Effect of a Machine Learning Algorithm to Guide Goal-Directed Therapy After Cardiac Surgery
Amanda Rea1, Alexandra Deasel2, Clifford Edwin Fonner3
1Amanda Rea is lead of advanced practice, clinical program manager, Division of Cardiac Surgery, University of Maryland St Joseph Medical Center, Towson, Maryland.
Background:
Goal-directed therapy allows clinicians to optimize perfusion and volume status in patients postoperatively.
Objective:
To evaluate the effect of a machine learning algorithm to guide postoperative goal-directed fluid therapy in cardiac surgery patients.
Methods:
A goal-directed fluid therapy program was implemented in a single center for coronary artery bypass patients with ejection fraction greater than or equal to 45% (implementation period: May 15, 2023, to May 31, 2024). Patient outcomes were compared with outcomes in matched historical control patients (control period: January 3 to October 31, 2022). The primary outcome was acute kidney injury.
Results:
A total of 479 eligible patients were evaluated (246 in the control group and 233 in the goal-directed therapy group). The incidence of acute kidney injury on postoperative day 2 (P = .01), on postoperative day 7(P = .02), and at discharge (P = .008) was lower in the goaldirected therapy group than in the control group.
Conclusions:
Patients in the goal-directed therapy program had a lower incidence of acute kidney injury compared with historical control patients. Incorporating a machine learning algorithm to guide goal-directed fluid therapy was a safe and less invasive way to monitor selected patients in the intensive care unit after cardiac surgery.