Risk modelling of outcome after general and trauma surgery (the IRIS score)

Abstract

Abstract Background A practical, easy to use model was developed to stratify risk groups in surgical patients: the Identification of Risk In Surgical patients (IRIS) score. Methods Over 15 years an extensive database was constructed in a general surgery unit, containing all patients who underwent general or trauma surgery. A logistic regression model was developed to predict mortality. This model was simplified to the IRIS score to enhance practicality. Receiver operating characteristic (ROC) curve analysis was performed. Results The database contained a consecutive series of 33 224 patients undergoing surgery. Logistic regression analysis gave the following formula for the probability of mortality: P (mortality) = A/(1 + A), where A = exp (−4·58 + (0·26× acute admission) + (0·63× acute operation) + (0·044× age) + (0·34× severity of surgery)). The area under the ROC curve (AUC) was 0·92. The IRIS score also included age (divided into quartiles, 0–3 points), acute admission, acute operation and grade of surgery. The AUC predicting postoperative mortality was 0·90. Conclusion The IRIS score accurately predicted mortality after general or trauma surgery.