
Journal of Korean Medical Science, Journal Year: 2024, Volume and Issue: 40(15)
Published: Dec. 12, 2024
Despite the considerable disease burden due to trauma injury, sufficient effort has not been made for assessment of nationwide care status in Korea. We explored feasibility a diagnosis code-based injury severity measuring method light its real-world usage. used datasets from National Emergency Department Information System calculate survival risk ratios (SRRs) and Korean Trauma Data Bank predict models, respectively. The target cohort was split into training validation using stratified random sampling an 8:2 ratio. established six major mortality prediction models depending on included parameters: 1) Injury Severity Score (TRISS) (age, sex, original Revised [RTS], [ISS]), 2) extended International Classification Diseases-based (ICISS) 1 RTS, ICISS international SRRs), 3) 2 SRRs based 4-digit codes), 4) 3 full-digit 5) 4 modified 6) 5 codes). estimated model fitted it datasets. measured area under receiver operating characteristic curve (AUC) discriminative ability. Overall performance also evaluated Brier score. observed though their slightly lower than TRISS (training cohort, AUC 0.936-0.938 vs. 0.949). Regarding SRR calculation methods, we did find statistically significant differences. alternative use Alert, Voice, Pain, Unresponsive Scale instead Glasgow Coma RTS degrade performance. availability practical expect our contribute strengthening Care by utilizing classification.
Language: Английский