TAEIR: Trust-Aware Edge Intelligence for Proactive ICU Risk Management

Intensive-care monitoring remains dominated by single-parameter threshold alarms, of which the large majority are non-actionable and which contribute to alarm fatigue and delayed response. Machine-learning early-warning systems trained on retrospective ICU databases have demonstrated strong discrimination for sepsis and circulatory failure, yet independent evaluations of widely deployed proprietary models show that discrimination in one cohort does not translate into calibrated, trustworthy alerts in another. Building on the authors' earlier IoT/Edge-AI Smart ICU Monitoring System, this paper proposes TAEIR (Trust-Aware Edge Intelligence for ICU Risk), a nine-layer framework addressing predictive uncertainty, calibration, explanation, drift, and the governance mechanisms clinicians need in order to trust an alert. This is a framework, methodology, and benchmark-design paper reporting no experimental results, hardware measurements, or clinical outcomes.

• Rakesh Kumar Agrawal

Preprints.org

Preprint

Agrawal, R. K. (2026). TAEIR: Trust-Aware Edge Intelligence for Proactive ICU Risk Management. Preprints.org. https://doi.org/10.20944/preprints202609.1303.v1

@article{agrawal2026taeir, title={TAEIR: Trust-Aware Edge Intelligence for Proactive ICU Risk Management}, author={Agrawal, Rakesh Kumar}, year={2026}, howpublished={Preprints.org}, doi={10.20944/preprints202609.1303.v1}}