Become an AAHPM Member to view PC-FACS
Artificial Intelligence for Better Goals of Care Documentation
Design and Participants: This retrospective study among adults admitted to an urban academic medical center assessed how an initiative to increase goals-of-care documentation rates is associated with goals-of-care documentation by race. The initiative embedded a mortality prediction score in the electronic medical record (EMR) to nudge clinicians to document goals-of-care conversations for patients with more than a 30% risk of 90-day mortality predicted by AI; patients with more than a 60% risk also received an EMR nudge to consult specialty palliative care to aid with goals-of-care conversations.
Results: Patients (N=3,643) were aged median 72 years (SD=13) and 87% White, with 42% admitted to an intensive care unit (ICU) and 15% dying during admission. Documentation was completed for 28%. By race, goals of care were documented for 30% Black, 28% White, and 24% other patients (P=.3933). There was no documentation rate difference among races over time.
Machine Learning for Targeted Advance Care Planning in Cancer Patients: A Quality Improvement Study
Design and Participants: This study implemented a mortality prediction algorithm for patients admitted from the ED to a dedicated solid malignancy unit at Duke University Hospital. Clinicians received an email when patients were identified as high-risk. ACP documentation and end-of-life care outcomes were compared before and after the notification intervention.
Results: Patients (n=88 pre- and n=77 postintervention) were aged median 66 years (range=28-92), and were 56% White, 38% Black, and 3% Hispanic/Latino. ACP conversations were documented for 2.3% of hospitalizations preintervention vs 81% postintervention (P<.001), and if the attending physician notified was a palliative care specialist (4.1% vs 85%) or oncologist (0% vs 76%) (P<.001). There were no between-group differences in length of stay, hospice referral, code status change, ICU admissions or length of stay, 30-day readmissions, 30-day ED visits, and inpatient and 30-day deaths.
Commentary: Two innovative quality improvement studies leveraged AI-driven mortality prediction tools to nudge ACP discussions and EMR documentation. In both studies, ACP documentation was strikingly infrequent prior to implementing the nudge approach, improving postintervention. Yet, patterns of end-of-life care did not substantially change. Findings suggest that AI can be harnessed to identify patients for whom ACP discussions should be prioritized and increase ACP documentation. However, content, cadence, and quality of the ACP discussions are likely variable, contributing to the lack of impact on distal outcomes. Next steps might entail characterizing the quality of these crucial discussions, lending greater insight into why end-of-life outcomes remain unchanged and how we might intervene to ensure goal-concordant end-of-life care.
Bottom Line: AI tools offer great promise in optimizing delivery and documentation of goals-of-care and ACP discussions. Yet, holding and documenting ACP discussions may be insufficient to impact downstream outcomes and ensure goal-concordant end-of-life care.
Reviewer: Prasanna Ananth, MD MPH, Yale School of Medicine, New Haven, CT
References:
1. Waldrop DP, McGinley JM. “I want to go home”: How location at death influences caregiver well-being in bereavement. Palliat Support Care. 2020;18:691-698.
2. Lenko R, Voepel-Lewis T, Robinson-Lane SG, et al. Racial and ethnic differences in informal and formal advance care planning among U.S. older adults. J Aging Health. 2022;34:1281-1290.
3.Huang IA, Neuhaus JM, Chiong W. Racial and ethnic differences in advance directive possession: role of demographic factors, religious affiliation, and personal health values in a national survey of older adults. J Palliat Med. 2016;19:149-156.
Sources:
Piscitello G, Schell JO, Arnold RM, Schenker Y. Artificial intelligence for better goals of care documentation. BMJ Support Palliat Care. 2024;spcare-2023-004657. doi:10.1136/spcare-2023-004657.
Access this article on PubMed.
Patel MN, Mara A, Acker Y, et al. Machine learning for targeted advance care planning in cancer patients: a quality improvement study. J Pain Symptom Manage. 2024;68(6):539-547. doi:10.1016/j.jpainsymman.2024.08.036.
Access this article on PubMed.