The CompliMind team recently visited Norton Lea, a purpose-built 18-bed mental health unit and one of the newest in the country. Delivered as part of a dormitory elimination programme following the 2017 CQC decision, it is a £36m build that has already won awards for both engagement and mental health design. We wanted the whole team to see what considered mental health estate design looks like in practice and understand how much of the real decision-making is not captured in guidance.
First and foremost: A massive thank you to Mark Swain for enabling the visit and welcoming us, and to Alan Pattison for taking hours out of a busy few days before opening to guide us through the unit and share your expertise with us.
After some reflection, three things stood out.
It would be easy to assume that a new-build unit starts from a blank page. Norton Lea showed us the opposite. The Trust had owned the land for over 40 years, and at the centre of the site sits a 150-year-old cork oak tree with a protection order. That single tree dictated the entire site layout. The building was designed around it, not in spite of it. You cannot move a protected oak, and the Trust could not sell the land without it, so the clinical environment had to accommodate the tree's root protection zone and canopy spread.
Situated just 1.5 kilometres from the coast, flood risk also shaped the infrastructure from the outset. Power distribution was specifically designed to mitigate flood failure risk, with resilience built into the electrical systems.
The team was candid about what they would have done differently given more space: a second de-escalation suite, more car parking, more toilets. But the learning is clear: site constraints drive clinical design decisions. Every mental health new-build will face its own version of the oak tree. The earlier those constraints are understood and accepted, the less costly the redesign later.
The 12-month design and 24-month build phase was described as genuinely collaborative, and it showed. The team was very intentional in involving lots of "experts by experience", alongside ex-patients, mental health nurses, clinical experts, estates teams, designers and contractors from IHP. People who had been through acute mental health crises directly shaped decisions about room layout, door design, colour, noise and the feel of communal spaces. Their perspective caught risks and opportunities that a purely clinical or engineering-led team might have missed.
The Trust mental health team also delivered toolbox talks to the construction workers, partly so that everyone on site understood the environment they were creating and why a seemingly minor detail like the gap above a door or the weight of a curtain rail carries clinical significance. But there was a second reason: construction is one of the sectors most affected by mental health issues, and the team was intentional about the wellbeing of the people building the unit, not only the people who would eventually live in it. The care extended to the process, not just the result.
Learning was drawn from other trusts in Lincolnshire and Manchester, though this was limited to a few trusts in the region. If you are commissioning a mental health new-build today, your ability to learn from others depends largely on who you know and whether they have time to talk to you. There is clearly potential to make that easier.
The team also worked with a local charity to run the on-site café 'Norton Tea' under a revenue share model, with the charity providing training as well. They are exploring a parcel locker to increase public footfall on the site. In this way, the unit is community-integrated rather than isolated, somewhere that local people have a reason to visit. A multi-faith room reflects the same thinking: this is a place designed for the whole community.
This was the takeaway that stuck with me most. Mental health estate design depends heavily on tacit institutional knowledge, which accumulates over years of frontline work and lives in the heads of experienced practitioners rather than in published standards. The density of design decisions at Norton Lea that fall outside formal guidance surprised all of us.
Anti-ligature measures make this concrete. Breakaway curtains are calibrated to release at a specific weight threshold. Smoke detectors and cameras use magnetic fittings so they cannot become anchor points. Doors open both ways to prevent barricading. Water outlets, sinks and showers are all anti-ligature rated. Beds are fixed with integrated drainage and airflow. Mattresses and chairs are sealed to prevent concealment. Anti-ligature kits are positioned with clear location signage so staff can respond immediately. Courtyard walls and fixed outdoor furniture are designed to prevent climbing. The team flagged a "door behind a door" configuration as a particularly high-risk arrangement, a finding born from operational experience rather than any published standard. All of these items are significantly more expensive than their standard equivalents, and each one has to be justified against budget constraints where the cost case is rarely clean.
Anti-barricade design follows similar experiential logic. Doors that open both ways, furniture that cannot be moved to block entry, and sightlines that allow staff to assess a room before entering are all shaped by practitioner knowledge of how crisis situations actually unfold.
Sensory regulation is treated as a safety intervention in its own right. Acoustic resonance panels reduce noise levels across communal areas. The colour palette is deliberately muted. Bedroom doors carry local architectural imagery to support orientation, a detail that matters for patients who may be disoriented or distressed. Sleep hygiene is prioritised through control of light, noise and even caffeine availability. A dedicated sensory room has been installed in response to increasing autism presentations, recognising that the sensory needs of this patient group differ.
Then there are the purposeful derogations. HTM guidance requires fire-rated materials for mattresses, but the Trust derogated to softer mattresses as they offer a genuine therapeutic benefit for patients in acute mental distress. A harder, fire-compliant mattress meets the standard but does not support recovery. To manage the additional fire risk, certain items were banned from the ward and overall fire load was reduced. This kind of controlled trade-off between compliance and patient experience is common in mental health estate design, but the reasoning behind it depends on the people who made the decision. When they move on, that reasoning goes with them.
Many of these decisions sit in options appraisals, in conversations between clinicians and designers, and in the lived experience of people who have been through acute mental health crises themselves. The team estimated that with the knowledge they now hold, the 12-month design phase could be compressed to roughly six months. The build itself, around 24 months, would remain similar. That compression comes entirely from not having to rediscover what someone else already figured out.
Visiting Norton Lea sharpened our thinking about where AI could make a practical difference in mental health units.
The most obvious application is in capturing tacit knowledge and derogation decisions. The reasoning behind thousands of individual design choices at Norton Lea currently lives in the memories of a small number of people. AI that can review options appraisals, support structured post-project stakeholder interviews, and extract insights from meeting minutes and day-to-day design decisions could give this knowledge a permanent, searchable home. The next trust commissioning a mental health unit could draw on a structured body of experience rather than starting cold.
There is also work to do on the regulatory layer. HBN 03-01, Royal College of Psychiatrists guidance, and ligature and seclusion best practice all need to be accessible in a structured way. This matters most in areas that are principle-based rather than prescriptive, where the guidance tells you what to consider rather than what to do. An AI system that can surface relevant principles alongside real-world precedents from previous builds would be far more useful than one that returns the clause and nothing else.
We came away convinced that compliance tools need to move beyond individual user personalisation towards trust-type customisation. Whether you work in an acute, mental health, community or ambulance trust changes the operational context around any given regulation. A water safety technician in a mental health trust asking about outlet flushing faces different considerations from the same technician in an acute hospital. The regulatory text may be identical, but the escort requirements, tool control protocols and patient access restrictions around that task are not.
And the visit made a strong case for embedding a patient experience lens into compliance review. Prompting practitioners to consider sensory load, privacy impact, sleep hygiene and dignity is what separates mental health estate work from a purely engineering-led compliance exercise. That perspective is what "experts by experience" bring to the design process, and it is the perspective most likely to fade once their direct involvement ends.
We left Norton Lea with pages of notes, a better understanding of what mental health estate design demands, and a sharper sense of the gap between written guidance and operational reality. Closing that gap, even partially, could save future teams months of rediscovery.
We are currently collaborating with the University of Cambridge on research into the application of AI to capture tacit knowledge and post-project learnings in healthcare estates. If you would like to be involved in the experimental phase of the research in April/May 2026, please get in touch with fw416@cam.ac.uk or carl@complimind.co.uk.