Does unemployment contribute to poorer health-related quality of life among Swedish adults?

Unemployment is in general something that has a negative effect on health [1–3], not only at the time of becoming unemployed, but also in a longer time span [4–8]. This relationship has previously been shown for different dimensions of self-assessed, mental and physical health and for depressive symptoms [2, 3]. In previous research, priorities have often been focused on determining the magnitude of the effect at the population level, e.g. through the meta-analyses by McKee-Ryan et al. [1] and by Paul and Moser [3]. However, the study context is highly important to consider because the health consequences from unemployment vary between groups of individuals, e.g. men and women, but also between and within countries and over time [2]. For some groups of individuals, e.g. for Spanish women [9] and for Swedes with only a primary-school education [10], there are even indications of a positive or no effect on health from unemployment. Therefore, policy decisions attempting to reduce unemployment from a health perspective need to consider for whom it is most important to create new job opportunities. Despite a growing literature about the health aspects from becoming unemployed, there is still not much knowledge about how unemployment affects groups of individuals in relation to education level, marital status, previous health, and gender [2].

Despite many health measures being used to study the effect on health from unemployment, studies using health-related quality of life are lacking, and probably the most commonly used instrument in public health research, the EuroQol 5 dimensions (EQ-5D) [11], has as far as we know, not previously been used in unemployment research. The EQ-5D instrument consists of a descriptive system with five questions, a visual analogue scale (EQ-VAS), and a value set. An advantage with the EQ-5D is that its value set, which is derived from responses to the descriptive system, allows it to be translated to so-called Quality Adjusted Life-Year (QALY) scores [12, 13]. The QALY score has two anchor points, 0 (death) and 1 (full health), which makes it possible to compare the magnitude of different public health problems. QALYs also enable comparisons between conducted and potential interventions in a systematic manner, which makes them attractive and commonly used in cost-effectiveness studies of new interventions.

In studies of unemployment and health, unemployment might at least partly be due to poor health, and for this reason one or more variables related to health before unemployment are therefore important as part of the statistical analyses. It is also important to add other potentially confounding variables. In previous research, gender, age, education level, marital status, household income, geographic location, and social network/support have been used most frequently in the statistical models and in presentations of stratified estimates [2]. Gender might be the most important variable to stratify for based on previous research and there are studies that show that women are more affected than men by unemployment as well as studies showing the opposite [2].

Our aim was to examine how unemployment impacts upon health-related quality of life among Swedish adults, and investigate these effects on population subgroups defined by education level, marital status, previous health, and gender.

In current study we have explored the impact on health-related quality of life among Swedish adults on unemployment. We show that unemployment is strongly related to a poorer health-related quality of life. The magnitude of the effect is large, with an absolute loss of QALY of 10% from unemployment according to our estimate. The effect on QALY is mainly explained by an increase in problems with anxiety/depression due to unemployment. In our study, 24% more of the unemployed than the employed had problems with anxiety/depression, despite there being as many as 35% of the employed having at least some problem. Also, for usual activities unemployment was shown to affect health negatively, while evidence of an increase in problems with pain/discomfort could not be statistically supported, though these were numerically more common among the unemployed than employed. For grouped analyses, after becoming unemployed married individuals, young individuals, and those who already had poorer health before their unemployment showed greater problems than people did in general.

That unemployment is negative for health has been shown in most previous studies [2, 3], and our study is well in line with these studies. Our study uses QALYs, which through its construction makes it possible to present the magnitude of health effects in both a comparable way with other public health problems and in terms of health-related quality of life [13]. For instance would our estimated effect of 0.096 mean that 9.1 years of employed years would be valued above living 10 years as unemployed. Other studies, such as meta-analyses have presented effect sizes [1, 3], but also effect sizes are difficult to translate in a similar way as QALYs.

It is interesting that it is mainly due to feeling anxious and/or depressed that unemployment causes significant problems in our study. In previous research on unemployment and health, the health measures and the health dimension in focus have varied to a large degree [2]. In the meta-analysis by Paul and Moser, which was published almost 10 years ago [3], effect sizes were presented for different aspects of mental health. Most of these measures showed a similarly sized negative health effect due to unemployment. It is thus difficult to draw firm conclusions about which health aspects are mainly affected by unemployment based on the literature. That our results clearly indicate more extensive problems related to anxiety/depression than for pain/discomfort might be explained by the fact that unemployment has both economic and social consequences. At the time of unemployment, some unemployed might have reduced their bodily burden after becoming unemployed and thus had lower problems with work-related pain/discomfort. Thus, pain/discomfort might be related to more of a problem than our study suggests, and it is advisable that longitudinal and qualitative studies investigate these health problems more in depth in relation to unemployment.

The results we present on the group level indicate, despite a higher estimated QALY for women, that health-related quality of life might be more negatively affected in men than in women by unemployment. Previous results have shown diverging results in a global perspective, as well as in Swedish studies [2]. Among previous Swedish studies, those using the General Health Questionnaire, which focuses on psychological symptoms, including anxiety, have indicated more problems due to unemployment for men than for women [23, 24], while those studies that have shown more problems for unemployed women than men have mainly used a question about self-rated health with 3 or 5 responses [6, 10, 25–27]. Thus, our results that indicate that men are more affected by problems with anxiety and depression than women when they become unemployed are in line with previous studies. It can be speculated that unemployment causes different health consequences for men and women, highlighting the need for looking at health in a broader view for determining who is most strongly affected by unemployment among men and women, but also for other groups of individuals. More research is needed to better understand differences in unemployment experiences for men and women because our study does not provide strong enough evidence to draw firm conclusions.

Our study suggests more problems for young people than others due to unemployment in the short-term perspective. Previous studies have shown varied results from the short-term perspective [2], and it is therefore difficult to draw firm conclusions. It has also been shown that there is a long-term effect on health from youth unemployment [4, 7, 8]. Even if the short-term negative health aspects might not be larger for the younger age groups, still, from the full time spectrum, the largest health consequences are among the youngest, at least if being long-term unemployed. Also, long-term consequences of unemployment have been shown for middle-aged persons despite being re-employed [6].

Few previous studies have presented results for different education levels [2]. Our study could not add much to this knowledge because the numbers of unemployed in the three education groups were too few for stable results. Our study suggests that married people who become unemployed fare worse than single people who become unemployed. Also for this group, the evidence base is weak mainly because results are rarely presented on the group level for marital status.

It is difficult to get good estimates for the effect on health from becoming unemployed, and this is mainly due to problems with health selection. In a cross-sectional study design such as ours, this is an even bigger hurdle to overcome than in the recommended longitudinal studies, which also face problems with showing causality. To improve the estimates, we asked respondents about their previous health status in order to get an idea about the contribution to health deterioration that is from the unemployment episode. Interestingly, despite expressing poor previous health, which was not to a greater extent likely to be related to unemployment, health was more affected for these individuals than for those who had good health before their unemployment. Targeting people with already poor health for labor market measures might therefore be even more important than previously considered, at least in the Swedish context. To our knowledge it is rare that results are presented stratified on a proxy for health before unemployment, such as in our case health 5 years prior to participation in the study. Our study focus is on the short-term perspective. A previous Swedish study, which focused on the long-term perspective of being unemployed, showed a similar long-term effect for those with good and poor health when becoming unemployed [6], thus, also indicating that those with poor health already at the time of unemployment were being affected by unemployment.

As already mentioned, our study has the limitation of being a cross-sectional study. In our study, we had a low response rate which also makes it difficult to draw firm conclusions. However, the number of responders was still large and conclusions for at least the group as a whole should therefore be valid. We defined unemployment as at least six months of unemployment during the past three years. Using current unemployment would have meant too few cases, and it would not have guaranteed that the person had an unemployment episode that would have meant a problem for him/her. Our definition of unemployment is common in similar previous studies, and we believe that it is a good way of capturing the exposure of being unemployed in both the short and long-term perspective.

A strength with our study is that it is the first to study health-related quality of life extensively. Our results on the group level might be weak, but they still provide important information that can be used by decision makers. They also provide important information to be used for future studies.

In our sensitivity analysis, we could show that those who responded as having extreme problems affected the results for gender. For the first scenario we had only small differences with our main scenario where we did not censor extreme responses. For our second scenario, we found some interesting differences, but because these depended on the fourth and fifth dimension where we had a large proportion of individuals, we think that this scenario should be seen as more of an illustration of how sensitive the results are for these responses, and not as a sign that the results from the main analysis are misleading. Still, our results highlight that the EQ-5D should rather be seen as an indicator of health and not as a measure that can produce a fine-tuned picture of an individual’s health status. In our view, no other set of questions than the EQ-5D, which measures self-assessed health, can more precisely estimate health-related quality of life on a scale from 0 to 1, so the EQ-5D should still be strongly advisable to use.

We used the United Kingdom value set for EQ-5D for our analyses [12]. There is a recent Swedish value set for EQ-5D, but it has rarely been used and this makes it difficult to compare QALYs with other populations [28], which is important from a health economics perspective. Our results could therefore be biased due to being based on a non-Swedish population. The United Kingdom value set for EQ-5D is over 20 years old, which might also bias results. However, we find no reason to believe that this would have more than a marginal effect on our findings.

In our study, we would have liked to have split the employed into permanent and temporary employees. However, our small sample size did not allow for such a discrimination of the respondents. Splitting the group of employed might have resulted in a larger gap between the permanently employed and unemployed than we observed in our study, which would, if true, have further emphasized the need for permanent positions instead of unemployment. The debate about stable positions in relation to non-stable positions is a hot topic politically, at least in Sweden, and there is evidence suggesting that job insecurity is as bad for health as unemployment. [29] Knowing more about this is important, but this was outside the scope of the present study.

In our study, we present a QALY difference of 0.096 between the unemployed and employed. As previously touched upon in the discussion, the interpretation of this QALY difference from the health economics theory is that 9.1 years of employment should be valued above 10 years of unemployment, i.e. that it is better to live 9% less lifetime if unemployment can be avoided. This highlights the importance of policy makers prioritizing from the public health perspective rather than from an economic growth perspective. However, the latter tends to dominate discussions about lowering unemployment. We might have overestimated the negative health effect of unemployment, but even so the effect is nevertheless likely to be negligible. In line with a previous review [2], we recommend that more focus should be put on group-wise analyses in order to determine for whom it is most important to prioritize efforts to avoid unemployment, both in regard to political priorities and for future research.

An added value with our study is that, besides increased knowledge about how unemployment relates to health-related quality of life, our results can be used for health economic studies because QALYs are the most common measure used in cost-effectiveness analyses. We also intend to use our results for a future health economic evaluation where we study the cost-effectiveness of increased staffing within home care.

Our study was performed in Sweden. The results are likely to be representative for other countries, e.g. other Nordic countries, that have a similar labor market. Our results should add valuable information also for other European countries even if their labor market system more or less differ from the Swedish.

In our study, we show that the health deterioration from unemployment is likely to be large, as our estimated effect implies almost 10% poorer health-related quality of life (in absolute terms) from being unemployed compared to being employed, and the problems are most apparent for the anxiety/depression scale of the EQ-5D instrument. Our results show, just like previous research, that unemployment hits groups of individuals differently, and measures directed to specific groups of individuals are therefore recommended. For future research, it is important to put more emphasis on groups of individuals to get a better basis for prioritizations within labor market measures in order to lower the public health impact of unemployment.