Risk modelling for carotid endarterectomy

Background: A 30‐day stroke or death rate of 3–10 per cent has been reported in patients undergoing carotid endarterectomy (CEA). The aim of this study was to identify the risk factors that influenced the outcome and to develop a logistic regression model that can aid in the comparative audit of CEA. Methods: Some 836 CEAs performed by four vascular surgeons from 1992 to 1999 were analysed. The median age of the population was 69 (range 38–86) years, and the male: female ratio was 1·6. Multiple logistic regression was used to model the effect of 15 risk factors on the 30‐day stroke and death rate. The outcome after risk adjustment was compared for four surgeons and two vascular units. Results: The overall 30‐day stroke and death rate was 3·9 per cent. Regression modelling identified heart disease, diabetes and stroke at presentation as significant risk factors. A risk score of 1 was assigned to each of these risk factors to produce a logistic regression equation (where P is the probability of 30‐day stroke or death): ln( P /1 − P ) = 0·9294(risk score) − 4·2726.Risk score 30‐day stroke or death risk (%)0 1·4 (0·9–2·1) 1 3·4 (2·2–5·2) 2 8·3 (2·9–21·5) 3 18·8 (3·5–58·8)Values in parentheses are 95 per cent confidence intervalsThe observed 30‐day stroke and death rate for four vascular surgeons was 3·0, 4·0, 4·2 and 4·3, and for two vascular units was 3·9 and 3·9. Differences between the surgeons and the vascular units were not statistically significant after allowing for risk adjustment using this model. Conclusion: Multiple logistic regression models can successfully identify patients at higher risk from CEA, and aid in the comparative audit of surgeons and vascular units. © 2001 British Journal of Surgery Society Ltd