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

OBJECTIVES: Cardiac surgery is associated with perioperative complications, some of which might be attributable to hypotension. The Hypotension Prediction Index (HPI), a machine-learning-derived early warning tool for hypotension, has only been evaluated in noncardiac surgery. We investigated whether using HPI with diagnostic guidance reduced hypotension during cardiac surgery and in the ICU. DESIGN: Randomized clinical trial conducted between May 2021 and February 2023. SETTING: Single-center study conducted in an academic hospital in the Netherlands. PATIENTS: Adults undergoing elective on-pump coronary artery bypass grafting, with or without single heart valve surgery, were enrolled if a mean arterial pressure (MAP) greater than or equal to 65 mm Hg was targeted during the surgical off-pump phases and ICU stay. After eligibility assessment, 142 of 162 patients approached gave informed consent for participation. INTERVENTIONS: Patients randomized 1:1 received either diagnostic guidance in addition to standard care if HPI reached greater than or equal to 75 ( n = 72) or standard care alone ( n = 70). MEASUREMENTS AND MAIN RESULTS: The primary outcome was the severity of hypotension, measured as time-weighted average (TWA) of MAP less than 65 mm Hg. Secondary outcomes encompassed hypertension severity and intervention disparities. Of 142 patients randomized, 130 were included in the primary analysis. The HPI group showed 63% reduction in median TWA of hypotension compared with the standard care group, with a median of differences of -0.40 mm Hg (95% CI, -0.65 to -0.27; p < 0.001). In the HPI group, patients spent a median 28 minutes (95% CI, 17-44 min) less in hypotension, with a measurement duration of 322 minutes in the HPI group and 333 minutes in the standard care group. No significant differences were observed in hypertension severity, treatment choice, or fluid, vasopressors, and inotrope amounts. CONCLUSIONS: Using HPI combined with diagnostic guidance on top of standard care significantly decreased hypotension severity in elective cardiac surgery patients compared with standard care.