The implications of using truncated Medicare definitions of avoidable hospital visits.

Abstract

11087 Background: As part of the Outpatient Quality Reporting program, Medicare reports on potentially avoidable, acute hospital visits after chemotherapy, using only fee-for-service (FFS) claims. This measure aggregates emergency department visits, observations, and inpatient admissions for 10 avoidable conditions (e.g. pain, vomiting), within 30 days of a chemotherapy infusion and is known as the OP-35 measure. The literature largely applies the diagnosis codes without reference to chemotherapy: it can be impractical to apply clinic-level changes only to Medicare FFS enrollees, and large, nationally-representative, hospital visit datasets do not contain chemotherapy infusion dates. There has been little scrutiny of the implications of using such truncations of this policy tool. Methods: We used a population-based cohort of incident cancers (2015-2023) from two sites: an academic medical center and safety-net health system, identifiably linking patients to their comprehensive hospital use from a regional health information exchange (all non-federal hospitals within a 150-mile radius of Dallas, TX). We tracked the changes in avoidable hospital visits for each measure specification: using diagnosis codes alone; narrowing to within 30 days of chemotherapy; counting only the first visit in a 30-day span; narrowing to Medicare FFS. We used mixed-effects (clustered to patient) multivariate logit to model avoidable visits occurring within or outside of a chemotherapy 30-day window; and avoidable vs. non-avoidable hospital visits across payors, adjusting for clinical and demographic variables. Results: We linked 31,305 incident cancer diagnoses (mean age 64; 50.6% female; 57.7% Black or Hispanic; 23.6% advanced stage cancer; 19.2% gastrointestinal cancer, 12.2% breast, 9.2% lung) to 190,967 visits across 76 hospitals. Although 28.0% (n= 54,233) had an avoidable diagnosis coded, only 24.3% (n=13,179) of them were within 30 days of chemotherapy. After excluding multiple visits in the 30-day span, 9.1% (4,917 of the 54,233 avoidable conditions visits) remained, with 14.0% (n=689, or 1.3% of 54,233) under Medicare FFS (1.6% Medicare Advantage, 30.2% commercial, and 54.4% uninsured/Medicaid). In adjusted analyses, avoidable conditions were significantly more likely to occur within 30 days of chemotherapy than outside of it (aOR 1.34, 95% CI: 1.27-1.40, p < 0.001); and compared to Medicare FFS, commercially-insured encounters had higher odds of avoidable hospital visit (aOR 1.31, 95% CI: 1.12–1.50, p < 0.001). Conclusions: Three-quarters of hospital visits for avoidable conditions occurred outside of a 30-day span after chemotherapy, though in adjusted analysis, avoidable visits were more likely to occur after chemotherapy. As little as 1.3% of visits for avoidable conditions were captured by the Medicare definition. Further investigation of the best uses of this measure is warranted.