Introduction: Preoperative hypoalbuminaemia is a risk factor for surgical site infection (SSI) in the South African (SA) setting. However, the predictive accuracy of preoperative hypoalbuminaemia has not been tested against established SSI risk stratification models in our setting, which could have important implications for SSI prevention strategies. With reference to SSI in SA settings, the study objective was to compare the overall predictive accuracy of preoperative hypoalbuminaemia with that obtained for the SENIC/NNIS risk scores.
Method: This was a sub-analysis of a pre-existing laparotomy patient registry (N = 439). Variables collected as part of the registry included preoperative serum albumin measurements and all parameters of the SENIC/NNIS risk scores. Preoperative hypoalbuminaemia was defined as preoperative serum albumin of < 30 g/L. The study outcome was SSI up to 30 days postoperatively. Overall predictive accuracy was determined through a receiver operator-characteristic (ROC) curve analysis, with results presented as C-statistics (95% confidence intervals [CI]).
Results: The C-statistics obtained for preoperative hypoalbuminaemia, the SENIC risk score, and the NNIS risk score were 0.677 (CI: 0.609–0.746), 0.652 (CI: 0.582–0.721), and 0.634 (CI: 0.563–0.705).
Conclusion: All three methods display similar predictive accuracy for SSI. However, preoperative hypoalbuminaemia has several practical advantages over the SENIC/NNIS scores which must be considered.