Impact in an age of constraint: why neighbourhood health needs better evaluation, not just better intentions

Neighbourhood health is quickly becoming one of the defining ideas in the future of the NHS. The ambition certainly sounds easy to support, even if many aspects aren't exactly new: more care closer to home, better integration between services, greater emphasis on prevention, and a more proactive offer for people with complex needs. It speaks directly to the three 'left shifts' set out in the 10 Year Health Plan for England.

But as neighbourhood health moves from policy ambition into practical implementation, a harder question comes into view: how will we know whether it is working? And if it does work, will it work everywhere?

This is not just a technical question for analysts. It is a strategic question for systems, providers, commissioners and the communities they serve. Neighbourhood health is being developed in a context of intense financial and operational pressure. The NHS is being asked to recover access, improve productivity, reduce avoidable demand, tackle inequalities, and shift activity out of hospital settings. At the same time, local systems are being asked to build new models of care across organisational boundaries, often with limited new resources and significant variation in starting points.

In that context, impact evaluation cannot be an afterthought. Nor can it be reduced to a narrow exercise in counting activity.

If neighbourhood health is to become more than another failed flagship, systems will need to understand not only whether new models are being delivered, but whether they are creating meaningful value for patients, staff, communities and the wider health and care system.

Measuring what matters, not just what moves

One of the biggest risks in evaluating neighbourhood health is that we measure what is easiest, rather than what matters most. It is relatively straightforward to count the number of multidisciplinary team meetings held, care plans completed, people referred into a neighbourhood service, appointments delivered in community settings, or our old friend, emergency admissions. These are useful measures. They tell us something about mobilisation, reach and delivery. But they do not, on their own, tell us whether care has actually improved.

A neighbourhood model might generate a large volume of contacts without reducing fragmentation. It might move activity from one part of the system to another without improving patient outcomes. It might help identify unmet need without having the capacity to respond to it. Or it might improve patient experience, staff collaboration and build the kind of trusted relationships that underpin real and sustainable change in ways that are not immediately visible in headline activity data.

This is why impact evaluation needs to start with a clear theory of change. What is the model expected to do? Which groups of people is it designed to benefit? What mechanisms are expected to drive change? Is the intended impact fewer hospital admissions, better supported self-management, earlier intervention, improved access, reduced duplication, better staff experience, or all of the above?

Neighbourhood health is not a single intervention. It is a way of organising care. That makes evaluation more complex, but also more important.

The value question

In the current climate, it is inevitable that neighbourhood health will be judged partly through a value for money lens. That should not be seen as a threat. Done well, value for money analysis can help make the case for sustained investment in prevention, integration and community-based care.

If the only question asked is whether neighbourhood health rapidly reduces hospital activity, many promising models may appear to disappoint. Some benefits take time to emerge. Some interventions may initially increase demand by uncovering unmet need. Some may improve quality, safety or experience without generating that holy grail, immediate cash-releasing savings.

A more mature approach would ask: what value is being created, for whom, over what time period, and at what cost? Value for money is not the same as 'cheap'. It should mean that resources are being used in ways that generate meaningful, equitable and sustainable benefit, even if that benefit is not immediately cash-releasing.

Getting serious about attribution

Another tricky challenge is attribution. Neighbourhood health models are being introduced into complex systems where many other changes are happening at the same time. Waiting list initiatives, urgent care pressures, workforce changes, digital transformation, winter pressures and local improvement programmes may all affect the same outcomes.

A simple before-and-after analysis will often be inadequate. If emergency admissions fall after a neighbourhood model is introduced, was that because of the intervention? Or because of wider trends, coding changes, seasonal variation, changes in hospital thresholds, or other local programmes? Equally, if admissions do not fall, does that mean the model failed? Or did it prevent an increase that would otherwise have happened?

This is where impact evaluation needs to bring stronger counterfactual thinking into routine practice. In some cases, it may be possible to compare neighbourhoods that receive the intervention earlier with those that receive it later. In others, matched comparator populations may be used. For some outcomes, interrupted time series analysis may help distinguish a change associated with implementation from pre-existing trends. Difference-in-differences approaches may be useful where comparable areas are available and the timing of implementation varies.

These methods are not magic. They rely on assumptions and need careful interpretation. But they are usually stronger than simply comparing performance before and after implementation.

For neighbourhood health, the question is not whether every local evaluation can be a perfect academic study. It cannot. The question is whether systems can move towards more credible, transparent and decision-useful analysis.

Implementation is the intervention

There is sometimes a temptation to separate 'impact' from 'implementation', as if the former is the serious analytical question and the latter is a softer add-on. For neighbourhood health, that distinction does not hold.

Implementation is the intervention.

The success of neighbourhood health depends on relationships, trust, data sharing, shared governance, workforce models, referral routes, clinical leadership, community engagement and the ability to work across organisational boundaries. The NHS England neighbourhood health guidelines emphasise the role of integrated working between ICBs, local authorities and providers, as well as the need to progress neighbourhood models ahead of wider reform.

If those enabling conditions are weak, the model may not deliver its intended impact. If they are strong, outcomes may improve even before major structural changes are visible in activity data.

This means evaluation needs to combine quantitative and qualitative methods. Routine data can show patterns in activity, outcomes and utilisation. Interviews, focus groups, observation and case studies can help explain why those patterns are emerging. They can also identify whether the model is working differently for different groups of patients, staff or communities.

This matters because neighbourhood health is, at it's heart, relational. It is about how services connect around people, and that needs an evaluation approach that reflects the 360-degree nature of the offer.

What this means for data professionals

This all has important implications for data professionals. Neighbourhood health will not be evaluated well by analysts working in isolation, however technically skilled they are.

The strongest evaluation teams will be multidisciplinary, bringing together quantitative analysts, qualitative researchers, clinicians, operational leads, health economists, population health specialists and people with direct experience of services.

Data professionals have a vital role to play in making sense of routine datasets, identifying variation, testing assumptions and tracking outcomes over time. But their work needs to be grounded in an understanding of how care is actually delivered, how patients move through local systems, and what staff and communities experience on the ground.

In this context, the data professional's role shifts from reporting performance to supporting learning, helping neighbourhood teams understand not only what is changing, but why, for whom, and, critically, what should happen next.

Building learning into neighbourhood health

Neighbourhood health is a compelling direction of travel. It reflects many of the things that patients, staff and communities have been asking for: more joined-up care, more support closer to home, and a stronger focus on prevention rather than crisis response.

But good intentions will not be enough. In an age of constraint, neighbourhood health will need to demonstrate not just activity, but impact; not just ambition, but value; not just local innovation, but learning that can be used to improve decisions across the system.

That requires evaluation that is rigorous but pragmatic. It needs to use routine data intelligently, bring stronger counterfactual thinking into real-world settings, capture implementation honestly and ultimately broaden the analytical skills required to analyse and interpret quantitative, qualitative and economic data.

The future of neighbourhood health will not be determined only by the models that are designed, but by how well systems learn from them.

Impact evaluation should not sit at the end of that process. It should be part of how neighbourhood health is built.