A year of community health worker support did not prevent diabetes in prediabetes

Design

Pragmatic cluster-randomised controlled trial

Population

Adults aged 18–75 with HbA1c 5.7–6.4% in two New York primary care systems

Primary outcome

Incident type 2 diabetes up to 6 months after a 12-month programme

Effect

5.9% vs 4.3%; risk difference 1.6% (95% CI −3.68 to 4.10)

A pragmatic cluster-randomised trial in New York primary care, published in PLOS One on 30 October, allocated care teams to a 12-month community health worker programme or usual care for adults aged 18 to 75 with HbA1c 5.7 to 6.4%. The programme used health action plans, motivational interviewing, education and community referrals.

Incident diabetes was 5.9% with the programme against 4.3% with usual care, a difference that was not significant. HbA1c, weight, blood pressure and lipids did not differ. The programme did increase primary care visits (78% vs 63%) and, less welcome, emergency department visits (35% vs 29%).

The evidence that prevents diabetes in prediabetes comes from structured, intensive lifestyle programmes with weight-loss targets, and from metformin in selected people. A lighter, engagement-focused model did not reproduce that effect. When referring someone with prediabetes, choose a programme with a defined weight and activity target over general support.

This is relevant to India, where community health worker models are widely used for non-communicable disease. Engagement is a means, not the outcome.

• Refer people with prediabetes to a structured lifestyle programme with explicit weight-loss and activity targets.
• Do not assume more clinic contact will lower progression to diabetes on its own.
• Consider metformin for higher-risk prediabetes where guidelines support it, for example a BMI of 35 or more or previous gestational diabetes.
• Recheck HbA1c yearly in prediabetes whatever programme is offered.

Why it matters

It separates engagement from effect, which matters for anyone designing or relying on community programmes.

Don't overread it

Follow-up was short, only six months after the programme ended, so a later benefit cannot be excluded.

The statistics, in plain English

The risk difference was 1.6 percentage points, with a 95% confidence interval from −3.7 to +4.1. The interval crosses zero, so the trial cannot tell whether the programme helped, harmed or did nothing; it can say a large benefit is unlikely.

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.