Presentation (CTI)

OBJECTIVES: Controlling healthcare expenditures is critical for public health systems as costs continue rising. Understanding territorial determinants is essential for optimizing resource allocation and reducing health inequalities. Open data remains underexploited in health economics research. This study aims to identify environmental, socioeconomic, demographic, and clinical determinants, known as Exposome, of total healthcare expenditures across French metropolitan departments using open-source databases.
METHODS: We analyzed 96 French departments using national databases (2019-2024): Open-Medic, ScanSanté, Cépi-DC, sick leave data, and Pathologies-prevalences. Total expenditures per 10,000 inhabitants were computed as: pharmaceutical costs + Medecine-Surgery-Obstetric hospitalizations costs + At-Home Hospitalization stays costs + general practitioner consultations costs + sick leave indemnities. We examined 31 territorial exposome indicators derived from open-source data, including altitude, healthcare accessibility, hospital density, population remoteness, fast-food density, demographic structure (age groups, household size), socioeconomic factors (poverty rate, median income, social housing rate), environmental factors (organic farms, polluting industries, radioactive waste), and tourism infrastructure as explanatory variables. Several factors were derived merging databases and aggregating at department level. The dependent variable was normalized by log1p transformation to better handle near-zero values. After outlier removal, multicollinearity was addressed through iterative VIF-based elimination, followed by bidirectional AIC-based stepwise selection.
RESULTS: The final regression model demonstrated strong explanatory power. Cancer prevalence emerged as the strongest predictor (p<0.001), followed by organic farming density (p=0.008). Negative predictors included healthcare access (p=0.018), hospital density (p=0.006), median living standard (p=0.029), and polluting industries (p=0.016). Marginally significant predictors included population remoteness (p=0.072), 0-19 years percentage (p=0.063), household size (p=0.090), and sick leave personnel (p=0.062). All regression assumptions were satisfied.
CONCLUSIONS: Cancer prevalence drives territorial healthcare expenditures in France. Negative associations between living standards and expenditures, alongside positive organic farming effects, reveal complex patterns combining disease burden, socioeconomic factors, and rural-urban disparities. Future SNDS studies could enhance granularity and precision.