An externally validated robust risk predictive model of adverse outcomes after carotid endarterectomy.

BACKGROUND: The goal of this study was to construct and externally validate a risk prediction model for patients who undergo carotid endarterectomy (CEA).

METHODS: The Vascular Study Group of New England (VSGNE) and Vascular Quality Initiative (VQI) databases were queried for patients who underwent CEA. Data on perioperative variables, comorbidities, and medications were entered into a logistic regression model as predictors of the composite adverse outcomes in the VSGNE sample. Adverse outcomes were defined if the patient experienced any of the following events: postoperative stroke, myocardial infarction, mortality, or discharge to rehabilitation facility. Backward elimination (α level of .2) was then used to select a more parsimonious model. Calibration was performed to measure how closely predicted outcomes agreed with observed outcomes. To assess calibration we used Hosmer-Lemeshow test, the predictive value of the model was assessed via the C-statistic. The external validation was then performed using the VQI sample after excluding those in the VSGNE sample (VQI-VSGNE) following a similar method.

RESULTS: The constructed model showed a substantial predictive capability of adverse outcomes (C = 0.711) with goodness of fit (Hosmer-Lemeshow lack of fit test, P = .494). A significantly higher rate of adverse outcomes was noted for the VQI sample (5.2%; n = 12,075) compared with the VSGNE sample (4.49%; n = 8661; P = .017). The discriminating ability of the VSGNE model remained substantial in the external data (C = 0.702).

CONCLUSIONS: The internally validated VSGNE CEA risk model, which robustly predicted adverse outcome after CEA, was externally validated by testing it against the remainder of VQI patients who underwent CEA by a diverse array of physicians. This tool provides a simple and reliable method to risk-stratify CEA patients using only their preoperative conditions. A risk score based on this model can reliably stratify patients according to their risk of adverse outcomes after CEA.