Crisis, Suicide, Aggression and De-escalation

This is the chapter where the whole book gets tested. A person in crisis has run out of the coping strategies that used to work, and the nurse standing in front of them is either a stabilising presence or another thing happening to them. Three situations dominate acute mental health nursing: the person whose usual life has come apart, the person who is thinking about ending their life, and the person whose distress is escalating toward physical aggression. All three are managed with the same underlying moves — slow down, assess before you intervene, use the least restrictive option that keeps everyone safe, and treat the person as a partner rather than a problem. You will learn to ask about suicide in plain words, to build a safety plan that a person can actually use at three in the morning, to read the escalation curve early enough that talking still works, and to know the law that governs restraint and seclusion when talking has not.

Section 1Crisis: when the usual coping runs out

The anatomy of a crisis

A crisis is a state of acute emotional disequilibrium in which a person’s habitual problem-solving methods have failed to resolve a threat. Notice what that definition does not say. It says nothing about the objective severity of the event. Two people can lose the same job on the same day; one reorganises the budget and starts applying for work, and the other stops eating and stops answering the phone. The first person is stressed. The second is in crisis. What separates them is not the stressor but the balance between the perceived threat, the coping repertoire available, and the support the person can reach. Crisis is always a relationship between the event and the person, which is why you must never rank a patient’s distress against your own sense of how bad the situation ought to feel.

Gerald Caplan described crisis as unfolding in phases. In the first phase the person meets the stressor and reaches for familiar coping strategies, and anxiety rises modestly. In the second phase those strategies fail, anxiety climbs, and the person begins trial-and-error attempts. In the third phase, tension is high enough to mobilise emergency resources — novel problem-solving, requests for help, sometimes redefinition of the problem. If the crisis still does not resolve, the fourth phase brings major disorganisation: profound anxiety, impaired concentration, disrupted sleep, and behaviour that others find alarming. Suicidal thinking, substance use and aggression frequently appear here, not because the person is fundamentally different from who they were last month, but because the person has run out of options.

Crises are time-limited. Left alone, the acute disequilibrium usually resolves within four to six weeks, one way or another. That time limit is the entire clinical opportunity. A crisis resolves upward — with new coping skills and a higher baseline of resilience — or downward, into a lower level of functioning, a new psychiatric diagnosis, or death. The intervention you provide in those weeks disproportionately determines which way it goes. This is why crisis work is short, focused and present-tense. You are not doing insight-oriented therapy. You are helping a person get through the next seventy-two hours with their life and their relationships intact.

Three kinds of crisis

Crises are conventionally sorted into three types, and the distinction matters because it changes what you look for. A maturational (developmental) crisis arises at a normative life transition — leaving home, becoming a parent, retiring, the death of the last surviving parent that moves a person into the oldest generation. These events are expected, which is precisely why the person often gets no support for them: everyone assumes it is manageable because everyone goes through it. A situational crisis follows an unanticipated personal event — job loss, divorce, a new diagnosis of cancer, a miscarriage, eviction, the arrest of a family member. An adventitious crisis follows a disaster affecting many people at once: a hurricane, a mass casualty event, a war, a community-wide loss.

These categories overlap and stack. A nursing student in the middle of a maturational transition who then loses housing and is also living through a regional disaster carries three loads at once. When you assess, ask about all three layers rather than settling for the first one the person names. The presenting complaint in an emergency department is frequently the smallest of the person’s problems — it is simply the one with a phone number attached.

Culture shapes what counts as a crisis and what counts as help. Loss of face, immigration status, a religious community’s response to a divorce or a disclosure of sexual orientation, or the shame attached to psychiatric care in a given family can be the central threat while the chart records something else entirely. Ask what the event means to this person, in their family and community, before you assume you understand the size of it.

Crisis intervention and psychological first aid

Crisis intervention is deliberately narrow. Its goal is the restoration of the person’s pre-crisis level of functioning or better, not personality change. The sequence is reliable: establish contact and safety first; help the person put the problem into words and shrink it to its actual boundaries; identify the coping they still have; generate a small number of concrete, doable next steps; and arrange concrete follow-up with a named person and a time. Roberts described a seven-stage model that expands this, but the compressed version — contact, boil the problem down, mobilise coping — is what you will actually use at a bedside.

Psychological first aid is the disaster-context cousin of crisis intervention and follows the same logic: ensure physical safety and basic needs, offer calm and non-intrusive comfort, help the person connect with social supports, provide accurate information, and link to services. Note what psychological first aid deliberately excludes. It does not push people to recount the traumatic event in detail, and single-session mandatory psychological debriefing of trauma survivors has been shown in trials not to prevent post-traumatic stress disorder and may worsen outcomes for some. Presence, practical help and connection are the evidence-based intervention. Forced narration is not.

Directiveness in crisis work is calibrated to functioning. The more disorganised the person, the more structure you provide — short sentences, one instruction at a time, concrete choices. As the person reorganises, you hand decisions back. Handing decisions back is not a courtesy; it is the therapeutic mechanism. A person who leaves a crisis contact having made two of their own decisions has evidence that they are still an agent in their own life, which is exactly what the crisis had called into question.

Trauma-informed care as the default posture

Assume that a trauma history may be present in every person you meet on a behavioural unit, because the base rate is high and the history is often not in the record. Trauma-informed care is not a therapy; it is a way of running a service so that care does not re-injure. The Substance Abuse and Mental Health Services Administration frames it around six principles: safety; trustworthiness and transparency; peer support; collaboration and mutuality; empowerment, voice and choice; and attention to cultural, historical and gender issues.

Translated to the floor, this means telling people what is going to happen before it happens, explaining why a search or a blood draw is being done, offering genuine choices wherever a choice exists, knocking, using the person’s preferred name, and never using power casually. It means recognising that being held down by several people, having clothing removed, being confined to a locked room, or being watched while using a toilet reproduce the conditions of prior abuse with uncanny precision. That recognition is not a reason to never restrain anyone; it is a reason to treat every restraint as a treatment failure worth examining, and to plan around known triggers before they are pulled.

Trauma-informed practice also changes how you interpret behaviour. A person who becomes intensely distressed when a male staff member stands between them and the door is not being unreasonable. A person who freezes and cannot answer a direct question is not refusing to cooperate. Ask “what happened to you?” rather than “what is wrong with you?”, and record what the behaviour communicates rather than a judgement about the person’s character.

Section 2Suicide risk: asking the question directly

What the epidemiology actually tells you

Suicide is among the leading causes of death across most of the adult lifespan and the rate has risen in many countries over the past two decades. Some patterns are stable enough to shape practice. Older men, particularly older white men living alone, have the highest rate of death by suicide. Women attempt more often; men die more often, largely because of method lethality. Firearms account for roughly half of suicide deaths in the United States and are the single most lethal common method, with case fatality far above overdose. Rates are markedly elevated among Indigenous populations, veterans, people who identify as lesbian, gay, bisexual, transgender or queer — especially youth in unsupportive environments — people in the first weeks after psychiatric discharge, and people recently released from incarceration.

Language matters here and it is assessed. A person dies by suicide or ends their life. The phrase “committed suicide” carries the residue of an era when suicide was a crime and a sin, and families hear it that way. Similarly, an attempt is never “failed” or “successful”; describe it as a suicide attempt and describe the medical seriousness separately. Avoid “gesture” and “manipulative” entirely — those words tell you nothing about lethality and everything about the writer’s frustration, and they have preceded deaths. A self-injurious act made in front of others still kills people.

One more correction to a persistent myth: asking a person directly about suicide does not plant the idea and does not increase risk. Randomised and observational studies of screening have found no increase in ideation or distress from being asked, and often a reduction. What increases risk is a person who has been thinking about killing themselves for weeks, is asked nothing, and concludes that the subject is unspeakable even here.

Risk factors, protective factors and why prediction fails

Static risk factors are historical and unchangeable: a previous suicide attempt (the single strongest predictor), family history of suicide, history of childhood abuse, chronic medical illness, male sex, and a diagnosis of a mood disorder, psychotic disorder, borderline personality disorder or a substance use disorder. Dynamic risk factors are the ones you can act on today: current ideation with plan and intent, access to lethal means, acute intoxication, agitation, insomnia, hopelessness, recent loss, acute psychosis with command hallucinations, recent discharge, and abrupt social disconnection. Protective factors buffer but never neutralise: reasons for living, responsibility for children or animals, a therapeutic relationship, religious or cultural prohibition, engaged treatment, and restricted access to means.

Now the uncomfortable truth. Meta-analyses covering fifty years of research show that risk factors predict suicide barely better than chance at the level of the individual. There is no score that tells you who will die. Tools such as SAD PERSONS were built for teaching and perform poorly as triage instruments, and using a low score to justify discharge is indefensible. What replaces prediction is formulation: a narrative statement of this person’s risk that combines chronic vulnerability, current acute drivers, protective factors and what is modifiable now — and that names the plan. “Chronic elevated risk from prior attempt and alcohol use disorder; acutely elevated by job loss four days ago and access to a firearm at home; protective factors are custody of two children and willingness to engage; plan is means restriction with his sister tonight, safety plan completed, contact in 24 hours.” That paragraph is defensible care. A number is not.

Two clinical patterns deserve specific mention. A person emerging from severe depression may be at increased risk as psychomotor retardation lifts before mood does, restoring the energy to act. This is why the first weeks of antidepressant treatment are an observation priority rather than a reassurance: antidepressants take roughly two to six weeks for full mood response, energy and volition often return earlier, and regulators require a boxed warning about increased suicidal thinking and behaviour in children, adolescents and young adults under twenty-five during early treatment and after any dose change. Increase contact and check ideation in that window; do not read “he is up and about now” as improvement. And a person who has been agitated and ambivalent for days who suddenly becomes calm and settled has not necessarily improved — sudden serenity can mean a decision has been made. Both patterns should raise, not lower, your level of attention.

How to ask: the words that work

Ask plainly, without preamble that signals your own discomfort, and ask in a sequence that moves from the general to the specific. Normalising the question lowers the barrier: “When people are carrying as much as you are, they sometimes have thoughts of ending their life. Have you had thoughts like that?” Then follow the ladder — passive ideation (“wish you would not wake up”), active ideation, method, plan, intent, preparatory behaviour or rehearsal, and access to means. Ask what has stopped them so far; the answer gives you the protective factors in the person’s own words, which are the ones that will work at two in the morning.

Silence after the question is your friend. Count to five in your head before filling it. If the person discloses, respond to the content and not to your own alarm — thank them for telling you, then keep gathering detail. Visible panic in the nurse teaches the patient that this information is too much for professionals to hold, and the next person they tell will be no one.

Structured screening: the C-SSRS and what it captures

The Columbia Suicide Severity Rating Scale is the most widely implemented structured screen. Its screening version asks about a wish to be dead and non-specific active thoughts, then, if either is positive, about method without intent, intent without a specific plan, and a specific plan with intent, followed by lifetime and past-three-month preparatory or attempt behaviour. The value of the instrument is that it forces the specific questions that clinicians skip under time pressure, and it produces a shared vocabulary across an organisation. The Ask Suicide-Screening Questions tool is a four-item screen validated for use with youth in medical settings, and the Patient Health Questionnaire item nine is a case-finding prompt only — a positive item nine mandates a full assessment, it is not itself an assessment.

A structured screen never replaces the interview. It tells you what to ask next. Run it as a conversation rather than a recitation, and record the patient’s own words in quotation marks; “I keep thinking about my father’s gun in the closet” carries clinical information that a checked box destroys. The Joint Commission’s National Patient Safety Goal on suicide prevention requires screening of all patients being evaluated or treated for behavioural health conditions, an evidence-based assessment when screening is positive, documented risk mitigation, and environmental risk reduction in areas where these patients are cared for.

Finally, screen everywhere, not only in psychiatry. A large proportion of people who die by suicide have contact with a health service in the preceding month, most often primary care or an emergency department, and often for a physical complaint. Chronic pain, new cancer diagnosis, traumatic brain injury, and delirium in an older adult are all contexts in which the question belongs.

Section 3Safety planning and the safer environment

The safety plan, step by step

The Safety Planning Intervention developed by Stanley and Brown is a brief, collaborative, written plan that has evidence for reducing subsequent suicidal behaviour and improving treatment engagement, particularly when paired with follow-up contact. It has six steps and it is built with the patient, in the patient’s language, in a single sitting of fifteen to forty-five minutes. Step one is warning signs: the specific thoughts, images, moods, situations and behaviours that tell this person a crisis is beginning. Step two is internal coping strategies the person can use alone, without contacting anyone — a shower, a specific playlist, walking the dog, a video game — chosen because they have worked before, not because they sound healthy.

Step three is people and social settings that provide distraction: not confidants, just company — a brother’s house, a coffee shop, a cousin who talks about football. Step four is people the person can ask for help directly, with names and numbers written down. Step five is professionals and agencies, including the clinic number, the on-call service, the crisis line (988 in the United States) and the nearest emergency department. Step six — and this is the step with the strongest independent evidence — is making the environment safe by restricting access to lethal means.

Two implementation details separate a real safety plan from a form in a chart. First, test each step out loud: “It is eleven at night and the thoughts are getting louder. What is step two? Say it back to me.” If the person cannot retrieve it, the step is wrong. Second, decide where the plan physically lives — a photograph on the phone, a card in the wallet, a copy on the refrigerator, a copy with a family member with the patient’s consent. A plan the person cannot reach at the moment of crisis is a plan that does not exist.

Why no-suicide contracts are not care

A “no-suicide contract” or “contract for safety” asks the patient to promise, verbally or in writing, not to harm themselves. There is no evidence that it reduces suicide or suicide attempts, and there is reason to think it causes harm. It shifts responsibility for safety onto the person who is least able to carry it. It provides false reassurance to the team, and documented contracts have appeared in the records of people who died the same week. It also filters the population who will talk to you: a person who is genuinely unsure whether they can stay safe simply declines to disclose rather than break a promise. Legally, the contract confers no protection and can be evidence that the assessment stopped where it should have continued.

The replacement is the safety plan, and the difference is structural. A contract asks for a promise about a feeling the person cannot control. A plan lists behaviours the person can perform. A contract ends the conversation. A plan generates specifics — which cousin, which number, which drawer the pills are in, who is holding the firearm this week. If your institution still uses contract language, the professionally defensible move is to document a collaborative safety plan and the means restriction discussion regardless of what the form is called.

Lethal means counselling

Suicidal crises are often brief and highly ambivalent. Studies of survivors of near-lethal attempts find that many describe minutes rather than hours between decision and act, and the large majority of people who survive an attempt do not go on to die by suicide. Because the crisis is time-limited and the ambivalence is real, putting time and distance between a person and a highly lethal method changes outcomes. This is not a theoretical claim; it is supported by natural experiments in which the removal of a specific method — barriers on bridges, restriction of the most toxic pesticides, changes in analgesic pack sizes, detoxification of domestic gas — produced sustained falls in overall suicide deaths without full substitution to other methods.

How to have the conversation: be matter-of-fact, do not moralise, and frame it as temporary and collaborative. “While things are this hard, I want to make it harder for a bad ten minutes to become permanent. Who could hold your firearms for a while?” Options include off-site storage with a trusted person, a licensed dealer or a law enforcement agency where permitted, locking devices with the key held by someone else, and separating ammunition from firearms. For medications, dispense smaller quantities, use blister packs, and have a family member store and dispense. Remove sharps and cords named by the patient, secure alcohol, and address vehicle access if that is the plan. Document the specific agreement, who is holding what, and when it will be reviewed.

Involve family with the patient’s consent where possible, but note that most jurisdictions permit disclosure without consent when there is a serious and imminent risk of harm. You can always receive information from a family member without breaching confidentiality, and telling a family member what to remove from the house is a permissible disclosure in an emergency in most legal frameworks. Know your own state or national rule before you need it, not while you are standing in the corridor.

Three legal ideas govern what happens next when a person at high risk does not want to stay. The first is admission status. A voluntary patient consents to admission and retains the right to request discharge, usually in writing and subject to a short statutory period in which the team may evaluate and, if criteria are met, convert the status. An involuntary admission (an emergency hold or civil commitment, named differently in every jurisdiction) requires that the person have a mental disorder and meet a statutory threshold — typically danger to self, danger to others, or grave disability — and it is authorised by a designated professional for a limited number of hours or days before a hearing. Say “involuntary admission” or “placed on a hold,” not “we committed him”; the older phrasing is both imprecise and freighted. Involuntary admission also does not by itself remove the right to refuse a specific treatment, and it never removes the rights to be informed, to be treated with dignity, to communicate with counsel, or to the least restrictive environment that will keep the person safe.

The second is capacity, which is decision-specific and fluctuating, not a global label. A person has capacity for a given decision if they can understand the relevant information, retain it, weigh it against their own values, and communicate a choice. Refusing what you recommend is not evidence of incapacity, and a psychiatric diagnosis is not evidence of incapacity; intoxication, delirium, severe depressive hopelessness or florid psychosis may be. Assess capacity for the specific decision in front of you, document the four elements, and reassess — a person who lacks capacity at two in the morning may have it back by noon.

The third is the limit of confidentiality. Patient information is protected, and the exceptions are narrow and named: a serious and imminent risk of harm to the patient, a credible threat to an identifiable third party, suspected abuse or neglect of a child or vulnerable adult, and a valid court order. The Tarasoff line of cases established a clinician duty to warn or, more broadly, a duty to protect an identifiable potential victim, which in practice may be discharged by warning the person, notifying police, arranging admission, or intensifying treatment — the statute in your jurisdiction defines what is required and permitted. Tell patients the limits at the start of the relationship rather than invoking them as a surprise, and disclose the minimum necessary to the people who need it.

Observation levels, ligature risk and the transitions

Inpatient safety rests on the environment, the observation level and the transitions between settings. Ligature risk is the environmental hazard that matters most on an inpatient behavioural unit, because hanging and strangulation are the leading inpatient method. A ligature point is anything to which a cord can be attached; the standard is a ligature-resistant environment in areas where patients at risk are cared for without continuous observation — breakaway rails and hooks, sloped tops, tamper-resistant fixtures, no exposed pipes or cords, weighted or fixed furniture, and no free plastic bags. Belts, shoelaces, drawstrings, phone charging cables, headphone cords and lanyards are the everyday items that cause deaths. Environmental rounds are everyone’s job, including yours on every shift.

Observation levels typically run from routine checks, through fifteen-minute checks, to continuous line-of-sight observation, to one-to-one arm’s-length observation including in the bathroom. Two rules: checks must be at irregular intervals within the specified window so that they cannot be timed and worked around, and the observer must actually see chest rise and skin, not a shape under a blanket. Documentation of a check that did not happen is falsification of a medical record, and this is one of the most common findings in inpatient suicide investigations.

The highest-risk periods are transitions: the first days of admission, any change in observation level, pass or leave from the unit, the days around discharge, and above all the first week and first month after discharge, when the suicide rate is many times the rate in the general population. Discharge planning is therefore a clinical intervention, not paperwork. A follow-up appointment within seven days, a caring contact by phone or message in the first days, the safety plan in the patient’s pocket, the means restriction confirmed with the person who is holding, and a warm handover to the receiving clinician are the elements that reduce post-discharge deaths.

Section 4Aggression: the escalation curve and the matched response

Reading the curve

Aggression is almost never sudden. It follows a curve with recognisable stages, and the entire skill of de-escalation is intervening at the earliest stage you can detect. At baseline the person is settled and interventions are relational: know their triggers, know their preferred calming strategies, build rapport when nothing is happening. In the triggering phase something happens — a denied pass, a phone call, a room change, a delayed medication, a misunderstood instruction — and anxiety rises: pacing, tapping, louder speech, repeated requests. The response is supportive and practical: approach, ask, listen, solve the solvable problem. Most escalations end here if anyone notices.

In the escalation phase the person becomes defensive: challenging, arguing, refusing, blaming, invading space, swearing. Rational argument now fails because the person’s capacity to process complex language is falling. This is the phase for limit-setting delivered as choice, reduced stimulation, and a shift to short sentences. In the crisis phase there is loss of control — throwing, striking, self-injury, or flight. Talking has a much lower yield here and the task becomes protection of the person, other patients and staff with the least restrictive effective intervention. The recovery phase brings gradual settling with residual arousal, where re-escalation is easy and demands should be minimal. Finally, post-crisis depletion brings exhaustion, remorse and often shame — the moment for repair, not for a lecture.

Structured tools help teams see the curve earlier than intuition does. The Brøset Violence Checklist scores six observable items — confusion, irritability, boisterousness, verbal threats, physical threats and attacking objects — and predicts violence within the next twenty-four hours well enough to be useful for shift-level planning. The Dynamic Appraisal of Situational Aggression is a similar structured shift-based tool. Use them as prompts to act, not as scores to record.

One correction that belongs in every chapter on this subject: people with mental illness are far more likely to be the victims of violence than the perpetrators of it. The great majority of violence in society is not attributable to psychiatric illness, and the risk attributable to mental illness alone is small, rising mainly with untreated acute psychosis combined with substance use. Say this to students, to families and to journalists, and do not let unit culture teach otherwise.

Verbal de-escalation in practice

Verbal de-escalation is a teachable, structured skill, best summarised by the Project BETA consensus domains. Respect personal space — stay at least two arm’s lengths away, stand at an angle rather than squared up, and make sure both you and the patient have an unobstructed path to the door with neither of you blocked in. Do not be provocative: hands visible and open, relaxed posture, no pointing, no crossed arms, no staring, no laughing or side conversations with colleagues. Establish verbal contact: one person speaks. A single spokesperson is not a stylistic preference; multiple voices are experienced as a crowd and they escalate.

Be concise. Short sentences, simple words, repeat as needed. Identify the person’s wants and feelings, then say them out loud — the free information the person gives you (“nobody tells me anything in here”) names the intervention. Listen closely to what the person is saying and reflect it back. Agree where you truthfully can, agree in principle where you cannot agree in fact, or agree to disagree. Set clear, respectful limits, framed as information about consequences rather than threats. Offer choices and optimism, because a person who has choices is not cornered. And debrief afterward, with the patient and with the staff.

The physical environment does half the work. Reduce noise, turn off the television, dim overhead lights if you can, move onlookers away, and remove objects that can be thrown. Time is an intervention: sitting down when the patient sits, and being willing to spend twenty minutes talking, is dramatically cheaper than a restraint episode in every currency — injury, trauma, staff time and legal exposure. Offer food, water, a phone call, a cigarette substitute such as nicotine replacement, a blanket, or a quiet room, because unmet basic needs drive a large share of unit conflict.

Dialogue: what lands and what detonates

Scenario: a man on an inpatient unit has just been told his weekend pass is cancelled. He is standing close to the nursing station, voice raised, fists clenched, and other patients are watching.

Milieu, delegation and the team

The therapeutic milieu is an active intervention: predictable structure, posted schedules, groups that actually run, staff visible in the day area rather than behind glass, and consistent limits applied by everyone. Most unit violence is preceded by conflict about rules, waiting, denied requests, or a patient feeling ignored — all of which are milieu problems. Community meetings, clear communication about what is happening and when, and honouring small commitments (“I will come back at three” — then come back at three) prevent more aggression than any technique.

Delegation on a behavioural unit follows the same rule as everywhere: you may delegate tasks, never the nursing process. Trained assistive personnel may perform one-to-one and interval observations and document them, conduct environmental safety rounds and searches of belongings under RN direction, take vital signs, escort patients, assist with hygiene and meals, run structured recreational activities, provide reorientation and, when trained and competent, participate in a restraint hold. They must report to the RN promptly and repeat patient statements verbatim — “he said he wants to go be with his brother who died” is far more useful than “he seemed down.”

What assistive personnel may not do: perform or interpret a suicide risk assessment or mental status examination, decide or change an observation level, initiate or discontinue restraint or seclusion on their own judgement, administer medication, perform the required one-hour face-to-face evaluation, conduct patient teaching, or evaluate a response to an intervention. If an unlicensed staff member tells you a patient “is fine now” and asks to reduce observation, the assessment is yours to perform. Note also the practical staffing rule: the person conducting a one-to-one observation is not available for anything else, and quietly reassigning them is how inpatient deaths happen.

Section 5Restraint, seclusion and what happens afterwards

The least restrictive rule and the law

Every patient has the right to be free from restraint and seclusion imposed as coercion, discipline, convenience or retaliation. Restraint is any manual method, physical or mechanical device, material or equipment that immobilises or reduces the ability of a patient to move their arms, legs, body or head freely, or a drug used to control behaviour or restrict freedom of movement that is not a standard treatment or dosage for the patient’s condition. Seclusion is the involuntary confinement of a patient alone in a room or area from which they are physically prevented from leaving. A voluntary time-out that the patient can end is neither.

Restraint and seclusion may be used only when less restrictive interventions have been determined to be ineffective to protect the patient or others from harm, must be discontinued at the earliest possible moment regardless of the time remaining on the order, and require an order from a physician or other licensed practitioner permitted by state law and hospital policy. PRN or standing orders for restraint are prohibited — every episode requires its own order. If restraint is applied in an emergency before an order exists, the order must be obtained during the intervention or immediately after.

Under the United States Centers for Medicare and Medicaid Services conditions of participation, restraint or seclusion used for violent or self-destructive behaviour carries order time limits of up to four hours for adults aged eighteen and older, two hours for children and adolescents aged nine through seventeen, and one hour for children under nine. Orders may be renewed within these limits for up to twenty-four hours total, after which the practitioner must see and assess the patient in person before any new order. A physician, other licensed practitioner or a trained registered nurse must conduct a face-to-face evaluation within one hour of initiation, assessing the patient’s immediate situation, reaction to the intervention, medical and behavioural condition, and the need to continue or terminate. When a trained RN performs it, the attending practitioner must be consulted as soon as possible afterwards. Simultaneous restraint and seclusion is permitted only with continuous in-person monitoring by trained staff, or in-person monitoring for the first hour followed by continuous audio and video monitoring in close proximity.

Documentation must show the behaviour that necessitated the intervention, the less restrictive alternatives attempted and why they failed, the order, the type of restraint, the one-hour evaluation, monitoring at the required intervals, the patient’s response, care provided, criteria for release, and the time of release. Deaths associated with restraint or seclusion must be reported to CMS within the required timeframes. State mental health law, the Joint Commission standards, and your hospital policy may all be stricter than the federal floor — follow the strictest applicable rule.

Monitoring the restrained patient

A restrained patient is a physiologically unstable patient in a forced position, frequently after a struggle, often having received sedating medication. Assessment includes at minimum signs of injury from the application, respiratory and circulatory status, skin integrity and colour, distal pulses, capillary refill and sensation in restrained limbs, correct application and tension of devices, range of motion, vital signs, level of consciousness, hydration, nutrition, toileting, and readiness for release. Typical practice is continuous observation with documented assessment at fifteen-minute intervals, limb release and range of motion at least every two hours, and offering fluids and toileting on a regular schedule.

Prone restraint is the single most dangerous position because of positional asphyxia, and it is prohibited or heavily restricted in most modern policies. Weight applied to the back, chest or neck, obesity, intoxication, agitated delirium and recent physical exertion all compound the risk. If any hold must be used, the supine position with the head free and no weight on the torso is safer, and the hold should last only long enough to apply a device or administer medication. Never place anything over the face. Sudden quiet during a hold is a respiratory emergency until proven otherwise, not compliance.

Restraint is contraindicated or requires exceptional caution in a patient with a suspected medical cause for agitation. Before you assume a behavioural cause, check the glucose, oxygen saturation, temperature and vital signs, and consider delirium, head injury, hypoxia, hypoglycaemia, infection, postictal state, alcohol or benzodiazepine withdrawal, anticholinergic toxicity and thyroid storm. An older adult who is newly agitated at night is delirious until proven otherwise, and restraining a delirious patient worsens the delirium, increases falls and prolongs hospitalisation. Look also for akathisia — the inner restlessness caused by antipsychotics — which is routinely mistaken for worsening agitation and then treated with more antipsychotic, making it worse. Akathisia is treated by reducing the offending agent and adding a beta blocker such as propranolol or a benzodiazepine.

Chemical restraint and emergency medication

The word “chemical restraint” is regulatory, not pharmacological: it means a medication used to control behaviour or restrict movement that is not a standard treatment for the patient’s condition. Giving a scheduled antipsychotic to a person with schizophrenia is treatment. Giving a sedative to a person because the unit is short-staffed is restraint, and it carries every requirement described above. Offering oral medication first, and offering it as a genuine choice, is both more respectful and more effective than most people expect — many agitated patients accept oral medication when it is offered rather than announced.

Commonly used agents include a second-generation antipsychotic such as olanzapine, risperidone or ziprasidone; a benzodiazepine such as lorazepam, which is preferred when the agitation is due to alcohol or sedative withdrawal or to stimulant intoxication; and haloperidol, often combined with lorazepam. Key monitoring: check for over-sedation and respiratory depression, particularly with a benzodiazepine in an intoxicated patient or an older adult; monitor the QT interval with haloperidol, droperidol and ziprasidone; and do not give intramuscular olanzapine within one hour of a parenteral benzodiazepine because of reported deaths from cardiorespiratory depression. Watch for the extrapyramidal side effects of antipsychotics: acute dystonia in the first hours to days, treated urgently with an anticholinergic such as benztropine or diphenhydramine and a genuine airway emergency if the larynx is involved; akathisia within days; drug-induced parkinsonism over weeks; and tardive dyskinesia, the late-onset involuntary orofacial and limb movements that may be irreversible and are screened for with the Abnormal Involuntary Movement Scale rather than found in an emergency. In a delirious older adult the rule runs the other way from the usual instinct: benzodiazepines generally worsen delirium and are reserved for the case where alcohol or sedative withdrawal is the cause, so a low-dose antipsychotic is preferred otherwise. Remember too that benzodiazepines can cause paradoxical disinhibition — more agitation, not less — particularly in older adults, children and people with cognitive impairment or a brain injury, and that giving more of the same drug when that happens compounds it. Any patient given repeated antipsychotic doses needs monitoring for neuroleptic malignant syndrome — hyperthermia, lead-pipe rigidity, autonomic instability, reduced consciousness and a rising creatine kinase, developing over one to three days — which is distinguished from serotonin syndrome, where the picture is hyperreflexia, clonus (worst in the lower limbs), tremor, dilated pupils, diarrhoea and an onset within hours of a serotonergic drug or dose increase. Both are hyperthermic emergencies; neuroleptic malignant syndrome means stopping the antipsychotic and supporting the patient (with dantrolene or bromocriptine in severe cases), while serotonin syndrome means stopping the serotonergic agent, with cyproheptadine as the specific antidote. Rigidity plus slow onset points to neuroleptic malignant syndrome; clonus plus fast onset points to serotonin syndrome.

An SBAR escalation for the fourth row of that table sounds like this. Situation: “Dr. Okafor, this is Priya, RN on Unit 4. Mr. Delgado in room 12 has severe restlessness and I think this is akathisia, not agitation.” Background: “He is a 34-year-old man admitted three days ago with a first episode of psychosis, started on haloperidol 5 milligrams twice daily, dose increased yesterday.” Assessment: “He is pacing continuously, cannot sit for more than thirty seconds, says his legs feel like they are crawling, and he is asking for help rather than being threatening. Vital signs stable, temperature 36.9, no rigidity, glucose 96.” Recommendation: “I am requesting you review him now for akathisia, and I would like an order for propranolol or lorazepam and consideration of reducing the haloperidol. He should not receive another antipsychotic dose until you have seen him.”

Debriefing: the patient and the staff

Debriefing after a restraint, seclusion or serious aggressive incident is a standard expectation, and it happens twice. The patient debrief takes place once the person is settled and able to engage — usually hours later, not during post-crisis depletion. Its purpose is to repair the therapeutic relationship, hear the patient’s account, identify what triggered the escalation and what would have helped earlier, and update the care plan and any advance preferences accordingly. Ask directly: what set this off, what did we do that made it worse, what has worked for you before, what should we do next time. Acknowledge the experience honestly. “Being held down is frightening and I am sorry it came to that” is not an admission of liability; it is trauma-informed practice.

The staff debrief is a separate meeting, ideally within twenty-four hours and involving everyone present, including assistive personnel and any student. It reviews the timeline, whether the earliest warning signs were noticed and acted upon, whether the least restrictive options were genuinely tried, whether roles and the single-spokesperson rule held, whether monitoring and documentation met standard, and what the unit will change. It is also where staff distress is acknowledged. Nurses are assaulted at high rates and often normalise it; injuries and near-misses must be reported through the formal system, not absorbed. Note the distinction drawn earlier: an operational review of a clinical incident is appropriate and useful, whereas mandatory single-session emotional debriefing of trauma survivors is not supported by evidence — offer support and follow-up rather than compelling anyone to narrate their feelings.

Finally, the wider concept of postvention: the coordinated response after a suicide death. When a patient dies, the surviving patients, the family and the staff all need a planned response — accurate information, avoidance of sensational detail or method description, identification of others at elevated risk through exposure, access to support, and a review that is conducted to learn rather than to blame. Clinicians who lose a patient to suicide experience grief, self-doubt and sometimes post-traumatic symptoms, and a culture that treats that as weakness loses good nurses. Plan the postvention response before you need it.

Words worth knowing