Limitations and strengths
Prior to proceeding with the discussion of the results of this study, certain important methodological issues must be discussed.
A major limitation of this study is that it is a cross-sectional survey, making causal interpretations hazardous. We thus refer to associations between variables, instead of emphasizing the concept of determinants. Our study does not allow for conclusions on the effects of unemployment on health and is only a descriptive analysis of variable patterns related to perceived health in a specific group of long-term unemployed persons in Germany. Second, a valid estimate of the true response was hindered by several selection biases which restrict the applicability of the findings: besides a health-related selection into unemployment, a selection bias due to voluntary participation and the non-randomized selection of two settings of welfare-to-work organizations is likely. Study participants may constitute a subgroup of more motivated long-term unemployed persons, and so a generalization of the findings is consequently not valid. Third, as self-reported health data may lead to under- or over-estimation of the real health status of a given individual, we cannot objectively describe the health status of our sample. Moreover, the list of symptoms and conditions is not complete. The use of perceived health as a dependent variable in our model is methodologically grounded in recent studies [
20‐
22], but remains to be validated in longitudinal approaches. Finally, the number of participants in the two settings was unequal (setting Munich, n = 289; setting Hanover, n = 76) and the respective settings appear to be geared toward the needs of distinct sub-groups of the long-term unemployed.
The primary advantage of our study stems from the fact that it is setting-based, and that we analyzed a target group rarely addressed by health research, mainly due to the expected non-compliance of individuals to study conditions. Apart from a reasonable number of participants, it was also possible to include and better characterize important subgroups, such as participants with personal migration experiences. The use of validated, standardised questionnaires to assess perceived health, anxiety, and depression is a further strength of this study. The SF-12 questionnaire was developed for use in general populations (i.e. not for express use in clinical settings), and is traditionally used to measure functional health status. It has also been used in unemployment research previously [cf., 20, 21, 22], where patients’ perception of health was discussed as an important obstacle to a successful return to work. The HADS questionnaire is a clinically validated instrument which has also been used in unemployment research [
21,
36]. Furthermore, health behavior and socio-economic variables relevant for both health and the ability to work were documented.
Interpretation
Compared to general population data and to other reports on the health status of unemployed persons in epidemiological studies [
21] or intervention trials [
22], participants in this cross-sectional study showed signs of reduced health with respect to the following parameters:
1) Perceived health in our sample differed significantly not only from the norm values of the general population, but also from findings in other unemployment studies. The mean value of perceived physical health in our group of long-term unemployed persons was 44.6 (SD = 10.0). As an illustrative example, a recently published Norwegian sample of unemployed persons [
21] yielded a value of 49.62. The difference in perceived mental health was even more acute, with a mean value in our group of 44.0 (SD = 11.3), as opposed to 51.23 in the Norwegian sample. A five point difference in the SF-12 scores is considered indicative of a clinically relevant change or difference [
30]. The description of the Norwegian sample did not provide additional information on the length of unemployment of their participants, although the authors did note that long-term sickness absentees were analyzed separately. The health differences between the two samples illustrate the need for more differentiation and subgroup analyses in unemployment research.
2) The low levels of perceived health observed in this study, with particularly pronounced deviations from the reference value for the mental SF-12 sub-score, are mirrored by HADS-Anxiety and HADS-Depression scores which are also significantly lower than the corresponding reference values. Compared to other samples of unemployed persons, the HADS results differ according to the HADS dimension: The mean value of the HADS-Depression score in our sample was 6.4 (SD = 4.2), as compared to 4.9 (SD = 3.0) in a sample of younger unemployed persons (mean age 29 years) in Germany who had experienced several instances of unemployment [
36]. The mean value of the HADS-Anxiety sub-scale in our sample was 7.3 (SD = 4.2), similar to the 7.44 (SD = 3.32) reported in the younger cohort of unemployed persons [
36]. In a meta-analysis by Paul and Moser [
14], the average number of persons with psychological problems among the unemployed was 34%, whereas in our sample of very-long-term unemployed persons 55% showed evidence of depression and/or anxiety disorders.
3) The prevalence of obesity in our study was 29%, as compared to 11.7% in women and 12.9% in men in the general German population [
39]. In a Dutch multidisciplinary health program for unemployed persons with health complaints, 29.4% of individuals were classed as obese [
22]. In our sample, 34% of participants were shown to be suffering from hypertension, whereas in the general population the one-year-prevalence of hypertension was only 7.3% among women and 9.8% among men [
39]. Health-risk behavior was also widespread in our sample: 54% of participants were smokers, whereas the prevalence rates for smoking in the general German population stand at 33.4% among women and 42.2% among men [
39].
The results of the linear regression models indicate that mental and physical perceived health in long-term unemployed persons are both significantly associated with mental health status. Not surprisingly, mental health status as measured by HADS was the main predictor of perceived mental health as measured by the SF-12 mental component score (p < 0.0001). The fact that mental health status as measured by HADS was also the main predictor of perceived physical health is consistent with the findings of the Whitehall II study, which demonstrated a moderate correlation between mental and physical health among people with a low socioeconomic status [
40]. Schutgens et al. [
22] conclude that “perceived health may be influenced by cognitions, for example the way people cope with their health problems”, which might explain the impact of mental health status on reduced PCS and MCS scores in our sample.
Perceived mental health was also significantly associated with individuals’ level of physical activity, providing some evidence of the importance of physical activity to this particular group. Previous studies have reported contradictory evidence concerning the impact of physical activity interventions on the health status of the unemployed: A Dutch health program, focused on changing health behaviour—e.g. by offering sessions of physical activity—and teaching coping strategies for health complaints, failed to show positive effects on perceived health [
41], whereas Watson and colleagues [
42] reported some positive effects stemming from physical activity interventions in combination with cognitive behavioral interventions among unemployed persons.
The fact that women reported significantly lower levels of perceived mental health in our sample of long-term unemployed persons supports calls for more gender-focused research into the long-term health effects of unemployment. In the meta-analysis of McKee-Ryan et al. [
4], the changing role of women in the workplace is discussed as a potentially plausible explanation for the differences in association between health and unemployment among men and women.
Interestingly, the setting of the welfare to work organization (the Hanover job centers as opposed to the non-profit organizations of the secondary labor market in Munich) appeared to play a more important role in differentiating between unemployed persons with different levels of perceived health than did the personal socio-demographic characteristics of participants. The exact reasons for the differences between the two settings can only be hypothesized: In Hanover, the average age of participants was higher than in Munich; the setting variable remained significant, however, even after adjusting for age and gender. From a salutogenic perspective the setting in Munich offers more structure, increased levels of social support, and a better perspective for their clients than does the Hanover job center. These setting specific differences may contribute to improvements in perceived health, especially in a subgroup possessed of severe mental health problems. An additional difference between settings is the fact that the labor market in Hanover is less dynamic than that in Munich. The results concerning differences in setting should be interpreted with caution, however, as the two settings may address different sub-groups of unemployed persons, and both the recruiting process and requirement of voluntary participation may have biased the results.
Surprisingly, we could not validate the influence of some factors mentioned in the literature which might have influenced the perceived health of the individuals comprising our sample. According to the meta-analysis of McKee-Ryan and colleagues, sample type—in particular, recent graduates as compared to mature unemployed persons—is an important predictor of mental health among the unemployed [
4]. In contrast, educational level and personal migration experience were not found to contribute to the final regression models used in this study, and did not show significant associations with perceived health in the bivariate analyses of the cross-sectional analyses presented here.
In Paul and Moser’s meta-analyses [
14], a curvilinear relationship between unemployment duration and mental health was identified, with a stabilization of mental health during the second year of unemployment and increasing psychological problems among very-long-term unemployed persons. In our selected sample, more than 50% of participants had been unemployed for five or more years. The bivariate analyses of the SF-12 mental sub-score did not show significant differences between the different categories of unemployment duration. The equivalent analysis of differences in the physical health score, however, showed that participants with a history of five or more years of unemployment reported significantly reduced physical health scores.
Although the regression analyses identified important factors associated with self-perceived health in our sample, the findings also suggest that additional factors not measured in our study may explain key differences in perceived health among the long-term unemployed. The fact that only 20% of the variation in the SF-12 physical component scores was explained by the regression model—which included age, gender, setting, and BMI, along with mental health status according to HADS—illustrates the need to gain a better understanding of those factors other than mental health status or setting effects influencing the perceived health of long-term unemployed persons. Due to the fact that only a fairly select few health variables were assessed in our analyses, our findings here are preliminary and limited. Additional methodologically elaborated research, particularly longitudinal designs, will help elucidate the relationship between perceived health, long-term unemployment, and both identified and as-yet-unidentified explanatory factors.