Journal: NPJ digital medicine
This publication is a systematic review and meta-analysis evaluating whether automated EHR-based algorithms improve palliative care delivery compared with usual care.
Study design and methods
- • Systematic search of PubMed, Embase, and Cochrane Library through December 2025.
- • Included randomized trials testing automated EHR algorithms designed to identify patients with high palliative care needs and trigger palliative involvement.
- • Seven trials comprising 125,666 patients were pooled.
- • Random-effects models were used to estimate risk ratios (RRs) and weighted mean differences with 95% CIs.
- • Risk of bias was assessed using Cochrane RoB 2.0, and certainty of evidence with GRADE.
Key outcomes
-
Palliative care consultations
- • Noncancer populations: RR 2.19 (95% CI, 1.12–4.28).
- • Cancer populations: RR 5.31 (95% CI, 3.49–8.09).
- • Automated algorithms substantially increased palliative care consult rates, particularly in oncology.
-
Advance care planning
- • Do-not-resuscitate documentation: RR 1.22 (95% CI, 1.17–1.28).
- • Algorithms meaningfully improved DNR documentation.
-
End-of-life utilization and outcomes
- • Hospice use and in-hospital mortality: only marginal improvements.
- • No significant effects on ICU admission, hospital length of stay, or family-reported psychological outcomes.
Clinical implications
- • Automated EHR-based identification tools reliably increase both palliative care consultations and advance care planning documentation.
- • Downstream effects on resource utilization, mortality, and family-centered outcomes are limited or neutral based on current evidence.
- • For oncology practice, integrating such algorithms may improve the timeliness and frequency of palliative involvement, but additional strategies are likely needed to translate this into more substantive changes in utilization and patient- or family-centered outcomes.