Comparison and interpretation
Those with severe and diagnosed mental disorders are less likely to be employed [
40‐
42]. However, milder and self-reported mental symptoms are prevalent also among employed people, with repercussions on their functioning, work ability and even mortality [
4]. In our study, all-cause mortality, after considering covariates, showed no associations with either prior GHQ or MCS. Disease-based natural mortality neither showed any associations with GHQ or MCS.
In some previous studies, associations between similar mental symptoms and all-cause mortality have been found before and after considering sociodemographic and health-related covariates [
18,
20,
23,
25], whereas in some others the associations have not survived adjustments for prior ill health [
19,
21]. The associations for all-cause mortality have been relatively weak and observed, in particular, for the highest scores of mental symptom scales [
18,
23]. Examining mental symptoms longitudinally, their stability and adverse changes have shown associations with all-cause mortality [
28,
29]. We measured mental symptoms at baseline and the stability and remission over the follow up remains an open question. In a British study, social class was a modifier and the lowest class had the greatest mortality risk due to mental symptoms [
25], whereas in our study social class had no effects on the studied associations. Future studies would benefit from measuring symptomatology longitudinally as well as considering a broad range of sociodemographic and health related covariates.
Mental symptoms have further shown associations with cardiovascular, respiratory, liver disease and cancer mortality [
18‐
21,
23,
43 ]. These associations have been confirmed primarily for the highest scores of mental symptoms. The associations have been relatively weak and they have attenuated or even rendered non-existent after considering sociodemographics and prior ill health.
In our study, the association of mental symptoms with unnatural mortality was strong and consistent before and after adjustments. Half of the unnatural deaths were accidental, such as traffic accident, drowning or poisoning. However, the evidence of an association between mental symptoms and mortality from unnatural causes is limited [
18,
23,
44], with one study finding no association for mortality from injuries [
20]. Substance abuse, in particular alcohol abuse, is related to accidental deaths, and to mental symptomatology [
45].
The other half of unnatural deaths was due to suicides that we found to be particularly strongly associated with mental symptoms. Overlap between accidents and suicide is possible, but unknown to us in these data. We acknowledge that the number of unnatural deaths in our study was limited. Nevertheless, the models converged and the associations remained stable and statistically significant throughout the analysis for unnatural mortality in general and suicidal mortality in particular. Similar associations were found for GHQ-12 as well as SF-36 MCS before and after adjustments. While suicide has been previously studied as a cause of death among those with severe and diagnosed mental disorders [
4,
23,
44], our understanding of the links between mental symptomatology and subsequent suicide has been much poorer.
That the two instruments of mental symptoms were in a similar way associated with accidental and suicidal mortality provided stronger evidence than using one instrument only. Previous studies have not included simultaneously multiple measures of mental symptoms, and thus comparing the mutual consequences for natural, accidental and suicidal mortality of symptomatology measured in two or more ways has not been possible. Further studies with multiple measures of mental symptoms and mortality are warranted.
Among the risk factors for suicide, prior severe mental disorders play an important role [
35,
46,
47]. Of such disorders, depression is the most prevalent among people who die by suicide [
5]. The evidence comes largely from clinical samples and uses diagnostic information on mental disorders. Major depression and other diagnosed mental disorders contributing to suicide mortality suggest the importance of mental pathways to suicide. The strong associations between the two instruments of mental symptoms and subsequent suicide mortality suggest that the instruments are markers of such pathways. Thus, mental symptoms are likely related to various comorbid mental disorders, such as depression, anxiety, psychotic and personality problems as well as substance abuse. It has been shown that mental symptoms, such as those indicated by GHQ, are comorbid, in particular, with depressive and anxiety disorders [
48]. In order to provide a fuller picture of the contribution of mental symptomatology to suicide and other unnatural causes of death, further studies need to consider multiple comorbid disorders in conjunction with mental symptoms.
The current evidence on the associations between mental symptoms and mortality is based on a limited number of studies using divergent study populations, sample sizes, study designs, measures of mental symptoms, causes of death and sets of covariates. This contributes to the heterogeneity emerging from the existing research base, showing both associations and no associations. Our study, suggesting associations between mental symptoms and unnatural mortality, in particular suicidal mortality, highlights the need for valid measures and strong study designs.
Methodological considerations
Participation to our baseline survey was acceptable. Consenting to data linkage was high but contributed further to the non-participation. We were able to conduct extensive non-response analyses for both participation and consenting to data linkage [
31]. These analyses suggest that major bias in the data is unlikely. However, we acknowledge that the non-participation remains a potential source of bias.
An advantage was that reliable and complete register based data on causes of death could be individually linked to our baseline survey data, containing a range of sociodemographic and health-related covariates. Nevertheless, residual confounding cannot be ruled out.
We were able to employ two often-used instruments of mental symptoms, i.e. the General Health Questionnaire [
9] and the Short Form 36 mental component summary [
10]. Both are reliable and well-validated inventories, but we acknowledge that self-reports are subject to potential reporting bias, and clinical and diagnostic tools are equally needed. Our instruments reflect generic mental symptomatology and we were unable distinguish between key types of mental symptoms. We were also unable to consider the comorbidity of mental symptoms with more severe mental disorders.
The 12.5-year follow up among employees included relatively few deaths, and the number of unnatural deaths and suicides was limited. Irrespective of this, our analysis was able to produce meaningful and stable estimates, but analyses that are more detailed could not be done.
Our data were derived from a cohort of Finnish midlife employees. The data were thus non-clinical and not based on self-selection. Nevertheless, occupational cohorts tend to be healthier than general populations due to the healthy worker effect. Such selection may well concern our cohort, although milder mental symptoms are less likely to cause exit from work than more severe clinically significant mental disorders.
The female majority in our cohort is a potential limitation as there are sex differences in the causes of death among the working age population, with women dying more often from cancers and men more often from cardiovascular and accidental causes. However, in this midlife employee cohort, male and female mortality was relatively similar and we found no sex interactions in the studied associations between mental symptoms and mortality. Nevertheless, direct generalisations of our findings to broader age groups, general populations or further national contexts are not warranted.