Unnatural and suicide deaths

In stark contrast to previously reported forensic^{Reference Hoang, Stewart and Goldacre6,Reference Fazel, Fimińska, Cocks and Coid22} and mainstream psychiatric findings^{Reference Chesney, Goodwin and Fazel2,Reference Olfson, Wall, Wang, Crystal, Liu and Gerhard23} we did not observe exceptional rates of suicide and accidental/unnatural deaths. We described a sixfold (SMR = 625) increase in suicide and an almost fourfold (SMR = 375) increase in unnatural deaths against the general population. In comparison, Clarke et al ^{Reference Clarke, Davies, Hollin and Duggan12} reported a SMR for suicide of 3231, and for all unnatural deaths of 1898 in relation to a cohort followed for a maximum of 20 years (n = 595, 5593 PYAR) after first admission to a medium secure unit. Including the few drug/alcohol deaths within accidental (unnatural) deaths we only returned an almost ninefold increase. No suicide/accidental deaths occurred within high secure care, and only one of our eight suicide/accidental deaths happened within the psychiatric hospital environment. Of the n = 31 suicide/open verdicts and accidental deaths reported by Clarke et al ^{Reference Clarke, Davies, Hollin and Duggan12} almost half (n = 15) the preventable deaths occurred within high/medium security and the wider hospital environment.

We considered a number of factors in exploring this finding of a low rate of suicide and accidental deaths within our cohort. First, was the Scottish suicide rate lower than elsewhere in the UK? The 2018 Scottish suicide rate (16.1/100 000 persons) was greater than the comparably reported English rate (10.3/100 000).²⁴ Therefore Scotland does not have a naturally low rate of suicide.

Second, was our low rate of suicide reflected in the general adult psychiatry population in Scotland? This was refuted by a historical Scottish population study of discharged, long-term (>1 year) psychiatric in-patients that reported an increased suicide risk 13 times the population rate.^{Reference Stark, MacLeod, Hall, O'Brien and Pelosi25}

Third, was the diagnostic profile of this Scottish cohort impactful? Unlike other UK regions,^{Reference Coid, Hickey and Yang13} within Scotland offenders with a primary diagnosis of personality disorder generally remain within the criminal justice system. At baseline only 5% received a primary diagnosis of personality disorder whereas 70% attracted a schizophrenia diagnosis. This compares with 25.8% of individuals with personality disorder in a Swedish forensic cohort,^{Reference Fazel, Wolf, Fimińska and Larsson11} 26.6% in an English medium secure cohort^{Reference Clarke, Davies, Hollin and Duggan12} and 13.5% in an English community cohort.^{Reference Coid, Hickey and Yang13} Only the Clarke et al^{Reference Clarke, Davies, Hollin and Duggan12} study splits suicide between mental illness and psychopathy at 72.2% and 16.6%, respectively, which is reduced in the personality disorder group given that it was 26.6% of their cohort. The different diagnostic profiles may be a factor in explaining our findings but this is not supported by the Clarke et al^{Reference Clarke, Davies, Hollin and Duggan12} study and would not account for the extent of the difference. Suicide rates are generally high among individuals experiencing personality or psychotic disorder^{Reference Arsenault-Lapierre, Kim and Turecki26} with risk of accidental death higher for personality disorder compared with schizophrenia.^{Reference Crump, Sundquist, Winkelby and Sundquist8} Contextualising our suicide deaths within our diagnostic profile we observe a similar pattern; 60% of people who died by suicide had schizophrenia at baseline, one person had ASPD and another alcoholism. A total of 80% were male and the same proportion had a history of heavy/abusive alcohol use, all displayed polysubstance drug misuse, known risk factors for non-natural death.^{Reference Arsenault-Lapierre, Kim and Turecki26} That description differs from the overall cohort, 47% of whom reported substance misuse at baseline and 29% had comorbid ASPD.

Fourth, does the year the cohort was established or the length of follow-up influence findings? Compared with other studies^{Reference Fazel, Wolf, Fimińska and Larsson11–Reference Coid, Hickey and Yang13} our cohort was followed for the longest time therefore increasing rather than decreasing suicide likelihood. Alternatively, have more recently established cohorts observed behavioural change, making suicide/accidental in-patient death more likely? Only one suicide has been noted in our cohort source hospital (The State Hospital) since 1996 (email from Health Records, The State Hospital, tsh.Health_Records@nhs.net, November 2019).

Finally, is engagement with the Scottish forensic mental health system protective against suicide or accidental death? An examination of suicides (n = 14) occurring within high secure care in Scotland (The State Hospital, 1972–1996)^{Reference Ramsay, Gray and White27} noted treatment-resistant schizophrenia and having committed violent offences as suicide risk factors. Greater liberalisation of the hospital, increased activities and reintroduction of clozapine were suggested for the suicide reduction observed over time. This continuing developmental process alongside estate renewal may influence our observed low suicide/accidental death rate while patients resided in high secure care. At death two individuals were under mental health legislation/restrictions. Examining CDR figures, restrictions made little difference to the premature death rate. Also considered was if a longer period of contact and mandated relationship with services may protect against suicide; alternatively, lack of autonomy may be a negative influence. Neither was apparent within our findings.

We hypothesise that factors within the Scottish forensic in-patient environment; physical, procedural and/or relational are protective in terms of suicide prevention or in deterring behaviour leading to accidental death and that these may have an ongoing effect on patients’ relationships with services in the community. There may also be organisational factors that reduce avoidable deaths. Scottish forensic services have advantages in terms of size and cohesion, with few independent secure beds, and the strategic lead of the Forensic Network.

Risk of premature mortality

A large proportion of our cohort died (36.9%) demonstrating an almost fourfold increase (SMR = 397) in all-cause mortality. This represents a slightly higher CDR than reported for forensic services in England/Wales,^{Reference Fazel, Fimińska, Cocks and Coid22} mirrors an English general adult community-recruited cohort experiencing psychosis^{Reference Reininghaus, Dutta, Dazzan, Doody, Fearon and Lappin28} and is lower than reported internationally.^{Reference Fazel, Fimińska, Cocks and Coid22} In contrast, 91% of our cohort died of natural causes four times (SMR = 401) the rate of the general population. Natural deaths within the discharged Scottish general adult psychiatry in-patient population has been reported at SMR = 169^{Reference Ajetunmobi, Taylor, Stockton and Wood29} indicating a difference in mortality between Scottish forensic and general adult in-patients.

Overall, 70.8% of deaths were premature and naturally occurring at an average age 55.6 years. Respiratory disease/cancer was the underlying cause of almost a third of premature and 36% of all deaths, with cardiovascular disease related to 19% of premature and 21% of all mortality. Study recruitment occurred during 1992/3 when almost all patients smoked or were passive smokers. Since December 2011, our cohort source hospital, The State Hospital, Carstairs has been entirely smoke free.

Rates of respiratory related deaths, in excess of fourfold the expected level have been noted in an early population study of discharged Scottish general adult psychiatry long-term (>1 year) in-patients^{Reference Stark, MacLeod, Hall, O'Brien and Pelosi25} with rates of death related to cardiovascular disease being slightly raised at SMR = 160. A later examination^{Reference Ajetunmobi, Taylor, Stockton and Wood29} capturing all discharged general adult psychiatric in-patients reported that cardiovascular disease, despite displaying only a small rise in relative risk (SMR = 170) was responsible for 67% of total mortality and 54% of the total years of potential life lost. A similar English study examining death within a year of discharge of patients with schizophrenia reported SMR of 470, and 250 for respiratory and circulatory disease deaths, respectively, with increased rates from 1999 to 2006.^{Reference Hoang, Stewart and Goldacre6}

The cohort of 28 females represents around 50% of Scottish female forensic patients, reported as n = 56 in 2004^{Reference Black and Robertson30} and n = 60 in 2017.³¹ There is an obvious disparity between the male and female groups. Almost all women died prematurely aged under 48 years with mean loss of 24 years of potential life. Their SMR for natural death was three- to fourfold the male rate and that held for females with schizophrenia, rising to fivefold for a primary diagnosis of ASPD. Receiving a comorbid diagnosis of ASPD was also associated with premature death. This suggests that ASPD may have a unique impact on the physical health of women in a manner not reflected in risky behaviour or suicide completion.

There is a lack of published literature regarding gender differences in mortality among patients located within forensic services with papers and reviews generally commenting on the percentage of males within the cohort.^{Reference Coid, Hickey and Yang13,Reference Fazel, Fimińska, Cocks and Coid22} Where data does exist,^{Reference Clarke, Davies, Hollin and Duggan12} our results replicate those findings, namely that SMR for death because of natural causes, unnatural causes and suicide are higher for the female than the equivalent male cohort. However, as with those presented here, findings must be interpreted with caution because of the extreme gender imbalance evident within forensic psychiatric services and reflected in small numbers of females in research cohorts.

A similar picture has, however, been observed within mainstream services, specifically in relation to individuals with schizophrenia. A population follow-up of all in-patients admitted with schizophrenia diagnosis from 1980 to 2006 indicated that up to 1992 the SMR for males in relation to all-cause, unnatural and natural deaths exceeded that of females; however, for individuals admitted post 1992 the opposite was observed: females displayed greater all-cause, unnatural and natural SMR.^{Reference Høye, Jacobsen and Hansen32} More recently an American population study reported that female SMR for all-cause mortality exceeded the male equivalent; however, the SMR for unnatural death was higher among males.^{Reference Olfson, Gerhard, Huang, Crystal and Stroup5}

Almost 68% of males died prematurely at mean age 50 years. Males detained under civil measures were significantly more likely to die prematurely than those males under criminal procedures. Similarly, males in receipt of depot antipsychotics at baseline were significantly more likely to have died prematurely. It is conceivable that as civil patients transfer to high secure care because of acute morbid positive symptoms and/or locally unmanageable levels of violence or aggression^{Reference Thomson, Bogue, Humphreys and Johnstone16} and by utilising depot medication as a proxy for illness severity, we may be observing those who experienced the severest symptoms and psychological distress, dying prematurely. Specifically, that is, individuals who may be more greatly influenced by psychophysiological factors.

Scottish forensic services acknowledge that female patients represent a heterogeneous group, often more chaotic and challenging with differing needs to male patients,^{Reference Black and Robertson30} with higher rates of mortality^{Reference Davies, Clarke, Hollin and Duggan33} and generally poorer outcomes.^{Reference Sahota, Davies, Duggan, Clarke, Huband and Owen34} Although Scotland lacks female high secure care and exclusively single gender medium secure provision (with some services provided within England), because of the general complexity of female patients, providing appropriate relational security and support can be more important than physical security.³¹ Regardless of shortcomings in the Scottish female forensic estate, the high levels of premature mortality of both genders cannot be overlooked.

Despite spending on average longer as high secure in-patients, with exposure to the health benefits that offers, there was little difference in CDR for those subject to restrictions on discharge and those not. Fazel et al^{Reference Fazel, Fimińska, Cocks and Coid22} pointed out that mortality rates among forensic in-patients are high but more closely reflect the general adult psychiatry population than prisoners. They postulate that greater than any factors negatively influencing mortality within forensic environments are the poor lifestyle choices evident among general psychiatric populations that can compound psychotropic medication side-effects. Given our respiratory findings, however, the impact of a heavy smoking environment must be acknowledged. Although poor lifestyle choices have an impact on morbidity/mortality among individuals experiencing SMI again the increased morbidity risks and resultant mortality are not fully explained by behavioural patterns.^{Reference Penninx, Milaneschi, Lamers and Vogelzangs35,Reference Darrow, Verhoeven, Révész, Lindqvist, Penninx and Delucchi36}

Protective factors

Within this cohort some protective factors appeared evident. Although both genders died at early ages the males lived on average almost 8 years longer compared with the females. Males with a primary diagnosis of intellectual disability appeared less likely to die (SMR = 147, 95% CI 53.9–326) and it could be that those patients with intellectual disability received less psychotropic medication and accompanying side-effects. One reason suggested for high rates of morbidity among individuals with mental ill health is the propensity for diagnostic overshadowing to occur. This is when disparities occur in the treatment and diagnosis of physical disorders as a result of misattribution of physical symptoms to mental illness.^{Reference Shefer, Cross, Howard, Murray, Thornicroft and Henderson37} Although we interpreted experiencing primary intellectual disability as a protective factor, diagnostic overshadowing has been presented as a particular problem within the intellectual disability population with symptoms related to physical or mental ill health being misattributed to their intellectual disabilitiy.^{Reference Javaid, Nakata and Michael38} We suggest that in line with our assertion that engagement with the Scottish forensic mental health system may be protective against unnatural death and death by suicide, for individuals with intellectual disability, location within services almost exclusively under National Health Service operational control may confer advantages to forensic patients with intellectual disability in terms of staffing, their training and support and patient services offered. Higher levels of in-patient and community support to avoid offending and foster appropriate behaviour may also reduce stress and encourage patients with intellectual disability to adopt heathier lifestyles, with a potential reduction in drug and alcohol use. More attention may be paid to their physical health by staff, and general practitioner referrals encouraged and supported, leading to earlier physical health intervention and a reduction in diagnostic overshadowing.

It is also acknowledged that individuals with intellectual disability represent a particularly vulnerable subpopulation within the prison environment, being subject to high rates of mental disorder^{Reference Chaplin, McCarthy, Underwood, Forrester, Hayward and Sabet39} and possibly greater risk of suicidal ideation than the general population.^{Reference Giannini, Bergmark, Kreshover, Elias, Plummer and O'Keefe40} We propose that a contributing factor to this poor outlook is a lack of equivalence, equality and equity for people with intellectual disability within the UK prison system.^{Reference Herrington41} Again we propose that our intellectual disability cohort were protected from premature mortality precisely because they were, where appropriate, diverted from the prison environment and supported by specialist forensic psychiatric services within hospital and community settings, designed to promote and provide equality of life experience. Further research is required to address the paucity of literature robustly identifying and exploring the journey of individuals with intellectual disability through services.

Morbidity

Antipsychotic medications, a reliable mechanism for easing symptoms, reducing distress and therefore enhancing recovery carry with them a diversity of side-effects: activating, sedating and metabolic. Unsurprisingly antipsychotics have been targeted as a possible reason for the premature mortality associated with schizophrenia.^{Reference Saha, Chant and McGrath1,Reference Weinmann, Read and Aderhold42} Population studies^{Reference Tiihonen, Mittendorfer-Rutz, Torniainen, Alexanderson and Tanskanen43} indicate that individuals with schizophrenia treated with antipsychotics or antidepressants have a lower risk of death compared with individuals not receiving such medications. A primary diagnosis of schizophrenia was applied to 70% of this cohort, and receiving depot neuroleptics was significantly associated with premature death; however, there were no significant differences in mean diagnoses of cardiovascular or endocrine, nutritional and metabolic disease applied to the living participants and the prematurely deceased. What is apparent is the significant difference in terms of respiratory disease, with those dying prematurely receiving more diagnoses. Although undoubtedly the legacy of an era, smoking remains an issue for forensic patients after discharge from controlled in-patient environments.

Those who died prematurely attracted significantly more diagnoses related to injury, poisoning and other external causes; however, these did not translate into high numbers of traumatic deaths or completed suicides. Overall, those dying prematurely received more physical health diagnoses across all listed ICD-10 blocks and spent significantly more days as general hospital in-patients than those who remained alive evidencing the poorer physical state of that group.

Scotland does not have the healthiest national population, indeed the ‘Scottish effect’ is much examined with 17 hypotheses proposed to account for excessive premature mortality.^{Reference Cowley, Kiely and Collins44} We suggest that something akin to the ‘Scottish effect’, specifically in terms of biopsychosocial stress and the resultant physiological stress response, is being observed among mainstream and forensic psychiatric populations. High levels of adverse life events observed among Scottish forensic patients^{Reference Karatzias, Shevlin, Pitcairn, Thomson, Mahoney and Hyland45} together with the dose–response association observed between adverse life events/psychological distress and a negative impact upon subjective and objective physical health^{Reference Karatzias, Yan and Jowett46} may be evidenced within our reported cohort. There is increasing physiological evidence of heightened levels of oxidative stress and inflammation within anxiety, depressive, bipolar disorder and schizophrenia.^{Reference Darrow, Verhoeven, Révész, Lindqvist, Penninx and Delucchi36,Reference Salim47} These processes provide mechanistic pathways by which leucocyte telomere length can be shortened. Meta-analysis has evidenced shortened telomere length across a range of psychiatric disorders^{Reference Darrow, Verhoeven, Révész, Lindqvist, Penninx and Delucchi36} with length being an indicator of cell ageing and short length associated with age-related morbidities, for example immune dysregulation, cancer, diabetes and cardiovascular disease.^{Reference Blackburn, Epel and Lin48} Interpreting the morbidity and excess mortality observed in cohorts such as this through the lens of the ‘Scottish effect’ may lead to better targeted local interventions.

Limitations

This study has several limitations. Follow-up physical health information could only be requested for deceased or consented participants. The wider study adopts a gatekeeper approach therefore all cohort members whose gatekeeper denied access because of their inability to provide ‘fair’ consent⁴⁹ or they were not physically/mentally well enough to be approached were excluded. This removed individuals with the greatest physical health needs and/or the severest psychiatric symptomatology. The mortality status of eight individuals remained unknown although of these seven resided in Northern Ireland. We could not locate or engage with appropriate Northern Ireland services to confirm their status or locate their current care team. It proved impossible to access information without consent. Accessing cohort member gatekeepers within England was equally difficult but mortality information moved between NHSCR and their English counterparts. Some individuals with mortality status unknown could have died because of suicide or accidental causes, raising our mortality profile, but it would remain lower than the reported studies. We were also as reliant on the general hospital clinicians accurately recording and/or applying the most appropriate ICD-10 codes as we were of the cohort member and their accompanying staff providing a precise physical health history.