A Medical Embodiment of the Mena Dominance Law: Cross-Dataset ICU Mortality Modeling, Organ-Structured Deficit, and SOFA Comparison

This paper presents a medical embodiment study of the Mena Dominance Law (MDL) in critical care mortality modeling. The manuscript is intentionally narrower than the general MDL theory papers: it does not restate the law as a new universal claim, but evaluates whether MDL-derived representations carry clinically useful signal in intensive care unit (ICU) outcome prediction. The embodiment is studied across three major ICU resources-eICU, HiRID, and MIMIC-IV-together with cancer-focused subcohorts. The principal benchmark is SOFA, with emphasis on whether MDL-derived models, alone or in combination with SOFA, improve discrimination. Across the main strict six-organ cohorts, the combined MDL+SOFA models outperform SOFA-only models in every dataset. In eICU, strict six-organ AUC improves from 0.7533 for SOFA-only to 0.7954 for MDL-only and 0.8120 for MDL+SOFA. In HiRID, the corresponding values are 0.7759, 0.7805, and 0.8062. In MIMIC-IV, they are 0.6912, 0.7393, and 0.7562. Cancer subset analyses generally preserve the same directional pattern, although inference weakens materially in the smallest subgroup, the HiRID hematologic cancer-proxy strict organ-complete cohort. Beyond aggregate performance, the medical embodiment reveals structure that a pure summed severity score does not directly expose. Across datasets, non-survivors show both an upward shift in deficit burden and a redistribution of dominant-organ identity. In HiRID, survivors are most often respiratory-dominant, whereas deaths show relatively greater concentration in central nervous system, renal, and liver dominance. In MIMIC-IV, deaths are markedly more cardiovascular-dominant than survivors. These findings support the view that the embodiment captures structured multiorgan deterioration rather than undifferentiated severity alone. The strongest claim supported by the results is practical: in the declared critical-care embodiment studied here, MDL-derived features add clinically meaningful predictive information beyond SOFA and provide a more structured view of deficit organization across organ systems.