Design: Cluster randomised trial, 10 clinics
Population: 791 Black and White adults aged 65 or over with serious illness
Primary outcome: Advance care planning completed within 12 months
Effect: Black: 71% vs 62%, RR 1.15 (0.83-1.59); White: 68% vs 78%, RR 0.88 (0.65-1.18)

EQUAL ACP was a cluster randomised trial in 10 US primary care and geriatrics clinics, enrolling 791 Black and White adults aged 65 or over with serious illness. Clinics were randomised to a facilitated advance care planning programme or a self-guided written one, both delivered by trained lay facilitators.

Within one year, 69% overall completed some form of advance care planning, mostly informal conversations with family; only 130 of 789 had a new formal document or order. Within each racial group the two approaches did not differ significantly: among Black participants 71% vs 62% (RR 1.15, 95% CI 0.83 to 1.59), among White participants 68% vs 78% (RR 0.88). The facilitated approach did relatively better among Black than White participants (interaction RR 1.32). Matching facilitator and participant by race made no difference.

For clinicians outside the USA, the transferable finding is that lay-led programmes, whether facilitated or self-guided, got most seriously ill older patients talking with their families. Turning those talks into a recorded plan is the step that mostly did not happen.

Lay facilitators can deliver advance care planning; it does not have to wait for a doctor's appointment.
Most completion was informal family discussion; ask whether preferences have been written down or recorded.
Record any stated preferences in the notes, since that is where formal planning most often failed.
One approach did not suit every group equally; offer a facilitated conversation to patients who struggle with written materials.

Why it matters

Families talked, but the plan rarely reached the medical record, which is where it is needed in a crisis.

The statistics, in plain English

Both within-group risk ratios have confidence intervals that cross 1, so neither approach was shown to be better. The interaction term is significant, but interactions are secondary findings and should be read as a signal for further study.