In this analysis, people with mental disorder diagnoses who had had contact with secondary mental healthcare services, had substantially higher mortality than expected in all diagnostic groups examined. The calculated SMRs varied modestly by gender and substantially by age, and also fluctuated markedly across different ethnic groups, which might be caused by small size in specific populations.
People with SMI have substantially higher than expected mortality in all age groups. The raised risk of mortality in these cohorts is consistent with previous studies, indicating that mortality rates among individuals with SMI are higher than that of the general community [6
]. A previous longitudinal register-based study with the maximum follow up of 18 years in the UK investigated 'any psychiatric diagnosis' as an exposure and found this to be associated with a 65% higher than expected total mortality [7
]. In our analysis, the SMRs of those with SMI suggested a more than two-fold higher mortality, although this may reflect a focus on the more severe mental disorders represented within the SMI label. Causal pathways between mental disorder and mortality have yet to be fully elucidated and are likely to be multiple. While suicide and accidents/violence are important considerations for clinical services, research has tended to show mortality increases across all major causes, including cardiovascular diseases (heart attack and stroke) [11
]. These may be influenced by direct effects of mental disorder symptoms (for example on suicide and accidents). Pathways may also reflect adverse lifestyle factors influenced by the presence of mental disorders and themselves responsible for associations with mortality; these include worse nutrition, physical inactivity, alcohol use, smoking, and illicit drug use [2
]. Adverse effects from psychotropic (particularly antipsychotic) medication have also received increasing consideration in terms of their role in raised mortality as an outcome [17
]. Specifically, antipsychotic agents are often prescribed long term and may increase the risks of diabetes mellitus and cardiovascular diseases with events including QT-interval prolongation, ventricular arrhythmias, pulmonary embolus, atherosclerosis, and sudden cardiac death [4
]. The analyses presented here were not intended to elucidate causal pathways, but rather to constitute the first stage in a series of investigations of these issues with future studies clarifying further the role of socioeconomic status, education and cognitive abilities which are all associated with higher mortality and might confound the reported associations [21
Higher mortality in people with schizophrenia has been recognised since the 1930 s [11
]. Nonetheless, as stated earlier, this mortality gap has remained stubbornly unchanged [10
], and may even have increased over time [5
], despite developments in mental health services and the introduction of better tolerated antipsychotic agents. Previous studies suggest that around two-thirds of the excess mortality among people with schizophrenia may be attributable to natural causes, as discussed above [23
], with the remaining one third being due to suicide and other unnatural causes [23
]. Regarding bipolar disorder, a non-significant raised SMR for all causes of death was reported in another study performed in South London with a more limited sample size (239 cases with 42 deaths over 19 years) [27
]. Regarding SMRs for cases with depressive episode or recurrent depressive disorder, a particularly high mortality risk was identified among the younger age stratum (15-44 year olds, Table ). However, the SMRs for depression were relatively low [28
]. This finding may be due to the fact that people with depression known to secondary care are not representative of those with the disorder in the community and, in particular, that referral bias for secondary care favours those with relatively good health or higher social class [29
Our findings are, we believe, novel in the presentation of SMRs stratified by age, gender, and ethnicity. Effect modification, where demonstrated, may provide at least some supportive evidence regarding causal pathways since it indicates uneven distribution of risk; however, such conclusions can only be drawn tentatively and require confirmatory investigation. Gender differences in the associations of interest might reflect different levels of environmental support (for example, lower social support for men with schizophrenia might account for the higher SMR in that group), or might reflect gender differences in the severity of the condition in question or in the level of comorbidity with other physical, mental or personality disorder [31
]. The substantial differences in SMRs between ethnic groups again suggest that some of the associations between SMI and mortality may be socially mediated, although confidence intervals were wide for many groups and negative findings should be viewed with caution. The diminution of mortality risk with age may reflect survival effects, with those surviving with serious mental disorder to age 65+ being relatively healthier in other respects. Alternatively, people with SMI in older age ranges may be more likely to remain in contact with mental health services and adhere to treatment, whether for mental or physical disorders. However, the age diminution could also reflect differences in the nature of the underlying disorders, such as symptomatic differences between early- and late-onset schizophrenia, or differences in substances misused in younger and older adults. Further, it may reflect the excess risk of SMI being additive rather than multiplicative, so it is obscured by deaths from other causes in older age groups [7
This investigation has a number of strengths. We were able to draw on a large numbers of case records from the largest single provider of secondary mental healthcare in Europe. The National Health Service presents additional contextual advantages because of its near-total coverage of all aspects of healthcare in the UK. This investigation was able to draw on complete electronic clinical records of more than 31,000 patients diagnosed with SMI, depression or substance use disorders providing the statistical power to be able to differentiate between disorders as well as explore subgroup-specific mortality risk.
However, potential limitations should also be considered. First of all, confounders other than age and gender might still exist. Case registers derived from secondary healthcare offer particular advantages for investigating "high penetrance" disorders - i.e. those like schizophrenia and bipolar affective disorder where the chances of secondary care contact are high. For lower penetrance disorders, inferences need to be more cautious and both depressive and substance use disorders fall into this category - i.e. cases known to secondary care may reflect more severe primary disorders, comorbidity, environmental disadvantage or referral biases. Prevalence bias is an additional issue which needs consideration, in that the cases known to a service within a given time period are likely to be dominated by those with long and relapsing clinical courses - they therefore cannot be taken to generalise to incident cases. A further methodological challenge was that service data were principally (but not entirely) derived from a single catchment area whereas expected deaths were derived from national data. However, a sensitivity analysis using London-specific data, described above, did not reveal any meaningful differences. Finally, we have not included data on causes of death and further research is required for this, as well as for investigating specific risk factors for mortality within particular disorders. Diagnostic categories overlapped since a proportion of individuals suffer from more than one mental disorder, but no attempt in this analysis was made to consider comorbidity.