Understanding how risk works

by TheEditor

Categories: Investigative, Law, Reporting

To ordinary people risk is something felt and seen. If something is dangerous you expect normally to be aware of it, like the edge of a cliff or an open manhole on some pavement. This post should be read together with Risk – how is that to be managed?

If a large dog is growling near your leg, you hear it, see it and if you wait too long you’ll feel it. Do you wait for your leg to be eaten to say, “Oh dear – there is a risk of people’s legs being eaten by this dog” ? One would hope not. Or in psychiatric circles, should we wait until a fork is stuck into somebody’s eye or arm to say “There is a risk of that patient sticking forks into people’s bodies.“? Or should we say, “Well, he’s only ever used pens to attack people, so no evidence that he would use a fork.“? I have heard such statements in real practice. In a previous post I touched briefly on what is risk. I dig deeper in this post.


I cover a few hurdles in risk assessment, evaluation and control. I will look into assessment of risk and how I think it ought to be used in psychiatric circles. I have seen hundreds of risk assessments over the last 25 years. I have never seen sound evaluation of risk or controls exerted over the risks identified.  It sounds like a strange thing to say when at the end of every HCR-20 is a set of scenarios that contextualise risk. I am aware that it may not be in the risk assessment itself. But if it is not in the risk assessment and therefore not used, it is useless.  In other words I have seen no true evidence-based use of risk assessments.  My perception is that risk assessments like the HCR-20 are done mainly to enable the CQC  and organisations they regulate, to tick a box. That’s my honest perception.

[AC_PRO id=913]

Psychological hurdles

Prediction and uncertainty

People know that the future is unknown to them. Human nature though, yearns for someone to foresee the future with great accuracy. Why? Uncertainty creates anxiety. People do not like uncertainty or anxiety. Although no one will admit to it, what people want is for someone to say “Mr X will stab someone on the corner of North and West Streets between 10AM and 11AM on Monday 15th August 2024.” That sort of prediction may not be believed but people still want it. Why? It is clear, specific, bold and provides certainty. Certainty? Yes – it provides ‘certainty’ at a psychological level, even if unrealistic.

Contrast the above with, “I can’t say when Mr X will stab someone. It could be some time 6 months after he stops taking his depot medication“. How useful is that? Some will say the statement is not useful. That’s because they’d be focusing on the ‘stabbing of someone’ at some unclear time in the ‘future’. There is no clarity or psychological certainty about when risk may materialise. Instead workers need to focus on supervision to encourage and educate Mr X not to stop his depot for any length of time. It is how risk-statements are used, that is more important.


I may have said the following before. Prediction is taken by most people to mean that some thing will happen. That immediately sets up what I call ‘the silent binary’. It is silent in the minds of others.  It is a binary because either something in the prediction happens or it does not. In reality nobody should relish a prediction such as “Mr X will stab someone on the corner of North and West Streets between 10AM and 11AM on Monday 15th August 2024.” The reason should be obvious – if the prediction comes through, there will be serious trouble for everybody – and the person making the prediction should feel no sense of accomplishment.

Words such as predict and prediction are used very loosely in psychiatry. The words are conflated and confused with probability (see below).


It is almost impossible to make probability estimates about ‘an individual’ carrying out some risky act in psychiatry e.g. jumping off a high building in a suicidal act*. If you ask any psychiatrist what is their percentage risk estimate for an attempted suicide, you would probably get an answer such ‘It depends‘. What they mean is ‘they can’t answer’.  I said almost because the era of quantum computing is upon us – and maybe a suitable quantum computer (if one had access to it) could give some figures for ‘an individual’. Conventional high powered computers can give estimates for groups of individuals with the characteristics of ‘an individual’ in various scenarios. But almost no one outside of a major research centre has access to such computing power. Probabilities that apply to groups are not the same as probabilities that apply to individuals. [*To be clear, I am not saying that if some one jumps off a high building (without a parachute) that there is not a 99.99% chance of death.]

An individual – the individual  – is what we deal with everyday. The individual may resemble a large group of similar individuals but minor circumstances may affect outcomes. For example, in one situation I know of, a chap dispassionately told his brother “You shouldn’t use drugs.” That triggered an assault and major scuffle involving use of a knife. The fracas was broken up and a murder avoided by their mother throwing a chair between them.  No psychiatrist sitting in their office could contemplate  warning anyone not to say “You shouldn’t use drugs…” in order to avoid the probability of a murder.

Obviously if a patient has serious anger management problems – and has never been amenable to therapy – then knowing and avoiding known triggers may be a plan.


As I said before ‘risk is simply the probability of adverse outcome’. Whilst most people will accept that , the definition does not say many things that people need to know.

Problems with risk assessment

  1. How is the probability to be quantified or estimated?
  2. How reliable are such risk estimates?
  3. Lack of consensus on risk estimates.
  4. How can risk estimates be acted on?
  5. What is the precautionary principle?
  6. What are the economic, legal and social impacts with acting on risk estimates or principles?

Asteroid impact

Obviously ‘asteroid impact’ on the earth has little to do with psychiatry. But how the risk is conceptualised is important in exploring this topic. Everybody knows that the risk of an asteroid – one large enough – striking the earth and wiping out all or most of humanity, is pretty small. What ordinary people might want to know is the chance of  such a catastrophic event. If they hear something like (for example) ‘1 in 100 million over the next 30 years‘, that’s reassuring as they can move on swiftly. But unordinary people – such as leaders of major economic powers – might look deeper. For them 1 in 100 million over 30 years wouldn’t mean that such an asteroid cannot fall in the next 5 or 10 years. What this means is that the value of a risk assessment is different based on ‘who you are’ and the nature of impact you might conceptualise. Joe Bloggs walking down the street is simply concerned that an asteroid won’t fall on his head any time soon. If you’re the leader of a Nation, the life of your Nation and its people is a broader matter that is considered over a longer time frame than ‘any time soon’. Bloggs is mainly concerned that his house won’t be flattened. Nation Leader is concerned that his country’s economy and people are not flattened at any time over the next few years . Individuals with lower levels of responsibility engage with risk differently compared to those with much higher levels of responsibility.

The problem with this is that people with higher levels of responsibility are in well in the minority. In a world where ‘numbers’ matter, where group perceptions and opinions matter most – in political decision-making for example – high level decision makers tend to have a more difficult time. Would tax-payers accommodate a flat tax of 1% on all income to deal with the risk of catastrophic asteroid impact? You could imagine the heated debates that would surround that. Who would feel the tax most?

Virus and tsunami impact

I won’t dwell too long on this. Most of what I said above applies. The COVID pandemic was not ‘predicted’. But it is impact could have been much less if the world was prepared. The world was warned to prepare for a virus like COVID (see video below). Likewise there are (separate) sound references to the world being unprepared for tsunamis. These sorts of risks are seen as ‘too small’ or too far away. What really matters was not the size or perceived remoteness but the impact. To be clear – I am not saying that all potential catastrophic outcomes should lead to spending like no tomorrow. I’m not saying that, because someone might infer I am implying that ‘if patients ‘may be suicidal’, means you put everybody on 1:1 observations‘. It is not what I mean. What I am saying is that there must be a well reasoned strategy to control risk. The absence of that well reasoned strategy is what courts disaster. Preparedness means readiness for a sound response. Unpreparedness is readiness to be a victim of chaos.

Quantification/Estimation/Evaluation (QE2)

QE2 in scenarios of asteroid impact are subject to mathematical analysis. QE2 in everyday practice of psychiatry is not subject to mathematical analysis. The business is done by gathering a set of facts in instruments such as the HCR-20, and a range of other risk assessment tools that cover aggression, violence, suicide and so on. Experience is absolutely necessary.  Trained professionals then think about the facts and various scenarios, and come to some ideas about how high the risk is. An inexperienced member of staff may report that a patient is not suicidal because “He said he wasn’t.” Or it is with some frequency heard from the inexperience or not well trained, “No suicidal ideas reported“. Those sorts of words should always lead to an expert re-evaluation of risk.

Over the years – after using many risk assessment tools – I still want to know one main thing: “Over the next few days, is the risk high, medium or low? [In whatever domain I’m looking at]” It is one of the simplest questions but the hardest for professionals to answer. Normally when I’ve asked it, I tend to get further consideration of variables not followed by a categorisation into ‘high medium or low’. In other words I tend to get a formulation. But I’ve already passed formulation. I’m into the stage of risk-control. So, if the risk is high I need to know what we’re doing to control it.

I can’t think of a single person who would say that crossing a busy road with the eyes closed, is low risk. What’s different in psychiatric circles? Crossing a busy road is about ‘things’ one can see and take physical action to avoid. In psychiatric risk estimation, there are no equivalents of ‘the road’, the ‘vehicles’, ‘their speed’ and ‘density of traffic’. Tangibles like those are not easy to come by in psychiatric work. In short term risk evaluation, what we may have is frequency of incidents, degrees of emotional upheaval, and disturbed mental state. When the latter are serious it is easy to say ‘high risk’ of say violence.

But there are other problems, for example there are some patients who say little, conceal their mental states, have told untruths in the past, and there may have been infrequent incidents. Low frequency and lack of evidence does not mean low risk. It could mean high risk. Why? All it means is that the probability of an adverse event happening in the nebulous future is disconnected from an evidence-base. There is an unseen risk lacuna. How is this realised? Many a hospital has been caught out especially around winter time with patches of suicides. It is well known from suicide investigations, that the quietest unassuming patients are the ones to watch for suicide risk. In pre-hospital medicine – where trained medical professionals go out to rescue injured people – it is well known that those who are not shouting in pain are the ones most at risk, if they’re not already dead (obviously). It is pretty standard practice to prioritise those who are alive and in a semi-conscious or unconscious state. That’s where lives will be lost first.

Fear and anxiety

At a practical level I need to know – with all the variables we currently know – ‘Is the risk over the next few days high, low or medium?’ I suggest that hesitance in categorising the risk is about avoidance of finding an estimate that was wrong. Human beings do not like to be wrong. If you say the risk of suicide is low, after considering everything, and then a patient ends up dead on a train track a few days later – the worry is ‘What will happen to your career?’

Health professionals rarely express that sort of worry but it is there in the backdrop. If you agree that that happens, what it means is that the risk to careers is a barrier to expression of risk statements about patients. To be clear, I am not saying that this happens all the time or most of the time. I cannot say how often it happens or what is the evidence base. What I am saying is experience-based over the course of the last 20 years. It happens often enough. How do I know? I see it among professionals who – covering their anxiety intellectualise non-commitment to categorisation. You do not need to be a psychiatrist to perceive when anybody is keeping a straight face, in covering some emotion. You know there is anxiety when occasionally there is mention of ‘being sued’. In those scenarios I would ask, “Sued on what grounds?” – and I’d often not get an answer, or would get “I do not know.. just being sued.

Never events

These are things that must ‘never happen’ again. This means 100% never again.

According to the National Quality Forum (NQF), “never events” are errors in medical care that are clearly identifiable, preventable, and serious in their consequences for patients, and that indicate a real problem in the safety and credibility of a health care facility. There are other kinds of never events in each industry.

The reality is that no system involving the work of people or products made by people will be perfect. Systems can be corrected where there was an obvious flaw but the correction itself does not mean that the system will not fail again.

A nurse who is on constant suicide watch over patients must never fall asleep. That seems reasonable enough. However, it continues to happen in a minority of instances. The issue seems simple enough but it rather more complicated – and I am unable to go into that level of complication here.

Cost first, patient last.

This is anxiety about cost. All services are concerned about cost. In controlling risk of suicide for example – in inpatients – there is a cost to special observations (often known as 1:1). In 2005 a certain organisation serving the UK’s NHS clearly stated in writing that they “…had a projected overspend of  £8 million and that they could not afford to continue paying extra costs of £4000/week for special nursing observations.” (for the record, the organisation is not named or identifiable in this post but the author has hard evidence in hand). ‘Price Agreements’ in the existing contract required for authorisation to be sought within 72 working hours of starting the special nursing observations. Authorisation was, “assumed unless a written response in objection is received from yourselves within 72 hours of notification.” But unilaterally the fund holders decided, “no further special nursing observations would be approved, or paid for, above the agreed daily rate, where prior approval has neither been sort or given.” The material change may not be obvious to those who have not studied the words carefully. In effect funding was subject to the fund holder being given 72 hours notice of when special observations would be required. How would any psychiatrist know 72 hours in advance, when a patient would need special observations and/or an increase in special observations? Words on this page do not fully give context because there was a lot more going on. Therefore parsing the sentences and interpreting the words may come up with different meanings. The author is the only person who knew exactly the full contexts and meanings.

The above cost pressures are not isolated – but evidence of how they affect health care are hard to come by. On the above occasion it was by chance that the fund holders were probably dim enough, to put such things in writing. Make no mistake, funding of the prevention of suicide is under cost pressures all the time. The general public and professionals outside of health services will not know the grain of the detail.

Risk control is costly. Cost causes anxiety, which impairs judgement about controlling risk. Clinicians have a duty to use resources of any institution carefully. However, cost should not enter into any equation about providing life-protective care. Treatment under S145 Mental Health Act 1983, includes care (inter alia).

All health care providers have a duty to put the patient first. It is written. However, no one enforces the rule – so it is a useless thing. There is lip service about ‘patient first’. In reality it is cost first.

Risk hierarchy

The infographic shows one example of how risk is considered at different tiers in a hypothetical business.

Workers at the shop floor sense different risks compared to those higher up the tree. Shop floor workers are usually the lowest paid but have the greatest need to protect their jobs.

Those above shop floor management in most organisations can ‘pick up’ and move to other organisations with less discomfort. They are not as concerned about getting the sack. They’ll jump or move before that happens.

Senior management is concerned mostly about the whole organisation. They’re not too bothered about sacking people lower down the tree once the CEO and shareholders are happy. It is well known from statistics that company shares tend to rise when there are mass lay offs and redundancies. So – in general high level folk in many organisations aren’t too bothered about policy breaches at the shop floor, unless those have potential for fines or legal action.

You may not agree with the above. Perhaps the organisation you work for functions differently. The example here even if a bad one, is only about the hierarchy and how risk is perceived.

In health services – say in an A&E department – the trauma consultant is concerned primarily with saving a life. The pharmacist may be concerned with local policy and the ‘correct dose’ of a medication. The nurse may be concerned with sterile equipment. The consultant in a difficult trauma case, may have to bend or disregard policy to save a life. If he does and saves a life, everybody is happy. If he doesn’t the guillotine may fall on him because the breaches have a probability of becoming unacceptable.

Risk control

This is about doing something. The obvious issue here is that doing something must be about a probability that has been identified in a specific set of emerging circumstances.

Controlling risk is about intervening in a set of circumstances so that the probable adverse outcome has less chance of materialising.

But there are major problems here:

  1. Reaction time – will the controlling response be in time or timely.
  2. Will the controls be followed up and adjusted to meet dynamically changing circumstances.

To put this into perspective, consider someone responding to the shot of a fast moving tennis ball with their racket. The primary risk is ‘missing the ball’. The tennis player has to make that swing whilst drawing on a lot of experience and training. There is not time to mathematically compute (in the human brain) the physics and mathematics of the ball’s trajectory and the response of the tennis racket.

But in mental health risk management, the objects we are controlling are quite often not tangible or measurable. Nobody I know over the last 30 years of psychiatric practice (except me) would measure the degree of psychosis or depression with rating scales at important turning points in a patient’s case.

Dynamically controlling risk of ‘a disturbed patient’ means well-qualified assessment of mental state and risk factors, and responding with appropriate and proportionate measures.

Slipping past the racket

I use the word racket, in preference to ‘net’ because above, I used the analogy of a tennis player with a racket aiming at a fast moving ball. You could conceptualise this in cricket or football if you prefer analogies there.

So the racket has missed the ball? Why?

  1. The ball – it was moving too fast.
  2. Wind conditions were terrible.
  3. Something is wrong with the racket.

The above is what happens in many situations where there has been failure to catch and control risk.

No one says:

  1. It was my/our fault – we reacted too slowly.
  2. We were disorganised and lacking coordination.
  3. We lacked expertise.
  4. Our team was overworked and several were sent on annual leave at the same time.

Requirements for good risk control

  1. Situational awareness.
  2. Expertise and experience.
  3. Well resourced, organised and coordinated teamwork.
  4. Sound responsivity (like reflexes).


Other posts that may interest you...

Estimating cost of expert reports

This post will deal with psychiatric reports, mainly for the courts. These reports  often appear simple but in most cases based on experience, they involve complex and/or demanding underlying issues. ...

Read more

Car repairs v human repairs

Today I was at my Citroen garage to have a malfunctioning driver’s side door handle replaced. I had problems getting in and out of the car. The problematic door handle ...

Read more

What is a treatment plan?

Experience over the last 30-odd years has led to an evidence-based view that treatment plans are woefully deficient across health services. The evidence seen is that treatment plans tend to ...

Read more