Number of Adverse Job Conditions and Mental Health
Initially, three random intercept models were fitted to examine the association between the total number of adverse job conditions and mental health in different age groups. The models included variables indicating the three job adversity categories, i.e., jobs with one, two, and three or more adverse conditions (jobs with no adverse conditions served as the reference group). The covariates described previously were included in the models. Mental health was the dependent variable. The results indicate a statistically significant decline in mental health for all age groups as the number of adverse job conditions increased. However, the decline in mental health of younger workers ( of age) who reported one adverse job condition was not statistically significant compared with jobs with no adverse conditions. In addition, the magnitude of decline in mental health when working in a job with two adverse conditions was greater for midaged [, 95% confidence interval (CI) = to , and ] and older workers (, to , and ) compared with younger workers (, to , and ).
This initial finding was systematically investigated in further random-intercept models, which were fitted in steps as presented in Fig.
1. The coefficient values and their corresponding standard errors (in parentheses) for the models are listed in Table
3.
Model A included variables reflecting age groups (younger than 25 years old served as the reference group) and job categories (jobs with no adverse conditions was the reference group). The model indicated a significant decline in mental health among workers who reported that they experienced adverse job conditions compared with those who experienced no adverse conditions. Moreover, the magnitude of mental health decline increased with the number of adverse job conditions. The poorest jobs (three or more adverse conditions) were associated with an average 6.93-point decline in mental health scores. A four- or five-point difference on the mental health scale is considered to be “clinically relevant” (
Butterworth et al. 2011b).
Controlling for covariates in Model B slightly reduced the magnitude of the coefficients; however, the relationships identified in Model A remained significant. Apart from job quality, the factors of being a female, having a long-term disability or health condition, experience of adverse life events in the last 12 months, having parenting responsibilities, being a single parent, and being a single person were all significantly associated with poorer mental health. The inclusion of survey wave numbers did not indicate any significant systematic change in mental health scores over time.
Model C included interaction terms between job quality categories and age groups to further examine the combined effect of age and total adverse job conditions on mental health. The significant associations previously identified in Model B remained significant in Model C, except for the decline in mental health for jobs with one adverse condition. Furthermore, Model C indicated that the decline in mental health experienced by midage workers who experience two adverse job conditions was significantly larger than the decline in mental health of younger workers in the same category of jobs. This was evident by the statistically significant () interaction term between Age group 2 and jobs with two adverse conditions. In addition, the interaction between jobs with two adverse conditions and Age group 3 was borderline significant (), suggesting a weaker but similar effect for older workers.
Model D included an additional variable to control for any experience of unemployment or not being in the labor force during the survey period. Previous research has indicated that periods of unemployment can negatively impact individuals’ mental health (
Paul and Moser 2009;
Backhans and Hemmingsson 2012). Controlling for this potential effect in Model D slightly changed the coefficients, but the associations previously identified in Model C remained significant. In addition, any experience of unemployment or not being in the labor force was negatively and significantly associated with mental health.
As a sensitivity test, another model was fitted that regressed mental health on all the variables included in Model D and included random effects for both slope and intercept. Thus, by allowing a separate slope for each individual, this model accounted for the differences between individuals in terms of their mental health score changes over the survey waves. The significant effects in this model were identical with those found in Model D, and the coefficient differences were minor. Comparison of the goodness of fit between this model and Model D, using the Akaike information criterion and the Bayesian information criterion, indicated only a small improvement when including a random effect for slope in the model. Therefore, it was decided to proceed with Model D.
Model D indicated that workers in all age groups whose jobs have two or more adverse conditions have lower mental health scores than workers whose jobs have no adverse conditions. Moreover, the inclusion of interaction terms (between job categories and age groups) indicated that the mean decline in mental health for midage workers who worked in jobs with two adverse conditions was greater than younger workers who worked in similar jobs, and the difference in means was statistically significant (). A similar effect was observed for older workers, but the coefficient was borderline significant (i.e., ).
Fig.
2 illustrates the moderating effect of age on the association between job quality and mental health based on the estimated coefficients in Model D. The mean mental health scores and the 95% confidence intervals were estimated for each category of jobs. The slope of lines connecting the mean values indicate the mean rate of decline in mental health scores for each age group as the number of adverse job conditions increased.
The rate of decline in mental health scores was relatively steady for older workers across different adverse conditions, and the rate of decline was slightly slower when moving to jobs with three or more adverse conditions. Midage workers’ mental health had a sharp decline when working in jobs with two adverse conditions and the rate of decline remained similar when moving to jobs with three or more adverse conditions. Younger workers experienced a slower decline in mental health scores compared with other groups when working in jobs with one adverse condition or two adverse conditions, but experienced the greatest decline among all the age groups when moving to jobs with three or more adverse conditions.
Job Quality Components and Mental Health
Models were fitted for each age group regressing mental health on the five job quality components (Table
4). The association between experiencing different types of adverse job conditions and mental health was explored separately for each age group (jobs with no adverse conditions served as the reference category). The models were adjusted for the covariates included in Model D.
Model E indicated that low job security and unfairness of effort and reward (as perceived by participants) were significantly negatively associated with mental health scores in younger workers. Models F and G indicate that for midage and older workers, in addition to low job security and unfairness of effort and reward, high job demand and complexity and high job intensity were also significantly negatively associated with lower mental health scores.
Comparing the magnitude of the coefficients suggested a larger impact of low job security on younger workers’ mental health (, to , and ) compared with other age groups. For older workers, the impacts of job demand and complexity (, to , and ) and job intensity (, to , and ) on mental health were larger compared with midage workers. For midage workers, any experience of unemployment or not being in the labor force was negatively and significantly associated with mental health.