What’s the Numbers Needed to Treat (NNT) for Community Health and Wellbeing Workers (CHWWs)?

Commissioners often ask: “How many residents need to be supported by a Community Health and Wellbeing Worker for one person to benefit?”.
It’s a fair question. With pressure on services and tight budgets, systems need clear evidence that early-intervention models reduce demand.

CHWWs are local residents employed to support around 120–150 nearby households. They visit regularly, build long-term relationships, and help people with a mix of health, social and practical issues. Because they work upstream, they often uncover unmet needs long before people appear in clinical settings.

Why does NNT come up?

NNT (Numbers Needed to Treat) is a clinical concept that works well when a single intervention leads to a single outcome, for example, anticoagulation to prevent stroke. A very strong NNT might be 20–30 for anticoagulation in atrial fibrillation, or 50–100 for antihypertensives in high-risk hypertension.

What IS the NNT of CHWWs in Westminster based on real-world data?

Using cohort data from North West London ICB (see picture), CHWWs would support:

• 119 residents for one fewer non-elective admission
• 56 residents for one fewer ED attendance

These figures sit comfortably within commissioning ranges for accepted preventative interventions and reflect only a fraction of the effects CHWWs generate.

Why NNT undervalues CHWWs

NNT is built for single, tightly defined pathways. It assumes one exposure, one mechanism and one outcome. CHWWs don’t operate in that way.
CHWWs work across whole households and multiple determinants of health at once. One relationship influences many outcomes: emergency admissions, ED attendances, GP demand, medication adherence, social stability, mental health and more. They are multi-component, relational and cross-sectoral. Benefits are cumulative and often indirect. For example, a CHWW supporting a grandmother to get her cataract fixed may be the enabling step that allows her daughter to return to work and stabilise the household, but no clinical NNT captures that chain.

Prevention typically has higher NNTs than acute care, and models like CHWWs are penalised by design.

But lets go with this concept for a moment: What if CHWWs were a “poly-pill”?

If a CHWW were a pill given to a high-risk population (such as the most deprived 20%), then over five years a typical CHWW caseload might identify:

• ~10 previously undiagnosed hypertensives changing their lifestyle and/or starting medication
• ~5 people starting or rescuing statin therapy
• ~1 additional case of silent atrial fibrillation detected and anticoagulated.

Applying standard trial effect sizes, the combined impact of just these three channels results in an NNT of around 45–50 high-risk people over five years to prevent one major cardiovascular event.

That is essentially the same range as widely accepted primary-prevention drugs and excludes the gains in admissions, mental health, stability and long-term condition control that CHWWs deliver on top.

Enhanced adherence

Detecting risk and prescribing a drug however is only useful if the medication is taken correctly. This is where CHWWs work their magic: CHWWs improve adherence because they understand the household context and build trust. In deprived populations, baseline risk is not limited to a clinical subgroup, the whole population behaves like a high-risk cohort. That is why models acting across multiple pathways at once generate meaningful reductions in demand.

What about formal NNT evidence?

Randomised trials for community-based prevention are rare but we have one underway in Westminster. The cluster randomised controlled trial of CHWWs will report next summer and will be able to generate a formal NNT if required.

But is NNT the measure that matters?

For relational, place-based, whole-household models, measures like connection, trust, activation and quality of care may tell us far more than a single-event NNT ever could.

If CHWWs were a pill, most of us would be on it.