Concurrent external validation of various out-of-hospital cardiac arrest prognostication scores in a Korean population: a multicenter retrospective cohort study
- Author:
Sun Bom KWON
1
;
Jeong Ho PARK
;
Sang Do SHIN
;
Young Sun RO
;
Jun Hee LEE
;
Dae Han WI
Author Information
1. Department of Emergency Medicine, Wonkwang University Sanbon Hospital, Gunpo, Korea
- Publication Type:Original Article
- From:Journal of the Korean Society of Emergency Medicine
2023;34(6):487-497
- CountryRepublic of Korea
- Language:English
-
Abstract:
Objective:Various out-of-hospital cardiac arrest (OHCA) prognostication scores have been developed. However, the application of these scores is often limited owing to missing predictor variables. This study aimed to compare and validate various OHCA prognostication scores using simple imputation methods that can easily be applied in clinical situations.
Methods:Adult patients presenting with OHCA with a sustained return of spontaneous circulation (ROSC) between October 2015 and June 2020 were the subjects for the analysis. We evaluated six OHCA prognostication scores: the ROSC after cardiac arrest (RACA) score, CaRdiac Arrest Survival Score (CRASS), NULL-PLEASE, predictive score (PS), cardiac arrest hospital prognosis (CAHP) score, and the OHCA score. For missing predictors, median imputation for continuous variables and mode imputation for categorical variables were performed before the analysis. We evaluated the discrimination and calibration powers of each prognostic score for good neurological recovery at discharge. The area under the receiver operating characteristic curve (AUC) was used to assess the discrimination power, and a calibration plot and the Hosmer-Lemeshow test were used to assess the calibration power.
Results:Of the 12,321 patients, 5,191 were subjected to analysis. Among them, 924 (17.8%) had good neurological recovery. Certain predictors often had missing values-no-flow time 1,107 (21.3%), low-flow time 862 (16.6%), pH 1,104 (21.3%), lactate 1,820 (35.0%), and creatinine 2,257 (43.5%). After imputing the missing variables, the CAHP score showed the highest AUC (0.957; 95% confidence interval, 0.950-0.963), and the CRASS and PS also presented excellent discrimination power (AUC 0.914 and 0.942, respectively). However, the CAHP and NULL-PLEASE scores were well calibrated (Hosmer-Lemeshow test, P>0.05).
Conclusion:Among the six prognostic scores, the CAHP score showed the highest discrimination and calibration powers.
