Customization and external validation of the Simplified Mortality Score for the Intensive Care Unit (SMS-ICU) in Brazilian critically ill patients

Purpose: To customize and externally validate the recently proposed Simplified Mortality Score for the ICU (SMS-ICU, a simple score for 90-day mortality that has no need for ancillary testing results) for in-hospital mortality and to compare its performance to SAPS 3. Material and methods: We used data from two distinct large cohorts of adult Brazilian patients with unplanned ICU admissions to perform a first-level customization (43,017 patients admitted to 78 ICUs) of the original SMS-ICU score for in-hospital mortality and, sequentially, externally validate it (313,365 patients admitted to 99 ICUs). Performance of SMS-ICU was assessed through measurements of discrimination and calibration and compared with SAPS 3. Results: In the validation cohort, median SMS-ICU was 13 (IQR 8-16) points and median SAPS 3 was 44 (IQR 36-51). Discrimination of SMS-ICU was good (AUC 0.817; 95% CI 0.814-0.819) but slightly lower than of SAPS 3 (AUC 0.845; 95% CI 0.843-0.848;). The customized SMS-ICU predictions were comparable to SAPS 3 in terms of calibration. Conclusion: In this external validation of the SMS-ICU in a large Brazilian cohort, we observed good discrimination of SMS-ICU and acceptable calibration after first-level customization. SMS-ICU can be used as a measure of illness severity for acutely admitted ICU patients in clinical studies.