Purpose : The purpose of this study was to compare the predictive perfor mance of the APACHE II and APACHE IV scoring systems for mortality a nd length of stay among patients admitted to a medical intensive care uni t of a tertiary hospital, and to ev...
Purpose : The purpose of this study was to compare the predictive perfor mance of the APACHE II and APACHE IV scoring systems for mortality a nd length of stay among patients admitted to a medical intensive care uni t of a tertiary hospital, and to evaluate their clinical utility in the Korean clinical setting while providing a basis for future calibration discussions. Methods : This retrospective observational study included all patients admi tted to the medical intensive care unit of a tertiary hospital located in D metropolitan city between January 1 and December 31, 2024. Among 617 eligible patients, 301 were included in the final analysis after applying exc lusion criteria. APACHE II and APACHE IV scores were calculated based o n the worst values recorded within the first 24 hours after ICU admission. APACHE II scores were automatically calculated using the hospital electro nic system, whereas APACHE IV scores were obtained using the official I ntensive Care Network calculator. Descriptive statistics, Pearson correlatio n analysis, and receiver operating characteristic (ROC) curve analysis wer e performed using IBM SPSS Statistics version 29.0. Comparisons of the ar ea under the curve (AUC) were conducted using the DeLong test in R. L ength of stay was compared using the Wilcoxon signed-rank test accordin g to the results of the normality test. Results : The results of this study are as follows. 1. A total of 301 patients were included. There were 186 males (61.8%) a nd 115 females (38.2%), with a mean age of 69.37 ± 14.82 years. The ad mission routes were the emergency department in 179 patients (59.5%) an d general wards in 122 patients (40.5%). The distribution of departments i ncluded pulmonology in 178 patients (59.1%), cardiology in 24 patients (8. 0%), and gastroenterology in 30 patients (10.0%), among others. Re-admiss ion occurred in 31 patients (10.3%), and emergency surgery was performe d in 5 patients (1.7%). Mechanical ventilation was applied in 160 patients (53.2%). Oxygen delivery devices applied included mechanical ventilators in 157 patients (52.2%), high-flow nasal cannula in 80 patients (26.6%), and n asal cannula in 40 patients (13.3%). Regarding life-sustaining treatment pla ns, active treatment was selected by 224 patients (74.4%). 2. The mean APACHE II score was 22.49 ± 7.05, with a median of 22 (I QR, 17–27), and the mean APACHE IV score was 89.75 ± 25.50, with a median of 88 (IQR, 71–105.5). The predicted mortality was 44.63 ± 21.4 0% (median, 42.4%) for APACHE II and 43.28 ± 26.09% (median, 40.2%) f or APACHE IV. The observed intensive care unit length of stay was a me an of 10.76 ± 11.21 days, with a median of 6 days (IQR, 3–14), whereas the APACHE IV–predicted length of stay was a mean of 5.81 ± 5.71 day s, with a median of 5.4 days (IQR, 4.2–6.7). The duration of mechanical v entilation was a mean of 6.32 ± 10.55 days. 3. Correlation analysis between APACHE II and APACHE IV scores demons trated a positive correlation (r=.80, p<.001). 4. For prediction of intensive care unit mortality, the AUC of APACHE II was 0.71 (95% CI, 0.65–0.78) and that of APACHE IV was 0.76 (95% CI, 0. 70–0.82), with a statistically significant difference between the two models (p = .001). For prediction of hospital mortality, the AUC of APACHE II wa s 0.70 (95% CI, 0.64–0.76) and that of APACHE IV was 0.76 (95% CI, 0.70 –0.82), with APACHE IV showing a significantly higher discriminative perf ormance(p<.001). 5. The median observed intensive care unit length of stay was 6 days (IQ R, 3–14), which was significantly longer than the APACHE IV–predicted le ngth of stay of 5.4 days (IQR, 4.2–6.7)(p<.001). Conclusion : In this study, a positive correlation was identified between A PACHE II and APACHE IV scores, indicating that both scoring systems sim ilarly reflect patient severity. In terms of mortality risk stratification, APA CHE IV demonstrated superior discriminative performance compared with APACHE II by incorporating a broader range of clinical contextual variabl es. In contrast, APACHE IV tended to underestimate the actual length of stay. Therefore, combined use of APACHE II, which is simple to calculate, and APACHE IV, which provides higher predictive accuracy, may improve prognostic assessment and resource management in the intensive care uni t. Furthermore, the development of a Korean-adjusted APACHE model tha t reflects the characteristics of Korean patient populations and healthcare systems may contribute to improved predictive accuracy and more efficie nt ICU operation.