Artificial Intelligence-Generated Daily Care Plans for Family Members of Cardiovascular ICU Patients AI-Generated Family Care Plans in the CVICU

Background Family members of patients in the cardiovascular intensive care unit (CVICU) have a critical need for timely, accessible, and accurate information regarding their loved one's care. However, communication between care teams and families remains inconsistent and represents a persistent unmet need. This study assessed the feasibility and acceptability of AI-generated daily written care plans for family members of CVICU patients. Methods This pilot quality improvement project used a phased design alternating intervention and control periods among family members of CVICU patients at an academic tertiary care center. During intervention phases, participating families received physician-reviewed AI-generated daily care plans, generated from audio recordings of bedside rounding discussions or the team's written medical note, by email. Feasibility was assessed by care plan accuracy, care plan generation time, and physician review time. Acceptability was measured by family-reported usefulness and physician feedback. Results A total of 494 care plans were generated for 142 intervention families (236 total participants). Mean generation time was 31.9±9.4 seconds and physician review time was 23.9±9.0 seconds. Only 6.7% of plans required editing, with edits decreasing significantly over time (p=0.001). Plans were generated in English (65.5%), French (30.3%), and other languages (4.2%). Among intervention families with follow-up data available, 92.9% found the plans useful, and physician acceptability was high across all domains. Conclusions AI-generated daily care plans for CVICU patients' families were feasible to implement and perceived as useful by both families and physicians. Further studies are needed to evaluate their impact on communication quality and patient- and family-centered outcomes.