Geospatial Risk Mapping and Predictive Modeling of Incident Heart Failure in an Urbanized Regional Cohort: An Epidemiological Study

Abstract

Background: The global rise in incident Heart Failure (HF) is inextricably linked to accelerating urbanization, demanding sophisticated primary prevention tools that integrate clinical risk with environmental determinants. Traditional risk stratification models often fail to account for the local impact of factors like air quality, temperature variation, and noise pollution, which are increasingly recognized as non-traditional cardiovascular stressors. This investigation sought to develop a comprehensive predictive model using geographically indexed data to identify asymptomatic individuals at the highest risk of new-onset HF within the next three years.

Methods: This large-scale retrospective cohort study analyzed electronic health records (EHR) from 15,000 asymptomatic, non-HF adults over a 5-year period. Patient data were geocoded and linked with high-resolution, time-variant regional environmental exposure metrics (e.g., particulate matter, thermal fluctuation). A Random Forest survival model was trained on the combined clinical, demographic, and geocoded environmental features to predict the primary endpoint: incident, de novo diagnosis of HF (HFrEF or HFpEF). Model performance was evaluated using the concordance statistic (C-statistic) and Relative Risk Quotient (RRQ).  

Results: The integrated Geospatial-Clinical Risk Score (GCRS) model demonstrated excellent discrimination for predicting incident HF (C-statistic: 0.86; 95% CI: 0.83–0.89), significantly improving upon clinical factors alone (C-statistic: 0.73; p<0.001). Feature importance analysis confirmed that average daily exposure to ambient fine particulate matter (PM 2.5) was the third strongest predictor of incident HF, surpassed only by age and baseline hypertension. Individuals classified in the highest risk quartile by the GCRS exhibited an adjusted Relative Risk Quotient of 3.1 for developing HF over the study period (p<0.001). The study identified precise urban micro-environments responsible for a disproportionate burden of incident cases.

Conclusion: This methodology facilitates the strategic allocation of public health resources and supports the critical regional objective of advancing advanced risk and disease detection to enhance the integrated care delivery model.