Comparison with previous studies
To the best of our knowledge, no previous studies have investigated the association between conflicts at work and the risk of dementia. In most cases, clinical dementia is preceded by a decline in cognitive function, and although a decline in cognitive function does not always lead to a dementia diagnosis (Richard and Brayne
2014); we also include results from studies of cognitive functioning in the following discussion.
Our findings are in contrast to results from two previous studies that reported an adverse effect of negative aspects of social relations on cognitive function (Liao et al.
2014; Xu et al.
2016). Yet, another study showed no effect (Seeman et al.
2001), and another one reported better cognitive function among participants reporting higher social interactions (Hughes et al.
2008). Similar to our study, some of these past studies estimated negative social relations in midlife (Liao et al.
2014; Seeman et al.
2011). Nevertheless, the contradictions between our findings and findings from past studies suggest a differential role of social relations at work versus social relations in private life on cognitive function. Social relations in private life could be more important to people, and the impact may often be more prolonged, whereas people are likely to leave their job if they experience major conflicts at work.
Our findings raise the question whether there is a sex difference in the association between conflicts at work and dementia. Having conflicts with a supervisor and colleagues appear to be associated with a higher risk of dementia among men, although, none of the estimates were statistically significant and the results were based on few cases. Only a few studies stratified their analyses by sex (Liao et al.
2014; Seeman et al.
2001). However, these studies did not reveal any difference between men and women regarding the association between negative social relations and cognitive function. Because of gender segregation (both horizontal and vertical) at the Danish labor market in the 1980s and 1990s (Bloksgaard
2011), the psychosocial working environment may have differed between male- and female-dominated jobs. Such differences may partly explain the difference between men and women regarding the association between conflicts at work and dementia. However, we cannot support this hypothesis with empirical data, as we do not have information about the participants’ working environment.
We hypothesized that as a severe stressor, conflicts at work might be linked to the risk of dementia in old age through various potential mechanisms. For example, conflicts could cause prolonged stress that leads to brain atrophy and Aβ deposition in the brain, which are considered hallmarks of Alzheimer’s disease. Prolonged stress also influences cardiovascular health, mental health and health-related behaviors, which could eventually lead to dementia. However, our data did not support our hypothesis. There may be several reasons for the negative results, including the possibility that (1) the follow-up time was too long resulting in a dilution of the effect, (2) we had too few cases, (3) conflicts were only assessed at one point in time, (4) the applied questions assessing conflicts at work were insufficient to assess actual conflicts, or (5) that our hypothesis was not valid.
Strengths and limitations of the study
The main strength of our study is the long-term follow-up and the exclusion of the first 5 years of follow-up after the baseline assessment, which was done to reduce the risk of reverse causation, although reverse causation might still be a concern (Sperling et al.
2011). Nevertheless, with the long follow-up time, the association is likely to be diluted. This might be difficult to avoid when investigating long-term effects of midlife exposures on an outcome occurring primarily in old age, such as dementia. An additional strength of the current study is the adjustment for a wide range of covariates including various indicators of private life social relations, e.g., satisfaction with private life social relations and having a confidant or not. Furthermore, there was a negligible loss to follow-up due to emigration.
One of the main limitations of the present study is the low prevalence of participants who reported prolonged or serious conflicts at work (8.5%). The prevalence was somewhat lower than observed in a recent survey in Denmark, in which conflicts at work were operationalized as “experiencing any quarrels or conflicts at work within the past 12 months” (The National Research Center for the Working Environment
2016). Conflicts at work have been investigated within the past decades, within the areas of organizational management and occupational medicine. Some scales for assessing conflicts at work were developed in the 1980s and 1990s, including the Interpersonal Conflict at Work Scale, the Organizational Constraints Scale, and the Quantitative Workload Inventory (Spector and Jex
1998). In our data, conflicts were assessed by two questions, and our measures might not include all the variations that the more comprehensive scales capture. Future studies should investigate the association between conflicts at work and dementia using a validated scale. As our measures of conflicts at work did not address the number, duration, severity, and timing of conflicts at work, we are likely to miss information that is important for the hypothesized association between these conflicts and dementia risk. Also, our measures cannot differentiate between the duration and severity of conflicts, as it combines “prolonged” and “serious” conflicts as one entity although the two could have different effects. Another limitation related to our measures of conflicts is that they were assessed at only one point in time. This could have led to exposure misclassification, which may have diluted the observed association between conflicts and dementia.
Furthermore, self-reported information on conflicts at work and private life social relations can be influenced by individual characteristics. For instance, people who score high on the personality trait neuroticism may report more conflicts and negative social relations. According to a recent review, neuroticism has been shown to increase the risk of dementia, while conscientiousness protects against dementia (Low et al.
2013). Moreover, other studies have suggested that neuroticism increases the risk of interpersonal conflicts at work (Appelberg et al.
1991) and decreases the number of confidants (Kendler et al.
2003). Therefore, our results could be confounded by unmeasured personality factors. Lifestyle-related factors, e.g., smoking, physical activity during leisure time, and alcohol consumption, can reflect people’s behavior and thereby also certain personality traits. However, we cannot confirm this hypothesis with our data. Adjustment for lifestyle-related factors in our analysis, however, did not change the results substantially.
Among other unmeasured confounders, work-related factors (job control and job complexity) and cognitive ability at baseline may have influenced our results. Low job control and low job complexity in midlife are associated with lower cognitive function and a higher risk of dementia in old age (Then et al.
2014). We adjusted for educational attainment in our analyses, as educational attainment influences what type of occupation a person will have. Furthermore, a recent study suggested occupational complexity strongly mediates the cognitive gain associated with higher levels of education (Fujishiro et al.
2017). Also, educational attainment has been shown to be strongly correlated with cognitive ability in both young adulthood and later life (Mortensen et al.
2014). Yet, adjustment for education did not considerably change the results in the present study. Therefore, it is unlikely that lack of adjustment for job complexity and baseline cognitive ability has substantially influenced our results. Furthermore, we did not have full information on the number of years of employment before baseline, which could be a confounder and may have affected our results.
We adjusted for aspects of social relations in private life in our analyses, but not for other aspects of social relations at work than prolonged or serious conflicts. Both low social support at work (Andel et al.
2012) and a lack of contact with colleagues during working hours (Ishtiak-Ahmed et al.
2017) have been shown to be associated with higher risk of dementia. The buffering effect of social support on stressful life events in relation to health outcomes is well documented in the literature (Umberson et al.
2010). Also, individuals with social support at work may react differently to prolonged or serious conflicts at work compared with individuals without social support.
A validation study reported that the positive predictive value of the registered dementia diagnoses in the Danish national hospital registers was 86% (Phung et al.
2007). However, the hospital registers only capture about two-thirds of all cases in Denmark (Phung et al.
2010). This underreporting could have affected our results by diluting the association between conflicts and dementia in the present study. It is unlikely that missing information has influenced our results as only 1.1% had at least one variable with missing information among the covariates included for adjustments in the main analyses.
Another concern is that 10% of the 10,135 participants in the third survey of CCHS did not respond to the questions on conflicts at work. Non-responders have increased mortality and morbidity in most epidemiological studies (Rothman et al.
2008). If they had had more conflicts at work compared to responders, it can blunt our results. The percentages of dementia among the excluded and included participants were 16% and 11% respectively, and the incidence rates were 156 and 119 per 10,000 person-years, respectively. Furthermore, among the 837 participants (mean age: 70 years) who were excluded due to death within the first 5 years since baseline assessment, 4.2% of them reported experiencing conflicts at work compared to 8.5% of the included participants, indicating that the participants’ status of reporting conflicts may not have influenced these deaths. Thus, the exclusion based on mortality within five years from baseline are unlikely to affect our results.