Physical frailty is common in patients with cirrhosis and is strongly associated with health outcomes. While tools have been developed to assess frailty within administrative datasets, these instruments are neither specific to patients with cirrhosis nor validated against established measures of physical frailty. Adult cirrhosis patients evaluated for liver transplantation with in-person frailty assessments using the Liver Frailty Index (LFI) were identified from a single center, 2017-2024. The electronic health record was searched to identify ICD-10 codes associated with encounters in the two years prior to frailty assessment. Lasso regression was used to select ICD-10 codes associated with LFI and generate a cirrhosis administrative frailty index (CAFI). Model performance was assessed using discrimination (AUC) and calibration (CITL, slope). Competing-risks regression was used to determine the association between CAFI and pretransplant mortality (death or delisting due to illness), adjusting for MELD score. Among 2,377 patients, 22% were frail by LFI>4.4. The CAFI includes 29 predictors, strongly associated with LFI>4.4, with an AUC of 0.73. In comparison, the Hospital Frailty Risk Score weakly discriminated LFI>4.4 (AUC=0.58). In multivariable models, each point increase in CAFI was associated with 29% increased risk of pre-transplant mortality (SHR=1.29, 95% CI: 1.11, 1.50). We developed the CAFI, the first ICD-based frailty index specific to cirrhosis patients and validated against an in-person frailty measure. The CAFI will enable investigators to account for frailty as a confounder or mediator in administrative data-based studies, improving the validity of population-based research in patients with cirrhosis.
Calthorpe et al. (Mon,) studied this question.