Using a large database, we evaluated the characteristics of the four severity scores between the patients in the postoperative overnight ICU stay and prolonged ICU stay groups. In the overnight ICU stay group, there was a poor correlation between SOFA and APACHE III scores or SAPS II. In the prolonged ICU stay group, the correlation between SOFA and APACHE III scores or SAPS II was moderately accurate. Moreover, the predictive values of SOFA scores for in-hospital survival showed poor or low accuracy in the overnight ICU stay group and moderate accuracy in the prolonged ICU stay group; the accuracy of the other three scoring systems was high in the prolonged ICU stay group. These results indicate that APACHE II and III scores or SAPS II would be more useful than SOFA scores at ICU admission for the prediction of in-hospital mortality, especially for patients with prolonged ICU stay.
A systematic review by Minne et al. showed that the AUC of SOFA scores for predicting survival was 0.61–0.88 [3]. They summarized that AUCs of SOFA scores were worse than those of APACHE II/III scores. In a study in the United Kingdom, the prediction values were as follows: APACHE II score (AUC = 0.62) and SOFA score (AUC = 0.61) [4]. Additionally, they described that mortality prediction with SOFA and APACHE II scores was not significant for medical patients; however, for surgical patients, it was significant with APACHE II score (AUC = 0.71) and SOFA score (AUC = 0.62). Our results are in line with the findings of these previous studies, highlighting the different predictive values among postoperative and other patient groups.
The authors have been interested in the admission, triage, and discharge (ADT) criteria of the ICU [5] owing to economic and financial concerns. In Japan, the government mandated a diagnosis-based insurance payment system. Recently, the national provision regarding ICU admission was revised. Hence, hospitals are required to report the SOFA scores at ICU admission and data of patients with sepsis. The ADT guideline documented that any scoring system alone cannot determine the level of care (recommendation 2C) [5]. This was based on a study that showed that a physician’s prediction of patient survival was more accurate than that of scoring systems [6]. Moreover, physicians’ and nurses’ predictions improved the discriminative accuracy of in-hospital survival [7]. Concerning the SOFA score beyond a threshold value not being included in the ICU fee, we aimed to illustrate how the SOFA score alone was unlikely to predict the need for ICU stay.
The demand for ICU stay varies from rigorous monitoring to intensive mechanical life support. The real necessity of ICU stay is difficult to anticipate because ICU admission is subjective, and the severity of a disease does not clearly show a threshold for deciding admissions [8]. Additionally, the admission and triage criteria are occasionally not followed in cases of mass disaster or infection pandemics [5]. In the present study, we considered patients with postoperative overnight ICU stay, the aim of which is mainly concerned with patient monitoring. Irie et al. summarized general patient characteristics in the JIPAD data and analyzed the severity of disease among those requiring monitoring and critically ill adults [9]. They showed the difference in the standardized mortality ratio, APACHE III scores, SAPS II, and in-hospital mortality between the groups. Regarding patient classification criteria, we used the same grouping and the group names were changed to overnight and prolonged ICU stay; however, we also included the SOFA data and focused on the accuracy of the scoring system.
Our study has some limitations. First, the data in the JIPAD were from a limited number of ICUs in Japan. As half of the patients in the database were postsurgical patients, it is difficult to generalize the results to other types of ICUs. Second, the severity scores entered into the database were calculated in various ways. In some institutes, they utilized a semi-computerized method, wherein the physiological values in the electronic medical records were entered onto the data-acquisition platform and expert intensivists authorized by the JIPAD committee proved the total values. At other ICUs, the value was entered manually. Quality control was checked by the JIPAD; however, the quality may need to be validated carefully in each case. Moreover, SOFA scores were first included in the JIPAD in 2018, and our data comprised the SOFA values in the first year after inclusion. It is possible that the quality of SOFA score data would increase over time. Third, there is a possibility of errors in the database. Fourth, the category of postoperative overnight ICU stay patients might not be a well-defined grouping. The decision of a 1-day ICU stay depends not only on the mild disease status but also on various other factors.