Rofecoxib (Vioxx) voluntarily withdrawn from market

ABSTRACT Objectives To compare the accuracy of the SOFA and APACHE II scores in predicting short-term mortality among ICU patients with sepsis in an LMIC. Design A multicentre, cross-sectional study. Setting A total of 15 adult ICUs from 14 hospitals, of which 5 are central hospitals, and 9 are provincial, district, or private hospitals, throughout Vietnam. Participants We included all patients aged ≥18 years who were admitted to ICUs for sepsis and who were still in ICUs from 00:00 hour to 23:59 hour of the study days (i.e., 9th January, 3rd April, 3rd July, and 9th October of 2019). Main outcome measures Short-term mortality was the main outcome, including hospital and ICU mortality. Results Of 252 patients, 40.1% died in hospitals, and 33.3% died in ICUs. SOFA (cut-off value ≥7.5; AUROC: 0.688 [95% CI: 0.618-0.758]; p<0.001) and APACHE II score (cut-off value ≥20.5; AUROC: 0.689 [95% CI: 0.622-0.756]; p<0.001) both had a poor discriminatory ability for predicting hospital mortality. However, the discriminatory ability for predicting ICU mortality of SOFA (cut-off value ≥9.5; AUROC: 0.713 [95% CI: 0.643-0.783]; p<0.001) was better and greater than that of APACHE II score (cut-off value ≥18.5; AUROC: 0.672 [95% CI: 0.603-0.742]; p<0.001). A SOFA score ≥8 (OR: 2.717; 95% CI: 1.371-5.382) and an APACHE II score ≥21 (OR: 2.668; 95% CI: 1.338-5.321) were independently associated with an increased risk of hospital mortality. Additionally, a SOFA score ≥10 (OR: 2.194; 95% CI: 1.017-4.735) was an independent predictor of ICU mortality, in contrast to an APACHE II score ≥19, for which this role did not. Conclusions Both SOFA and APACHE II scores were worthwhile in predicting hospital and ICU mortality among ICU patients with sepsis. However, due to good discrimination for predicting ICU mortality, the SOFA was preferable to the APACHE II score in predicting short-term mortality. Strengths and limitations of this study An advantage of the present study was data from multicentre, which had little missing data. Due to the absence of a national registry of intensive care units (ICUs) to allow systematic recruitment of units, we used a snowball method to identify suitable units, which might have led to the selection of centres with a greater interest in sepsis management. Due to the study’s real-world nature, we did not make a protocol for microbiological investigations. Moreover, we mainly evaluated resources utilized in ICUs; therefore, the data detailing the point-of-care testing and life-sustaining treatments were not available. To improve the feasibility of conducting the study in busy ICUs, we opted not to collect data on antibiotic resistance and appropriateness. The sample size was relatively small, which might have led to overfitting in the multivariable prediction model.