H. Kadri1, A. Hider1, W.G. Henderson1,2, M.R. Bronsert1, T.N. Robinson1, R.C. McIntyre1, R. Meguid1 1University Of Colorado School Of Medicine, Aurora, CO, USA 2Colorado School of Public Health, Aurora, CO, USA
Introduction: An increasing number of operations are being performed on older adults, and assessing their risk of postoperative complications is important to inform patients and providers. The Surgical Risk Preoperative Assessment System (SURPAS) is an extensively validated parsimonious surgical risk calculator, but its accuracy in geriatric patients has not been evaluated. We hypothesized that SURPAS predicts postoperative complications similarly between geriatric (age≥65) and younger (age<65) patients.
Methods: Data from the ACS-NSQIP PUF, 2009-18, were used to develop the SURPAS models; data from 2019-20 were used in testing. Nine surgical specialties were included in analysis. Multiple logistic regression analysis was used in model development and testing, with the independent SURPAS variables (primary operation CPT code, wRVU, ASA class, functional health status, surgeon specialty, emergency status, in/outpatient, and age) and complications as dependent variables. 30-day outcomes included mortality, overall morbidity, wound infection, urinary tract infection (UTI), venous thromboembolism (VTE), pulmonary, cardiac, renal, stroke, and bleeding complications, unplanned readmission, and non-home discharge. SURPAS was evaluated using c-indices, Brier scores, observed vs. expected complication rates, and Hosmer-Lemeshow calibration graphs.
Results: The developmental and testing data had sample sizes of 6,924,709 and 1,929,287, respectively. In the testing data, 37.7% of patients were aged ≥65 years. The mean c-index across all 12 outcomes for geriatric patients was 0.800 (range 0.690 for readmission to 0.888 for mortality) vs. 0.846 for younger patients (range 0.747 for readmission to 0.938 for mortality). The c-indexes were similar for overall morbidity (0.795 for geriatric vs. 0.792 for younger) and infection (0.786 vs. 0.773). The average Brier score was slightly worse for geriatric patients (0.036, range 0.004 for stroke to 0.106 for overall morbidity) vs. younger patients (0.019, range 0.001 for stroke to 0.069 for overall morbidity). Hosmer-Lemeshow graphs showed good calibration between geriatric and younger patients (Figure). SURPAS slightly overestimated risk in geriatric patients for morbidity and UTI, and slightly underestimated it for infection and cardiac complications. SURPAS more significantly overestimated risk in geriatric patients for bleeding and non-home discharge.
Conclusion: The performance of SURPAS preoperative prediction of postoperative outcomes was slightly less in geriatric patients compared to younger patients, but still clinically acceptable. This highlights that SURPAS provides accurate risk assessment in older patients, even though the risk calculator was not developed specifically for them.