Future impacts of colectomy healthcare pathways on quality of care in bundled payment experiments, a national retrospective cohort in France

Objectives: To perform an ex-ante evaluation of French bundled payment experiments to evaluate the potential effects on hospital readmission and length of stay (LOS), and whether it could be used as a lever for improving quality of care after initial stays for surgery. Design: A retrospective cohort analysis was performed using data from the French National Health Data System (SNDS). Setting: We used hospital and ambulatory data for colectomy between 2014 and 2016 (exhaustive French national data). The national database included 42,603 cancer colectomy stays during the study period. Participants: The inclusion criteria were stays coded in the database with a principal diagnosis of colon cancer and a colectomy procedure excluding total colectomy, identified from the diagnosis related group. All partial colectomies performed in France from the 1st January 2014 to the 30th June 2016 were included, except those that met exclusion criteria, which were: admission to emergency departments, having already undergone a total colectomy (no time limit), or having undergone a partial colectomy in the previous year. Patients who died during the bundled payment period (45 days before the index stay and up to 90 days after) were excluded from the analysis. These criteria were selected and validated by the stakeholders who developed the bundled payment specifications. Primary and secondary outcome measures: The main outcome was the variation in readmissions as a function of the initial LOS using a segmented regression method, and controlling with variables used by the health authorities. We also produced models by sector (public/private) and practice (Enhanced Recovery After Surgery [ERAS] model or not). Results: We obtained a J-shaped curve including two distinct parts, with a breakpoint at five days for patients without major comorbidities. Before the breakpoint, increased LOS was associated with a lower probability of readmission (coefficient = −0.016, 95%CI [−0.011;-0.021], p