Numerous studies have consistently reported that individuals are at high risk of suicide after discharge from psychiatric inpatient care (1–7). Although risk is apparent up to the first year after hospital discharge (8–10), it is particularly high in the first few weeks and months (3,5,6,11,12). Our previous work identifying risk factors for suicide in the postdischarge period found that 43% died in the first month after discharge (7). Particular characteristics of suicides that occur soon after discharge compared with later have been documented. For example, a national record linkage study in Finland showed that patients who died by suicide within a week after hospital discharge had more severe psychopathology, poorer levels of functioning, and used more lethal suicide methods compared with suicides that occurred later (13).
Most studies examining the social and clinical risk factors associated with postdischarge suicide have focused on suicide that has occurred up to a year after discharge. However, the incidence of postdischarge suicide is highest the closer to the time of discharge and decreases thereafter. Our earlier controlled study found a number of clinical and service-related risk factors for suicide within three months of discharge, including a diagnosis of affective disorder, missed service appointments, a secondary psychiatric diagnosis, and clinical symptoms at last contact with mental health staff (7). A further case-control study in Hong Kong identified previous self-harm, compulsory admission, and living alone as risk factors for suicide in the two-month postdischarge period (14). However, whether these predictors relate to those who die in the first few weeks after discharge, when risk may be greatest, remains to be clarified.
With the shift from hospital-based treatment to care in the community, and recent evidence that the incidence of postdischarge suicide may be increasing in the United Kingdom (15), further knowledge of risk factors for suicide during this important transition period is needed. To our knowledge there have been no national controlled studies assessing suicide risk in the two weeks after discharge. Yet, identifying predictors of poor outcome soon after a psychiatric hospitalization may help service providers to improve discharge planning and care arrangements, thereby potentially reducing the incidence rate. We therefore conducted a detailed retrospective analysis of a consecutive series of individuals in England who had died by suicide within two weeks of discharge from psychiatric care. This short time frame was selected because the early period after discharge represents the highest risk and may help in identifying service-related risk factors (16). We aimed to compare cases of suicide with a control group of surviving postdischarge patients to identify clinical and psychosocial risk factors, including variation in aftercare received.
The study was conducted within the National Confidential Inquiry Into Suicide by People With Mental Illness (NCI), a clinical survey that collects data on all suicides among people with recent contact with mental health services. The method of the NCI is described fully elsewhere (16,17). Briefly, data collection on cases of suicide has three stages—first, the collection of a comprehensive national sample, irrespective of mental health history; second, the identification of people in the sample who had been in contact with mental health services in the 12 months before death; and third, the collection of clinical data about these people. The NCI achieves a suicide questionnaire response rate of over 95%.
The NCI received ethical approval from the North West Research Ethics Committee and from the Research and Development departments at all but two of the NHS Trusts in England providing secondary mental health care for adults. These two trusts were excluded from the study.
Information on all deaths in England receiving a verdict of suicide or an open verdict at coroner’s inquest was obtained from the Office for National Statistics (ONS). Open (also known as undetermined) verdicts are received in cases where the coroner cannot establish suicidal intent. We have included these cases because most are thought to be suicides and are conventionally used in suicide rate estimation in the United Kingdom (18). These suicides and probable suicides are referred to as suicides in this article. Data provided by ONS included method of death, verdict, age, and sex. Detailed clinical information was supplied by the relevant clinician responsible for the patient, who was sent a questionnaire and asked to complete it after discussion with other members of the mental health team.
The cases were a consecutive series of individuals aged 18 to 65 who died by suicide between January 1, 2004, and December 31, 2006, within two weeks of being formally discharged from psychiatric inpatient care in England (N=120). Discharged patients were defined as those who were fully discharged from their inpatient admission and did not include patients on leave or on trial discharge from an inpatient stay. We excluded suicides of individuals older than 65 because risk factors are known to differ for older people (19) and inpatient care for this age group is provided by specific services in most areas in the United Kingdom.
Identification of living control patients
Patients for the control group were selected from the Hospital Episode Statistics database, which collects deidentified patient data on all psychiatric National Health Service (NHS) inpatient admissions in England. The data set consists of person-level data for each admission, including patient hospital number, NHS number, age, sex, date of admission, admission end date, outcome (discharge, death, or hospital transfer), consultant code, and hospital code. Because patient identities were anonymous, clinicians were sent the patient hospital number, NHS number, and date of birth to identify control group patients. One living patient per case was randomly selected to be a control and satisfied the following criteria: the patient had been in psychiatric inpatient care and had been discharged on the same day as the corresponding suicide case. No other matching criteria were used because we wanted to include commonly matched variables such as sex in the analysis. To minimize information bias, data were collected up to the index date, that is, the day of death for the corresponding case.
The suicide questionnaire completed by clinicians consisted of sections on sociodemographic characteristics; clinical history, including exposure to adverse life events in the preceding three months; further details of the suicide (such as substances used in cases of self-poisoning or whether the suicide was part of a pact); aspects of care; and details of the preceding inpatient admission and discharge. The sociodemographic, psychosocial, and clinical items reflect many of the more frequently reported risk factors for suicide. Identical items for the patients in the control group were collected with the use of an adapted version of the suicide questionnaire that referred to the index date rather than the date of suicide.
Potential predictors were selected on an a priori basis by using results from a clinical survey of suicides (16). These were separated into the following domains for analysis: sociodemographic (age, sex, marital status, employment status, and living circumstances), clinical (primary psychiatric diagnosis, any secondary diagnosis, number of previous admissions due to any psychiatric illness [including acute psychiatric settings, drug and alcohol services, and long-term psychiatric ward admissions], duration of current psychiatric illness, medication adherence, recent adverse life events, and enhanced aftercare received under the Care Programme Approach [CPA]), behavioral (previous self-harm, violence, and substance misuse), last psychiatric admission (whether detained under the Mental Health Act [MHA], duration of last admission, whether it was a readmission within three months), last contact with services (contact within a week of the index date, face-to-face contact with a member of staff, and presence of psychiatric symptoms), and last formal discharge (patient-initiated discharge [defined as self-discharge, patient request despite medical advice, or request after breach of ward rules], no follow-up appointment arranged, and missed last appointment).
The CPA is a system of delivering community mental health services in England and Wales. It provides supervision by a care coordinator and regular multidisciplinary case reviews to patients with complex health and social care needs. Those receiving care under the CPA are typically vulnerable patients with severe mental illness who are at risk of suicide or self-harm and who have misused drugs or alcohol. The care coordinator is usually a psychiatric nurse, social worker, or occupational therapist whose role is to review the patients’ mental health care needs through a formal written care plan.
The MHA 1983 provides the legislation by which people with a mental disorder can be detained via hospitalization or police custody for their own or others’ safety and can be assessed or treated against their wishes.
We based our power calculations for this study on risk factors identified in an earlier study on suicide occurring within three months of discharge (7). This prior study indicated the presence of at least four of the seven independent risk factors for postdischarge suicide in 36% of cases and 13% of control group patients (7). The intended sample size for this study was 120 cases of suicide and 120 control patients, which would have provided over 95% power to detect a statistically significant (two-sided α=.05) difference in exposure prevalence of 36% in the suicides and 13% in the control group (odds ratio=3.8). Because of nonresponse in the control group, the achieved sample of suicide cases (N=100) and patients in the control group (N=100) was still sufficient to detect this difference in exposure prevalence with over 95% power.
Conditional logistic regression was selected as the analytical model because of time matching of cases and control group patients on discharge date. Stata 11.0 software (20) was used to calculate odds ratios, 95% confidence intervals, and associated p values. Univariate analyses were first performed to assess the effect of individual factors on the risk of postdischarge suicide. A multivariate model was then fitted to identify independent predictors of poor outcome. Backward elimination procedures were performed within each domain, with explanatory variables carried forward from each domain that were either significant (p<.05) or borderline significant (.05<p<.1). The final regression model included only variables that were independently significant at the 5% level. Tests for multicollinearity among the independent variables were performed with the collin command in Stata (21).
Of the 120 patients who died within two weeks of discharge from psychiatric inpatient care, 85 (71%) had suicide verdicts returned and 35 (29%) had open verdicts returned (both groups referred to as “suicides” herewith). A majority (N=77, 64%) were male. One hundred control questionnaires were returned—a response rate of 83%. Data from 20 control group patients could not be collected because of missing case notes or erroneous patient identity numbers. The following analyses relate to the 100 suicide cases and their time-matched control group patients.
Characteristics of the postdischarge suicide cases
The most common method of suicide was hanging (N=39), followed by self-poisoning (N=23). Death by jumping from a height or in front of a vehicle occurred in 19 cases. In eight cases the method was drowning, and in five cases suicide was by cutting or stabbing. The most common primary diagnoses were major affective disorder (N=35), personality disorder (N=14), adjustment disorder (N=13), and schizophrenia (N=11). A majority of all cases (N=67) also involved a secondary diagnosis, most commonly depressive illness (N=12).
Fifty-five suicides took place within a week of discharge; the highest number of deaths within this period was on the second day after discharge (N=13, 24%). Nearly half (N=24 of 49, 49%) had died before their first follow-up appointment. Those who died within the first week of discharge were more likely than later suicides to be receiving care under the CPA (58% versus 36%; p=.02) and less likely to have been detained under the MHA at the last admission (7% versus 22%, p=.045).
Univariate models of suicide risk
Tables 1–3 show the univariate analyses assessing factors associated with suicide after psychiatric discharge. There was no difference by sex between the suicide and control groups, but those age 40 and over were at greater risk of suicide (Table 1). Factors associated with postdischarge suicide included a history of self-harm, a comorbid psychiatric diagnosis, and recent (<12 months) illness onset (Table 2). Adverse life events within three months of the index date occurred significantly more frequently to those who died by suicide than control group patients, most often relationship breakups (breakups occurred for 11 suicide cases versus three control group patients; p=.05). Nonadherence concerning medication in the month before the index date was a factor among fewer patients who died by suicide than control patients, and fewer of those who died by suicide were receiving enhanced aftercare under the CPA.
Table 1Sociodemographic characteristics of suicide cases within two weeks of hospital discharge and a matched control groupa
| Add to My POL
|Suicide cases||Control group|
|Risk factor||N||%||N||%||OR||95% CI||p|
|Male (reference: female)||62||62||49||50||1.67||.93–2.99||.09|
|Employment status (reference: employed)|
| Unemployed or long-term disability||72||73||76||82||.80||.35–1.87||.30d|
|Unmarried (reference: married or cohabiting)||73||74||75||79||.81||.39–1.69||.58|
|Living alone (reference: living with others)||46||46||47||52||.78||.44–1.38||.39|
|Homeless (reference: not homeless)||1||1||4||4||.25||.03–2.24||.22|
Table 2Behavioral and clinical characteristics of persons who died by suicide within two weeks of hospital discharge and a matched control groupa
| Add to My POL
|Suicide cases||Control group|
|Risk factor and category||N||%||N||%||OR||95% CI||p|
|Behavioral features (reference: no history)|
| History of self-harm||78||78||42||43||4.09||2.12–7.91||<.01|
| History of violence||30||32||32||35||.95||.52–1.74||.88|
| History of substance misuse||65||66||51||56||1.50||.87–2.57||.14|
| Diagnosis (reference: all others)|
| Affective disorder||35||35||35||36||.95||.51–1.78||.87|
| Co-occurring disorders (reference: none)c||24||24||21||21||1.19||.61–2.31||.61|
| Any secondary psychiatric diagnosis (reference: none)||67||67||42||45||2.57||1.39–4.77||<.01|
| >5 previous admissions (reference: ≤5)||21||21||24||25||.78||.39–1.56||.48|
| Illness onset <12 months (reference: ≥12 months)||32||33||18||19||2.08||1.07–4.03||.03|
| Past-month medication nonadherence (reference: adherence)||11||11||22||26||.41||.19–.89||.02|
| Recent (<3 months) adverse life events (reference: none)||56||58||27||30||4.43||1.95–10.06||<.01|
| Received care under the CPA (reference: did not receive)d||48||48||58||64||.50||.27–.91||.02|
Table 3Contact with psychiatric services among suicide cases within two weeks of hospital discharge and a matched control groupa
| Add to My POL
|Suicide cases||Control group|
|Risk factor and category||N||%||N||%||OR||95% CI||p|
| Detained under the Mental Health Act (reference: not detained)||14||14||21||23||.55||.26–1.15||.11|
| Index admission <1 week duration (reference: ≥1 week)||41||41||25||27||2.00||1.05–3.80||.03|
| Readmission within 3 months (reference: no readmission)||26||26||17||17||1.64||.85–3.19||.14|
|Last contact with services|
| Contact within a week of the index date (reference: >1 week)||86||87||76||79||2.00||.81–4.96||.13|
| Face-to-face contact with a member of staff (reference: telephone contact)||82||84||91||96||.18||.04–.82||.03|
| Psychiatric symptoms at last contact (reference: none)|
| Depressive symptoms||25||26||12||13||2.22||1.01–4.88||<.05|
| Suicidal ideation||15||16||8||9||2.14||.87–5.26||.10|
| Patient initiated (reference: planned)c||35||35||21||21||1.56||.86–2.81||.14|
| No follow-up (reference: follow-up)||7||7||10||10||.67||.24–1.87||.44|
| Missed last appointment (reference: kept appointment)||20||21||9||9||1.73||.82–3.63||.15|
Persons who died by suicide were more likely to have had a short (less than one week) index admission, but there were no significant differences between suicides and the control group in the number of previous admissions, having a history of violence or substance misuse, co-occurring psychiatric and substance use diagnoses, or being detained under the MHA during their last admission.
More patients who died by suicide than patients in the control group had been in contact with a member of the mental health team within a week of the index date, although this did not significantly increase risk (Table 3). The last contact, however, was more likely to be by telephone call than face to face among those who died by suicide compared with those in the control group. Those reporting depressive symptoms or hopelessness were more likely to die by suicide, and although recent self-harm and suicidal ideation were more common among suicides, these did not predict suicide. Similarly, more suicide patients had initiated their own discharge (through self-discharge, request against medical advice, or breach of ward rules) and missed their last appointment with psychiatric services, but these differences did not reach statistical significance.
Multivariate model of suicide risk
The final multivariate model (Table 4) included six independent predictors of suicide: age 40 or over, male sex, history of self-harm, psychiatric comorbidity, recent adverse life events, and a short (less than one week) index admission. The prevalence of multiple risk factors among those who died by suicide was high; significantly more had at least four of the six risk factors compared with control group patients (57% versus 9%; p<.001). Being under enhanced CPA was independently protective of suicide risk in this population. Diagnostic tests showed that multicollinearity was not a major issue in the final model (21). All of the variables showed variance inflation factor values between 1.02 and 1.22 and tolerance values between .82 and .98.
Table 4Independent predictors of postdischarge suicide risk in the final multivariate conditional logistic regression modela
| Add to My POL
|Adjusted OR||95% CI||p|
|Age ≥40 (reference: <40)||5.30||1.55–18.09||<.01|
|Male (reference: female)||3.07||1.06–8.89||.04|
|History of self-harm (reference: no history)||6.80||2.12–21.83||<.01|
|Any secondary psychiatric diagnosis (reference: none)||4.37||1.24–15.41||.02|
|Recent (<3 months) adverse life events (reference: none)||4.60||1.49–14.21||<.01|
|Index admission <1 week duration (reference: ≥1 week)||3.42||1.14–10.25||.03|
|Received care under the CPA (reference: did not receive)b||.27||.09–.78||.02|
This was the first national case-control study to investigate predictors of suicide that occurs within two weeks after discharge from psychiatric inpatient care. We found a greater incidence of suicides in the first week after discharge, with nearly half occurring before the first outpatient appointment. We identified a number of important risk factors for suicide in the immediate postdischarge period, including a short admission and adverse life events. We found that enhanced aftercare for those in receipt of CPA services was a protective factor. We also found that social risk factors for suicide in the general population (such as living alone or being unemployed) did not increase risk in our clinical sample of those in mental health service contact (22,23), although other conventional risk factors (such as older age, male sex, a lifetime history of self-harm, and psychiatric comorbidity) showed a positive association with suicide.
Our results confirm findings from Europe (6) and the United States (24) of an increased suicide risk for patients with a short hospital admission. This may be indicative of premature discharge and emphasizes the importance of careful discharge planning despite apparent recovery by the patient. Few studies have demonstrated that an aspect of service provision may be protective of suicide, although the use of compulsory treatment has been found to reduce risk among current psychiatric inpatients (25). The CPA brings enhanced supervision and increased intensity of case management, including regular risk assessments. Dennehy and colleagues (26) noted that most discharged patients who died by suicide had no identifiable key worker or care coordinator (an important component of the CPA, which provides a designated clinician or therapist to coordinate a patient’s treatment), but this factor did not distinguish between suicides and living control patients. Our results are the first, to our knowledge, to suggest that this higher level of aftercare may be independently protective.
Although we found comorbid psychiatric disorders to be a risk factor, we did not find co-occurring disorders (severe mental illness plus concurrent substance misuse) to be associated with suicide. We also did not find medication nonadherence to be associated with increased risk of suicide; indeed, nonadherence was more common among control group patients than suicide cases. These results may be a reflection of mental health service recommendations in England to target particular clinical groups, including nonadherent community patients and those with co-occurring disorders; both groups have seen a fall in suicide rates after specific policy implementation (27).
The association between adverse life events and suicide has been reported previously and implies that such experiences may be potent indicators of suicide risk (11,19,28). Whether these events occurred before or during admission is unknown, but the return to a potentially stressful situation after the relative stability of an inpatient setting may be a trigger for suicidal behavior. Knowledge by staff, therefore, of situations that patients face after discharge, including any exposure to detrimental life experiences, should form a vital element in the discharge planning process and may require the involvement of family and friends to supplement support.
Our study has the strength of being a national case-control study. For the first time, we were able to quantify risk on a relatively large sample over a restricted period, which allows us to assess factors related to care. Further strengths include the representative random sampling of control group patients from the population at risk, low levels of missing data in the explanatory variables, and the ability to assess a wide number of potential risk factors. However, limitations to the study included the retrospective data collection from clinicians based on their knowledge of the patient, clinical judgment, and case records. Some misclassification in risk factor measurement may have occurred as a result. Clinicians who provided the information were not blind to suicide case versus control status and may have been biased by their awareness of outcome. However, most of the questionnaire items are factual, and information was retrieved from case notes completed at a time when the outcome was unknown; therefore, our results should not have been greatly affected by any reporting bias. Further, a number of suicide studies have relied on similar methods (1), and NCI questionnaire data have been shown to be reliable (29). Finally, this study was based in England only, which may limit the generalizability of some of our findings. However, some of the risk factors for postdischarge suicide are universal (3,6) and we believe relevant to mental health services internationally.
The period after hospital discharge should be a time of recovery, not increased suicide risk. Our results underline the importance of careful discharge planning for patients viewed as high risk, particularly older patients, males, those with a history of self-harm, and those with comorbid psychiatric diagnoses. The potential role of adverse life experiences means that mental health clinicians need to be aware of the social circumstances patients are facing after hospital discharge. The first week postdischarge should be recognized as the period of highest risk, and follow-up arrangements should reflect this. Aspects of care, such as use of the CPA with the involvement of a care coordinator, may help reduce risk at this vulnerable time.
The mechanism by which the CPA and other components of aftercare may be protective of suicide is complex, but future studies may be able to elucidate elements of the CPA that are specifically involved in lowering risk. Quantitative studies could, for example, assess the impact of length of CPA participation or quality of the therapeutic relationship on outcome. Qualitative research has the potential to better delineate the processes by which the CPA reduces suicide risk.
Previous research has shown that case-management approaches are known to increase service contact and improve patient satisfaction, although any association in reducing clinical symptoms is less clear (30,31). However, evidence of the positive effects of providing more intensive support to the most vulnerable patients has recently been shown in England, where mental health service improvements, such as introducing 24-hour crisis care and patient follow-up within seven days of discharge, were both associated with reduced suicide rates in particular groups (27). It has also been suggested that even low-intensity contact by services in the first few days after discharge, such as via a postcard (32) or mobile telephone message (33), may be helpful, although the evidence is far from clear (34). Indeed, our results have highlighted the importance of face-to-face contact with patients after discharge, which may be protective.
This study was funded by the United Kingdom National Patient Safety Agency. The views expressed in this article are those of the authors and not the funding body. The authors acknowledge the help of district directors of public health, health authority and trust contacts, and the clinicians for completing the questionnaires. The authors thank the other members of the research team and David While, Ph.D., for his statistical advice.
The authors report no competing interests.