TrialsgrambetaRCTs are beautiful

Parikh algorithm-based palliative care

2025

Algorithm-Based Palliative Care in Patients With Cancer: A Cluster Randomized Clinical Trial

In 562 patients with advanced lung or gastrointestinal cancer, default palliative care orders with opt-out increased completed palliative care consultations within 12 weeks compared with peer-comparison reports with referral at clinician discretion.

Result

Efficacy: Completed palliative care consultation within 12 weeks of enrolmentaOR 8.9 (95% CI 5.5-14.6) · P < 0.001

Below 1 favours peer-comparison reports with referral at clinician discretion, above 1 favours default palliative care orders with opt-out.

Event rates
Absolute difference
35.6 points
NNT
3

Absolute difference and NNT are derived from the event rates.

Also reported: Among 179 decedents, systemic therapy within 14 days of death 6.5% vs 16.1% (aOR 0.3, 95% CI 0.1–0.7; P=0.05). No differences in quality of life, feeling heard and understood, or late hospice referral.

Original publication
Subtopic
Default orders
Design
Two-arm cluster RCT, 15 community oncology clinics, Tennessee, USA
Relevance
Important: one of several pillars.
Source
Parikh RB, et al. JAMA Netw Open. 2025;8(2):e2458576.doi 10.1001/jamanetworkopen.2024.58576

Discussion & critique

A scalable implementation strategy for guideline-recommended early palliative care in community oncology, where access lags: defaults with accountable justification lifted consultation rates from 8.3% to 43.9% and cut systemic therapy near death. The 'algorithm' is a guideline-derived EHR rule rather than a learned model, quality of life and hospice timing did not change, and the trial sits within a single oncology network. Shares its behavioural-economics lineage with Manz ML nudges (SIC).