A footprint decides who is inside the service

Every neighbourhood health plan begins with a boundary, and a boundary is a decision about inclusion before it is anything else. The Neighbourhood Health Framework asks integrated care boards and health and wellbeing boards to "agree neighbourhood footprints around natural communities for the future development of INTs" during 2026 to 2027 (Neighbourhood Health Framework, 2026). It never says how large a natural community is, and it refuses to say: the government will "not mandate nationally the size of neighbourhood health geographies". That refusal is honest about local difference and unhelpful about the one question every board must still answer. This week argues that the answer is a judgement, that it can be judged against evidence, and that the evidence points in two directions at once.

The framework offers a working scale even as it declines to require one. A single neighbourhood provider contract "would work for a population of around 50,000", and a multi-neighbourhood provider contract for "a population of around 250,000 or more" (Neighbourhood Health Framework, 2026). The neighbourhood health centre guidance repeats the lower figure, planning centres for "a population footprint at a scale in line with the Neighbourhood Health Framework (around 50,000, recognising the need for local flexibility)", with general practice at the core (Neighbourhood Health Centre Guidance, 2026). The 10 Year Health Plan had already set the direction of travel: two new contracts "to encourage GPs to work over larger geographies" (10 Year Health Plan, 2025).

The sizes actually in use

Behind the framework's round numbers sits a wider range. Primary care networks were recommended to serve 30,000 to 50,000 people. Morciano and colleagues tested whether they did. In January 2020, 6,758 practices had formed 1,250 networks; only 726 of them, 58 per cent, were inside the recommended range. Eighty-four networks, 7 per cent, held fewer than 30,000 people, and 440, or 35 per cent, held more than 50,000. Thirty-four networks, 3 per cent, were a single practice. The authors found substantial variation in composition as well as size, and concluded that "this high variability between PCNs is a risk to their future performance" (Morciano et al., 2020).

The wider architecture has its own ladder. Sanderson and colleagues describe the English system as a three-tier model, with systems of 1 to 3 million people, places of 250,000 to 500,000, and neighbourhoods of 35,000 to 50,000. Their qualitative study of three integrated care systems found that in practice these scales "vary considerably" and "may elude neat containment within coherent territorial geographies" (Sanderson et al., 2023).

Each size implies a different caseload. On the framework's own proportion, people with frailty, care home residents and people at the end of life are 3 to 5 per cent of the population yet represent over 25 per cent of non-elective admissions and 50 per cent of bed days (Neighbourhood Health Framework, 2026). A neighbourhood of 50,000 people therefore holds 1,500 to 2,500 such people, which a named team can know by name; a multi-neighbourhood footprint of 250,000 holds 7,500 to 12,500, which no single team can hold and which has to be divided again. A network below 30,000 holds fewer than 1,500, and the smaller the footprint the harder it is to sustain any specialist capacity across it.

List size, continuity and outcomes

Hull and colleagues measured continuity across all 126 practices in three East London boroughs, using the usual provider of care score: the proportion of consultations with the most frequently seen doctor. The mean was 0.52, with a range from 0.32 to 0.93, and smaller practices had higher scores, in a mobile and multi-ethnic population (Hull et al., 2022). They found the score correlated strongly with patients' own reports of seeing the doctor they preferred.

Barker, Steventon and Deeny linked primary and secondary care records for 230,472 patients aged 62 to 82 across 200 practices. Average continuity was 0.61, but it was "0.59 in large practices versus 0.70 in small practices", and higher continuity was associated with fewer admissions for conditions considered manageable in primary care: on their model, an increase of 0.2 in continuity across all patients would reduce those admissions by 6.22 per cent (Barker et al., 2017).

The mortality evidence runs the same way. Pereira Gray and colleagues reviewed 22 studies from nine countries and found that 18 of them, or 81.8 per cent, reported statistically significant reductions in mortality with increased continuity, 16 for all-cause mortality (Pereira Gray et al., 2018). The evidence is observational and the authors could not combine the studies, but the direction is consistent and it survives adjustment for multiple confounders in most of the studies reviewed.

The list size effect survives adjustment too. Kayira and colleagues clustered 6,024 English practices by list size and workforce composition and found that continuity of care was 34.5 per cent worse in large multidisciplinary practices than in small GP-reliant ones, 27.0 per cent against 41.2 per cent, with medium-sized practices in between (Kayira et al., 2024). Their conclusion is a warning for a policy that expects practices to grow: as England moves towards larger multidisciplinary practices, efforts should be made to preserve good patient experience.

Where scale helps, and why the trade-off is real

The same body of work shows what is lost by staying small. Kayira and colleagues found that larger multidisciplinary practices had higher cancer detection rates, and that medium-sized practices with multidisciplinary input did better on confidence and trust and on immunisation coverage (Kayira et al., 2024). Zhao, Meacock and Sutton analysed monthly data from 6,149 practices between August 2022 and July 2024 and found that additional clinical staff improve patient access more when employed in larger practices, estimating that at median administrative staffing a doctor adds 223 appointments a month and other clinical staff 152 (Zhao et al., 2026). A footprint too small to employ the skill mix cannot capture that gain.

This is the week's contested question in plain form. If continuity of relationship is the goal, the evidence favours smaller teams: continuity falls as team size rises, and continuity is associated with fewer admissions and lower mortality. If the aim is a pooled budget and population health analytics, the evidence favours larger footprints: the multi-neighbourhood provider and the integrated health organisation both assume a population large enough to carry risk and to be segmented. A neighbourhood drawn at 50,000 can hold the relationship; a footprint drawn at 250,000 can hold the analytics and the budget. No single size maximises both, and the honest work is to say what the footprint is for, then to nest one scale inside the other so that neither collapses.

Geographic access, and the people a boundary misses

Travel time is the criterion that most easily exposes an arbitrary line. Isaksson and colleagues modelled driving times for 9,946,509 Swedish residents and 1,167 primary care centres. The mean travel time to the nearest centre was 5.6 minutes, but 15.8 per cent of residents could not reach one within 10 minutes and 0.9 per cent faced more than 30 minutes; in very rural municipalities the mean was 15.1 minutes, against 3.0 minutes in large cities (Isaksson et al., 2026). A footprint drawn on population alone would merge those two populations and hide the difference between them.

Deprivation does not behave as an intuitive map would suggest. Adams and White found that, contrary to the Inverse Care Law, more deprived wards in the North East of England were geographically closer to general practices than affluent wards, while cautioning that "geographical proximity to services does not necessarily ensure that services will be accessed or that they are of good quality" (Adams and White, 2005). Proximity is not access.

The populations most likely to fall outside any footprint are the ones least visible in routine data. Humphrey and colleagues surveyed 49 people from inclusion health groups, including people with lived experience of homelessness, asylum seeking, addiction and irregular immigration. Despite formal entitlement to free primary care, 17 of them, or 35 per cent, reported obstacles to seeing a general practitioner, and participants described being refused registration, lacking digital access, and using emergency services because they could be reached without registering (Humphrey et al., 2023).

Boundary-crossing is the rule rather than the exception. Clarke and colleagues used patient registration patterns to build 165 data-defined networks across Greater London, grouping 1,291 practices. The median network list size was 53,490, but a median of only 70.1 per cent of patients attended a practice inside their allocated network, ranging from 44.6 per cent to 91.4 per cent (Clarke et al., 2020). Even a footprint optimised on the data leaves, at the median, nearly three patients in ten receiving their care outside it.

How a system should define a footprint

The guidance says what a defensible answer looks like without giving it. By 28 May 2026, regions working with integrated care boards had to set out "the latest thinking on how they will define neighbourhoods geographically in their area" (Neighbourhood Health Centre Guidance, 2026). The framework lists the considerations: the local health economy, access requirements, local governance structures such as area committees and parish councils, Pride in Place neighbourhood boards, and alignment with local authority boundaries where that is possible (Neighbourhood Health Framework, 2026).

Built into that list is the data a system needs. Patient flows, the method Clarke and colleagues used, show where people actually receive care and where a boundary would cut them. Deprivation and ethnicity distribution show whether the footprint is internally coherent or averages a deprived estate with an affluent one and serves neither. Travel time, measured as Isaksson and colleagues measured it, shows what the boundary costs the people at its edge. Community assets and existing organisational boundaries show what already exists and what the footprint would have to build or duplicate.

Each criterion has an honest limit. Patient flow data reflects the people who use services and misses the ones who do not, a limitation Clarke and colleagues accept of their secondary care data. Deprivation and ethnicity data are ecological, and a ward average can hide a street. Travel time by car is not travel time by bus. Community asset maps record buildings, not capacity. And organisational boundaries are the easiest criterion to use and the one the guidance itself says neighbourhoods should not be built around, since the model is meant to organise services "around population needs rather than organisational or contractual boundaries" (Neighbourhood Health Centre Guidance, 2026).

A wrong-scale footprint does not merely underperform. It reproduces the fragmentation it was meant to fix: too small, and each neighbourhood rebuilds the silo at a lower level, unable to hold a specialist or a budget; too large, and the team loses the relationships that the continuity evidence shows carry the outcomes, while the map looks tidy. The framework concedes that "neighbourhoods are not currently single organisations" and that in many cases they will not need to be. The footprint is the decision that determines whether the arrangement that follows can work. It should be written down, tested against the criteria above, and revisited when the data says the natural community sits somewhere else.