Hindsight bias and the overestimation of suicide risk in expert testimony

In Rabone v. Pennine Care NHS Foundation Trust the Supreme Court examined the duties that the European Convention for the Protection of Human Rights and Fundamental Freedoms might place on hospitals caring for informal psychiatric patients.¹ We have grave concerns about the quality of the expert evidence presented to court in this case.²

Melanie Rabone was 24 when on 4 March 2005 she attempted suicide by tying a pillowcase around her neck and was admitted to hospital diagnosed with ‘a severe episode of a recurrent depressive disorder’. By 14 March she ‘had shown sufficient signs of recovery’ to be allowed overnight leave, and on 18 March she was discharged to accompany her family on a week-long trip to Egypt. On 11 April she was readmitted voluntarily after tying lamp flex around her neck. By 19 April Ms Rabone had again shown some signs of improvement. She requested leave and, following a meeting with her psychiatrist and mother where she agreed not to self-harm, 2 days’ leave was granted. She spent most of the next day with her mother, but in the afternoon said she was going to see a friend. She hanged herself from a tree in a local park sometime after 5 pm.²

The court sought expert evidence as to whether there was a ‘real and immediate’ risk to the life of Ms Rabone on the day she was granted leave. The expert psychiatrist engaged by the claimants estimated that Ms Rabone's ‘immediate risk’ of suicide on 19 April was ‘of the order of 70%’.² The Trust's expert was more conservative. He expressed the view that ‘the risk was approximately 5% on 19 April (after leaving hospital) increasing to 10% on 20 April and 20% on 21 April’.² The written judgments do not record how these figures were arrived at, but it is hard to see how they could have been based on what is actually known about the likelihood of suicide by psychiatric in-patients on approved leave.

The suicide of psychiatric in-patients (including those on approved leave) was the subject of a systematic review and meta-analysis.³ Its results suggest that Ms Rabone's depressed mood and previous suicide attempts would have meant that she was more likely to die by suicide than another in-patient without those features. It is possible, using these empirical data and making an assumption of the base-rate of suicide among all in-patients, to calculate the probability of such a ‘high-risk’ patient's admission ending in suicide. Such a calculation, even with an extremely pessimistic base-rate assumption, reveals that the likelihood of a ‘high-risk’ patient dying by suicide while an in-patient is probably no more than 1.2%. Since Ms Rabone's admission lasted 10 days, it is hard to see how a realistic estimation of her risk of suicide on any particular day could have been much beyond one tenth of that – 0.12%. The experts’ estimates, the more conservative of which was accepted by the court,¹ were between 40 and 600 times that figure.

We can only speculate as to how the experts arrived at their estimates, however, the most obvious possibility is that they utilised their clinical judgement based on reviews of Ms Rabone's file. Clinical judgement about the likelihood of future events is known to be affected by a range of well-established weaknesses including the failure to consider known risk factors, an inability to consider co-variation between risk factors, underutilisation of base-rate data, and a range of cognitive biases including confirmatory bias supporting an initial hypothesis.⁴ In this case though, the most potent influence was probably the tendency to see events that have already occurred as being more predictable than they were before they took place. This is referred to as hindsight bias and is one of the strongest and most ubiquitous of the cognitive biases.⁵

The Pennine Care NHS Foundation Trust was found to have failed to avoid a ‘real and immediate’ risk of death by allowing Ms Rabone home on leave when, the court reasoned, her doctors should have refused that leave. The court also reasoned that had she insisted on leaving against advice, her doctors could have, and should have detained her using the coercive treatment provisions in the Mental Health Act 1983. This failure, the court held, amounted to a breach of her human rights. The psychiatrists’ risk calculations formed the basis of the court's finding that there was a duty to protect Ms Rabone's ‘right to life’ under Article 2 of the European Convention, and the breach lay in the hospital's failure to detain her against her will.

The decision in Rabone v. Pennine Care NHS Foundation Trust means that risk calculations have the potential to affect the rights of all psychiatric patients to access leave or to refuse hospital admission or lengthy hospital stays where their suicide risk is thought to be ‘real’ – that is ‘significant’ and not ‘remote’ or ‘fanciful’ – at the time they request to leave the hospital.

The principal duty of the expert witness is to provide accurate, objective and unbiased testimony about complex matters before the court. Now that hospitals have a clear responsibility to protect psychiatric patients thought to be at immediate risk of suicide, if necessary by invoking coercive powers to detain and treat, psychiatric experts must make realistic estimations of the likelihood of suicide based on the facts of the case, knowledge of the literature and careful avoidance of hindsight bias.