Abstract
<jats:p>Background: Palliative care is recommended by clinical practice guidelines for patients with advanced heart failure (aHF), yet specialty palliative care (SPC) remains substantially underutilized in this population. We sought to quantify between-facility variation in SPC receipt among people with aHF and determine how much variation is explained by patient case mix and facility structural characteristics versus residual unmeasured factors. Methods: This retrospective cohort study included 23,991 Veterans with prevalent aHF identified through administrative data across 133 VA Medical Centers (VAMCs) with ≥20 aHF cases, from January 2022 to December 2023. Variation was assessed using multilevel logistic regression with facility random intercepts, the intraclass correlation coefficient (ICC), and the adjusted median odds ratio (aMOR). Facility-specific risk-standardized SPC rates were used to estimate SPC encounters attributable to facility performance better or worse than the national rate. Results: Of the sample, the mean patient age was 72.3 years (SD = 10.0), and 97.6% were male. The national observed rate of SPC was 17.5%, with risk-adjusted rates varying approximately 14-fold across facilities (3.3% to 45.6%). The adjusted ICC was 11.5%, and aMOR was 1.87 (95% Confidence Intervals 1.70-2.06). Measured patient case-mix and facility structural characteristics explained only 18.9% of between-facility variation (proportional reduction in the ICC, fully adjusted vs. null model). Facilities performing better than the national rate delivered 791 more SPC encounters than expected (17.8%), while those performing worse than the national rate delivered 480 fewer encounters than expected (10.8%). Conclusions: In the context of a national mean rate of SPC that reflects substantial underuse, delivery varied 14-fold across VAMCs, with most variation unexplained by patient complexity or measured facility resources. These findings suggest that potentially modifiable organizational factors, beyond patient preferences or facility structures alone, may contribute to current utilization gaps and represent actionable targets for quality improvement.</jats:p>