Deconstructing the Modified Sequential Organ Failure Assessment (mSOFA) Score: A Decision Tree Analysis Reveals the Pathophysiological Hierarchy of Organ Dysfunction Predictors in Prehospital Critical Care.

BACKGROUND: Early recognition of multiple organ dysfunction syndrome (MODS) in the prehospital setting remains challenging and is critical for reducing mortality among critically ill patients. The modified Sequential Organ Failure Assessment (mSOFA) score was developed for prehospital use with point-of-care testing; however, the internal decision architecture of this score remains unexplored. METHODS: A prospective, multicentre, observational study was conducted across three emergency medical systems (EMS) in Spain, included adults (n = 12,212) with acute illness transferred to emergency departments with high priority. A decision tree (DT) model was developed using the five components of the mSOFA score, lactate, Glasgow Coma Scale (GCS), oxygen saturation-to-fraction of inspired oxygen ratio (SaFi), mean blood pressure (MBP), and creatinine, to predict in-hospital mortality at 2 days. The study followed the STROBE and TRIPOD+AI guidelines. RESULTS: Cohort mortality was 5.0% (n = 609). Lactate emerged as the primary decision variable, with a threshold of 3.8 mmol/L, followed by SaFi, GCS, and creatinine in patients with hyperlactataemia, and by GCS and creatinine when lactate was below the threshold. CONCLUSIONS: The DT deconstructs the mSOFA from a static score into a dynamic, interpretable clinical decision framework. By revealing the pathophysiological hierarchy underlying the score, with lactate as the central stratification axis, this approach enables EMS clinicians to identify distinct clinical phenotypes and anticipate deterioration based on the sequential evaluation of organ dysfunction rather than a single numerical threshold.