The aim of the present study was to examine the role of managerial leadership in the association between psychosocial work stressors (high psychological demands, low decision authority and exposure to workplace violence) and sickness absence among health and social care workers. We found support for the hypothesis that poor perceived leadership of the closest manager was associated with a higher risk of sickness absence over time, and that good managerial leadership buffered the effect of workplace violence on sickness absence. However, we could not find any support for the hypothesis that psychosocial work stressors mediate the association between poor managerial leadership and register-based sickness absence.
The finding that poor leadership of the closest manager was associated with higher risk for sickness absence two years later supports our first hypothesis and is in line with earlier studies (17). The finding of the present study strengthens the evidence for an association between leadership and sickness absence by the fact that the reversed association—from sickness absence to managerial leadership—was controlled for, but showed no significant association, and that we used a register-based measure of sickness absence.
However, our second hypothesis was not confirmed. The association between poor perceived leadership of the closest superior and sickness absence was not mediated by high psychological demands, low decision authority or exposure to workplace violence. More specifically, even though there was a statistically significant association between low decision authority and sickness absence over time, managerial leadership was not associated with decision authority. Interestingly, we found a statistically significant association in the reverse direction from decision authority to managerial leadership, such that lower decision authority was related to poorer managerial leadership two years later. This may indicate that the employees’ perceptions of their managers’ leadership skills may be affected by the levels of decision authority they have been given regarding their work tasks over time.
With regards to psychological demands and workplace violence, no significant associations were found with either managerial leadership or sickness absence over time. These findings are not in line with two cross-sectional studies on nurses that did find significant associations between the closest managers’ leadership and decision authority (23, 38) and psychological demands (23) (to be noted Malloy and Penprase (38) did not support the latter association). One plausible reason for the inconsistency may be that these earlier studies were cross-sectional whereas the present study had a more robust research design, using a prospective approach with variables being measured two years apart, controlling for cross-sectional correlations and reversed longitudinal associations (i.e., from psychosocial work stressors to managerial leadership). Perhaps any causal effect of managerial leadership on psychosocial work stressors plays out rather contemporarily (our results also support cross-sectional relationships); hence, a two-year time span may be too long, and might fail to capture the mechanisms in question. In line with such a claim is, for instance, a study by Nielsen, Randall (24) where a mediating mechanism of work characteristics could be found when measuring leadership behaviour and work characteristics contemporarily, but not with measures 18 months apart. Thus, there is a need for more research scrutinising the mechanism utilising different and shorter time intervals.
Concerning managerial leadership as a potential moderator in the associations between psychosocial work stressors and sickness absence, managerial leadership only seemed to matter in the association between workplace violence and sickness absence. Thus, partial support was found for our third hypothesis. More specifically, only for those reporting poor managerial leadership was there a small association between workplace violence and sickness absence over time, meaning that in cases of experiencing workplace violence, poor leadership may increase the risk for subsequent sickness absence. Put differently, good leadership may protect the worker from suboptimal health outcomes due to workplace violence. The effect was rather small, but as violence is more present in the health and social care sectors compared to other sectors (2, 13), it may be of practical significance. Earlier studies examining potential moderator effects of leadership on the associations between psychosocial work stressors and health outcomes are rare (26) and results are mixed. For instance one study—using nationally representative work environment studies from two Nordic countries—did not support a buffering effect of good leadership compared to poor leadership on the association between emotional job demands and antidepressant treatment (28), whereas another study found an interaction effect of job strain and supportive leadership on poor well-being ten years later (27). However, the direction of the effect found in the latter study was unexpected as those who reported low job strain together with a lack of supportive leadership had poorer well-being compared to those with high job strain. As has been acknowledged by scholars, more research is warranted on the possible stress buffering effects of good leadership (26).
In the present study we utilised an assessment of managerial leadership, which rather broadly measures rudimentary leadership behaviour. In other studies, lack of such leadership has been acknowledged as a psychosocial stressor which increases the risk of suboptimal health (19, 31). Nevertheless, there are a number of leadership theories and measurements in the leadership research field, focusing on different dimensions of leadership styles and behaviours. The present measurement scale was developed in the Swedish context (31), but has certain resemblances with several other scales (19). For example, a number of items pertain to task-oriented behaviours or to the contingent reward subscale of transactional leadership, which assesses to what extent the employee perceives that “the leader clarifies expectations and sets up constructive transactions for meeting these expectations” (39, p 26). To a smaller extent, the present measurement also resembles relation-oriented behaviours and transformational leadership, where the latter during the last decades have dominated the leadership research field (26). Finally, low values may pertain to more laissez-faire leadership behaviours (the absence of leadership). Many leadership scales have been found to correlate highly to each other. For instance, Zwingmann, Wegge (39) in a large multinational sample found that transformational leadership and contingent reward were highly correlated (about .90 on average), which indicates that these two often go hand in hand. Also, laisse-faire leadership showed strong negative correlations (-.76 and − .72) with both transformational leadership and contingent reward (39). This implies that leaders usually do not lead by only adopting one style but by combining several. Nevertheless, the importance of different leadership styles and their impact on certain job characteristics and on sickness absence may vary in different industries on the labour market. In future research, different leadership scales should therefore be tested with regard to particular work settings within the health and social care industry.
Strengths and limitations
The present study had several strengths; it was based on a sizeable longitudinal cohort study, and was largely representative of the Swedish working population. Four time points were utilised and all our models were cross-lagged panel data models (controlling for cross-sectional associations, autoregressive associations and for reversed associations). A major strength with this method is that it measures within-individual variance, hence many unmeasured confounding variables, such as personality, are taken into account. However, we do not know whether the choice of two years between the measurement points is optimal. Perhaps, some mechanisms are faster or take longer to evolve.
Sickness absence was register-based, in contrast to the predictor and mediator variables which were self-reported, thereby decreasing the risk for common-methods bias (40, 41). Another strength of the use of register data for measuring sickness absence is that it should be accurate as it is linked to benefits for the individual. On the other hand, we lack information on shorter absence (1–14 days), which means that in the comparison group (0 days) there may be individuals who have many sickness absence days (spread over one year) that are never registered. Thus, the reported associations between our studied factors and sickness absence are likely to represent underestimations of the true situation.