Clinical heterogeneity in performance of a commercial prediction model for clinical deterioration - Takeaways - MDSpire
Coming Soon: Introducing MDSpire News. Learn more
Conexiant’s news site is now MDSpire News. Learn more

Performance of a Commercial Clinical Deterioration Prediction Model Varies Across Patient Care Settings

  • By

  • Sandra Taylor

  • Christine Williams

  • Sili Fan

  • Irene Cortes-Puch

  • Sarina Fazio

  • Anna Liu

  • Rachael Callcut

  • Bruce Hall

  • Jason Y Adams

  • October 13, 2026

Share

  • 1

    This retrospective study evaluated EDI performance across clinical services and hospital locations among adult inpatients at a Northern California academic hospital from 2021 to 2024.

  • 2

    The analysis included 77,478 encounters and 10,579 deterioration events, with an overall AUROC of 0.763 (95% CI, 0.759-0.768).

  • 3

    EDI discrimination was highest in hematology, oncology, and bone marrow transplant services (AUROC, 0.815) and lowest in cardiology (0.670) and neurosciences (0.677).

  • 4

    The study defined deterioration using intensive care transfer, rapid response team activation, comfort-care initiation, or death.

  • 5

    The authors reported substantial variation in predictive performance across clinical subpopulations and concluded that subgroup performance should be understood for inpatient model deployment.

Original Source(s)

Clinical heterogeneity in performance of a commercial prediction model for clinical deterioration

  • by Sandra Taylor, Christine Williams, Sili Fan, Irene Cortes-Puch, Sarina Fazio, Anna Liu, Rachael Callcut, Bruce Hall, Jason Y Adams

  • October 13, 2026

Related Content

Cumulative training load may signal musculoskeletal injury

Wearable sensor measures showed associations with musculoskeletal injury, but evidence certainty remained very low.

Timing and dose do not tell the same story: a reply to Wieruszewski and Khanna

Rethinking diaphragmatic ultrasound for AECOPD weaning prediction: considerations on parameter selection, thresholds, and outcome definitions