Estimation of hospital catchment populations using data on patient hospital use in France

BackgroundEstimates of disease burden from hospital data require well-informed estimates of the size of the catchment population. Data on patient flows from residential areas to a hospital can be used to estimate detailed catchment populations by age, year and type of hospital visit. MethodsCatchment populations were estimated for hospitals throughout France using a proportional flow approach. Data on hospital use and patient residence were accessed from the Agence Technique de lInformation sur lHospitalisation (ATIH). For patients coming from each administrative area, we calculated a preference for each hospital, and combined this with population data for the area to estimate the catchment population of each hospital. For one hospital group, we compared this with data on emergency visits, and data from a retrospective cohort study. ResultsEstimated catchment population by hospital group ranged from 4 million per year for Assistance Publique - Hopitaux de Paris (AP-HP) downwards, with the catchment population strongly reflecting geographic proximity and hospital scale. The type of hospital substantially impacted the size of the catchment area. In the analysis of a single hospital group, the size of the catchment population varied widely with the diagnostic categories associated with the hospital visit. Emergency visits represented a smaller catchment population. The estimated proportional contribution of different departments to the estimated catchment population was similar to their contribution to observed hospital admissions. Incidence rates for a respiratory virus using this catchment population estimation method were consistent with national incidence rates. ConclusionsThis study demonstrates the consistency of the proportional flow framework when applied to appropriate data on hospital usage. The study provides catchment populations for each hospital in France which can be used for burden estimates such as incidence rates, as well as providing insight into the catchment populations served. Analysis at the department geographic level provided an appropriate balance between detail of analysis and the need to mask data for anonymisation. Further analysis should explore how the size of the catchment area corresponds to the associated travel time to the hospital in question.