Introduction
Despite overall work injury rates declining in most high-income countries [
1,
2], equivalent improvements in return-to-work (RTW) rates (i.e. percentage returning to work within certain disability duration windows) have not been observed. In Australia and New Zealand, the latest data indicate RTW rates have remained static for 15 years [
3]. Canadian-wide statistics comparing the percentage of wage loss claims at specific durations (e.g., 30 or 180 days after injury) indicate that disability duration has remained constant or increased between 2000 and 2008 [
4]. Societal changes are making improvements in RTW more difficult to achieve. The ageing workforce poses particular challenges given findings that older workers take longer to RTW than younger workers and are more likely to ‘relapse’ into a period away from work following an initial return to work [
5]. Similarly, there is a growing trend in precarious employment relationships (e.g., workers with short-term contract arrangements). Workers with precarious job arrangements also take longer to RTW than those with secure employment relationships [
6].
There is now a substantial research literature on RTW interventions delivered in the workplace. This diverse literature contains relatively few high quality intervention studies. One systematic review of workplace based interventions published in 2004, for workers with musculoskeletal (MSK)- and pain-related conditions, identified ten good quality intervention studies after completing a search that retrieved 35 relevant studies [
7]. The review found strong evidence that time away from work (work disability duration) is reduced by work accommodation offers and contact between healthcare providers and the workplace, and moderate evidence that other disability management interventions were effective. There was limited or mixed evidence of the impact of these interventions on health related quality of life.
The complex nature of interventions in this field poses a direct challenge for researchers. Conducting high-quality work disability research, and in particular, evaluating return-to-work interventions which have many socio-legal aspects and often requires the endorsement and cooperation of stakeholders with competing interests (e.g., employers, insurers, labour unions, provider networks, compensation authorities, etc) is difficult [
8]. Still, in the decade since the review’s publication, and other studies by the same research team [
9], there has been steady growth in the volume and scope of RTW intervention studies published. RTW or work disability research has emerged as a stand-alone field of endeavour encompassing multiple disciplines, with a rapidly growing evidence base [
10].
This is true for both MSK and pain-related conditions; and more recently mental health (MH) conditions. The growth in literature focused on interventions to manage depression in the workplace has grown substantially over the last 5 years. In 2010, several authors from this research team published a systematic review [
11] on interventions to manage depression in the workplace, finding 12 high quality studies. Recently, this team has sought to update findings on this question and have found the body of relevant literature to have more than doubled in the last 5 years (unpublished data).
Consistent with the best practice of updating systematic reviews as new evidence emerges [
12], we sought to update and extend the previous review of workplace based RTW interventions that was limited to MSK and pain-related conditions. The primary objective of this review was to synthesize evidence on the effectiveness of workplace-based RTW interventions that assist workers with MSK, mental health (MH), and pain-related conditions to return to work after a period of work absence. The focus of this update was expanded to include MH conditions, based largely on input from our occupational health and safety (OHS) stakeholders given that the burden associated with managing the effects of mental health conditions in the workplace is extensive [
13‐
16]. A particular strength of the Institute for Work & Health (IWH) systematic review program is the unique process of stakeholder engagement adopted throughout the review process [
17]. Our stakeholders provide guidance to ensure the review question is relevant, the search terms are comprehensive and the targeted literature identified is up-to-date. But more importantly, stakeholders helped us examine the findings from this review to determine the best wording for our key messages to facilitate uptake and dissemination of these evidence-based approaches for OHS practitioners and other workplace parties This paper focuses on the evidence on RTW outcomes. A future paper will address the evidence from this review on recovery outcomes.
Methods
The systematic review followed the six review steps developed by the Institute for Work & Health (IWH) for OHS prevention reviews [
18]: (1) question development, (2) literature search, (3) relevance screen, (4) quality appraisal, (5) data extraction, and (6) evidence synthesis. The review team consisted of 17 researchers from Australia, Canada, Europe and the United States. Reviewers were identified based on their expertise in conducting epidemiologic or intervention studies related to work-related conditions, their experience in conducting systematic reviews or their clinical expertise. Review team members had backgrounds in epidemiology, ergonomics, kinesiology, physical therapy, psychology, social sciences, and information science. All 17 team members participated in all review steps.
The IWH Systematic Review program follows an integrated stakeholder engagement model during reviews [
17]. Stakeholder meetings were held on multiple occasions through the review process in Toronto, Canada and Melbourne, Australia. Stakeholders were selected from injured worker advocacy groups, unions, workplaces, and health and safety associations and provided valuable input on search terms, inclusion/exclusion criteria, operational definitions, terminology, other search considerations, how findings of the review might be used, potential audiences, how the finalized review could be presented, how the review findings could be disseminated, and stakeholder information and communication needs throughout the review process.
Question Development
The review team and stakeholders participated in a meeting to discuss the review update research question, and proposed search terms. The review question and search terms from the original review were used as a starting point and were updated through this process of question development. The inclusion of MH conditions to the final research question was an addition driven largely in response to stakeholder feedback through this process.
Literature Search
Search terms were developed iteratively by the research team in consultation with a librarian, content area experts and stakeholders. Search terms were identified for three broad areas; population terms for workers and for injury/conditions, intervention terms, and outcome terms. Both database-specific controlled vocabulary terms and keywords were included. The terms within each category were combined using a Boolean OR operator and then terms across the three main categories were combined using a Boolean AND operator. The complete list of terms used in our search is reported in Supplementary Table 1.
The following electronic databases were searched; Medline, EMBASE, CINAHL, PsycINFO, Sociological Abstracts, Applied Social Sciences Index and Abstracts (ASSIA), and ABI Inform (American Business Index) from 1990 to April 2015. Research prior to 1990 was considered informative from a historical perspective but less relevant to current personal injury-illness compensation and other health care system and therefore excluded from this review. As the controlled vocabulary and the ability to handle complicated multi-term searches differ across the databases searched, search terms were customized for each database as required. All peer-reviewed literature was included, including non-English citations.
In addition to the database searches, the review team identified, from their own holdings and via contact with international content area experts, a list of studies that were in press or otherwise forthcoming in the published peer review literature.
References were loaded into commercially available review software (DistillerSR
®) [
19], which was also used for all remaining review steps. DistillerSR
® is an online application designed specifically for the screening, quality appraisal and data extraction phases of a systematic review.
Relevance Screen
The review team devised five screening criteria to exclude articles not relevant to our review question: (a) commentary/editorial, (b) study was not about RTW or disability management/support, (c) non-intervention studies or interventions that did not occur as part of a system, program, policy or work practice change, (d) interventions that were not workplace-based, and (e) study population included greater than 50% of any of the following excluded conditions: severe traumatic brain injury, spinal cord injury, severe lower limb traumatic injuries including amputations; MSK disorders secondary to cancer, cancer-related pain or osteoporosis; and severe mental disorders (i.e. bipolar disorder, chronic severe depression or schizophrenia).
First, titles and abstracts of references were screened by a single reviewer. To limit the possibility of bias, a quality control (QC) step was implemented. A QC reviewer independently assessed a randomly chosen set of 329 titles and abstracts (approximately 5% of references from the search). Comparing the QC reviewer responses directly to review team responses, 27 conflicts (8%) (i.e. where the QC reviewer disagreed with the assessment of the original reviewer) were found. However, only four (1.2%) were conflicts in which the review team excluded references and the QC reviewer included them. The small (1.2%) number of consequential discrepancies suggests that reviewers had a similar understanding and application of the screening criteria.
Second, the full text of articles that advanced through the title and abstract screening process were screened using the same criteria, with two reviewers independently reviewing and coming to consensus. When consensus could not be reached, a third reviewer was consulted.
Quality Appraisal
Relevant articles were appraised for methodological quality. The team grouped multiple articles associated with a single study, designating one article as the primary article. Study quality was assessed using 25 methodological criteria within the following broad headings: Design and Objectives, Level of Recruitment, Intervention Characteristics, Intervention Intensity, Outcomes, and Analysis (see Supplementary Table 2).
Methodological quality scores for each study were based on a weighted sum score of the quality criteria (with a maximum score of 96). The weighting values assigned to the 25 criteria ranged from ‘‘somewhat important’’ (1) to ‘‘very important’’ (3). Each study received a quality ranking score by dividing the weighted score by 96 and then multiplying by 100. The quality ranking was used to group studies into three categories: high (>85%), medium (50–85%) and low (<50%) quality [
20].
Each study was independently assessed by two reviewers, who were required to reach consensus. Where consensus could not be achieved, a third reviewer was consulted. Team members did not review articles they had consulted on, authored or co-authored.
The quality appraisal represents an assessment on: internal validity, external validity, and statistical validity [
21]. A higher quality score increases the team’s confidence that an effect was an intervention consequence rather than the effect(s) of other workplace or external environment factors. Therefore, data extraction and evidence synthesis were only completed on high and medium quality studies.
Standardized forms based upon previous reviews were used for data extraction [
7,
11]. Extracted data were used to create summary tables sorted by intervention category and used for evidence synthesis. Data were extracted independently by pairs of reviewers. As in the relevance and quality appraisal stages, reviewer pairs were rotated to reduce bias. Team members did not review articles they consulted on, authored or coauthored. Any conflicts between reviewers were resolved by discussion. Stakeholders were consulted to determine relevant workplace-based RTW intervention categories.
Evidence Synthesis
The evidence synthesis approach [
18,
22] considers the quality, quantity and consistency in the body of evidence (see Table
1). First, the intervention categories created in the data summary tables were examined by the entire team. Once consensus was reached on the categories, the team moved to summarizing the evidence for each intervention category. Due to the heterogeneity of outcome measures, study designs and reported data, we chose not to calculate a pooled effect estimate. To determine individual study intervention effects, the following rules were applied: an intervention with a positive and no negative results was classified as a positive effect, an intervention with both positive and no effects was also classified as a positive effect intervention, an intervention with only no effects was classified as no effect, an intervention with any negative effect was classified as negative effect. Intervention effects were combined with the quality rating and number of studies to determine the level of evidence for each intervention category.
Table 1
Best evidence synthesis algorithm/algorithm for messages
Strong | High (H) | 3 | 3H agree; if 3+ studies, ≥3/4 of the M and H agree | Recommendations |
Moderate | Medium (M) | 2H or 2H and 1M | 2H agree or 2M and 1H agree; if 3+, ≥2/3 of the M and H agree | Practice considerations |
Limited | 1H or 2M or 1M and 1H | 2 (M and/or H) agree; if 2+, >1/2 of the M and H agree | Not enough evidence to make recommendations or practice considerations |
Mixed | 2 | Findings are contradictory |
Insufficient | Medium quality studies that do not meet the above criteria |
To generate practical messages, an algorithm developed by IWH along with OHS stakeholders was followed [
23]. A strong level of evidence leads to “recommendations”. A moderate level of evidence leads to “practice considerations”. For all evidence levels below moderate, the consistent message is: “Not enough evidence from the scientific literature to guide current policies/practices”. This does not mean that the interventions with limited, mixed, or insufficient evidence may not be effective; only that there is not enough scientific evidence to draw conclusions.
Discussion
The current review and evidence update gathers and synthesizes the scientific literature and presents practical messages for workplace parties and occupational health and safety practitioners. The review team consulted with these stakeholders to help ensure the messages were useful and applicable in practice.
The review identified 36 medium and high quality intervention studies that examined workplace-based RTW and disability management/support initiatives. The primary finding is strong evidence that multiple domain interventions are effective in improving RTW outcomes in workers with MSK, pain-related or MH conditions. In contrast, most single domain focused interventions have mixed or limited evidence to support their effectiveness. This result is aligned with one of the dominant theoretical paradigms in the work disability and return to work literature, the Sherbrooke model [
93]. This model proposes that multi-disciplinary and multi-factorial interventions that seek to address an array of individual and societal factors that influence RTW are likely to be effective.
Combining newer studies with those from the original review [
7] resulted in stronger evidence levels across a greater number of intervention categories. In addition, we were able to synthesize new evidence on intervention strategies to manage MH conditions in the workplace, which has emerged as an important area of concern for employers since the original review was published.
Our review identified that in most cases interventions were multi-faceted and included multiple intervention components, often operating across multiple domains (health focus, service coordination and work modification). This approach is different to the previous review [
7], which sought to evaluate the effectiveness of discrete intervention components; leading to a different interpretation of the literature.
For example, the original review, Franche et al. [
7], found a strong level of evidence for a positive effect of work accommodations, while in the current update only a moderate level of evidence was found. Of note, one of the interventions included in the original review examining work accommodation offers was reclassified in this review as a multi-domain intervention (of which work accommodations was only one of many components investigated) [
80‐
85]. Among the five studies in this review looking at the effect of work accommodation on its own, two were rated as high quality [
36‐
38,
58‐
60] and three were rated as medium quality [
67,
68,
87‐
89]. According to our evidence synthesis algorithm (shown in Table
1), a minimum of three high quality studies was necessary to assign a strong level of evidence, which contributed to the change in level of evidence.
Although the types of interventions evaluated were diverse across the 36 studies, they could be grouped into one of four major domains, and 12 intervention categories, based on a consensus view of the primary intervention objective (i.e. health-focused, service coordination, work accommodation or multi-domain). Nearly 60% of these studies (n = 21) included multi-domain interventions, indicating that they included at least two of the three intervention domains mentioned above. Ninety-four percent of the included studies (n = 34) used an estimate of lost time from work as their primary RTW outcome variable. This is consistent with the broader RTW research literature in which lost time is often the outcome used to assess return to work status, despite the inherent limitations of this approach [
94]. Other outcomes included work functioning and costs of work disability, but these were less commonly reported.
Our findings are consistent with other reviews that included workplace-based interventions [
7,
95‐
97]; although reviews that focused on RCTs only and conducted meta-analyses found only moderate levels of evidence for workplace interventions [
95‐
97]. While the current findings are consistent, our synthesis of workplace-based interventions for RTW in workers experiencing lost time from work due to MSK, pain-related and MH conditions includes practical messages for, and developed with, practitioners [
17,
23].
This review highlighted a number of features of the RTW literature, and of workplace-based intervention studies in particular, worthy of comment. Fourteen of the 18 high quality studies were randomized trials, while only five of the 18 moderate quality studies were randomized trials. The majority of moderate quality studies were cohort studies with comparison groups. Due largely to their design, these studies were unable to ensure the presence of important quality standards such as blinding of participants and personnel, and allocation concealment. These moderate quality studies also suffered from quality limitations in that they were subject to attrition bias (uneven attrition and substantial loss to follow-up) and did not routinely assess compliance with the intervention. The review identified a group of 19 published randomized trials, which demonstrates that it is feasible to conduct such trials in the field. We also identified three non-randomized trials and one cohort study that were rated as high quality, and five randomized trials that were rated as moderate quality. Moving forward, a strong focus on study quality in addition to trial design is warranted.
It is now accepted that the system of compensating work-related injury can exert powerful influences on injured worker RTW [
98]. Despite this, a recent systematic review identified that only a small proportion of studies including persons with compensable injury report on aspects of the compensation process [
99]. The authors proposed that research involving persons with compensable conditions should include a description of system level factors such as compensation system structure and administration (e.g., source of funding); scheme eligibility (e.g., workforce coverage, claim coverage, waiting periods); scheme benefits and entitlements (e.g., level and duration of wage-replacement benefits); and case management (e.g., work capacity review, role of physician). Descriptions of system factors were often absent in the studies included in the present review, despite the study samples being predominantly workers with compensable injuries.
Due to the substantial heterogeneity across studies regarding intervention components, workplace contexts and study designs, a meta-analysis was not conducted. Instead, a best evidence synthesis (BES) approach [
22] consistent with the original review [
7] was used. While this approach has been criticized for being at risk of producing biased results [
100], it is a transparent approach with clearly defined criteria to determine the level of evidence. This provides practitioners with useful information in addition to accessing the messages from the synthesis of studies. Practitioners can also more readily identify and consider relevant evidence from individual studies using this approach. This is especially practical when there are few studies available for a given intervention, as practitioners still need to act even when there is limited scientific evidence available to help guide their practice.
A particular strength of this review is the unique process of stakeholder engagement adopted throughout the review process. Our stakeholders provided guidance to ensure the review question was relevant, the search terms were comprehensive and the targeted literature identified was up-to-date. But more importantly, stakeholders helped us examine the findings from this review to determine the best wording for our key messages to facilitate uptake and dissemination of these evidence-based approaches for OHS practitioners and other workplace parties.
Conclusions
Our synthesis update of the scientific literature identified 12 different types of interventions from 36 studies examining three broad RTW outcomes (i.e. lost time, work functioning and costs associated with work disability). There were several intervention types that did not meet the criteria for high or moderate levels of evidence across these different outcomes. However, we note that this does not mean that these interventions are not effective, only that there is insufficient evidence to support recommending these interventions to address RTW outcomes based on the scientific evidence.
Graded activity programs and work accommodations had a moderate level of evidence for a positive effect in reducing lost time associated with work disability. Practitioners should consider implementing graded activity programs and work accommodations in practices if applicable to the work context.
Cognitive behavioural therapy (CBT) programs focused on work relevant solutions for MH conditions had a strong level of evidence for a positive effect on both reducing lost time and costs associated with work disability. Additionally, there was a moderate level of evidence that these work-focused CBT programs had a positive effect on work functioning after RTW. We recommend implementing work-focused CBT interventions to help reduce lost time and costs associated with work disability for MH conditions. Practitioners should also consider implementing these programs to help improve work functioning after RTW for individuals with MH conditions.
Alternatively, there was a strong level of evidence indicating that traditional cognitive behavioural therapy programs for MH conditions have no effect on reducing lost time from work. We recommend practitioners should seek alternative interventions (such as work-focused CBT programs) to improve RTW after illness for MH conditions.
There was a strong level of evidence to support multi-domain interventions that include multiple components aimed at service coordination, work modification and improving worker health for reducing lost time associated with musculoskeletal injuries and pain-related conditions. Additionally, there was a moderate level of evidence that these multi-domain interventions had a positive effect on improving work functioning after RTW and reducing costs associated with work disability. We recommend implementing a multi-domain intervention (i.e. with health-focused, service coordination, and work modification components) to help reduce lost time for MSK and pain-related conditions. Practitioners should also consider implementing these programs to help improve work functioning and reduce costs associated with work disability for people with MSK or pain-related conditions.