In short: NHS England's Community Diagnostic Centre programme has put scanning capacity into shopping centres, retail parks and community sites across England — over 170 of them. Each one produces hospital-grade imaging that has to reach a radiologist who is almost always somewhere else. The clinical model is sound; the connectivity model underneath it was rarely costed, and it is the part most likely to throttle throughput.
Key Takeaways
- A CDC is a data factory in a shop unit — a single CT or MRI study runs to hundreds of megabytes or several gigabytes, and a busy centre produces hundreds of studies a day that all have to leave the building.
- The radiologist is never on site — community diagnostics only works if images move quickly to a central or networked reporting team, so the connection out of the building is the actual clinical bottleneck, not the scanner.
- Retail-park leases come with retail-park broadband — CDCs sit on sites chosen for footfall and parking, not for fibre, and a single managed private 5G layer is often faster to stand up than a bespoke leased line.
In a nutshell
A hospital function in a high-street box
The Community Diagnostic Centre programme is one of the larger pieces of NHS infrastructure reform of the last few years, and one of the least understood from the outside. The premise, set out in Sir Mike Richards' 2020 review of diagnostics and adopted by NHS England, was straightforward: too much diagnostic capacity sat inside acute hospitals, where it competed with emergency work, was hard for patients to reach, and shut down whenever the hospital was under pressure. The answer was to move a large share of routine diagnostics — CT, MRI, ultrasound, X-ray, echocardiography, phlebotomy, lung-function testing — out of the acute site and into the community.
The programme delivered at pace. There are now over 170 Community Diagnostic Centres operating across England, in shopping centres, on retail parks, in repurposed high-street units and on the edges of town centres — sites chosen deliberately for the things that matter to patients: parking, public transport, opening hours, and not being inside a hospital. Between them they have delivered many millions of additional tests, and the model is widely judged a success on its own terms.
But a CT scanner does not care that it has been installed in a former department store rather than a radiology department. It produces exactly the same volume of data either way. And that data has to get from the retail park to a radiologist who, by the entire logic of the programme, is somewhere else. The clinical model decentralised the scanning; it did not, and could not, decentralise the reporting. That tension is where the connectivity problem lives, and it is the part we keep finding was never properly costed.
The data a scanner actually makes
It is worth being concrete about the volumes, because the abstraction "imaging data" hides how demanding this is. A single chest X-ray is a handful of megabytes. A CT study — and modern CT is the workhorse of a CDC — is a stack of hundreds or thousands of thin slices, routinely running to several hundred megabytes and, for a high-resolution multi-phase study, into the gigabytes. An MRI study is comparable. A busy CDC running multiple modalities across an extended day can generate several hundred studies in a single day, and the trend in imaging is relentlessly towards higher resolution and larger files, not smaller.
Every one of those studies has to travel from the scanner, through the local PACS or directly into the regional imaging network, to wherever the reporting radiologist sits — increasingly a networked or even outsourced reporting team that may be in another city or, for teleradiology, another time zone. The clinical value of the CDC is realised only when the report comes back. A scan that sits in a local buffer because the link out of the building is saturated is not a completed diagnostic; it is a patient still waiting, and a slot the centre cannot turn over.
This is the inversion that catches planners out. In a CDC the expensive, visible asset is the scanner, so that is what gets specified, sited and celebrated. But the actual throughput limit is frequently the unglamorous link between the building and the network. We have seen the pattern often enough to state it plainly: in a community diagnostic setting, the connection out of the building is the clinical bottleneck, and it is treated as an IT afterthought.
Why the site was never built for this
The difficulty is baked into the property strategy, and it is nobody's fault. CDCs were placed where patients could reach them — Sir Mike Richards' review was explicit that accessibility was the point. That meant retail parks, shopping centres and high-street units. Those locations are optimised for footfall and parking. They are emphatically not optimised for the kind of high-capacity, low-contention, clinically resilient connectivity a hospital radiology department takes for granted.
A retail unit comes with retail-grade broadband, or whatever the landlord's incoming service happens to be. Provisioning a dedicated high-capacity leased line into a leased commercial unit is possible, but it is slow — lead times of many months are common — expensive, and tied to a lease that may itself be relatively short. Meanwhile the centre is open, the scanners are running, and the imaging is piling up against whatever link is available. The mismatch between how quickly a CDC can be clinically commissioned and how slowly fixed connectivity can be delivered into a commercial site is a real and recurring operational problem.
There is also a resilience dimension that matters more in healthcare than in retail. A shop can tolerate its card terminals dropping for an hour. A diagnostic centre that loses its link mid-session has patients on tables and a reporting queue backing up. Single-circuit broadband into a leased unit is a single point of failure for a clinical service, and the clinical risk that creates is rarely visible on the property side of the project.
Where private 5G fits
A managed private 5G network is not the answer to every CDC, and we would be wary of anyone who claimed it was. But it fits a specific and common version of the problem unusually well.
First, speed of deployment. A private 5G layer can be stood up on a site in a matter of weeks, against the many-month lead time for a new leased line into a commercial unit. For a CDC that is clinically ready but connectivity-blocked — a situation we have seen more than once — that difference is the difference between opening on schedule and not.
Second, capacity where it is needed. A private network sized for the building gives the scanners, the PACS workstations and the patient-administration systems a high-capacity, low-contention path off site, with the studies prioritised over routine traffic through network slicing. The imaging gets out of the building while it is still clinically useful to move it.
Third, resilience and control. A private 5G layer can run alongside a fixed line as an active second path rather than a cold standby, so a circuit fault degrades the service instead of stopping it. And because the network is the operator's own, the imaging traffic stays under the trust's control and security posture end to end — which, for patient data, is not a small consideration.
Fourth, mobility — the part fixed connectivity cannot do at all. A meaningful share of community diagnostics is delivered from mobile units: the CT and MRI trucks that rotate between sites, the screening vans, the pop-up clinics in car parks. A mobile unit cannot wait three months for a leased line at every stop. A private 5G layer, or a managed connection onto one, gives the truck the same high-capacity path at every site it visits. The mobile fleet is where the case is strongest and least contestable.
The honest trade-offs
We try, in everything we write, to set out where our own answer is the wrong one, and here there are real limits worth naming.
A CDC sitting inside or adjacent to an existing acute hospital campus, already on the trust's fibre and network, does not need any of this — the connectivity problem it was built to have has already been solved by its location. A private network needs spectrum, which in the UK is straightforward through Ofcom's Shared Access licensing but is still a deliberate step. And it needs to be designed against the building and the modality mix, not dropped in generically; the right capacity for an ultrasound-and-phlebotomy centre is not the right capacity for a centre running two CT scanners and an MRI. As with any infrastructure, the case is strongest where the volumes are high, the site is genuinely off the hospital network, and the alternative is a slow, single, fixed circuit into a leased unit.
The wider point
The Community Diagnostic Centre programme did something genuinely useful: it pulled diagnostics out of overstretched hospitals and put it where patients could actually get to it. That decentralisation is the right direction of travel, and the throughput figures bear it out. But decentralising the scanner without decentralising the radiologist creates a hard dependency on the network in between — and that network was, too often, the line item that nobody owned when the property deal was signed and the scanner was ordered.
The lesson generalises well beyond imaging. As more of the NHS moves out of the acute hospital and into the community — virtual wards, community diagnostics, high-street clinics, mobile screening — the connectivity that used to be a free by-product of being inside a hospital becomes something that has to be deliberately designed, costed and owned. We think a managed private network is frequently the most practical way to provide it, particularly for the mobile and the hard-to-cable. Either way, the scan is only as good as the link that carries it. A diagnostic centre that can scan faster than it can send has simply moved the queue from the waiting room to the network.