Multicenter International Cohort Validation of a Modified Sequential Organ Failure Assessment Score Using the Richmond Agitation-sedation Scale

Abstract

Objective: In a multicenter, international cohort, we aimed to validate a modified Sequential Organ Failure Assessment (mSOFA) using the Richmond Agitation-Sedation Scale, hypothesized as comparable to the Glasgow Coma Scale (GCS)-based Sequential Organ Failure Assessment (SOFA).

Summary background data: The SOFA score, whose neurologic component is based on the GCS, can predict intensive care unit (ICU) mortality. But, GCS is often missing in lieu of other assessments, such as the also reliable and validated Richmond Agitation Sedation Scale (RASS). Single-center data suggested an RASS-based SOFA (mSOFA) predicted ICU mortality.

Methods: Our nested cohort within the prospective 2016 Fourth International Study of Mechanical Ventilation contains 4120 ventilated patients with daily RASS and GCS assessments (20,023 patient-days, 32 countries). We estimated GCS from RASS via a proportional odds model without adjustment. ICU mortality logistic regression models and c-statistics were constructed using SOFA (measured GCS) and mSOFA (measured RASS-estimated GCS), adjusted for age, sex, body-mass index, region (Europe, USA-Canada, Latin America, Africa, Asia, Australia-New Zealand), and postoperative status (medical/surgical).

Results: Cohort-wide, the mean SOFA=9.4+/-2.8 and mean mSOFA = 10.0+/-2.3, with ICU mortality = 31%. Mean SOFA and mSOFA similarly predicted ICU mortality (SOFA: AUC = 0.784, 95% CI = 0.769-0.799; mSOFA: AUC = 0.778, 95% CI = 0.763-0.793, P = 0.139). Across models, other predictors of mortality included higher age, female sex, medical patient, and African region (all P < 0.001).

Conclusions: We present the first SOFA modification with RASS in a "real-world" international cohort. Estimating GCS from RASS preserves predictive validity of SOFA to predict ICU mortality. Alternative neurologic measurements like RASS can be viably integrated into severity of illness scoring systems like SOFA.

FIGURE 1.

Graphical representation of development and validation of modified Sequential Organ Failure Assessment (mSOFA). The figure graphically represents the analysis utilized to develop and validate mSOFA. The analysis shown in black demonstrates the traditional method, including the Glasgow Coma Scale (GCS), to calculate the Sequential Organ Failure Assessment (SOFA), which is then used to predict mortality. The analysis shown in green shows the alternative method in the mSOFA, which uses the Richmond Agitation-Sedation Scale (RASS) to calculate GCS equivalents (GCS_eq) using a univariate proportional odds regression. GCS_eq are then used to calculate mSOFA. The ability of SOFA and mSOFA to predict ICU mortality is compared using respective multivariable logistic regression.

FIGURE 2.

Receiver operating characteristics (ROC) of Sequential Organ Failure Assessment (SOFA) score and modified Sequential Organ Failure Assessment (mSOFA) score for prediction of intensive care unit mortality. The figure represents the adjusted multivariable logistic regression model ROC for the prediction of ICU mortality by either the mean SOFA (black line) or the mean mSOFA (green line). Area under the curves (AUC) are presented with 95% confidence intervals (CI).