Background

Common mental disorder is a term incorporating depression, anxiety, adjustment disorders, and stress-related ill health, all of which have major consequences around the world. For the individual, CMD causes suffering, pose a risk of social isolation, and threaten the personal income. In medium- to high-income countries, depression has become the diagnosis with the highest societal burden due to disability, decreased workability, and years lost to premature death (https://osha.europa.eu/en/themes/psychosocial-risks-and-stress). Individuals on long-term sick leave due to CMD are also at a higher risk for other types of mortality such as cardiovascular disease and cancer, and depressed individuals have a higher risk for suicide (Bryngelson et al. 2013).

In Sweden, CMD have become the leading causes for sickness absence with benefits. In 2016, 38% of all new sickness benefit episodes were due to CMD (Social Insurance Agency of Sweden 2016). Further, sickness absence periods due to CMD are longer and recurrent compared to other diagnosis, and are also more prevalent among women (Swedish Social Insurance Agency 2014). In a report from the Organisation for Economic Co-operation and Development (OECD) from 2018, costs for CMD, summarizing costs for health care, sickness absence insurance and lost productivity, were estimated at 21 billion euro in Sweden in 2015 (OECD and EU 2018).

The risk of CMD is lower among individuals who are working compared to those who are not gainfully employed (Waddell 2006). However, some risk factors for CMD are work related, although different between professions and socioeconomic groups (Marmot 2017). Such risk factors include high demands, low control, poor social support, and an effort-reward imbalance (Theorell et al. 2015; Aronsson et al. 2017). Protective occupational factors, on the other hand, include experiencing fair treatment and having influence over one’s work. Moreover, 40% of employees in European workplaces feel that work-related stress is not handled well (https://osha.europa.eu/en/themes/psychosocial-risks-and-stress).

To date, many rehabilitative interventions for CMD, like cognitive behavioural therapy (CBT), physical activity, and relaxation have focused on symptoms (Nieuwenhuijsen et al. 2014). Further, such interventions are directed towards the worker, placing the responsibility of improving the health status on the individual. Less commonly, organizational change (Joyce et al. 2016) is used as interventions to improve workers’ mental health. Moreover, improvement in symptoms does not necessarily translate into return-to-work (RTW) (Ejeby et al. 2014). Adding a workplace-directed intervention to a clinical intervention seems to decrease sick leave and facilitate RTW (Nieuwenhuijsen et al. 2014). In a review from 2012, interventions successful also in RTW were work directed and based on problem-solving therapy (PST) (Arends et al. 2012). PST interventions include identifying the problem and possible solutions applied in the work environment, as well as developing strategies to address the identified problems.

The occupational health service (OHS) exists in most industrialized countries (Rantanen et al. 2017) and is, by definition, working to promote employee health. The OHS may act in the preventive field to ensure that ill health is prevented or minimized, as well as having a role in facilitating RTW through rehabilitation and work adaptations when ill health has occurred (Joosen et al. 2015). As the OHS is operating in the workplace setting, knowledge about the specific work situation is good, and investigations and interventions can be directed appropriately both on an individual, group, and organizational level. Due to the limited knowledge concerning effects on workability (for instance sickness absence, sickness presence, and RTW) from interventions given to employees with CMD, and that many interventions appear not to be work oriented, it would be important to synthesize intervention research where the OHS is involved (Verbeek 2007).

To the best of the authors’ knowledge, no systematic mapping of OHS interventions concerning CMD exists. Consequently, the aim of this systematic review was to synthesize the scientific research on OHS interventions targeting prevention or reduction of CMD at the workplace, using the outcomes of workability and symptoms of CMD.

Method

Eligibility criteria

Eligibility was assessed according to the studies’ population, intervention, control, and outcome (PICO system) as follows:

Population

Studies investigating employees (individuals, groups, or organizations) at risk for CMD (as identified by a risk tool) or diagnosed with CMD were included. Studies on employees at workplaces evaluating stress preventive interventions were also included. Studies investigating individuals with severe mental disorders (like schizophrenia) were excluded.

Intervention

Studies where the recruitment or the intervention was delivered by the OHS or OHS personnel were included. Despite diversity, we included all types of OHS if they were labelled as such. Any type of intervention to prevent or reduce the risk of CMD or consequences thereof on an individual or at the organizational level was included. Longitudinal studies with baseline and follow-up measurements were included (as described in the PICOs). Studies where it was not possible to clearly understand the intervention through reading were excluded.

Control

Individuals or groups who did not receive the target intervention.

Outcome

All types of outcomes concerning sickness absence, including RTW, and psychological health were included. In the following, we use the term “workability” as a summarizing term including sickness absence, RTW, and self-reported workability. This included workers who had the ability to work part time.

Eligible study designs were randomized controlled trials, quasi-experimental studies, and longitudinal studies with baseline and follow-up measurements. Systematic reviews were also included. Publications written in English, Danish, Norwegian, or Swedish were accepted.

Search strategy and information sources

A systematic search strategy was developed in collaboration with librarians and adapted to the following four electronic databases; Medline, PsycINFO, Web of Science, and Cochrane CENTRAL. The original search period was January 1990 to March 2014. A second search for papers published between April 2014 and May 2017 was added using the same search terms. Search terms were developed and refined in discussions between GB and the librarian. The final terms were “Occupational Health Services”, “Occupational Disease”, “Occupational Health”, “Occupational Medicine”, “Occupational Health Nursing”, “Occupational Health Physicians”, “Occupational Injuries”, “Return to Work”, “Workplace”, “Clinical Trial”, “Comparative Study”, “Evaluation Studies”, “Meta Analysis”, “Multicenter Study”, “Observational Study”, “Review”, “Systematic Reviews”, “Epidemiologic Studies”, “Intervention Studies”, “Mental Disorders”, “Stress”, “Burnout”, “Depression”, “Anxiety”, “Adjustment Disorders” and combinations thereof. Abstracts were downloaded onto an Endnote × 7 library. In a second step, the reference lists of the included studies were searched for additional studies, and articles included in reviews were also checked for eligibility.

Selection assessment and data extraction

Titles and abstracts of retrieved records were evaluated by two pairs of researchers, MV and AL for the first search in 2014, IA and EBB for the second search in 2017. In each round, the pair met and reviewed 100 abstracts together to calibrate their judgement, after which the remaining abstracts were divided between the reviewers. They each decided on inclusion and exclusion. If in doubt, the abstract was included.

Review of the full texts started in the same way by the pair meeting and reviewing 30 full texts together. In cases of disagreement, the articles were discussed until agreement was reached. Again, the list of articles was divided between the reviewers to individually decide on which to include and exclude. When in doubt, the study was discussed between the pair until agreement was reached.

Quality assessment

All reviewers independently completed a quality assessment of the included studies, using a review protocol developed by the Swedish Agency for Health Technology Assessment and Assessment of Social Services (SBU), (https://www.sbu.se/contentassets/997d90960dbf40bbb4a1121d32c8acda/bilaga_3_granskningsmallar.pdf) for systematic reviews based on AMSTAR (Shea et al. 2007). This checklist is in Swedish, which made communication between reviewers easy. The quality assessment focused on methodological limitations that may introduce bias and run the risk of jeopardizing the study results. In a second meeting, the pair’s assessments were compared and disagreements solved through discussion. The included studies were rated as low, medium, or high risk of bias. Disagreements were resolved by consensus with GB during the entire quality assessment process.

Data synthesis

Data concerning year of publication, country of origin, type of intervention and control, outcome measure, time of follow-up, and result (positive or negative) were summarized. All studies were categorized according to their outcomes; first according to their effect on workability and second according to their effect on CMD symptoms. Lastly, based on the quality assessment, the results from studies deemed as having low and medium risk of bias were collated in a narrative summary to assess if studies examining the same intervention reported similar effect on one of the outcome categories. Studies with a high risk of bias were not considered in the summary as results from such a study may be distorted (reflect effects from factors other than the intervention). If a minimum of two studies of medium or low risk of bias were evaluating the intervention, the scientific quality was deemed satisfactorily, and conclusions could be drawn. However, no heterogeneity between the included studies was a prerequisite for this judgement. If only single studies with medium risk of bias evaluated an intervention, no conclusions were made as the scientific quality was deemed uncertain.

The results across studies were simply synthesized according to (i) effect or (ii) no evidence of effect.

Results

The initial searches resulted in 1983 abstracts, and based on the first screening of abstracts, 1705 were excluded. Reviews of the remaining 278 full texts and an additional 5 found by searching the reference lists, led to the exclusion of another 239 articles, leaving 44 scientific articles, representing 33 original studies. The reason that the number of articles/publications was higher than the number of studies is that some studies were reported in more than one article. The search is summarized in Fig. 1.

Fig. 1
figure 1

The searches performed in 2014 and 2017, exclusions and final inclusions

Characteristics of the included studies

Seventeen studies were from the Netherlands, four were from Finland, three were from Sweden, three were from Japan, two were from the United Kingdom, and one study was conducted in each of the following countries: Denmark, Canada, Germany, and Korea.

In total, the studies included 11,504 study subjects, ranging between 24 and 3379 individuals. The quality assessment ranked 6 studies as low risk of bias, 16 as medium, and 11 as high risk of bias.

The majority of the studies (n = 18) reported rehabilitative interventions targeting individuals on sick leave due to CMD, 11 studies reported interventions directed towards individuals at risk of developing CMD, i.e. the intervention was based on a screening tool to identify workers at risk and directed towards this specific group, and a four studies were classified as preventive interventions directed towards individuals and organizations, without regarding individual risk status. Twenty-five of the studies were randomized control trials (RCTs), six were observational studies, one was a mixed method study (including an observational component) and one was a systematic review. Finally, the studies reported different outcome measures, usually measures of workability, mainly sickness absence and RTW, and symptoms (e.g. depression, anxiety, stress, insomnia). A summary description of the studies is presented in Appendix 1.

Rehabilitative interventions of individuals with CMD

A total of 18 studies reported rehabilitation interventions directed towards employees with CMD. In some of these studies, the employees were on sick leave for a CMD diagnosis. Six of the studies were classified as low risk of bias, eight as medium risk of bias, and four were assessed as having high risk of bias.

Cognitive behavioral (CBT) and problem-solving therapy (PST) interventions were used in six of these studies and in one meta-analysis. PST is based on CBT, and, therefore, we decided a priori to group these studies together. All of these studies included employees who were on sick leave due to CMD and also reported an outcome related to workability and are summarized in Table 1.

Table 1 Studies investigating cognitive behavioral (CBT) and problem-solving therapies (PST) provided by the occupational health services (OHS) as rehabilitative interventions for individuals on sick leave due to CMD

Based on studies with low or medium risk of bias, PST/CBT interventions were shown to be effective in decreasing days with sickness absence (Doki et al. 2015) as well as duration of sick leave (Klink et al. 2003) and recurrent sick leave (Arends et al. 2013). One study demonstrated a reduction of days to RTW (Volker et al. 2015), but these interventions were generally not found to improve full time RTW (Volker et al. 2015; Rebergen et al. 2009; Van der Klink et al. 2003; Vlasveld et al. 2012, 2013) after 12 months follow-up among individuals with CMD.

PST/CBT interventions did not significantly improve symptoms of CMD (Arends et al. 2013; Klink et al. 2003; Vlasveld et al. 2012, 2013) in most instances, only in one study (Vlasveld et al. 2012, 2013) where depression symptoms were reduced with > 50%).

Other studies examined interventions from the OHS physician (who were trained in guideline use or could consult a psychiatrist), a participatory intervention, multi-professional rehabilitation, a self-help manual coupled with PST, a workplace and pharmacological intervention, stress management, a gradual exposure to stress and occupational therapy. These studies are summarized in Table 2. Eight studies were assessed as low or medium risk of bias.

Table 2 Studies investigating different interventions [other than cognitive behavioral (CBT) and problem-solving therapies (PST)] for employees with CMD

Generally, these interventions did not significantly improve full time RTW (Van Beurden et al. 2015, 2017; Oostrom et al. 2009; Noordik et al. 2013; Hees et al. 2013; Feltz-Cornelis et al. 2010) after 3–18 months follow-up among individuals with CMD. The only exception was that educating the physician to work with the psychiatrist (Feltz-Cornelis et al. 2010) which showed a statistically significant effect on RTW after 3 but not after 6 months.

Depression symptoms were found to be significantly improved by occupational therapy (Hees et al. 2013) but not by any other investigated intervention (Oostrom et al. 2009, 2010; Noordik et al. 2013; Feltz-Cornelis et al. 2010). One study found increased self-efficacy beliefs in RTW after the OHS physician followed guideline recommendations (Beurden et al. 2015). Symptoms of stress were not improved by any of the investigated interventions (Oostrom et al. 2009; Noordik et al. 2013), and neither was anxiety (Oostrom et al. 2009; Noordik et al. 2013; Feltz-Cornelis et al. 2010), sense of coherence (Valtonen et al. 2015) or sleep problems and cognitive ability (Dalgaard et al. 2014).

Interventions directed towards employees at risk for CMD

In eleven studies, interventions were directed towards individuals labeled as “at risk” of future CMD based on results from screening questionnaires. Six of these were assessed as medium and 5 as high risk of bias.

Again, CBT was used in some of the interventions. These studies are summarized in Table 3.

Table 3 Studies investigating CBT directed towards individuals with a risk of CMD

Regarding workability; in studies of low or medium risk of bias where depressive symptoms improved, sickness absence improved after 12 months, but not after 18 months Valtonen et al. (2015).

Conclusions based on studies with low or medium risk of bias showed one study with positive effects on depression (Lexis et al. 2011), but the other failed to show statistically significant effects on emotional stress and insomnia (Yamamoto et al. 2016).

The remaining eight studies examined interventions including consulting a physician, advice, and personal feedback. These studies are presented in Table 4. Six studies were assessed as having medium risk of bias, and three as having a high risk of bias.

Table 4 Studies investigating physician consultation, stress handling with mindfulness, advice, and personal feedback directed towards individuals with a risk of CMD and psychoanalytical therapy

Sickness absence and presentism were reduced in one study (Noben et al. 2015), the risk of early retirement was reduced in another (Boer et al. 2004) and a reduction of total sickness absence was found in one study (Kant et al. 2008).

Five of the studies had medium risk of bias and investigated the effect on CMD symptoms but failed to show any improvements in these outcomes. However, some health benefits were seen in two studies (Karatzias et al. 2011; Peterson et al. 2008).

Preventive interventions

Four studies investigated preventive interventions directed towards individuals and organizations to prevent CMD. One was assessed as having medium risk of bias and three as having high risk of bias. These studies are summarized in Table 5

Table 5 Studies investigating interventions aimed at preventing CMD

In the study with a medium risk of bias (Vuori et al. 2012), a course in career development designed to prevent CMD, showed no effects on either wanting to retire, on symptoms of CMD, or on mental resources after controlling for depression at study start.

Discussion

A systematic review with a narrative synthesis was conducted in the specific setting of occupational health service, as this is a potentially important context for work-directed interventions to prevent or reduce symptoms or to improve workability among employees with CMD. Interventions based on PST and work-focused CBT were the most frequently studied interventions, used in rehabilitation and interventions directed at employees at risk.

The studies were rather heterogeneous concerning which interventions were used and which outcomes were measured, as well as the intent of the intervention; prevention, at risk or rehabilitation. The majority of the studies concerned rehabilitation of individuals who were diagnosed with CMD (Volker et al. 2015; Arends et al. 2013; Rebergen et al. 2009; Klink et al. 2003; Doki et al. 2015; Vlasveld et al. 2012,2013; Kroger et al. 2015; Beurden et al. 2015,2017; Oostrom et al. 2009,2010; Valtonen et al. 2015; Goorden et al. 2014; Dalgaard et al. 2014; Noordik et al. 2013; Hees et al. 2013; Feltz-Cornelis et al. 2010; Bender et al. 2016; Grossi and Santell 2009; Vente et al. 2008). In terms of workability, there were some evidence of improvement in time to first return-to-work among employees on sick leave for CMD receiving work-oriented PST and CBT, in comparison with treatment-as-usual. However, these interventions failed to improve full time RTW. Generally, symptoms decreased both among employees receiving interventions or treatment as usual, but usually there were no statistically significant differences between interventions and control in this respect.

An alternative or complementary strategy to rehabilitation can be to address the problem before it gets to the stage where large numbers of individuals suffer from CMD, i.e. to implement effective interventions directed towards people at risk of developing such problems. Using validated screening tools to identify these individuals, early interventions could be put in place. Eleven studies were directed towards employees at risk of developing CMD (Lexis et al. 2011; Grime 2004; Noben et al. 2014,2015; Kilfedder et al. 2010; Karatzias et al. 2011; Peterson et al. 2008; Kant et al. 2008; Boer et al. 2004; Kuoppala and Kekoni 2013; Gartner et al. 2013; Ketelaar et al. 2013; Salmela-Aro et al. 2004). The interventions included a great variety of measures, but the evidence is not sufficient to recommend a specific type or timing of interventions. We conclude that more research is needed.

Studies investigating preventive interventions were few (Vuori et al. 2012; Ahola et al. 2012; Vinberg et al. 2015; Kim et al. 2014; Kobayashi et al. 2008) and generally carried a high risk of bias. It would have been interesting to examine the effect of systematic work-environment measures aiming to create workplaces where people would get support, have control, and adequate time for recuperation. As only one study with sufficient quality was included in our review (Ahola et al. 2012), conclusions regarding effect of preventive interventions remain uncertain.

It is quite striking that most interventions do not affect both symptoms and workability. In some studies, only outcomes related to one of these were measured, but in the studies, both were reported (Volker et al. 2015; Arends et al. 2013; Klink et al. 2003; Vlasveld et al. 2012; Oostrom et al. 2009; Goorden et al. 2014; Noordik et al. 2013; Hees et al. 2013; Feltz-Cornelis et al. 2010; Lexis et al. 2011; Boer et al. 2004; Vuori et al. 2012), they would typically not follow each other. Thus, improving symptoms or diminishing the risk of CMD does not automatically improve workability (Vlasveld et al. 2012; Hees et al. 2013) and improvement in workability does not require an improvement in the patients’ CMD status greater than that of the control group (Arends et al. 2013; Klink et al. 2003). This finding may seem surprising, it is often inferred that when people are well, they will be able to work optimally. From this review we may conclude that for CMD, different mechanisms are determining symptoms and health than are determining workability. It seems that work-oriented CBT/PST interventions help individuals manage their work in such a way that they may be able to keep working despite still having CMD/ risk of CMD.

It is relevant to scrutinize the types of interventions offered in these studies. Commonly, they are targeting the individual, typically helping the individual to identify and manage stressors, largely ignoring any necessary action in the workplace. Even the interventions described as “workplace interventions” typically entails a meeting between the worker, the employer and a health care provider and targets barriers to RTW, i.e. individual adaptation (Vlasveld et al. 2012; Goorden et al. 2014; Dalgaard et al. 2014). Evidence is mounting that the psychosocial work environment is vital for the mental well-being of workers, yet many interventions fail to address this issue (Theorell et al. 2015; Aronsson et al. 2017).

There is clearly a need for evaluating theory-based preventive strategies with scientifically robust methods, such as the vital use of groups for comparison, to understand the effectiveness of such interventions. However, complex organizational interventions may also apply other methodological approaches to investigate their outcome, e.g. by a more detailed study of process mechanisms (Nielsen 2017). Only a few studies suggest that prevention among employees at risk is effective, but the interventions are diverse, so replication of these results is necessary. It may also be beneficial to combine these individually oriented interventions for employees at risk with more comprehensive organizational approaches (Kwak et al. 2019).

The review was done systematically in two rounds, in 2014 and again in 2017. Between the two searches, only 12 eligible studies were published. This alone is signifying the scarcity of evidence in the field but is also highlighting the fact that OHS is not an indexed term, making hand searches of reference lists necessary. Even though we used an experienced librarian and embarked on a systematic approach, searching the major databases, relevant studies may have been missed. A limitation of our review is that the first screening of abstracts in each round was done by one researcher only, although a calibration between the pair was undertaken and discussions ensued in cases of uncertainty. It is possible that the eligibility criteria were applied differently between researchers, possibly missing relevant articles. Further, conclusions from the grey literature were deliberately not included in our review, owing to the scientific weight of such evidence.

We purposefully only included studies where the OHS was involved in the intervention. This means that other important workplace interventions that were not labelled as such, or that were delivered by other parties, were not included in the review. The included studies also reflect the difference in OHS around the globe: nearly half of the eligible studies were from the Netherlands (Volker et al. 2015; Arends et al. 2013; Rebergen et al. 2009; Klink et al. 2003; Vlasveld et al. 2012,2013; Hees et al. 2013; Feltz-Cornelis et al. 2010; Boer et al. 2004), where the OHS is an important actor in worker health, supported by a legal framework. Only four included studies were from Asia (including a systematic review with a majority of articles from the Netherlands (Doki et al. 2015), and none were from the US, South America or Africa. This fact limits the generalizability of our findings outside the European and Asian setting. However, if similar services are offered in primary care in areas where OHS are not common, the results may be transferrable to these settings.

To conclude, there is some evidence that PST interventions or work-related CBT given at the OHS can improve workability among employees on sick leave due to CMDs but the evidence for these interventions concerning symptom reduction is uncertain.