J Am Heart Assoc. 2026 Mar 25:e045222. doi: 10.1161/JAHA.125.045222. Online ahead of print.
ABSTRACT
BACKGROUND: Hospital length of stay (LOS) among patients with heart failure (HF) is relevant for patients, payers, and hospitals. Risk adjustment of LOS supports fair and equitable reimbursement for facilities that may experience more complex cases, oftentimes serving marginalized populations. We aimed to assess factors contributing to HF LOS, building on commonly available information across a wide range of facilities.
METHODS: A Fine and Gray Cox proportional hazards model was fitted to assess hospital LOS using a large US cohort of 89 621 patients with HF hospitalized during the fourth quarter of 2023, controlling for censoring among patients leaving against medical advice and for competing risks of in-hospital all-cause mortality. In our primary aim, we risk-adjusted HF LOS for patient-level sociodemographic and clinical episode characteristics as well as facility-level factors. Model performance was assessed via concordance statistics across derivation and validation cohorts, and risk adjustments were reported as subdistribution hazard ratios. As a secondary aim, we explored facility-level risk-adjusted idiosyncratic differences in LOS.
RESULTS: Sociodemographic, clinical episode, and facility-level characteristics can explain differences in hospital LOS among patients with HF, with most variables being statistically significant. The model exhibited moderate performance with similar results across the derivation (C=0.686 [95% CI, 0.682-0.691]) and validation (C=0.691 [95% CI, 0.686-0.695]) cohorts.
CONCLUSIONS: Excess LOS can be attributed to multiple characteristics at the sociodemographic, clinical episode, and facility levels. We demonstrate a HF LOS risk-adjustment method that does not rely on, though can be expanded with, extensive patient clinical information, supporting more equitable assessments of facility performance and reimbursement.
PMID:41878833 | DOI:10.1161/JAHA.125.045222
