Diagnosis

Previous studies have clearly demonstrated the effect of substance misuse on increasing suicide risk.^{3,Reference Nielsen, Hjorthøj, Erlangsen and Nordentoft4,Reference Nilsson S, Hjorthoj, Erlangsen and Nordentoft10} Our study is consistent with these results. We found that homeless patients who died by suicide were more likely than those in stable accommodation to have a primary diagnosis of substance misuse, previous history of substance misuse, and to have been under drug and alcohol services at the time of their death. Clinicians also stated that improved access to a dual diagnosis service could have helped to prevent suicide in this group. Higher prevalence of substance misuse is also reflected in higher prevalence of self-poisoning deaths by opiates overdose in homeless patients, despite overall fewer deaths by self-poisoning; this is consistent with previous research.^{Reference Sinyor, Kozloff, Reis and Schaffer7,Reference Bickley, Kapur, Hunt, Robinson, Meehan and Parsons12}

The most recent ONS report on deaths of homeless people at the time of publication mentions substance misuse as an important factor in their increased mortality.³ Since substance misuse has been found to be associated with both suicidal ideation and attempts in homeless people,^{Reference Gentil17} it is important to involve homeless people in substance misuse treatment to reduce the risk of suicidal behaviours. Comorbid mental disorder and substance misuse often precedes suicide, and so close integration of substance misuse services with mental health services is an important step in risk prevention for this group.^{Reference Gentil17}

It is important to note comparable prevalence of schizophrenia and other delusional disorders in homeless and non-homeless patients who died by suicide in our sample; this is contrasting to the findings of a recent meta-analysis that has found significantly higher prevalence of schizophrenia and other delusional disorders in homeless people.^{Reference Ayano, Tesfaw and Shumet19} This difference was not as prominent when looking at data solely from high- and middle-income countries, which may explain our finding to some extent. Additionally, it is possible that this difference stems from the fact that the data for this study is only from people in recent contact with a mental health service who died by suicide. Finally, our sample comprised mental health patients who were more likely to be diagnosed using diagnostic criteria and instruments, rather than screening instruments, which could have contributed to lower overall prevalence of delusional disorders.^{Reference Ayano, Tesfaw and Shumet19}

Homeless patients who died by suicide had significantly lower prevalence of depression as the primary diagnosis compared with the non-homeless patients. Even though one would expect depression to be higher in this population given the stressors that being homeless entail, previous research also found the rates of depression in the homeless population to be lower than expected.^{Reference Fazel, Khosla, Doll and Geddes9} Fazel and colleagues hypothesised that this discrepancy may be because of suicide in homeless populations being mediated through other risk factors, such as substance misuse, rather than through depression.^{Reference Fazel, Khosla, Doll and Geddes9} This hypothesis is supported in our study by the significantly higher prevalence of a diagnosis of drug and alcohol misuse and personality disorder which, to some extent, could also be masking an underlying depression.

In-patient and post-discharge suicide

Homeless patients are more likely to die by suicide during an in-patient admission and the post-discharge period than non-homeless patients, as can be seen in Fig. 3. Increased risk of death by suicide in those that have had previous contact with mental health services has been recognised both in the general population^{Reference Qin, Agerbo and Mortensen20} and homeless people.^{Reference Nilsson S, Hjorthoj, Erlangsen and Nordentoft10} We found that the percentage of homeless people who died by suicide under psychiatric in-patient care was double that of non-homeless in-patients.

Even though the total number of in-patient suicides is falling,²¹ this may not be the case in this particular patient subgroup. It is also important to note that over 40% of homeless in-patients died while being off the ward without staff agreement. Previous research on in-patients who have died by suicide off the ward without staff agreement – so-called absconders – also noted higher rates of homeless patients dying under these circumstances, and recommended tighter observation of ward exits, individually tailored patient observation levels and improvement of the ward environment as prevention measures against absconding.^{Reference Hunt, Windfuhr, Swinson, Shaw, Appleby and Kapur22}

In addition to this, homeless patients’ percentages of post-discharge suicides are higher at all post-discharge time points than their non-homeless counterparts that have also died within the said time frame. It is noted that homeless patients are more likely to be discharged from in-patient care into housing, financial, and/or employment problems (i.e. being unemployed). Even though these areas are part of the standard risk assessment carried out prior to discharge from mental health services, it is probable that clinicians face significant limitations in addressing difficulties in these areas, thus being unable to reduce the risks. For patients, facing these problems again may contribute to the increased number of post-discharge suicides in the homeless patient group.

Contact with services

Homeless patients were more likely to be under drug and alcohol services or have been seen in the emergency department following self-harm than non-homeless patients. Clinicians felt that improved access to crisis facilities could have made the suicide significantly less likely at the time for homeless patients; this is consistent with the finding that homeless patient were less likely to be under a crisis team at the time of their death. Crisis teams often rely on good engagement from the patient although this may be difficult for someone who is homeless. This can also be the case for patients under community mental health teams, who may lack the resources to reach people who are homeless. In the UK, where the provision of homeless mental health services is variable across the countries with funding streams often being uncertain, the roles of community mental health teams in patient care can vary from assessment to a full provision of care.^{Reference Taylor23}

Additionally, a recent NCISH report states that the majority of homeless patients have been registered with a general practitioner in the year before their death by suicide.²¹ Even though homeless mental health patients may be seen as a hard to reach group, there is some evidence to suggest that they are in contact with services, although it is possible that their pattern of service use is different to that of non-homeless patients who died by suicide, namely, by coming to emergency departments more. It may be that because of sporadic contact or the isolated working of services crucial opportunities for suicide prevention have been missed. This highlights the necessity to develop services that cater to different patterns of contact by homeless patients and that can consider and address some of the relational difficulties that can interfere with meaningful engagement.

Limitations/methodological issues

As this is an exploratory, uncontrolled retrospective study, we cannot make any causal inferences, and therefore the descriptive findings should be interpreted with caution. There are some previously established limitations of the NCISH methodology, such as potential bias of the clinicians providing information.^{Reference Appleby, Shaw, Amos, McDonnell, Harris and McCann24} Even though we have minimised the possibility of missing certain suicide deaths by including those patients whose deaths were classified as open verdicts, it is possible that we missed some of the homeless patients that belong in the so-called ‘hidden homeless’ category such as those living in hostels and bed and breakfast. Also, many of the significant findings could be related to the state of homelessness patients in general, rather than being specific to homeless patients who die by suicide. Conversely, findings from this study cannot be generalised to the general population of homeless people, in which undiagnosed mental illness is often a problem. It is important to note that most homeless people do not contact mental health services unless this is actively sought by homeless treatment teams specialising in their care.^{Reference Folsom, Hawthorne, Lindamer, Gilmer, Bailey and Golshan25} The comparison shows instead whether homeless patients who died by suicide have a different pattern of risk from non-homeless patients and serves to highlight the suicide prevention measures that could have most effect for these individuals. As diagnoses were made and reported by the treating clinician, we cannot be sure which categorisation or diagnostic tools were used. Finally, we had no information about the length of homelessness for the deceased patients.

Clinical implications

There is already some work being done to help homeless people with mental health problems including the use of psychologically informed environments, an approach that helps service providers remodel services in order to better address psychological issues that homeless people face.^{Reference Keats, Cockersell, Johnson and Maguire26} Psychologically informed environments has been used in the creation of specialised homeless mental health teams with some positive initial results.^{Reference Taylor23} Further, the publication of standards for commissioners and service providers for people with complex problems²⁷ has helped commissioning of services for homeless people. These standards aim to support homeless people by improving access, engagement and liaising between services. Homeless people attending emergency departments while in crisis or being admitted to a physical health hospital may provide an opportunity for engagement with services and may require a proactive approach by mental health services. The Faculty for Homeless and Inclusion Health have created specialist teams of general practitioners that provide in-reach in physical health hospitals, a model that could be implemented in mental health hospitals.^{Reference Doyle and Hamlet28}

Providing appropriate help for mental health patients with substance misuse has become more challenging since the NHS reforms that have distanced substance misuse provision from mental health provision.^{Reference UK Drug Policy Commission29} This poses an additional risk as now patients with comorbidities need to access two separate services, one for mental health and one for substance misuse issues. This can be especially challenging for people without stable accommodation in an out-patient setting and can mean lack of substance misuse therapy in an in-patient setting. Services should therefore be well connected with each other and aware of the nature of engagement with services that homeless patients can have. Additionally, as evident from the high numbers of in-patient and post-discharge suicides among homeless patients, in-patient units should be aware of the increased risks of suicide during admissions and in the post-discharge period, and provide appropriate discharge planning and follow-up for homeless patients.

Research implications

Additional research is needed into homeless patients’ experiences of hospital admissions and discharge. Considering the high rates of in-patient suicides for this group, particularly when off ward without agreed leave, a deeper examination of their experiences during hospital admissions is warranted. Qualitative investigation with homeless patients and mental health staff, either through interviews or focus groups, might help to provide much needed insight into the particular difficulties they face during their hospital admission.