Care Team Structure & Evidence

Federal regulation (42 CFR §418.56) requires Medicare-certified hospices to provide care through an interdisciplinary group (IDG) that includes at minimum a physician, nurse, social worker, and pastoral or counseling services representative, with the patient and family treated as part of the team.

The Service Intensity Add-on (SIA; 42 CFR §418.302) provides additional Medicare payment for registered nurse and medical social worker visits in the last 7 days of life for patients on the Routine Home Care (RHC) level of care. This creates a financial incentive to increase skilled bedside presence at end of life.

CMS tracks two claims-based quality measures for hospice quality reporting: the Hospice Visits in the Last Days of Life (HVLDL) measure (whether patients received a skilled visit in the last 3 days) and the Hospice Care Index (HCI), a composite of ten care quality indicators. Both are publicly reported.

The Home Health Value-Based Purchasing (HHVBP) Model expanded nationwide in January 2023. Home health agencies receive payment adjustments of up to ±5% based on their Total Performance Score (TPS), incentivizing quality improvement across clinical and functional outcomes for Medicare home health patients.

Home health Conditions of Participation (42 CFR §484.55) require a comprehensive patient assessment at the start of each care episode, including functional status, medication management, and care coordination needs. This establishes the regulatory baseline for quality home-based palliative and post-acute care.

The SUPPORT trial (1995; 4,804 patients at five US teaching hospitals) tested giving physicians prognostic estimates plus a trained nurse who elicited preferences and supported communication. It found no improvement in communication, DNR timing, ICU or ventilator days, pain, or resource use. The result applies to that intervention and setting.

A meta-analysis of 43 randomized trials found palliative care was associated with better patient quality of life (SMD 0.46) and lower symptom burden at 1-3 months, with no association with survival (HR 0.90; 95% CI 0.69 to 1.17). In the 5 low-risk-of-bias trials the quality-of-life effect was smaller (SMD 0.20) and the symptom effect was not significant. The review did not test timing of referral.

In a retrospective comparison of 651 patients who died (82 in a home-based palliative care program, 569 receiving usual care) in one New York ACO, program enrollment was associated with lower costs in the last 3 months of life ($20,420 vs $32,420) and fewer hospital admissions. Enrollment was not randomized.

Workforce projections published in 2018 indicated a growing mismatch between the supply of hospice and palliative medicine specialist physicians and projected demand. The gap is expected to widen as the aging population increases need for specialist palliative care services.

In a single-center study of 2,031 patients with cancer who died 2018–2022, more comprehensive goals-of-care documentation was associated with less aggressive end-of-life care: 73.2% received aggressive care when none of eight GOC components were documented, versus 50.3% when seven were.

The most documented goals-of-care component was family involvement (75.0%); the least was fears (21.1%); only 5.4% of patients had all eight components documented.

In the CAPC Serious Illness Scorecard Michigan State Report, Michigan hospitals reporting palliative care by bed size: 67.4% (<50), 68.4% (50–149), 86.7% (150–299), 92.6% (300+); an estimated 92.1% of Michigan inpatients may have access to hospital-based palliative care.

Michigan has 286 certified palliative-care prescribers (2.9 per 100,000), no standalone Medicaid palliative benefit, and no legislatively established palliative care advisory council; it ranks 17/51 (C-TAC ACT Index) and 31/51 (AARP LTSS Scorecard).

A 2025 scoping review found 44 measures for goal-concordant care. No agreed criterion standard; measures vary in construct and validation.

An EHR-notes method for identifying goal-concordant care achieved strong inter-rater agreement (κ=0.92).

CANHELP is validated as a measure of patient and family satisfaction with end-of-life care, not as a measure of goal concordance.

In the ACP PEACE stepped-wedge cluster RCT (NCT03609177), a communication intervention increased advance-care-planning documentation by 6.8 percentage points (25.3% vs 20.8%; P<.001, primary outcome) but did not significantly change palliative care or hospice enrollment.

In a single-institution retrospective study (N=60 patients, Dana-Farber), GPT-4o identified advance-care-planning documentation in EHRs with sensitivity 0.85–1.0 and specificity 0.80–0.91 across ACP domains; hallucination index was low (<0.5). Limitations include small sample size and single-site design.

In a structured ethical stress test of 14 large language models on hospice and medical dilemmas, 7 of 14 endorsed ethically impermissible actions; 6.3% of outputs endorsed involuntary hospice transfer. Authors note these findings derive from a forced-choice stress test and do not generalize to routine clinical use.

A 2026 scoping review of 42 studies on AI in advance care planning found that use of generative AI is 'limited,' rule-based NLP dominates, and no randomized controlled trials of AI-delivered ACP conversations with patients have been published.

A 2026 scoping review of 15 studies on AI in palliative care characterized the evidence as 'geographically skewed, methodologically immature, and ethically underdeveloped'; only 2 of 15 studies measured quality of life directly.

Clinicians from MSK, MGH, Mayo Clinic, and Johns Hopkins published interdisciplinary practical guidance on AI for palliative care clinicians in 2026; no formal position statement from AAHPM, NHPCO, ASCO, or NCI on AI in palliative care has been identified in 2024–2026.