RESUMO
PURPOSE: The performance of the Intensive Care National Audit & Research Centre (ICNARC) model for predicting 1-month mortality was evaluated and compared with performances of APACHE II, adjusted APACHE (adj-APACHE) II, and SAPS II models. METHODS: All admissions except those <16-years-of-age (n=7) and with inadequate data (n=10) in our emergency intensive care unit (ICU) during one year (n=661) were retrospectively analyzed. Performance was assessed using Area under curve and Hosmer-Lemeshow goodness-of-fit. Customization was performed in randomly selected patients (n=324) and performance was evaluated in the remaining patients (n=337). RESULTS: Observed 1-month mortality was 25.6% (169/661). AUCs of ICNARC, APACHE II, adj-APACHE II, and SAPS II were .849, .845, .822, and .859, respectively. AUC of adj-APACHE II was smaller than the others (p<.05). All original models had poor calibrations (p<.05). After customization, AUCs of C-ICNARC, C-APACHE II, C-adjusted APACHE II, and C-SAPS II were .852, .849, .821, and .878, respectively. C-ICNARC and C-adj-APACHE II had fair calibrations (p=.251 and .074), but C-APACHE II and C-SAPS II had poor calibrations (p<.05). CONCLUSION: The ICNARC model demonstrates good performance for predicting 1-month mortality in patients of an emergency ICU in Korea.
Assuntos
Humanos , APACHE , Área Sob a Curva , Calibragem , Cuidados Críticos , Emergências , Unidades de Terapia Intensiva , Coreia (Geográfico) , Avaliação de Resultados em Cuidados de Saúde , Prognóstico , Estudos Retrospectivos , Índice de Gravidade de DoençaRESUMO
PURPOSE: The performance of the Intensive Care National Audit & Research Centre (ICNARC) model for predicting 1-month mortality was evaluated and compared with performances of APACHE II, adjusted APACHE (adj-APACHE) II, and SAPS II models. METHODS: All admissions except those <16-years-of-age (n=7) and with inadequate data (n=10) in our emergency intensive care unit (ICU) during one year (n=661) were retrospectively analyzed. Performance was assessed using Area under curve and Hosmer-Lemeshow goodness-of-fit. Customization was performed in randomly selected patients (n=324) and performance was evaluated in the remaining patients (n=337). RESULTS: Observed 1-month mortality was 25.6% (169/661). AUCs of ICNARC, APACHE II, adj-APACHE II, and SAPS II were .849, .845, .822, and .859, respectively. AUC of adj-APACHE II was smaller than the others (p<.05). All original models had poor calibrations (p<.05). After customization, AUCs of C-ICNARC, C-APACHE II, C-adjusted APACHE II, and C-SAPS II were .852, .849, .821, and .878, respectively. C-ICNARC and C-adj-APACHE II had fair calibrations (p=.251 and .074), but C-APACHE II and C-SAPS II had poor calibrations (p<.05). CONCLUSION: The ICNARC model demonstrates good performance for predicting 1-month mortality in patients of an emergency ICU in Korea.