Suicide by psychiatric patients: Nature of risk,risk categorisation and preventability

Clinical characteristics

Clinical characteristics of the sample are given in Table 2. Psychiatric care was coordinated by multidisciplinary input through the Care Programme Approach for 40.3% of the sample. Around 1 in 10 (11.5%) had missed their last appointment, and 20% were non-compliant with prescribed medication a month before they died. Psychiatric diagnoses are given in Table 3 and include both primary and co-morbid psychiatric diagnoses. All patients assessed had psychiatric diagnoses, and co-morbidity was common: 52.5% had at least two diagnoses, and 19.7% had three diagnoses. The majority (70.5%) had a history of psychiatric admission, and a small minority (6.3%) were inpatients at the time of their death.

OPEN IN VIEWER

Table 2.

Clinical characteristics of the sample.

Characteristics	Categories	Male (%)	Female (%)	Total (%)
Duration of psychiatric history	<12 months	22.2	6.3	18.0
	1–5 years	26.7	18.8	24.6
	>5 years	51.1	75.0	57.4
Presence of a major physical illness	No	71.4	75.0	72.4
	Yes	28.6	25.0	27.6
Number of psychiatric admissions	None	28.9	31.3	29.5
	1–5 admissions	53.3	43.8	50.8
	>5 admissions	17.8	25.0	19.7
History of deliberate self-harm*	No history	27.3	31.3	28.3
	Within 3 months of death	22.7	0.0	16.7
	More than 3 months before death	27.3	12.5	23.3
	Both within and more than 3 months of death	9.1	31.3	15.0
	Lifetime history but recent history not known	13.6	25.0	16.7
History of violence*	No history	60.5	93.8	69.5
	Within 3 months of death	11.6	0.0	8.5
	More than 3 months before death	11.6	0.0	8.5
	Both within and more than 3 months of death	0.0	6.3	1.7
	Lifetime history but recent history not known	16.3	0.0	11.9
History of alcohol misuse	No history	58.5	62.5	59.6
	Within 3 months of death	12.2	0.0	8.8
	More than 3 months before death	4.9	6.3	5.3
	Both within and more than 3 months of death	12.2	18.8	14.0
	Lifetime history but recent history not known	12.2	12.5	12.3
History of drug misuse	No history	69.8	81.3	72.9
	Within 3 months of death	7.0	6.3	6.8
	More than 3 months before death	2.3	6.3	3.4
	Both within and more than 3 months of death	11.6	6.3	10.2
	Lifetime history but recent history not known	9.3	0.0	6.8
Adverse life event within 3 months before suicide	No	53.8	66.7	57.4
	Yes	46.2	33.3	42.6

*p < 0.05.

OPEN IN VIEWER

Table 3.

Common diagnoses.

Diagnoses	Male (%)	Female (%)	Total (%)
Depressive disorders	57.8	55.6	57.1
Schizophrenia and other psychoses	18.9	11.1	23.8
Anxiety disorders	24.4	22.2	23.8
Personality disorders	17.8	11.1	15.9
Alcohol misuse and dependence	13.3	11.1	12.7
Drug misuse and dependence	11.1	11.1	11.1
Bipolar affective disorder	6.7	16.7	9.5
Adjustment disorder	4.4	11.1	6.3
Organic disorder	2.2	5.6	3.2

Almost one in five (22.7%) were seen in an accident and emergency department for self-harm a month before the suicide, following which 55.6% were referred to their own mental health teams and 11.1% to a new team for follow-up; 11.1% were provided consultation only. At the time of death, 25.5% of patients were reported to be not compliant with drug treatment. The main reasons for non-compliance were recorded as lack of insight (9.8%), side effects (3.9%) and others (11.8%).

Methods of suicide

Methods of suicide are given in Table 4. There were no significant differences between sexes. However, more females (44.4%) than males (33.3%) attempted hanging. Among the cases of self-poisoning, 21.6% used psychotropic drugs, 8.7% used benzodiazepines or hypnotics, 26.1% used paracetamol or paracetamol/opioid combination, 17.4% used opioids and 17.3% used other drugs or poisons. This information was not known in 8.7% of cases. Alcohol was consumed as part of the suicide act in 10.3% of cases. In 4% of cases, the death was part of a suicide pact. There were no cases of homicide before suicide.

OPEN IN VIEWER

Table 4.

Method of suicide.

Method of suicide	Male (%)	Female (%)	Total (%)
Self-poisoning	35.6	38.9	36.5
Hanging	33.3	44.4	36.5
Carbon monoxide poisoning	8.9	5.6	7.9
Jumping from a height	8.9	0.0	6.3
Jumping/lying before train/vehicle	4.4	5.6	4.8
Drowning	2.2	5.6	3.2
Suffocation	4.4	0.0	3.2
Slashing wrists	2.2	0.0	1.6

Month and season of suicide

The highest number of male suicides occurred in May and June (17.8% each month), and female suicides in September and October (16.7% each month). Based on the seasons, there was a clear downward trend, from the highest suicides in the spring to the lowest in the winter: March–May (spring) 34.9%, June–August (summer) 28.6%, September–November (autumn) 22.2% and December–February (winter) 14.4%.

Last psychiatric admission

Information about last admission was available for 90.7% of patients who had a psychiatric admission. Amongst these, the last admission was voluntary in 76.9% of patients, while the rest were detained under the Mental Health Act 1983. Admission was to the local hospital in 84.6% of cases; the duration of admission was less than seven days in 20.5% of cases, one to four weeks in 43.6%, 5–13 weeks in 10.3% and more than 13 weeks in 20.5%. It was a readmission within three months of a previous admission in 23.1% of cases. Most of the discharges were planned (84.6%), and in 5.1% it was a self-discharge or at the request of the patient.

Duration between discharge and suicide

The range of the duration following last discharge from psychiatric ward and suicide was vast, ranging from five days to 17.5 years. Further analysis suggested that 18.2% died within 30 days, 12.1% after one to three months, 18.2% after three to six months, 15.2% after 6–12 months, 12.1% after 12–24 months and and rest (24.2%) after two years. That means, cumulatively, that 63.6% died within the first year, while 75.8% died within the first two years, highlighting the risk in the two-year post-discharge period.

Last contact

A considerable proportion of patients (17.5%) had last contact within last 24 hours of suicide: 31.6% within one to seven days, 26.3% within one to four weeks, 10.5% within 5–13 weeks and 14% more than 13 weeks (three months) before the suicide. This contact was face to face in most (87.9%) cases and by telephone in 8.6%.

Clinicians who had last contact were from different professional backgrounds and roles. They included community psychiatric nurses (50%), junior psychiarists (patient’s own team; 37%), consultant psychiatrists (36.4%), ward nurses (13%), social workers (7.1%), occupational therapists (7.1%), junior psychiatrists on-call (7%), support workers (6.8%), alcohol workers (2.6%) and others (25%). Care coordinators were included in the last visit in 34.5% of cases.

Most of the last visits were routine or non-urgent (69%). Around 10.3% were requested urgently by patient or family and 15.5% by professionals; this also included assessment after deliberate self-harm in 3.4% of cases.

Information about psychopathology at last contact was available for 87.3% of patients. Among these, 21.8% had none, 20% had one and another 20% had two, with the mean being 2.1 ± 1.7. There were no sex differences. Commonly reported mental health problems at last contact were emotional distress (51.7%), depression (47.5%), hopelessness (26.8%), suicidal ideas (21.4%), delusions/hallucinations (18.6%) and recent deliberate self-harm (15.5%). In addition, there was deterioration in physical health in 13.6% of patients, increased use of alcohol in 7%, increased use of other substances in 8.5% and hostility in 5.2%.

At last contact, degrees of long-term and immediate risk of suicide were considered as no risk (9.8% and 38.5%, respectively), low risk (68.6% and 46.2%, respectively), moderate risk (19.6% and 11.5%, respectively) and high risk (2% and 3.8%, respectively). These risks correlated significantly (p < 0.001) with the number of symptoms at last contact. The presence of suicidal ideas did not influence risk categorisation significantly, whereas having hopelessness led to a higher (moderate and high) risk category for 40% compared to 11.4% with no hopelessness, which approached statistical significance (p = 0.053).

Views on prevention

Among all suicides, 66.7% were considered not preventable and 7.9% as preventable, with 15.9% unsure. There was no response regarding others. We compared preventable and non-preventable suicides for associated demographic and clinical variables. Factors which differentiated preventable with non-preventable suicides significantly were absence of depressive illness but presence of delusions or hallucinations at last contact, although the preventable proportion was considerably low. Factors that could have made suicides significantly less likely as expressed by the clinicians were closer supervision of the patient (34.5%), better staff training in risk assessment (17.9%), access to psychotherapy (16.7%), better patient compliance with treatment plan (14%), improved staff communication (10.7%), closer contact with patient’s family (7.4%), availability of dual diagnosis service (7.3%), availability of other treatment (6.5%), better liaison with different services (5.8%), access to alcohol services (5.5%), increased staff numbers (3.6%), more psychiatric beds/hospital facilities (3.6%) and higher dose of medication (1.8%).

Risk factors

Risk factors were common, and their multiplicity in the sample was striking. Historical risk factors, ¹¹ which are fixed and unmodifiable but are nonetheless indicative of suicide risk, were very common in the sample. For example, 71.7% had a history of self-harm, 31.7% within three months, with 22.7% being seen in accident and emergency after self-harm within a month. Furthermore, 70.5% had a history of previous psychiatric hospitalisation and 40.4% a history of alcohol misuse, while substance use (27.1%) and violence (30.5%) were also common.

Similarly, stable risk factors indicating long-term enduring risk were observed in a proportion of patients. Among the age groups, there were 12.7% in their 60s and 9.5% in their seventies, emphasising the well-known high risk of suicide in older adults. ¹² Almost half of the suicides were between 30 and 49 years of age, highlighting the continuing risk for young adults.

Considering sex, males were represented in a considerably greater proportion (71.4%) in this study, which is usually reported in suicide populations.^2,13 There was grave concern regarding marital or companionship status: 42.4% were single, 20.3% divorced/separated and 8.5% widowed. Only 28.8% were married or cohabiting. Lack of a companion and related loneliness and inadequate socialisation complicates the scenario. Providing opportunities or scope for socialisation for those who are single and lonely may help.

Personality disorder or traits are consistently being reported in suicides. Around 15.9% had diagnoses of a personality disorder; there was no scope to know how many more had dysfunctional personality traits in this audit. There is a greater understanding regarding patients with personality disorder being more vulnerable to suicide.^14,15 Although personality disorders are quite common in clinical populations and inpatients, there are concerns regarding the availability of appropriate services. ¹⁶ Identifying and providing appropriate therapeutic interventions to these patients may be helpful.

Modifiable risk factors

Current, dynamic risk factors which are modifiable and which can inform clinical intervention were identified in a considerable proportion of patients. For example, the psychopathologies in the last visit included suicidal ideation, hopelessness, emotional distress, depression and even psychotic features. There was increased use of substances. Treatment non-adherence and deterioration in physical health were also reported. A considerable proportion had current life events, which are known factors associated with suicidality.^17,18 All these are common in clinical populations, which make it difficult to predict, but measures can be put in place for these factors. However, it may be difficult to suggest that the effectiveness of the measures may be adequate to prevent the outcome in some cases.

Future risk factors which can also suggest intervention measures to avoid or decrease the intensity of their impact on suicidality were not specifically studied. However, access to the preferred method of suicide was available to many in the form of medication, with many available over the counter. Hoarding tablets is a particular issue in this regard. Access to sharp objects, high-rise buildings and scope for hanging are also methods that are difficult to avoid.

It is clear that the issue of risk factors is complex. They were multiple and concentrated in this population. Sociodemographic, historical, clinical and current dynamic risk factors were commonly observed. Single status, living alone and being without a job were common. There were multiple psychiatric diagnoses, co-morbidities (both physical and mental), substance use and stressful events in a considerable proportion of patients. For most of these people, risk factors continued in the long term. It has been reported that in most patient suicides, along with the chronicity of the illness, the risk of suicide continues indefinitely. ¹⁹

Risk and psychopathology at last contact and predictability

Most patients in the sample were seen by clinicians either in accident and emergency or in other places, including their home, and various known risk factors were observed. Whether any of these could have led to prediction of suicidal outcome leading to a change in the course of intervention deserves discussion.

Suicidal cognitions are common in the general population and stressful situations.^20–22 In a study in the general population, the lifetime prevalence of suicidal cognitions such as life not being worth living (11.5%), death wishes (8.2%), thoughts of suicide (4.8%) and seriously considering suicide (2.6%) have been reported. ²² Recently, in the WHO SUPRE-MISS community survey, which included eight countries, there was wide variation in prevalence rates of suicidal ideation (2.6–25.4%), plans (1.1–15.6%) and attempts (0.4–4.2%) by a factor of 10–14 across sites. ²⁰

Similarly, suicidal cognitions are extremely common in clinical populations. A considerably high proportion of people with current mental illness reported life not being worth living (59.1%), death wishes (40.8%) and suicidal ideation (30.6%). Suicidal plans and cognitions were high in people with a history of mental illness and in those with a current physical illness. ²³ Reported rates of suicide attempt for schizophrenia (23–34%), bipolar disorder (around 50%) and major depressive disorder (44%) are usually high.^24,25 In patients with obsessive compulsive disorder, reported current rates of suicidal ideation varied from 28% to 46.1% and from 59% to 62.3% during the lifetime, with a history of suicide attempt in 27% of cases.^26,27 Suicidal communications are also commonly observed before a suicide attempt, irrespective of the outcome. ²⁸

With a widespread prevalence of suicidal cognitions in both the general and clinical populations, predictability of a specific suicidal outcome in an individual patient has remained poor. In this audit, suicidal cognitions and risk factors were common. However, it appears that these could not be clinically linked prospectively to the possibility of completed suicide or to predict suicide.

Perceived risk of suicide at the last contact

Clinicians categorised short- and long-term risk of suicide at the last contact. In most cases, immediate risk was considered either nil (38.5%) or low (46.2%), which is similar to national figures, ¹ highlighting that even if people are perceived to be at a lower risk of suicide, they may still complete the act. Although clinicians’ assessment of psychopathologies and risk categorisation correlated well, it obviously could not identify individuals who would eventually die by suicide. So, it appears clinical assessment itself is unlikely to be able to predict suicides. While suicide risk can be generally apprehended, it is the predictability which has been uncertain. ²⁹ Based on these observations, it can be stated that prediction of suicide following clinical assessment is difficult.

Even risk scales do not help in predicting suicidality. Although the use of scales to predict risk were not evaluated in this audit, it has been reported recently that risk scales do not accurately predict repeated self-harm and suicide.^30,31 It is suggested that risk-assessment tools and scales should not be used to predict future suicide or repeated self-harm, and they should not be the sole criterion to decide on treatment options.^32,33 However, the scales can be used as a part of a holistic clinical assessment. ³⁴ The current evidence base is not robust enough to suggest any particular scale for routine clinical use to predict suicide.^30,35

A related area of risk assessment and predictability is staff morale following a suicide, especially when staff members had the last contact with the patient. Although staff morale was not evaluated specifically in this study, it is known that patient suicides affect clinicians both personally and professionally.^36,37 It may be a particular concern when the immediate risk was considered to be nil or low at the last visit. It is pertinent to highlight from the findings of this study that risk categorisation by clinicians is unlikely to predict suicide accurately. Studies are needed in order to evaluate staff and team perspectives following a suicide, their morale and how best to support it.

Method of suicide

Common methods of suicide in this population have been hanging and self-poisoning. Hanging in the community is known to have limited scope for prevention. ³⁸ The prevention of methods such as jumping from height or jumping before a vehicle or train is also difficult unless specific measures are taken at already identified hot spots. Restricting medication used for overdose (e.g. limiting the number of paracetamol packs that can be purchased, ³⁹ dispensing medication for fewer days at a time for vulnerable individuals, and prescribing medication that is comparatively less lethal in overdoses ⁴⁰ ) have been suggested as helpful strategies. However, these may be effective in only a minor proportion of cases, as observed in this study. It is difficult to remove the risk of self-poisoning completely, as people may hoard prescribed medication or buy many over-the-counter drugs.

Risk related to psychiatric admission and discharge

Can psychiatric admissions prevent suicides in the long term? Often, it is perceived that admission may be protective, which may be true in certain circumstances and acute states. However, there are contrasting observations about psychiatric admission and the risk of suicide. It has been observed that a history of psychiatric admission and psychiatric inpatient status are associated with an increased risk of suicide.^41,42 At one hand, risk of suicide and self-harm are considered factors for voluntary or compulsory psychiatric admission; ⁴³ on the other hand, it has been suggested that psychiatric admission may have a causal role in a proportion of inpatient suicides. ⁴⁴ However, there are complex issues surrounding compulsory detention and suicide. ⁴⁵

In this study, 70.5% of people had a history of psychiatric admission, while 6.3% died while in the hospital, and a considerable proportion (63.6% within one year and 75.8% within the first two years) died within a relatively short time period after discharge. It is known that suicide and self-harm risks remain high following discharge from a psychiatric unit. A meta-analysis suggested that during first three months after discharge, the suicide rate could be 100 times the global suicide rate, and it may be around 200 times higher than the global rate for patients admitted with suicidal thoughts or behaviours. Even many years after discharge, patients with a history of psychiatric admission have suicide rates around 30 times higher than typical global rates. ⁴⁶ It has been suggested that the re-emergence of the stressors leading to suicidal behaviour and stressors exacerbated by hospitalisation are important contributors. ⁴⁷ So, the issue of whether admission can prevent suicide effectively in the long term remains debatable. Irrespective of whether admission prevents suicide effectively, it definitely highlights those individuals who are at risk, and robust preventive measures may be considered for a period of time for these individuals. While on the ward, efforts may be expected to achieve zero suicide.

Preventability

While it is generally agreed that all suicides may not be preventable, ⁴⁸ a considerable proportion could be in most circumstances. However, only a small proportion of suicides in this study were considered preventable by the clinicians, which reflects the challenges in the prevention efforts in this population. Closer supervision, better staff training in risk assessment, access to psychotherapy, better treatment adherence by patient and improved communication are some of the frequently suggested factors that might help.

Considering the complex psychosocial issues associated with patient suicides, the importance of non-clinical resources should be realised, and strengthening them may help in prevention efforts. This would involve family and friends, ⁴⁹ along with the general public and other organisations. In this regard, there is a role for robust public mental health education.

Limitations

Detailed information on some of the variables was not available (e.g. the nature of stressful life events, personality traits, etc.). In a small proportion of cases, the exact nature of the coroner’s verdict was not available. Analysis on suicides in the inpatient setting could not be undertaken because of the small sample size. There was no scope to obtain information from the family or other caregivers. So, the observations are specific to clinicians only. There was no option of using scales for further assessments. The views regarding prevention possibilities were taken only from the clinicians. It may be better in future studies to include the suggestions from the relatives and the coroner’s verdict. Considering this is an audit, the findings are mostly relevant locally. However, Wolverhampton is a multicultural city, and the observations may be applicable to the patient population known to mental health services in similar multicultural areas in the UK.