The occult hypoperfusion index: a bedside formula to identify undetected critical illness

BACKGROUND: The prehospital modified Sequential Organ Failure Assessment (mSOFA) score identifies patients with organ dysfunction but cannot detect normotensive occult hypoperfusion. To characterize the physiological correlation structure of mSOFA misclassification groups and to derive, internally validate and externally validate a bedside Occult Hypoperfusion Index (OHI) targeting undetected critical illness in patients classified as low acuity by the mSOFA. METHODS: Prospective multicenter cohort study. Consecutive adults with acute illness attended by three physician-led emergency medical services. Prehospital mSOFA score computed from five point-of-care variables (mean arterial pressure, Glasgow Coma Scale [GCS], SpO₂/FiO₂ ratio, creatinine, and lactate). Primary outcome was all-cause mortality within 48 h of the index event. The index was then applied without refitting to an independent prospective cohort from Graz, Austria. RESULTS: Of 16,106 enrolled patients, 933 (5.8%) died within 48 h. Cross-tabulation of mSOFA threshold and mortality yielded 848 true positives (5.3%), 85 false negatives (FN, 0.5%), 2,136 false positives (13.3%), and 13,037 true negatives (80.9%). Correlation analysis revealed that the FN group was uniquely sparse, with a dominant lactate-GCS coupling of ρ=+0.616 (p<.001). This co-variation was formalized as OHI=lactate × (16 - GCS). Applied to the mSOFA < 5 subgroup (n = 13,122), the OHI achieved an area under the receiver operating characteristic curve of 0.851 (95% CI, 0.816-0.885), outperforming lactate alone (0.755; DeLong p = .004) and GCS alone (0.630; DeLong p < .001). In the external cohort (n = 802; 67 deaths), the OHI at the pre-specified threshold achieved an area under the curve of 0.861 (95% CI, 0.699-0.990), with sensitivity 80.0% and specificity 74.8%, and showed no evidence of miscalibration. CONCLUSIONS: In a large prospective prehospital cohort, a multiplicative index derived from correlation analysis of mSOFA misclassification identified patients who died within 48 h despite being classified as low acuity. These findings are hypothesis-generating: the index was derived in one health system and externally validated in another, but the external estimates rest on few events and prospective evaluation before clinical use remains necessary. Consistent discrimination across three geographically distinct EMS systems in a leave-one-region-out analysis supports, but does not establish, its generalisability; this has now been extended to an independent cohort from a different country.