Advanced artificial intelligence-guided hemodynamic management within cardiac enhanced recovery after surgery pathways: A multi-institution review.

ABSTRACT

Objective: The study objective was to report early outcomes of integrating Hypotension Prediction Index-guided hemodynamic management within a cardiac enhanced recovery pathway on total initial ventilation hours and length of stay in the intensive care unit. Methods: A multicenter, historical control, observational analysis of implementation of a hemodynamic management tool within enhanced recovery pathways was conducted by identifying cardiac surgery cases from 3 sites during 2 time periods, August 1 to December 31, 2019 (preprogram), and April 1 to August 31, 2021 (program). Reoperations, emergency (salvage), or cases requiring mechanical assist were excluded. Data were extracted from electronic medical records and chart reviews. Two primary outcome variables were length of stay in the intensive care unit (using Society of Thoracic Surgeons definitions) and acute kidney injury (using modified Kidney Disease Improving Global Outcomes criteria). One secondary outcome variable, total initial ventilation hours, used Society of Thoracic Surgeons definitions. Differences in length of stay in the intensive care unit and total ventilation time were analyzed using Kruskal-Wallis and stepwise multiple linear regression. Acute kidney injury stage used chi-square and stepwise cumulative logistic regression. Results: A total of 1404 cases (795 preprogram; 609 program) were identified. Overall reductions of 6.8 and 4.4 hours in intensive care unit length of stay (P = .08) and ventilation time (P = .03) were found, respectively. No significant association between proportion of patients identified with acute kidney injury by stage and period was found. Conclusions: Adding artificial intelligence-guided hemodynamic management to cardiac enhanced recovery pathways resulted in associated reduced time in the intensive care unit for patients undergoing nonemergency cardiac surgery across institutions in a real-world setting.