Development and validation of risk stratification tool for prediction of increased dependence using preoperative frailty after hepatopancreatic surgery

Background Despite the known association between frailty and postoperative morbidity, the use of preoperative frailty in surgical practice remains limited. We sought to develop a risk tool to predict postoperative increase in functional dependence. Methods Patients of ≥65 years in the National Surgical Quality Improvement Project database who had a primary hepatopancreatic surgery between 2015 and 2019 were used to identify predictors of increased dependence and development of a simplified tool to calculate the risk stratification score for increased discharge care level (https://ktsahara.shinyapps.io/care_discharge/). Results Among 31,338 patients who underwent primary hepatopancreatic surgery, 4,259 (13.6%) had an increased level of care at discharge compared to their preadmission care. Patients with increased discharge care had a higher proportion of patients with a modified frailty index of at least 2 (n = 1496; 35.1%) compared with individuals with unchanged care (n = 6,760; 25.0%). In addition, 12.3% (n = 3,858) were discharged to a skilled nursing or rehabilitation facility. Of note, the odds of increased care at discharge were increased by 1.41 (95% confidence interval: 1.32–1.50), 1.11 (95% confidence interval :1.11–1.12), and 1.95 (95% confidence interval:1.86–2.04) times with every unit increase in modified frailty index, age beyond 65 years, and the number of in-hospital complications, respectively. Area under receiver operative curve for the parsimonious model used to develop the risk calculator was 0.7486 (95% confidence interval: 0.7405–0.7566) (all P < .001). Conclusion Approximately, 1 in 7 patients required an increased level of care at the time of discharge compared with their preadmission status. A simplified web-based risk tool can be used in clinical practice as a surgical decision aid in post-discharge planning after complex elective surgery.