Simplifying Hospital Quality Comparisons for Vascular Surgery Using Center-Level Frailty Burden Rather Than Comorbidities

Background: Failure to rescue (FtR), or inpatient death following complication, is a publicly reported hospital quality measure. Previous work has demonstrated significant variation in the proportion of frail patients across hospitals. However, frailty is not incorporated into risk-adjustment algorithms for hospital quality comparisons and risk adjustment is made by comorbidity scores. Our aim was to assess the impact of frailty on FtR quality measurement and as a means of risk adjustment.Methods: Patients undergoing open or endovascular aneurysm repair or lower extremity bypass in the Vascular Quality Initiative (VQI) at centers performing > 25 vascular procedures annually (2003-2019) were included. Multivariable logistic regression evaluated in-hospital death using scaled hierarchical modeling clustering at the center level. Center FtR observed/expected ratios were compared with expected values adjusted for either standard comorbidity profiles or frailty as measured by the VQI Risk Analysis Index. Centers were divided into quartiles using VQI-linked American Hospital Association data to describe the hospital characteristics of centers whose ranks changed.Results: A total of 63,143 patients (213 centers) were included; 1,630 patients (2.58%) were classified as FtR. After accounting for center-level variability, frailty was associated with FtR [scaled odds ratio 1.9 (1.8-2.0), P < 0.001]. The comorbidity-centric and frailty-based models performed similarly in predicting FtR with C-statistics of 0.85 (0.84-0.86) and 0.82 (0.820.84), respectively. Overall changes in ranking based on observed/expected ratios were not statistically significant (P 1/4 0.48). High and low performing centers had similar ranking using comorbidity-centric and frailty-based methods; however, centers in the middle of the performance spectrum saw more variability in ranking alterations. Forty nine (23%) of hospitals improved their ranking by five or more positions when using frailty versus comorbidity risk adjustment. The centers in Quartile 4, those who performed the highest number of vascular procedures annually, experience on average a significant improvement in hospital ranking when frailty was used for risk adjustment, whereas centers performing the fewest number of vascular procedures and the lowest proportion of vascular surgery cases annually (Quartile 1) saw a sig-nificant worsening of ranking position (all P < 0.05). However, total number of surgical proced-ures annually, total hospital beds, for-profit status, and teaching hospital status were not significantly associated with changes in rank.Conclusions: A simple frailty-adjusted model has similar predictive abilities as a comorbidity-focused model for predicting a common quality metric that influences reimbursement. In addition to distilling the risk-adjustment algorithm to a few variables, frailty can be assessed preopera-tively to develop quality improvement efforts for rescuing frail patients. Centers treating a greater proportion of frail patients and those who perform higher volumes of vascular surgery benefit from a risk adjustment strategy based on frailty.