Background
Long-term sick leave is costly for society and has detrimental impacts on the individual [
1]. In order to reduce sickness absence rates several countries have implemented reforms promoting work-related activity for the long-term sick listed [
2]. One frequently used work activity is part-time sick leave (PTSL) where individuals work a percentage corresponding to their remaining work capacity [
3]. Continuing to work with sickness or disabilities (e.g., through PTSL) is believed to be beneficial for the worker as work meets important psychosocial needs and includes therapeutic elements that can be beneficial for mental health and well-being [
1,
4,
5]. Furthermore, returning to the workplace before full recovery is thought to be important for the return to work (RTW) process [
6,
7].
Several studies find that PTSL has favorable outcomes for RTW [
8‐
12], but others question its overall effectiveness [
13‐
15]. Reaching firm conclusions is difficult due to the selection effect related to the use of PTSL [
16]. That is, workers who are able to utilize PTSL have different personal or workplace characteristics to workers on full-time sick leave (FTSL), and therefore different probabilities of successful RTW. For example, studies have shown that those on PTSL and FTSL differ with regards to age, gender, education, diagnosis, manual or office work, and being in the private or public sector [
17‐
21]. Selection effects make direct comparisons between the groups difficult. There is also uncertainty around efforts to control for such confounding, for example by using propensity score matching [
22]. Using a randomized controlled trial design would eliminate bias caused by selection effects but is rarely feasible when examining PTSL as these arrangements are usually national schemes and thus eligible for everyone [
23]. In addition, randomizing individuals to more or less sick leave than they are capable of handling is ethically problematic. Due to these issues most evaluations of PTSL use observational register-based samples. One exception is a study by Viikari-Juntura et al. [
10], in which individuals with musculoskeletal disorders were randomized to PTSL or FTSL and the PTSL group showed improved RTW outcomes. In another study, Rehwald et al. [
12] included PTSL as one potential component of an intervention for newly sick-listed individuals. However, even though participation in the intervention group was randomized, the use of PTSL was decided by the administering job center, adapted for individual needs and local conditions. Thus, selection for PTSL could still play a role, for instance through work characteristics.
In previous register-based studies, most of the variability in PTSL selection seems to be explained by unobserved factors [
24]. Some of these unobserved factors have been proposed to include the health of the worker, and work-related and individual characteristics which may influence the propensity to use PTSL [
16,
19]. For instance, poorer health may prohibit work participation altogether [
25], and is the most commonly stated reason for not working after 8 weeks of sick leave in Norway [
26]. In a review of PTSL use, Kausto et al. [
3] also found that practical problems at work such as low flexibility, lack of control over work arrangements and poor collaboration were the main barriers for PTSL. Factors such as self-reported health, workplace adjustments, work autonomy and social support at work are also prognostic factors for RTW and hence influence sick leave duration [
25,
27‐
30]. Furthermore, work demands have also been shown to be a prognostic factor for RTW [
28,
31]. Balancing work demands and work hours has previously been reported to be a challenge when facilitating PTSL [
21,
32]. The above factors could impact both the propensity to use PTSL and the likelihood of RTW and confound the impact of PTSL on RTW.
Psychological resilience is a personal characteristic that may be relevant when examining PTSL selection. Psychological resilience consists of individual personal and social resources that help individuals adapt and function in the context of adversity or stress [
33,
34]. The concept has recently gained some attention in the RTW context but more research is needed [
35]. There are indications that resilience increases the functioning of those struggling with pain [
36,
37], and those with chronic disease [
38]. Thus, resilience could potentially also influence the propensity to remain at work despite ill health.
Differences between those on PTSL and FTSL with regards to the characteristics above are difficult to capture using registry data. Some of these factors have been examined in previous studies by self-report or using proxies, but the data are scarce and inconsistent. For instance, better health in those on PTSL has been proposed [
9], while other studies report poorer health, more previous sick leave, or chronicity [
3,
22,
39]. Some studies also report no differences regarding health [
17,
20]. Better psychosocial work environments and less conflict at work have also been found for those on PTSL [
15,
39]. However, knowledge regarding workplace adjustment latitude, work autonomy, the capacity to cope with work demands and psychological resilience is lacking.
In order to know more about the potential benefits of PTSL systems it is important to know what characterizes those who use the arrangements compared to those who do not [
3,
23]. Data on how health-, work-related, and personal characteristics influence PTSL selection are scarce and inconsistent. Further knowledge of how such factors are associated with PTSL could inform stakeholders such as general practitioners (GPs), employers, RTW coordinators, and social insurance services in developing solutions for work activation. Furthermore, unobserved confounding in evaluations of PTSL could contribute to incorrect conclusions which influence recommendations and policy. More knowledge of factors associated with PTSL could thus improve work activation strategies and the accuracy of and confidence in PTSL evaluations.
The objectives of the present study were to:
(1)
Explore whether a sample of long-term sick listed individuals on PTSL and FTSL differ in terms of previously known selection factors of PTSL (age, gender, education, private or public sector, diagnosis, and physical work demands).
(2)
Explore differences between the PTSL and FTSL groups in health, workplace, and personal characteristics that could influence the propensity for PTSL (current self-reported health, previous long-term sick leave, workplace adjustment latitude, psychosocial work environment, work autonomy, coping with work demands, and psychological resilience).
(3)
Examine whether identified differences in objectives (1) and (2) persisted independently of the known selection factors described in objective (1).
Discussion
This cross-sectional study of workers sick listed for 8 weeks aimed to describe differences between those part-time sick-listed and full-time sick-listed in terms of health-, workplace-related, and personal characteristics. We also wanted to investigate whether the differences between these groups persisted after taking into account known differences from previous research.
In line with previous research the results show that individuals on PTSL were more often women, had higher education and less physically demanding work, and more often worked in the public sector [
17,
19‐
21]. We also found less PTSL among the oldest and youngest workers, which is comparable to the findings of Ose et al. [
21]. The age-PTSL curve peaked at 42 years, meaning that those in the middle of their working life were the most likely to be part-time sick listed. The present study also found more individuals sick listed due to a psychological diagnosis in the PTSL group than the FTSL group, which has also been identified in Norwegian population data [
19]. After adjusting for the other covariates, however, the only significant associations on previously known selection factors were gender and physical work demands. The differences between the groups in terms of the above sociodemographic factors are usually captured in evaluations using register data. However, extra attention may need to be paid to physical work demands as information on this may not be directly available in register data and it is also a prognostic factor for RTW [
31]. Furthermore, even though more individuals with a psychological diagnosis were on PTSL in the present and previous studies, it has been proposed that the potential benefits of PTSL for RTW are less convincing for workers with mental health disorders [
12,
15,
51]. However, PTSL could be effective in combination with work-focused cognitive behavioral therapy [
52].
We found no differences between the groups in terms of previous long-term sick leave. This supports results from Finland [
17,
20], but contrasted studies which found associations with less PTSL [
9] or more PTSL [
22,
53]. Regarding current self-reported health we identified an inverse U-shaped curvilinear association with PTSL which largely did not change after adjustment for previously known selection factors. This association has been previously suggested due to the potential costs to the employer when facilitating PTSL [
8]. Costs associated with facilitating PTSL could contribute to the curvilinear association as individuals with the best health may be close to RTW and skip PTSL altogether, while those with the poorest health may be too ill to work at all [
8]. The association in the present study largely did not change after adjusting for known differences, indicating a robust curvilinear association.
The present study also found that workplace adjustment latitude and psychosocial work environment were associated with more PTSL. This could be related to more participants with higher education and sedentary work among those on PTSL, which may allow for more flexibility at work. However, adjusting for these covariates largely did not influence the associations. Workplace adjustments and psychosocial work environment also influence sick leave duration [
27,
29,
30], and could confound the suggested effect of PTSL on RTW. In the present study those on PTSL also report more work autonomy and poorer coping with work demands. In previous research those on PTSL reported that the reduction in work hours was not accompanied by a corresponding reduction in productivity expectations [
32], which could contribute to higher work demands for those on PTSL. However, the differences in work autonomy and coping with work demands were less convincing after adjustment which indicates that other characteristics captured by the covariates influence the relationship with PTSL (e.g., education). Finally, we found no differences between the groups with regard to psychological resilience. Resilience as measured here is a collection of psychological resources primarily relevant when faced with psychosocial adversity and may not be applicable to the diverse diagnostic sample in the current study.
Overall, there was a clear tendency that workers on PTSL had more flexible workplaces. Previous research has found that workers returning from sick leave need to have flexibility in order to successfully maintain health while meeting expected productivity demands [
54]. Vooijs et al. [
55] also argued that the most effective interventions to improve work participation were those focusing on changes at work rather than changing the individual’s abilities to meet work demands. However, previous research on RTW follow-up in Norway has suggested that facilitating PTSL through workplace adjustments could entail additional costs for the employer compared to hiring a substitute worker [
26]. In Norway, all wages are replaced for someone sick-listed and this could contribute to a lack of incentive for the employer to facilitate adjustments [
56].
Several previous studies have attempted to reduce the impact of selection when estimating the effect of PTSL on work outcomes. Markussen et al. [
8] used GP propensity to certify PTSL as an instrumental variable and found that GP’s with higher propensities to certify PTSL contributed to faster RTW. Similarly, Kools and Koning [
15] used Dutch insurance caseworkers’ propensity to assign PTSL as the instrumental variable, with more uncertain efficacy in terms of RTW. However, these professionals’ propensities to assign PTSL are likely also associated with their propensities to assign full RTW and possibly also their skill in facilitating RTW [
15,
23]. Furthermore, PTSL possibilities also need to be negotiated with the sick-listed’s employer [
15]. In this vein, Andren and Svensson [
57] found an effect on RTW by using the type of occupation as an instrumental variable. This study assumed that the job type (e.g., clerks, service and sales, managers, crafts and trades) has an impact on the possibility of PTSL, but not on the possibility of full recovery from sick leave in workers with musculoskeletal illness [
57]. However, it is likely that the different job types differ in work demands, which cast doubt on the assumption as work demands also influence RTW [
30,
31,
58].
Others have used propensity score matching between those on PTSL and FTSL to account for selection effects, and have largely found positive impacts of PTSL on work-related outcomes [
18,
20,
53]. Kausto et al. [
17] also adjusted for potential confounding in multivariate analyses and found that PTSL led to reduced future disability. However, the models in these studies are based on register data and did not include variables identified in the present study, such as work flexibility or self-reported health. Bosman et al. [
14], however, adjusted for such potential confounders. They included individual characteristics (age, gender and educational level), health-related factors (previous sick leave and diagnosis), and also work demands, work pace, and social support. They concluded that when adjustments were made, PTSL did not influence sick leave duration in their sample of workers with musculoskeletal disorders [
14].
The present study adds to the above literature by suggesting a general tendency towards selection where sick-listed workers with flexible workplaces more frequently were on PTSL. This could mean that previous studies that adjust for differences between PTSL and FTSL using common registry data variables may not capture all important workplace characteristics that influence PTSL use. Thus, future evaluations of PTSL should consider including more detail on self-reported health and workplace characteristics to account for differences between those on PTSL and FTSL.
Strengths and limitations
One strength of the current study is the extent of covariates which enabled us to investigate the associations between previously unexplored factors and the propensity to utilize PTSL. Registry data on the outcome variable can also be considered a strength as it helps avoid biases caused by self-report.
A limitation of the present study is the low recruitment rate of participants which could indicate participant selection. However, the present study largely found the same differences between PTSL and FTSL on the sociodemographic covariates as previous register-based studies which indicates that the current sample might be comparable to representative population studies. Another limitation in the present study is the use of single-item variables to examine workplace characteristics. Single-item variables make it difficult to determine exactly how these characteristics were different between the groups and may also lack construct validity. Future studies should use a broader set of validated questionnaires to investigate different aspects of work that could facilitate PTSL. The present study also lack data on the sick listing GP’s propensity to certify PTSL. Previous research has shown that GPs has some influence in determining sick listing percentage [
59] and could thus affect PTSL selection. Finally, the present study is cross-sectional, which limits any conclusions regarding causality. For instance, the mechanisms behind poorer coping with work demands are difficult to gauge in the present study as high demands is also a risk factor for sick leave [
60], and workers with manageable work demands may not be sick listed at all. Furthermore, mediation rather than confounding could explain the associations between the variables and PTSL. Further longitudinal research should thus investigate the causal pathways and potential mediation between work characteristics, covariates, PTSL, and RTW.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.