PRISM: multimodal clinical prognosis depends heavily on the data available prior to arrival

PRISM combines measurements, ECG data and population-based data. The study distinguishes its best scores from those obtained using the same input window as the systems under comparison.

PRISM brings together structured measurements, successive ECGs and similar patient representations for the clinical prognosis. The search for neighbours does not place any individual records found in the prompt: the system receives aggregated representations.

The authors emphasize that their best scores use 96 hours of additional pre-arrival history, which are absent from the published references. At a comparable window, the AUROC reached 0.8895, compared to 0.8937 for UniPACT. In an external cohort, supervised retraining improved transfer. These comparisons remind us that a medical score cannot be interpreted without specifying the data provided; no clinical autonomy is established.

Preprint filed on arXiv on October 9, 2026; results are those reported by the authors and remain to be independently confirmed.