Effect of a Machine Learning-Derived Early Warning Tool With Treatment Protocol on Hypotension During Cardiac Surgery and ICU Stay: The Hypotension Prediction 2 (HYPE-2) Randomized Clinical Trial

Schuurmans, J.

doi:10.1097/CCM.0000000000006518

Schuurmans HPI (cardiac surgery)

Effect of a Machine Learning-Derived Early Warning Tool With Treatment Protocol on Hypotension During Cardiac Surgery and ICU Stay: The Hypotension Prediction 2 (HYPE-2) Randomized Clinical Trial

Patient / Population Intervention / Exposure Comparison Outcome

In 142 adults undergoing elective coronary artery bypass grafting, Hypotension Prediction Index-triggered diagnostic guidance reduced severity of hypotension during surgery and ICU stay compared with standard care alone.

N

142patients

Design

Single-centre RCT, Netherlands

Endpoint

Time-weighted average of MAP <65 mmHg (severity of hypotension)

Relevance

3Supporting — context rather than must-know.

Result63% reduction in median time-weighted average hypotension; median of differences −0.40 mmHg (95% CI −0.65 to −0.27; P<0.001); a median 28 min (95% CI 17–44) less time in hypotension. No difference in hypertension severity or in fluid, vasopressor or inotrope amounts. 130 of 142 in the primary analysis.

Schuurmans J, et al. Crit Care Med. 2025;53(2):e328-e340. 10.1097/CCM.0000000000006518

Discussion & critique

Extends the HYPE result into cardiac surgery and the ICU, where hypotension is frequent and the index had never been tested. Single-centre and open-label, with a surrogate endpoint, hypotension measured only during off-pump phases and ICU stay, and 12 of 142 excluded from the primary analysis; whether fewer minutes of hypotension translate into fewer complications remains unanswered.