Development and validation of a predictive model for the risk of 30-day death in emergency department patients

Abstract

To explore the risk factors for 30-day death in emergency department patients, and then construct a prediction model and validate it using nomogram. A retrospective cohort study was conducted. The clinical data of 1 091 patients admitted to the emergency department of the First People's Hospital of Changde from January 1 to June 30, 2021 was collected, including 741 patients from January 1 to March 31 in the development group and 350 patients from April 1 to June 30 in the validation group. General information, first vital signs admitted to the emergency department, and laboratory results were collected, the modified early warning score (MEWS) was calculated, and 30-day outcomes were recorded. Univariate and multivariate Logistic regression analysis was used to screen out the risk factors of 30-day death. According to the results of multivariate analysis, the nomogram was used to construct a 30-day death prediction model. The receiver operator characteristic curve (ROC curve) was used to evaluate the consistency of the prediction model, the calibration of the prediction model was evaluated by the Hosmer-Lemeshow goodness of fit test. A total of 1 091 patients were enrolled. There were 741 patients in the development group, including 356 males and 385 females, aged (51.42±17.33) years old, and the 30-day mortality was 28.88%. There were 350 patients in the validation group, including 188 males and 162 females, aged (52.88±16.11) years old, and the 30-day mortality was 24.00%. The results of the univariate analysis showed that age, primary diagnosis on admission, consciousness, respiratory rate (RR), systolic blood pressure (SBP), heart rate (HR), pulse oxygen saturation (SpO2), MEWS score, erythrocyte sedimentation rate (ESR), procalcitonin (PCT) and body mass index (BMI) might be the risk factors for 30-day death in patients in the emergency department. The results of the multivariate analysis showed that the MEWS score [odds ratio (OR) = 14.22, 95% confidence interval (95%CI) was 1.46-138.12], ESR (OR = 46.71, 95%CI was 20.48-106.53), PCT (OR = 4.97, 95%CI was 2.46-10.02), BMI (24.0-27.9 kg/m2: OR = 37.82, 95%CI was 14.69-97.36; ≥ 28.0 kg/m2: OR = 62.11, 95%CI was 25.77-149.72) were independent risk factors for 30-day death in the emergency department (all P < 0.05). Using the four variables with the results of multivariate analysis to construct a nomogram prediction model, the area under the ROC curve (AUC) was 0.974 (95%CI was 0.753-0.983) for the development group, and the AUC was 0.963 (95%CI was 0.740-0.975) for the validation group. The Hosmer-Lemeshow test showed no statistically significant difference between the predicted outcome of the nomogram prediction model and the actual occurrence (χ2 = 1.216, P = 1.270). The prediction model developed by the MEWS score combined with BMI, ESR and PCT can scientifically and effectively predict the 30-day outcome of emergency department patients.