Unemployment and Early Cause-Specific Mortality: A Study Based on the Swedish Twin Registry

Margaretha Voss, PhD, MPH; Lotta Nylén, MPH; Birgitta Floderus, PhD; Finn Diderichsen, MD, PhD; Paul D. Terry, PhD


Am J Public Health. 2004;94(12) 

In This Article


The results from this study suggest that unemployment is associated with an increased risk of early death even after adjustment for several potential confounding factors, including socioeconomic status, lifestyle factors, and genetic and early childhood factors. In particular, unemployment was associated with increased mortality from suicide and external undetermined cause. Among unemployed men, an increased risk of death from malignant neoplasms was also suggested. Furthermore, the results indicate that the risk of early mortality following unemployment may be strengthened by social, health, and personality factors.

One way to assess unemployment's public health impact is to estimate the attributable fraction (the proportion of deaths that would be eliminated if mortality among the unemployed were reduced to that among the employed). The estimate depends on the strength of the association between exposure and outcome and also on the prevalence of the exposure. On the basis of our study results (24-year follow-up), and keeping the statistical imprecision in mind, the attributable fraction for suicides was 20% among women. For men, the attributable fraction for external undetermined cause was 38%.

Information about current or previous unemployment was collected at baseline in 1973 only. The number of unemployment episodes and the duration of each episode was not recorded. "Ever unemployed" may therefore apply to 1 or several occasions of different length, remote from or close in time to the start of follow-up. In addition, we had no information on unemployment occurring after 1973. Exposed individuals could repeatedly be unemployed during follow-up. This does not lead to misclassification of the exposure because we did not take different levels of exposure into account -- the exposure we analyzed was "ever unemployed." Individuals classified as never unemployed as of 1973 could encounter unemployment later on and therefore be misclassified. This limitation would lead to underestimated risk estimates if those who became unemployed after 1973 had the same mortality pattern as those reported as ever unemployed as of 1973. If the individuals who became unemployed after 1973 had a lower death rate than those earlier classified as unemployed, then the reported estimates could be biased upward; however, we have no reason to believe this to be the case. On the other hand, it is uncertain to what extent the results obtained are valid for more recent time periods, when worker groups other than those of the present study are facing unemployment.

We used a 24-year follow-up as well as a shorter period: the first 10 years of the follow-up (1973-1982). For several of the cause- specific diagnoses, higher mortality rate ratios were found for the 10-year period than for the longer follow-up. This finding could owe partly to increased misclassification of unexposed individuals over time. It could also be that unemployment entails an increased risk of death (during a limited time period) that gradually fades away. The results suggest that unemployment may have an impact on mortality, not only in a short-term but also in a long-term perspective.

In accordance with other studies,[12,21,25] several factors related to mortality were significantly more prevalent among the unemployed than among the never unemployed ( Table 1 ). These characteristics could either lead to or be a consequence of unemployment. The association between unemployment and mortality may therefore partly be a result of selection mechanisms, in that individuals with certain risk indicators are more likely to become unemployed than those without the risk indicators in question. However, unemployment may also contribute to the development of these risk indicators and, in turn, to poor health. The cross-sectional data did not permit a clear determination of the timing between risk indicators and unemployment in this study.

Adjustment for risk indicators that constitute links in the causal chain between an exposure and an outcome may inaccurately reduce an association and mask an actual effect or part of the effect attributable to the exposure. Low education, personality factors, and low socioeconomic status are comparatively stable over time and may in general precede unemployment. Shift work should also mainly (but not exclusively) precede unemployment, assuming that occupational mobility is comparatively low in terms of this occupational characteristic. Furthermore, we see no reason why unemployment should increase the probability of holding a job with shift work. Serious or long-lasting illness could also be a precursor, since subjects with poor health may be more likely to lose their job. Divorce may to some extent be a consequence of unemployment, and smoking and use of sleeping pills and tranquilizers could be a coping behavior owing to psychological stress caused by unemployment. In the extended multivariate analyses, the relative risks were often attenuated compared with the age-adjusted relative risks, and this attenuation may be partly attributable to unwarranted adjustment for factors in the causal chain.

A major advantage of our study was that it controlled for confounding owing to genetic and early social and environmental conditions by analyzing unemployment among discordant twin pairs. In general, twins have early social and environmental conditions in common; in addition, monozygotic twins are genetically identical and dizygotic twins have half of their genes in common. The twin analyses showed an increased risk of death for the exposed twin compared with his or her twin sibling. The confidence intervals were wide owing to a rather small number of deaths. Nevertheless, it is notable that this analysis, which controlled for many predisposing life conditions, yielded risk estimates comparable to the analyses based on the full cohort.

We found an association between unemployment and external causes of death for both women and men. This finding is consistent with those of other studies,[9,12,13] although few of these included women.[9] Suicide was clearly associated with unemployment among women, but not among men. On the other hand, men who experienced unemployment had an increased risk of death by external undetermined cause. This finding raises the possibility that suicide is less likely to be identified among men than among women. In some studies reporting an association between suicide and unemployment among men, deaths by external undetermined cause and suicides were combined.[19,20]

The results indicate that unemployment has an important impact on mental health. Unemployment may cause a deterioration of economic situation, downgrading of social status, broken social relations, changed risk behaviors, impaired psychological well-being, and depression, consequences that may develop into severe illness.[1,2,3,4,6,7,8,17,18,25]

Several studies have reported an increased risk of mortality from cardiovascular disease with unemployment.[1,21,22,23] Our study showed no such increased risk, which is in accordance with another recent Swedish study.[16] It has been suggested that inability to control for behavioral and medical parameters before and after unemployment, and a too short follow-up period, contribute to difficulties in showing a possible relationship between unemployment and cardiovascular diseases.[15] Our results do not support this view but rather indicate that an association between unemployment and death from cardiovascular diseases observed in some studies may be confounded by other risk factors.

Other studies have pointed out that the excess cancer mortality among the unemployed owes mainly to an increased risk of lung cancer.[2,11,12,21,24] Our results for men are consistent with this observation, although they lack precision owing to small numbers. As in our study, other studies have reported a higher prevalence of smokers among the unemployed, and they have shown that smoking habits stay quite stable during unemployment.[11,21] In our study, the prevalence of smoking among short-term and long-term unemployed women was similar, and there was only a slight difference among men, suggesting that, to some extent, smoking may be a precursor of unemployment. It is possible that smokers have a greater risk of losing their job either because of smoking or because of other factors related to smoking.

An increased mortality from alcohol-related diseases among the unemployed was reported by Martikainen.[12] Our data were consistent with such an association among men, although the results were based on few deaths and the risk estimate decreased in the full model. Morris et al. reported a higher prevalence of alcohol use among the unemployed.[21] In our study, increased alcohol use among the unemployed was seen only among women who experienced short-term unemployment in 1973.

As far as we know, no other study has focused on the question of interaction between unemployment and the characteristics of the individual relative to early death. It has been suggested previously that buffering effects from social support, for example, could reduce the negative effects of the stress of losing a job.[4,7,8,10] Our results indicate that modifying the effects of individual characteristics could strengthen the association between unemployment and mortality. For men, but not for women, low education seemed to enhance the association between unemployment and mortality. It is possible that unemployment means greater strain in economic and social terms, particularly for men with low education.

Furthermore, our results suggest that use of sleeping pills or tranquilizers may enhance the risk of early death among individuals who experience unemployment, particularly women. Use of these drugs may be caused by unemployment, and this type of coping may reflect a particularly strong reaction. Apart from the availability of drugs, which may be the direct cause of death, the interaction could also signify that women using these drugs are more vulnerable owing to different psychosocial problems or mental diseases, and that unemployment may augment a prevailing difficult situation. In addition, the synergistic effects indicated for unstable personality traits and for serious or long-lasting illness may indicate that unemployment among individuals already burdened by psychological disorders, somatic illness, or both may overwhelm the individual.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.