An EHR prompt took self-management referrals from 1% to 13%

Design

cluster-randomised controlled trial, 18 months, multilevel logistic regression

Population

eligible adults with type 2 diabetes attending 4 intervention and 2 control primary care practices in central North Carolina

Primary outcome

referral rate to diabetes self-management education and support

Effect

12.9% vs 1.2% at 12 months (OR 10.55, 95% CI 4.65-23.94); 13.7% vs 1.6% at 18 months (OR 9.44, 4.91-18.17)

The PROMPT study, a cluster-randomised trial across six primary care practices in central North Carolina, built a best-practice advisory into the electronic health record that fired when an eligible patient with type 2 diabetes attended. Four practices ran it for 18 months; two served as controls. At 12 months the referral rate to diabetes self-management education and support was 12.9% in intervention clinics against 1.2% in controls (P < 0.001). Adjusted for patient and clinic characteristics, the odds ratio was 10.55 (95% CI 4.65-23.94).

At 12 months standing orders were added so nurses and medical assistants could act on the advisory before the patient saw the clinician. At 18 months the rate was 13.7% against 1.6%, odds ratio 9.44 (4.91-18.17) — the gain held rather than decaying, which is not what usually happens to an alert left running for a year and a half.

The finding worth carrying is the one about who acts on the prompt. Alerts aimed at a clinician mid-consultation compete with everything else in the visit; the same alert routed to the person doing the vitals, with a standing order behind it, becomes a task rather than an interruption. That is transferable to any referral pathway that is quietly under-used — retinal screening, foot assessment, structured education — and it does not need new software, only a decision about whose screen the prompt lands on.

• Ask which of your referral pathways runs at low single-figure uptake — that is where a prompt pays
• Route the prompt to nursing or assistant staff with a standing order, not only to the prescriber
• Pair any new alert with a retirement date and a measure, or it becomes background noise
• A referral made is not education attended; measure the downstream attendance too

Why it matters

The barrier to structured education turns out to be the referral step, not patient willingness — and the referral step is fixable this month.

Don't overread it

Six clinics in one region, and the outcome is referrals made rather than education completed or glycaemia improved.

The statistics, in plain English

An odds ratio of 10.55 sounds enormous because the baseline is tiny — going from 1.2% to 12.9% is about 12 extra referrals per 100 eligible visits, which is the number that matters operationally. The wide confidence interval (4.65-23.94) reflects only six clinics: the direction is secure, the size much less so.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.