Racial Composition and Single-Input Dominance in the First Hospice Special Focus Program Cohort

Background Federal quality oversight increasingly uses composite algorithms built from administrative counts and caregiver-reported experience, but are rarely tested against actual enforcement cohorts. In December 2024, the Hospice Special Focus Program (SFP) selected 50 hospices for intensified oversight and possible Medicare termination; the program was paused in 2025 amid litigation. Objectives To assess associations between SFP selection and hospice size, patient racial composition, and independently measured quality, and whether selection reflected broad or narrow underperformance. Methods Cross-sectional analysis of all 7115 active hospices in the November 2024 CMS Provider Data Catalog. Selection was modeled with logistic regression adjusting for size, ownership, patient mix, and Hospice Visits in the Last Days of Life (HVLDL), a publicly reported measure not used by the algorithm. Composite scores for the 50 selected hospices were decomposed into standardized inputs. Results Selected and non-selected hospices did not differ in size (median, 152 vs 134; P = .64), although after adjustment for patient mix and quality, each doubling of size increased selection odds (adjusted odds ratio [aOR], 1.33; 95% CI, 1.03-1.71). Selectees served more non-White patients (median, 35% vs 17%; P < .001), an association persisting after HVLDL adjustment (aOR per percentage point, 1.03; 95% CI, 1.01-1.05) and carried by Black share. Lower HVLDL independently predicted selection (aOR, 0.97; 95% CI, 0.95-0.98). For 48 of 50 selectees, a single input accounted for most of the score-worsening contribution; a two-domain requirement would have excluded 29 (58%). Conclusion SFP selection identified hospices with worse independent quality but usually reflected a single input and disproportionately affected Black-serving hospices.