BACKGROUND: Frailty assessment adds important prognostic information during preoperative decision-making but can be cumbersome to implement into routine clinical care. We developed and tested an abbreviated method of frailty assessment using variables routinely collected by the Vascular Quality Initiative (VQI) registry. METHODS: An abbreviated frailty score (the simple Vascular Quality Initiative-Frailty Score [VQI-FS]) was developed using 11 or fewer VQI variables (hypertension, congestive heart failure, coronary artery disease, peripheral vascular disease, diabetes, chronic obstructive pulmonary disease, renal impairment, anemia, underweight, nonhome residence, and nonambulatory status) that map to recognized frailty domains in the Comprehensive Geriatric Assessment and the literature. Nonemergent cases registered in the VQI from 2010 to 2017 (n = 265,632) in seven registries (carotid endarterectomy, n = 77,111; carotid artery stenting, n = 13,215; endovascular abdominal aortic aneurysm repair, n = 29,607; open abdominal aortic aneurysm repair, n = 7442; infrainguinal bypass, n = 33,128; suprainguinal bypass, n = 10,661; and peripheral vascular intervention, n = 94,468) were analyzed using logistic regression models to determine the predictive power of the VQI-FS for perioperative and longer term (9-month) mortality. Nomograms were created using weighted regression coefficients to assist in individualized frailty assessment and estimation of 9-month mortality. RESULTS: The VQI-FS, using equal weighting of these 11 VQI variables, effectively predicted 9-month mortality with an area under the curve of 0.724 by receiver operating characteristic curve analysis. However, differential weighting of the variables allowed simplification of the model to only seven variables (congestive heart failure, renal impairment, chronic obstructive pulmonary disease, not living at home, not ambulatory, anemia, and underweight status); hypertension, coronary artery disease, peripheral vascular disease, and diabetes had relatively low predictive power. Adding procedure-specific risk further improved performance of the model with a final area under the curve on receiver operating characteristic curve analysis of 0.758. Model calibration was excellent with predicted/observed regression line slope of 0.991 and intercept of 5.449e-04. CONCLUSIONS: A differentially weighted abbreviated VQI-FS using seven variables in addition to procedure-specific risk has strong correlation with 9-month mortality. Nomograms incorporating patient- and procedure-adjusted risk can effectively predict 9-month mortality. Reliable estimates of longer term mortality should assist in preoperative decision-making for vascular procedures that often carry substantial risk of mortality.

Objective

Frailty is increasingly recognized as a key determinant in predicting postoperative outcomes. Centers that see more frail patients may not be captured in risk adjustment, potentially accounting for poorer outcomes in hospital comparisons. We aimed to (1) determine the effect of frailty on long-term mortality in patients undergoing elective abdominal aortic aneurysm (AAA) repair and (2) evaluate the variability in frailty burden among centers in the Vascular Quality Initiative (VQI) database.

Methods

Patients undergoing elective open and endovascular AAA repair (2003-2017) were identified, and those with complete data on component variables of the VQI-derived Risk Analysis Index (VQI-RAI) and centers with ≥10 AAA repairs were included. VQI-RAI characteristics are sex, age, body mass index, renal failure, congestive heart failure, dyspnea, preoperative ambulation, and functional status. Frailty was defined as VQI-RAI ≥35 based on prior work in surgical patients using other quality improvement databases. This corresponds to the top 12% of patients at risk in the VQI. Center-level VQI-RAI differences were assessed by analysis of variance test. Relationships between frailty and survival were compared by Kaplan-Meier analysis and the log-rank test for open and endovascular procedures. Multivariable hierarchical Cox proportional hazards regression models were calculated with random intercepts for center, controlling for frailty, race, insurance, AAA diameter, procedure type, AAA case mix, and year.

Results

A total of 15,803 patients from 185 centers were included. Mean VQI-RAI scores were 27.6 (standard deviation, 5.9; range, 4-56) and varied significantly across centers (F = 2.41, P < .001). The percentage of frail patients per center ranged from 0% to 40.0%. In multivariable analysis, frailty was independently associated with long-term mortality (hazard ratio, 2.88; 95% confidence interval, 2.6-3.2) after accounting for covariates and center-level variance. Open AAA repair was not associated with long-term mortality after adjusting for frailty (hazard ratio, 0.98; 95% confidence interval, 0.86-1.13). There was a statistically significant difference in the percentage of frail patients compared with nonfrail patients who were discharged to a rehabilitation facility or nursing home after both open (40.5% vs 17.8%; P < .0001) and endovascular repair (17.7% vs 4.6%; P < .0001).

Conclusions

There is considerable variability of preoperative frailty among VQI centers performing elective AAA repair. Adjusting for center-level variation, frailty but not procedure type had a significant association with long-term mortality; however, frailty and procedure type were both associated with nonhome discharge. Routine measurement of frailty preoperatively by centers to identify high-risk patients may help mitigate procedural and long-term outcomes and improve shared decision-making regarding AAA repair.