

As a next step, the analysis will be extended to a more detailed examination of diagnostic subgroups to better identify particularly vulnerable populations. These results will be presented at EGU2026. This study provides a quantitative assessment of heat-related impacts on mental health hospitalizations and contributes to the development of evidence-based indicators applicable for short-term applications (e.g., user specific impact-based weather warnings) as well as long-term climate risk assessments.