Predictive Risk Monitoring in Independent Mental Health

The technology is less complicated than the name suggests. 

The most widely deployed systems use infrared cameras or radar sensors, wall-mounted in patient bedrooms and seclusion areas. They don't require the patient to wear or carry anything, which matters in an acute mental health setting where compliance with wearable devices is unpredictable.

At their most basic, they track location and movement, flagging when a patient has been in a high-risk area for longer than a defined period, or when movement patterns change in ways that warrant a check. More advanced systems derive vital signs from the same sensor data: heart rate and breathing rate measured through photoplethysmography and actigraphy, the same underlying techniques used in contact-based monitoring devices. 

The AI layer processes that sensor data, identifies patterns, and generates alerts for clinical staff. What it doesn't do is make a clinical decision. The alert tells a nurse that something has changed, and the nurse then decides what to do about it. 

That's not a limitation of the technology. That's how it's supposed to work. AI detects the signal. The clinician decides what to do next. Any supplier who tells you otherwise is worth questioning. 

Half of NHS mental health trusts in England have now implemented a contactless patient monitoring platform on their inpatient wards. That level of adoption has generated a meaningful evidence base, and the picture is encouraging. 

The evidence base for contactless monitoring in inpatient mental health has matured quickly. Over the past three years, at least eight NHS provider inspection reports have cited positive feedback on the technology's contribution to safer care. The CQC, NHS England, the Health Services Safety Investigations Body, and the National Mental Health and Learning Disability Nurse Directors Forum have each published guidance on best practice for adopting these technologies in inpatient settings. That level of convergence across national bodies doesn't happen around something unproven. 

The patient experience picture is also worth noting. Qualitative research from a high-secure forensic psychiatric hospital found that the majority of patients felt safer when contactless monitoring was in use, and that therapeutic engagement with staff remained the same or improved. For independent providers, where patient experience shapes reputation and referral volumes, that finding carries commercial weight as well as clinical weight. 

The regulatory landscape has moved to reflect this. The CQC published a brief guide on digital contactless patient monitoring technologies in mental health inpatient services in 2025, setting out what inspectors will look for when these systems are in use. NHS England has published principles for using digital technologies in mental health inpatient treatment and care. The question for independent providers is no longer whether to engage with this technology, but how.

The Lampard Inquiry, which examined deaths under Essex mental health services, heard evidence about what happens when monitoring technology is over-relied upon. That evidence is worth taking seriously, and the right response is a clear-eyed understanding of what these tools can and cannot do. 

AI can't replace clinical observation. A system that tracks vital signs and movement isn't a substitute for a nurse who knows the patient, has read the care plan, and can read the room. The technology adds a layer of continuous monitoring that human observation can't provide. It doesn't replace the human judgement that observation requires. 

AI can't interpret context. A patient who has been in their bathroom for longer than the alert threshold may be in distress. They may also be taking a shower. The alert is a prompt for a human to check, not a diagnosis. Clinical staff need to understand this before they use the system, not after an incident. 

AI can't guarantee detection. Every system has a false negative rate: cases where risk goes undetected. Ask every supplier what their false negative rate is, under what conditions events are missed, and how those cases are reviewed. If they can't answer, they aren't ready for your environment.  

AI can't carry accountability. Clinical decisions informed by monitoring data remain the responsibility of the clinician. That accountability can't be delegated to the system, and your governance frameworks need to make that explicit before you go live. 

The regulatory picture is in transition, and independent providers need to understand where it currently sits. 

Some contactless monitoring systems are classified as medical devices in the UK. The CQC's brief guide notes that systems with advanced sleep monitoring functions have been cleared for use as regulated medical devices. Others aren't classified as medical devices at all, which affects what evidence the supplier is required to hold and what post-market surveillance obligations apply to them. Two products that look similar on a demo may sit in very different regulatory categories. 

The MHRA published guidance on Digital Mental Health Technologies in February 2025, setting out when these tools qualify as medical devices based on their intended purpose and functional impact. That guidance is the starting point for understanding where any given system sits. 

A new AI-specific regulatory framework is also coming. The National Commission into the Regulation of AI in Healthcare, chaired by Professor Alastair Denniston, launched a formal call for evidence in December 2025. Its recommendations, expected later in 2026, will directly shape the MHRA's new framework for AI as a medical device. Until that framework is published, the regulatory position for AI-powered monitoring tools remains unsettled. 

The practical implication for independent providers is straightforward. Ask every supplier whether their product is classified as a medical device, what evidence they hold for its safety and efficacy, and how they're preparing for the incoming regulatory framework. A supplier who can't answer those questions clearly isn't ready for a CQC-scrutinised environment. 

CQC assesses services against five key questions. The checklist below maps the governance questions you need to put to any AI monitoring supplier, and to your own organisation, against each one.  

Safe 

 Effective 

Caring 

Responsive 

Well-led 

Contactless monitoring generates data. The clinical value of that data depends entirely on what happens to it next: whether it reaches the right clinician at the right time, whether it's recorded in the patient record, and whether it informs care planning rather than sitting in a separate system that nobody checks between shifts. 

A monitoring alert that triggers a clinical check is only useful if that check, and its outcome, ends up in the electronic patient record (EPR). An alert pattern that repeats across three nights is only useful if the clinician reviewing the care plan on day four can see it. The EPR is what makes monitoring data clinically actionable rather than just technically present. 

For independent inpatient providers, this means the EPR conversation needs to happen before the monitoring supplier conversation. A system that doesn't integrate with your patient record creates exactly the kind of information gap that CQC inspectors look for: a safety tool that generates data nobody can evidence using. 

Access Rio supports structured clinical documentation, MHA management, and auditable records across inpatient mental health settings. For providers thinking about how AI monitoring tools would sit within their clinical infrastructure, the right starting point is understanding what your EPR can receive, record, and surface. Access is developing the next generation of of this software with Rio Evo for independent mental health providers. If you want to understand how that fits with the direction you're taking on patient safety technology, a consultation with our team is a good place to start.