RESUMEN
BACKGROUND: French Guiana faces singular health challenges: poverty, isolation, structural lag, difficulties in attracting health professionals. Hospital stays exceed the recommended durations. The present study aimed to model the impact of precariousness and geographic isolation on the hospital duration performance indicator and to recalculate the indicator after incrementing severity by 1 unit when patients were socially precarious. METHODS: Cayenne hospital data for 2017 were used to model the hospital duration performance indicator (IP-DMS) using quantile regression to study the impact of geographic and social explanatory variables. This indicator was computed hypothesizing a 1 unit increment of severity for precarious patients and by excluding patients from isolated regions. RESULTS: Most excess hospitalization days were linked to precariousness: the sojourns of precarious patients represented 47% of activity but generated 71% of excess days in hospital. Quantile regression models showed that after adjustment for potential confounders, patients from western French Guiana and Eastern French Guiana, precarious patients and the interactions terms between residence location and precariousness were significantly associated with IP-DMS increases. Recalculating the IP-DMSafter exclusion of patients from the interior and after increasing severity by 1 notch if the patient was precarious led to IP-DMS levels close to 1. CONCLUSION: The results show the nonlinear relationship between the IP-DMS and geographical isolation, poverty, and their interaction. These contextual variables must be taken into account when choosing the target IP-DMS value for French Guiana, which conditions funding and number of hospital beds allowed in a context of rapid demographic growth.