Evaluation of 6 scoring systems in emergency department

Mortality prediction models have always been an integrated part of medical research. Researchers are looking for developing scoring systems to evaluate the quality and quantity of care and to predict the outcome of patients. These models provide criteria and standard metrics to measure severity of the disease and evaluate mortality as the final outcome. In this study the validity of several prediction models were compared in terms of overall performance, discrimination, and calibration. Methods: Thorough a prospective cohort design, data from 2205 patients admitted in Imam Reza Hospital who based on Emergency Severity Index were triage level 1, 2, or 3 were used. Then for each patient scoring system were calculated including APACHE IV, APACHE II, SAPS II, SOFA, MSOFA and one other model that is introduced in this study and is called HOSP afterward. Patients were followed until the time of discharge from hospital and their status as dead or alive was regarded as the outcome. Next the predictive validity of these scoring systems were evaluated using Brier Score technique for overall performance, AUROC for discrimination, and Hosmer-Lemeshow (H-L) goodness-of-fit test for calibration. Besides, DeLong s test and standard mortality ratio were used for further evaluation. All analysis were performed using SPSS 21 and R 3.2.2 Results: A total of 2205 patients (53% males and 47% females) were included. 198 patients with 1 triage level (9%), 831 patients with 2 triage level (38%) and 1176 patients with 3 triage level (53%). Patients referring to ambulance consist of 1019 patients representing 46% of the total patients, and 1186 patients refer without ambulance, accounting for 54% of the total patients. The area under the curve was 0.847 for HOSP, 0.794 for APACHE IV, 0.754 for APACHE II, 0.752 for SAPS II, 0.739 for MSOFA, and 0.750 for SOFA. The mean and standard deviation for prediction of total mortality were19.31±22.23 for HOSP, 10± 5.6 for APACHE II, 49 ± 19.28 for APACHE IV, 29 ± 10.45 for SAPS II, 2.13 ± 2 for MSOFA, and 3±2.52 for SOFA. Conclusion: Among the previously established models, APACHE IV was significantly different from other models. According to the calculated Brier score, APACHE IV system with a value of 0.138 had the best performance of all other models and SOFA, MSOFA, APACHE II, and SAPS II are placed afterward. Regarding discrimination, the best performance was also seen in APACHE IV. Considering calibration, all of these models show an unsatisfactory performance. It is necessary to recalibration before using systems. However, the HOSP model could outperform the other models in all aspect of evaluation. So it can be used for benchmarking, decision making and organizational Planning.

کلیدواژه ها:

Emergency Department ، risk scoring systems ، APACHE ، SAPS ، SOFA ، MSOFA Calibration ، mortality prediction ، ESI triage algorithm