Prediction of Adverse Outcomes After Emergency General Surgery in Older Patients

Abstract

To the Editor: With great interest we read the recent article by Zattoni and colleagues assessing predictive ability of the Flemish Version of Triage Risk Screening Tool (fTRST) score for 30and 90-day postoperative morbidity and mortality in older patients undergoing emergency general surgery. Based on results from the multivariable logistic regression analysis and the area under the receiver operating characteristic curve (AUROC), they conclude that the fTRST score is an effective tool for prediction of postoperative short-term mortality and morbidity. Given that older patients undergoing emergency general surgery have the high risks of postoperative morbidity and mortality, findings of this study have potential implications for emergency surgical quality improvement initiatives. To avoid any optimistic interpretation of the results, however, there are some issues in this article that need further clarification and discussion. First, multivariable logistic regression analysis was used to determine preoperative and intraoperative risk factors for postoperative morbidity and mortality, and model discrimination was assessed using the AUROC. In the results, the authors described that multivariate analysis confirmed the roles of an fTRST score of 2 or greater and a Charlson Age Comorbidity Index score of 6 or greater in predicting postoperative mortality (AUROC = 0.85) and complications (AUROC = 0.82). However, the Tables 4 and 5 provided by the authors showed that the AUROC of both the fTRST score of 2 or greater and Charlson Age Comorbidity Index score of 6 or greater for postoperative mortality and complications was not more than 0.75. It is generally believed that AUROC of 0.90-1.0 indicates an excellent predictive value, AUROC of 0.75-0.89 suggests a good predictive value, while AUROC of 0.60-0.74 suggests a fair predictive value and AUROC of 0.50-0.59 indicates a low predictive value. Second, the fTRST score used in this study only included the five preoperative variables indicating preoperative health status and comorbidities (ie, presence of cognitive decline, living alone or no help frompartner/family available, reducedmobility or falls in the past 6 months, hospitalized in the past 3 months, and polypharmacy). It must be emphasized that besides preoperative health status and comorbidities, surgical burden of patients is another important determinant of short-termmortality and morbidity after emergency surgery. Thus, a risk score only including preoperative variablesmay notwell explain postoperative outcomes. The available evidence indicates that surgery type and time, intraoperative large blood loss, and blood transfusion are significantly associated with increased risks of mortality and morbidity after emergency surgery. In fact, the Surgical Apgar Score based on the estimated blood loss, lowest heart rate, and lowest mean artery blood pressure during surgery describes a combination of surgical complexity and the individual patient’s response to surgical stress. It has been shown that the Surgical Apgar Score is a powerful predictor of postoperative morbidity and mortality both in fit and frail older patients undergoing emergency abdominal surgery. As compared with the surgical risk score only based on preoperative factors, moreover, the models combining preoperative and intraoperative risk factors can provide the improved predictive ability for postoperative morbidity andmortality. Finally, in this study, the use of the same risk score to predict postoperative complications and mortality is not a suitable consideration. In fact, severe postoperative complications, such as acute kidney injury, sepsis, myocardial infarction, stroke, and pulmonary complications, have been significantly associated with short-term mortality after emergency surgery. Furthermore, there are interaction effects between certain postoperative complications to increase the risk of postoperative mortality. Specially, pulmonary complications, acute kidney injury, sepsis, and stroke are most likely to be involved in positive interactions. Thus, when establishing an effective risk score or model for prediction of postoperative short-term mortality, not taking postoperative complications into the account will undoubtedly decrease its discrimination ability. In view of above limitations existed in this study, we argue that the fTRST score only based on preoperative factors may be used as a simple preoperative screening tool for postoperative adverse outcomes of older patients undergoing emergency general surgery, but it not is an effective tool for prediction of short-term mortality and morbidity.