Enhanced observation sits at the heart of inpatient mental health safety policy. National guidance from the National Institute for Health and Care Excellence defines four escalating levels of monitoring: general checks roughly hourly, low-level intermittent checks every 30 to 60 minutes, high-level intermittent checks every 15 minutes, and constant observation involving uninterrupted visual contact. An estimated 13 to 16 percent of psychiatric inpatients are placed on constant observation at some point during their admission, at an annual cost to the NHS of approximately £35 million. But the evidence base for the practice is strikingly thin. Some studies link constant observation to reduced inpatient suicide; others find no meaningful relationship between observation intensity and self-harm, and some suggest constant observation may paradoxically increase violent incidents, as frightened and frustrated patients lash out at the staff watching them.

To understand why the practice persists despite these doubts, the researchers turned to rapid ethnography, a condensed form of anthropological fieldwork that combines immersive observation with interviews and document review under the time constraints of applied healthcare research. Over the course of the study, they accumulated 120 hours of non-participant observation across three acute female inpatient wards in two NHS mental health trusts, alongside 22 semi-structured interviews and informal conversations with psychiatrists, nurses, healthcare assistants, occupational therapists, ward managers and service users. Fieldnotes were structured using Spradley’s observation framework, recording space, actors, activities, objects, events, time, goals and feelings, and the analysis followed an iterative constant comparative method, with each ward treated as a naturalistic case study within a critical realist frame. Ethical approval came from the Yorkshire and The Humber research ethics committee, and a lived-experience advisory group helped shape the study design and recruitment strategy.

What the researchers witnessed and heard was a practice shaped less by clinical reasoning than by what they describe as a pervasive, fear-driven culture. Service users described night-time checks, delivered at intervals from every 15 minutes to hourly, as profoundly disruptive to sleep. Staff shone torches through vision panels, opened doors and switched on lights to confirm patients were breathing, waking people repeatedly in an environment already noisy with pacing and slamming doors. One service user described being woken every 15 minutes and never returning to sleep, ending up “in a worser position” because of exhaustion and irritability. Participants described sleep as essential to recovery and the interruptions as “inhumane,” feeding a vicious cycle in which medication was increased to force sleep and hospital stays lengthened. Beyond the bedroom, patients on constant observation reported having no privacy even in the bathroom, with some saying they withheld toileting to the point of kidney infection rather than endure a staff member’s foot in the door. For women with histories of sexual abuse, being observed by male staff was described as frightening and re-traumatising.

The staff side of the story was equally revealing. Again and again, clinicians acknowledged the harms of observation yet defended it as a shield against blame. Ward managers noted that newly qualified nurses prescribed enhanced observations out of fear of disciplinary action, keeping patients watched for longer than clinically necessary. Although organisations professed a “no blame culture,” staff described formal investigations after serious incidents that scrutinised the actions of frontline nurses and healthcare assistants, and many spoke of documenting meticulously so their decisions would “stand up in court.” A psychiatrist acknowledged the tension bluntly: sleep mattered for mental health, but “our main aim is to make sure they don’t do anything that can be life threatening… we have to make some small changes and sacrifices, and sleep is one of them.” The researchers read these patterns through the sociological lens of ritualisation, drawing on the work of Charles Bosk and on studies of surgical safety checklists, in which healthcare workers perform visible acts of compliance less because they demonstrably improve outcomes than because they signal diligence, protect reputations and reassure institutions.

The study also exposed a stark gap between policy and practice, echoing the safety-science distinction between “work as imagined” and “work as done.” Staff routinely deviated from local protocols: healthcare assistants disagreed over whether observation meant watching through a door window or entering the room; following a suicide on one ward, night staff imposed hourly checks on every patient regardless of prescribed risk levels; and experienced workers relied on gut feelings and subtle behavioural cues, such as a patient’s absence from communal areas, that no written procedure captures. In some cases this discretion served therapeutic ends, with staff giving distressed patients space to de-escalate or allowing unsupervised bathroom access for trauma survivors. In others it functioned as a private safety blanket, with “unofficial” observations sustained by intuition and anxiety rather than shared standards. The researchers note that frontline staff were rarely involved in writing the observation policies they were expected to follow, helping explain why procedures so often proved unworkable on the ward floor.

Perhaps the most damning findings concern the relationship between observation and recovery. Recovery-oriented care, endorsed by the UK, Denmark and Australia, prioritises autonomy, collaboration and trust. Yet service users described constant observation as isolating and dehumanising, with staff who “just follow you around” without engaging, and decisions about observation levels made over their heads. Paradoxically, the continuous presence of a person did not translate into human connection. When staff used observation as an opportunity for genuine engagement, initiating conversations, acknowledging distress, sharing activities, patients reported feeling safer; when it was reduced to mechanical surveillance, it eroded therapeutic alliance and sometimes worsened symptoms such as paranoia. A small minority of participants described a different path: one woman gradually tapered off more than a year of constant observation through collaborative safety planning, trigger identification and regular reviews, building confidence in managing her own risk even through occasional relapses of self-harm.

The evidence on whether any of this vigilance actually prevents death is sobering. The National Confidential Inquiry into Suicide and Homicide found that 91 percent of inpatient suicides between 2011 and 2016 occurred while patients were on intermittent rather than constant observation, suggesting that periodic checks may inadvertently create windows of opportunity rather than closing them. On that basis, some researchers, including the authors of this study, argue that night-time intermittent observations should be discontinued for people assessed at low or medium risk, with constant observation reserved for those at high and immediate risk. Yet the practice resists change. Observations are embedded in professional identity, particularly for healthcare assistants for whom the clipboard, even after it was formally abolished, functioned as a symbolic source of security. Staff who had responded to suicide attempts carried lasting trauma and requested higher observation levels than necessary to manage their own dread, often without voicing those fears to senior colleagues.

The authors frame enhanced observation as a prime candidate for de-implementation, the deliberate withdrawal of a low-value clinical practice, a field with growing literature but little application in mental health. Importantly, they found that patient resistance, a common barrier elsewhere, does not apply here: service users overwhelmingly preferred less observation, meaning reform could align with patient preferences rather than run against them. Practical openings identified in the study include regular multidisciplinary review of observation levels, genuine shared decision-making in safety planning, and routine questioning of whether observation is necessary, therapeutic and proportionate. The researchers also caution against simply substituting one form of surveillance for another, noting that video-based technologies such as Oxevision may reduce physical presence while intensifying feelings of being watched.

The study has limits the authors acknowledge candidly: fieldwork was relatively brief, the presence of an observer may have altered behaviour, and the focus on three female wards means the findings may not transfer to male, mixed-sex or forensic settings. Still, the ethnographic depth, capturing both the lived experience of being watched and the quiet anxieties of those doing the watching, offers something surveys cannot. The researchers conclude that institutions should reflect honestly on why patient observations are being used, refocus them where possible toward relational care, empower staff to challenge rituals, and co-produce safety strategies with the people who implement and endure them. As one service user put it, in words that capture the study’s central message: staff should “just be human.”