Prediction of suicide in a psychiatric hospital

Meta-analysis of the strength of exploratory suicide prediction models; from clinicians to computers

BACKGROUND

Suicide prediction models have been formulated in a variety of ways and are heterogeneous in the strength of their predictions. Machine learning has been a proposed as a way of improving suicide predictions by incorporating more suicide risk factors.

AIMS

To determine whether machine learning and the number of suicide risk factors included in suicide prediction models are associated with the strength of the resulting predictions.

METHOD

Random-effect meta-analysis of exploratory suicide prediction models constructed by combining two or more suicide risk factors or using clinical judgement (Prospero Registration CRD42017059665). Studies were located by searching for papers indexed in PubMed before 15 August 2020 with the term suicid* in the title.

RESULTS

In total, 86 papers reported 102 suicide prediction models and included 20 210 411 people and 106 902 suicides. The pooled odds ratio was 7.7 (95% CI 6.7-8.8) with high between-study heterogeneity (I2 = 99.5). Machine learning was associated with a non-significantly higher odds ratio of 11.6 (95% CI 6.0-22.3) and clinical judgement with a non-significantly lower odds ratio of 4.7 (95% CI 2.1-10.9). Models including a larger number of suicide risk factors had a higher odds ratio when machine-learning studies were included (P = 0.02). Among non-machine-learning studies, suicide prediction models including fewer risk factors performed just as well as those including more risk factors.

CONCLUSIONS

Machine learning might have the potential to improve the performance of suicide prediction models by increasing the number of included suicide risk factors but its superiority over other methods is unproven.

On Suicide in a South African General Hospital Psychiatric Inpatient Unit

Few studies have reported specifically on suicide in general hospital psychiatric inpatient units. Results of a study over a 10-year period of nine patients who committed suicide while being treated in such a unit are presented. Total admissions during the study period were 6 961 giving an average suicide rate of 1,4 suicides per 1 000 admissions. More males committed suicide than females. Most patients were unemployed with a history of parasuicide and psychiatric hospitalization. Psychiatric diagnoses of schizophrenia, affective disorder, and alcohol dependence predominated. Jumping from high-rise buildings near the hospital was the most common suicide method. Ward milieu and staff changes may have influenced suicide rates and the results are discussed in terms of psychological aspects with reference to system theory.

New Developments in the Understanding and Treatment of Suicidal Behaviour

Recent developments in the understanding, assessment and treatment of suicidal behaviour are reviewed. The accumulated social and demographic knowledge of suicidal populations has proved to be clinically useful in shaping the general level of concern, but at the individual level an understanding of the psychological processes involved in suicidal behaviour is required. Studies examining psychological processes involved in suicidal behaviour is required. Studies examining psychological processes in parasuicidal groups have revealed a number of deficits, including poor interpersonal problem solving, hopelessness about the future, and reduced ability to regulate affect. Research has also begun to look at some of the processes underlying these deficits, such as over-general retrieval of autobiographical memories and reduced anticipation of specific positive experiences. The clinical picture is now more optimistic, with therapies, such as Dialectical Behaviour Therapy, which focus on using problem solving strategies with those clients who are most vulnerable to repeat parasuicide episodes, producing demonstrable delays in parasuicide and reduced risk of repetition. The importance of the relationship between research and clinical practice is emphasized.

Suicide risk scales: do they help to predict suicidal behaviour?

Fifteen rating scales for the estimation of suicide risks are reviewed. The report focuses on methods of scale development and validation. Problems of predictive accuracy, and ways of improving the prediction of suicidal behaviour are discussed.

Preliminary field-testing of a risk estimator for suicide

A 15-item suicide risk estimator, empirically derived from a prospective study of 2,753 suicidal persons, was given a field trial in five clinical settings. Two hundred ninety subjects were included. Volunteer crisis workers needed an average of 4.5 minutes to complete the instrument, and rated its ease of administration 2.5 on a 5-point scale. When compared with clinical ratings, the instrument estimated suicide risk slightly higher than did the interviewers. Scale items that differed from the raters' intuition tended to be omitted more than others. The trial provided valuable information needed to prepare for a large scale evaluation.

Suicide among psychiatric patients of a district general hospital

Twenty-two psychiatric patients who committed suicide while receiving hospital treatment or within 3 months of discharge from psychiatric care were studied using general population statistics, a random sample of 100 psychiatric patients and a control group matched individually with the hospital suicides. The rate of suicide among psychiatric in-patients was over 50 times that in the general population. A higher vulnerability to suicide was detected among in-patients (v. other patient groups), men (v. women) and middle-aged patients (v. older and younger patients). Suicides were differentiated from controls in having suffered more losses (P less than 0.05), being psychiatrically ill for the first time (P less than 0.05), having a past history of deliberate self-harm (P less than 0.05), and not receiving a schizophrenic diagnosis (P less than 0.02). People with mid-life crises, patients who swing rapidly into depression while receiving treatment, and individuals who are judged to be depressed because of personal problems may carry a particularly high suicide risk during psychiatric treatment.

Family loss and hospital suicide

In 19 consecutive cases of inpatient suicide occurring over many years, 18 out of the 19 cases (95 percent) were judged to have a significant "family" loss issue affecting the suicide. In five cases (26 percent), there seemed to be a powerful institutional attachment and loss threatened that attachment. In 13 cases (68 percent), a major family issue was involved which included events that estranged the patient from the family (real or imagined); divorce or separation issues; death or illness of a significant other; a family that insisted on continued treatment (and/or growth). In 13 cases (68 percent), significant communication from or about the "family" occurred which seemed to have an impact on the suicidal behavior often within days of the final event. In 10 cases (53 percent), an actual family visit had occurred within two weeks of the death. These issues are discussed and case examples presented.