LAUSR

To the Editor The recent article by Sangji et al. [1] derivating and validating a novel physiological emergency surgery acuity score (PESAS) was of great interest. They show that the PESAS can assess the acuity of disease at presentation in emergency surgery patients and is strongly associated with postoperative mortality. The main strengths of this study are the use of the American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP) database including a relatively large sample of emergency surgery patients. Furthermore, the authors had applied properly statistical methods to derivate the PESAS. Given that accurate prediction of postoperative mortality is important for emergency surgery quality improvement, their findings have the potential implications. However, we noted the several aspects of this study that were not well addressed. First, in designing the PESAS, it was unclear why Sangji et al. only included the preoperative variables related to health status and comorbidities of emergency surgery patients, but not the intraoperative and postoperative risk factors affecting mortality risk of emergency surgery patients. Other than preoperative health status and comorbidities, the surgical burden is an important determinant for postoperative mortality. It has been shown that the Surgical Apgar Score based on the intraoperative blood loss, lowest heart rate and lowest mean artery pressure is a good predictor of mortality after emergency high-risk surgery [2]. Furthermore, postoperative complications, such as acute kidney injury, pulmonary complications and myocardial infarction, have been strongly associated with mortality after emergency surgery. Specially, postoperative acute kidney injury and pulmonary complications are common among emergency surgery patients and have been significantly associated with increased risk of early postoperative mortality [3, 4]. In fact, postoperative complications have been regarded as important targets for emergency surgical quality improvement initiatives. In available literature, predictive performance of the models only using preoperative variables for postoperative mortality has been questioned [5]. Thus, we believe that discriminating ability and predictive value of the PESAS designed by authors would have further been improved, if the model design had included intraoperative and postoperative risk factors affecting postoperative mortality of emergency surgery patients. Second, when validation of the PESAS was performed using 2012 data, only c-statistic was measured. The receiver operating characteristic curve was established, but the relative analysis was not carried out. To assess the predictive performance of the PESAS for postoperative mortality in a validation set, only providing the c-statistic is not enough, especially for low-risk patients. The authors should perform the calibration assessment by the Hosmer-Lemeshow goodness-of-fit test. The calibration assessment can test the predictive ability of the PESAS to match the number of actual events across deciles of risk-stratified subgroups. By providing the predicted probabilities and observed frequencies for postoperative mortality based on the PESAS, the readers can estimate whether there is a good overall agreement between predicted probabilities and observed frequencies in the development and validation sets. In the calibration assessment, a P value of\ 0.05 indicates that the null hypothesis—that actual observed events occur at a frequency similar to events predicted by the PESAS—is false, & Fu-Shan Xue [email protected]