Article Text
PDF
Abstract
Objectives Since 2014, the Federal German Pension Insurance has approved several departments to implement work-related medical rehabilitation programmes across Germany. Our cohort study was launched to assess the effects of work-related medical rehabilitation under real-life conditions.
Methods Participants received either a common or a work-related medical rehabilitation programme. Propensity score matching was used to identify controls that were comparable to work-related medical rehabilitation patients. The effects were assessed by patient-reported outcome measures 10 months after completing the rehabilitation programme.
Results We compared 641 patients who were treated in work-related medical rehabilitation with 641 matched controls. Only half of the treated patients had high initial work disability risk scores and were intended to be reached by the new programmes. The dose of work-related components was on average in accordance with the guideline; however, the heterogeneity was high. Work-related medical rehabilitation increased the proportion of patients returning to work by 5.8 percentage points (95% CI 0.005 to 0.110), decreased the median time to return to work by 9.46 days (95% CI -18.14 to -0.79), and improved self-rated work ability by 0.38 points (95% CI 0.05 to 0.72) compared with common medical rehabilitation. A per-protocol analysis revealed that work-related medical rehabilitation was more effective if patients were assigned according to the guideline and the minimal mandatory treatment dose was actually delivered.
Conclusions The implementation of work-related medical rehabilitation in German rehabilitation centres affected work participation outcomes. Improving guideline fidelity (reach and dose delivered) will probably improve the outcomes in real-world care.
Trial Registration Number DRKS00009780
- rehabilitation
- programme evaluation
- return to work
- employment
- health service research
Statistics from Altmetric.com
Key messages
What is already known about this subject?
Randomised controlled trials have shown that work-related medical rehabilitation improves work participation outcomes.
What are the new findings?
Our effectiveness study showed that the nationwide implementation of work-related medical rehabilitation programmes improved work participation also in routine care.
We observed consistent beneficial effects in favour of these programmes; however, the effects were small, and they were lower than the effects observed in the randomised controlled trials.
How might this impact on policy or clinical practice in the foreseeable future?
Our study identified two implementation failures: patient assignment and treatment dose were not in line with the guideline’s recommendations.
Reaching patients with high risks of not returning to work and accomplishing the treatment dose which is recommended will probably improve the outcomes in routine care.
Introduction
Musculoskeletal disorders are a leading cause of work absenteeism and work disability.1 In Germany, there are 150.4 million days of sickness absence and more than 21 000 new disability pensions caused by musculoskeletal disorders each year.2 3 To support work participation despite chronical illness in patients with chronic musculoskeletal disorders, rehabilitation is a key health strategy.4 In Germany, rehabilitation for working-age people is mainly provided by the German Pension Insurance. Rehabilitation by the German Pension Insurance is primarily approved to maintain and to restore work ability and to avoid disability pensions. It is provided as a 3-week programme. The treatment is multiprofessional and follows evidence-based therapy recommendations.5 6 The effects of German medical rehabilitation programmes on patients with chronic musculoskeletal disorders on work participation outcomes are, however, at best small.7 8 A randomised controlled trial by Hüppe et al did not show better work participation outcomes at the 12-month follow-up.9 Meta-analyses of observational studies without controls revealed a slight reduction in the number of days off work post programme7 8; however, particularly patients with poor work functioning had a high risk of not returning to work.10 11
To enhance the effectiveness of rehabilitation, work-related medical rehabilitation (WMR) programmes have been developed and tested in recent years.12 13 These programmes were designed for patients with a high risk of failing to return to work and enrich common medical rehabilitation programmes by adding four extra components: work functioning assessment, intensified social counselling, work-related psychosocial groups and work-related functional capacity training. The work functioning assessment uses a functional capacity evaluation to describe the patient’s work capacity and to identify discrepancies between the patient’s work capacity and job demands. In the counselling sessions, a social worker explores the job situation and determines the opportunity for a graded return to work. Moreover, the need for complementary services such as retraining, wage subsidies and ergonomic assistance is appraised and initiated if considered necessary. The work-related psychosocial groups are aimed at an individual’s return-to-work plan. Additionally, stress and conflict management strategies are introduced. Work-related functional capacity training executes job specific movements and postures according to the principle of graded activity. The patients also explore whether ergonomic devices might support them in handling the job demands. A meta-analysis of four randomised controlled trials which compared WMR programmes and common medical rehabilitation programmes for patients with chronic musculoskeletal disorders indicated higher rates of stable return to work and lower levels of sickness absence in the WMR rehabilitation programmes after 1 year.13 The absolute risk difference in terms of stable return to work favoured the WMR programmes by approximately 20 points.
To implement WMR programmes nationwide, a guideline was developed by the German Pension Insurance. This guideline describes the population that should be reached as well as the content and dose of the aforementioned core components of WMR. After a feasibility study,14 80 WMR departments were approved by the Federal German Pension Insurance in 2013 and early 2014. It was shown that the actual delivered dose of work-related components increased considerably due to the implementation of the guideline.6 However, the dose delivered varied between rehabilitation centres and deviated from the guideline’s recommendations in at least half of the centres.
Such implementation failure is a major concern when programmes from randomised controlled trials are put into practice. Interventions that work in efficacy studies with carefully selected patients and high treatment fidelity might not necessarily also perform well in real-world settings.15 Shortcomings in patient selection and treatment fidelity might reduce the effects, and an intervention might be useless due to such shortcomings. Therefore, our study was initiated to explore the effectiveness of WMR in real-world application. We assessed the guideline fidelity of the implemented WMR programmes, tested the effects of WMR programmes on return to work and several other outcomes and clarified whether better guideline fidelity might improve outcomes.16
Methods
Study design
The study was a cohort study under real-life conditions with two parallel groups. Participants received either a WMR or a common medical rehabilitation programme in one of the more than 200 rehabilitation departments that conduct rehabilitation programmes approved by the Federal German Pension Insurance. A detailed study protocol has been published elsewhere.16 The current utilisation of WMR is still considerably below the estimated number of patients who need it. Therefore, comparable controls who received a common medical rehabilitation programme were available, even though their severe restrictions in terms of work ability would also have justified participation in a WMR programme. We aimed to analyse 900 patients in each arm to identify small effects in favour of WMR. Baseline data were assessed after approval but before the start of the rehabilitation programme. Data on the dose received were assessed 3 months after the patients’ discharge from the rehabilitation centres. All other follow-up data were assessed 10 months after the patients’ discharge. Moreover, administrative data on the provided treatments were used.
Treatment
Control
Participants in the control group received a common medical rehabilitation programme lasting approximately 3 weeks according to current treatment standards and guidelines for the rehabilitation of patients with musculoskeletal disorders. The daily quantum of therapy was 3 to 4 hours. Following a multimodal approach, the programmes included sports and exercise therapy, physiotherapy, massage and other physical therapies, social and psychological counselling, patient education, pain management and relaxation training.
Intervention
Participants in the intervention group received a WMR programme according to the guideline for WMR,17 as well as to the current guidelines for the rehabilitation of patients with musculoskeletal disorders. The WMR programme comprised the common medical rehabilitation programme with the addition of the four components that were described above: work functioning assessment, intensified social counselling, work-related psychosocial groups and work-related functional capacity training. The guideline for WMR recommends a minimal dose of 1.5 hour work functioning assessment, 0.5 hour additional social counselling, 3 hour work-related psychosocial groups, and 6 hour functional capacity training per programme. If a participant receives all components with the minimal dose that is recommended in the guideline, this adds at least 11 hours of work-related diagnostic and therapeutic components per programme. The guideline also states that work-related psychosocial groups and work-related functional capacity training might not be necessary for all patients. However, all participants in a WMR programme should receive at least a work functioning assessment and social counselling, meaning a minimal additional dose of 2 hours per rehabilitation programme.
Setting and participants
The included patients were aged 18–65 years and had chronic musculoskeletal disorders, usually back pain. After patients had requested the Federal German Pension Insurance for a rehabilitation programme, their application documents were evaluated by the sociomedical service. As described in the guideline for WMR, the sociomedical service was advised to use the results of a standardised risk score for the assignment decision.10 This score (SIMBO, German: Screening-Instrument zur Feststellung des Bedarfs an medizinisch-beruflich orientierten Maßnahmen in der medizinischen Rehabilitation), which is part of the application documents, estimates the risk of not returning to work and ranges from 0 to 100 points.10 Scores of 30 points or more were recommended as an indicator of the need for WMR. Based on previous studies, we expected that the guideline’s recommendation on assignment would be followed only partially by the sociomedical service. All patients, that is, patients assigned to WMR and common medical rehabilitation, were then contacted by the researchers before the actual rehabilitation programme started and asked to complete the questionnaires. Rehabilitation was performed between March and September 2016.
Outcomes and other measures
Primary outcomes
The primary outcomes of this study were stable return to work, time to return to work and self-rated work ability. Stable return to work was defined as a minimum of 4 weeks of employment without sick leave.18 Time to return to work was assessed from discharge, and if necessary, censored to the time of follow-up. Self-rated work ability was assessed by the Work Ability Score (0–10 points).19
Secondary outcomes
The secondary outcomes were general health,20 physical functioning,21 pain intensity and pain disability,22 depression,23 anxiety,24 fear-avoidance beliefs,25 self-management skills,26 use of medications (pain medications, antidepressants, other medications), work disability expectations27 and unemployment. Higher general health, physical functioning and self-management skills scores indicated better outcomes, whereas higher scores for the other variables represented worse outcomes. A detailed description of all measures is given elsewhere.16
Dose delivered and received
To check guideline fidelity, we assessed the dose delivered and dose received. The dose delivered was assessed separately for the work functioning assessment, social counselling, work-related psychosocial groups and functional capacity training. The dose delivered was extracted from administrative records that record all the treatments provided during the rehabilitation programme. The dose received was measured using three patient-reported scores that assessed content (eg, having received advice about ergonomic design of the workplace), consistency (eg, that the whole rehabilitation team intensively dealt with the health problems in working life) and goal attainment (eg, being well prepared for returning to work).
Covariates
The following variables were assessed at baseline only: sociodemographic characteristics (sex, age, educational level, partnership, children and first language), health-related measures (somatisation,28 pain generalisation, psychosocial strain,29 type of musculoskeletal disorder, number of secondary diagnoses and mental comorbidity), the intention to deal with work-related problems during the rehabilitation programme,12 characteristics of the job situation (employment state, sickness absence within the preceding 12 months, current sick leave, job position and economic sector) and the risk of failing to return to work using the SIMBO.10 The SIMBO was assessed independently from the score that was part of the application documents.
Statistical analysis
Descriptive statistics were used to characterise the samples treated by WMR and common medical rehabilitation. To compare the outcomes of both programmes, we identified for each patient treated in a WMR programme one similar patient who was treated in a common medical rehabilitation programme. Matching was realised using propensity scores (nearest-neighbour matching without replacement). The propensity score is the conditional probability to receive the treatment, for example, WMR instead of a common medical rehabilitation programme. We used a logistic regression model with 32 variables to calculate the propensity scores.30–32 The assessment of variables as outlined in our study protocol16 and their inclusion in the logistic regression model was driven by the idea to consider the most relevant factors which may be associated with both assignment and outcomes. The propensity scores were calculated from five multiple imputed data sets and then averaged. Multiple imputation was chosen as restricting the sample to complete cases only in a comprehensive regression model can reduce sample size substantially even if the amount of missing data in single variables is low. Matching was performed with the average scores.33 The matched control group had similar characteristics to the group of WMR patients, thus allowing an unbiased estimation of the treatment effect for patients who actually were treated in a WMR programme. To examine whether the beneficial effect of WMR could be higher if guideline fidelity was improved, we performed a per-protocol analysis and included only patients with SIMBO scores of at least 30 points and only WMR patients who actually received the minimal treatment dose of 2 hours per programme that is described as mandatory for all patients treated by WMR. We assumed that deviations from the guideline’s recommendations would weaken the effects in favour of WMR, and that the effect in patients who were intended to be reached and treated as recommended would be more similar to the effects known from the randomised controlled trials.
The outcomes of both groups were compared using linear or logistic regression models. The median time to return to work was estimated with a parametric regression survival-time model. Since participants were clustered within centres, we used random-effect models. We report adjusted predicted estimates and standard errors for each group and the absolute differences between WMR and common medical rehabilitation programmes and their 95% CIs. The results of the statistical tests were regarded as significant if the two-sided p value of a test was less than 0.05. All calculations were performed in Stata SE V.15.0.
Results
Recruitment and participants
We contacted 16 823 patients per postal mail who had been approved for rehabilitation due to musculoskeletal disorders. A total of 9761 patients completed the baseline questionnaire between March and July 2016 before starting their programme. We excluded 2547 patients since they did not consent to linkage of the questionnaire and administrative data, or the patients started neither a common rehabilitation programme nor a WMR programme. Another 1815 patients were excluded since they were younger than 18 years of age or did not report their baseline employment status, the duration of the rehabilitation programme or the treatment dose was incomplete or inconsistent, or they started the rehabilitation too late to ensure the follow-up at 10 months. Finally, 3831 patients (71%) completed both the 3-month and the 10-month follow-up questionnaire. Of these, 3190 patients were treated in a common medical rehabilitation programme and 641 in a WMR programme. The final matched sample included 1282 patients, 641 in each arm. In our per-protocol analyses, we compared 274 patients treated in WMR with 274 matched controls (figure 1).
Flow of participants. MR, medical rehabilitation; WMR, work-related medical rehabilitation.
Baseline characteristics
The mean age of the final matched sample was approximately 52.4 years (SD=7.7) and 75.3% were women. Sickness absence before starting the rehabilitation programme was approximately 13 weeks. The SIMBO score was, on average, nearly equal to 30 points, and only 46.6% had scores of 30 points or more. The samples of participants in WMR and the common medical rehabilitation programmes were balanced for all the variables that were included in calculating the propensity scores (table 1 and online supplementary file 1).
Supplemental material
Baseline characteristics
Dose delivered and received
Treatment dose of the work-related components was, on average, in accordance with the guideline. The mean dose of work-related components per programme was approximately 11.1 hours in WMR but only 2.9 hours in common medical rehabilitation (table 2). The heterogeneity was high between facilities, ranging from 2.8 hours to more than 20 hours. Eight per cent of the patients who were assigned to WMR did not even receive the minimal mandatory dose of 2 hours per programme. The content, consistency and goal attainment scores were all in favour of the WMR.
Implementation of work-related components
Primary outcomes
The return to work rate improved from 75.3% in common medical rehabilitation to 81.1% in WMR (p=0.033), that is, by 5.8 percentage points. The time to return to work decreased from 20.30 to 10.84 days (p=0.033). The self-reported work ability increased by 0.38 points (p=0.024) (table 3).
Primary and secondary outcomes
Secondary outcomes
The beneficial effects of WMR were also seen in several secondary outcomes. Patients felt less depressed, reported less fear avoidance, had better self-management skills, were less likely to be unemployed and reported less frequent use of other medications (table 3). However, these effects were only small.
Per-protocol analysis
The likelihood of returning to work was much lower in both groups than in the full sample; however, the difference increased to nine points in favour of WMR (p=0.036; table 4). The time to return to work was longer than in the primary analysis. The advantage of WMR was approximately 100 days (p=0.045). Self-reported work ability improved by approximately 0.7 points (p=0.036).
Primary outcomes from per-protocol analysis
Discussion
WMR improved work participation when used in routine care. We observed beneficial effects of WMR; however, these effects were small, and they were lower than the effects observed in the randomised controlled trials. The per-protocol analysis indicated that the effect of WMR in real-life care depends on guideline fidelity, in particular, reaching the patients as described in the guideline, for example, patients with a high risk of failing to return to work, and assuring the recommended treatment dose is provided. The effect in real-world care was probably reduced since, first, half of the patients had low risks of failing to return to work, and second, several patients were not treated as recommended by the guideline.
Effectiveness studies of nationwide programme implementation have attracted increasing attention in the last decade in occupational and rehabilitation medicine.31 34 35 A rigorous approach to identify implementation failure can help to overcome gaps between efficacy and effectiveness. Our study is important since it shows that the effect of WMR was reduced in routine care, and we identified two important weaknesses in the programme implementation, the removal of which will probably improve work participation outcomes.
First, half of the patients were assigned to WMR even though they had rather low risks of not returning to work. These patients do not seem to benefit from a more complex intervention. Such misassignment is a common implementation failure and was recently also observed in evaluations of the Danish return-to-work programme35 and the Finnish vocationally oriented multidisciplinary early rehabilitation programme.31 Guideline-appropriate assignment to WMR was hampered by the fact that the assignment was primarily carried out by the sociomedical service of the Federal German Pension Insurance. The decision by this service is based on the patient application documents only. The documents that assess the risk of failing to return to work were incomplete in approximately one-third of the patients, and other assignment strategies were probably used in these instances. Physicians may have considered work demands (eg, shift work) or specific professions (eg, nurses) as particular suited to be assigned to WMR. Though these factors also determine returning to work, their prognostic accuracy in predicting return to work is poor compared with standardised risk scores as the SIMBO. Moreover, in several cases, patients were assigned to WMR despite the application documents clearly indicating a low risk of not returning to work. This might be partly driven by an administrative fixation on easy measurable goals. The inclusion of patients who very likely succeed will increase the return-to-work rates. This is, however, a waste of money since the outcomes are not due to programme participation.
Second, the recommended treatment dose was not accomplished for several patients. Treatment delivery is a key condition for programme effectiveness. Dose deviations from the plan and dose variations among facilities are common when implementing complex interventions.36 Implementation of the new components of WMR needs up-to-date professional skills, novel equipment and tools, and additional staff.6 Though the rehabilitation centres were differently prepared for the challenge of this implementation, the dissemination strategy of the Federal German Pension Insurance did not explicitly address such potential differences between the facilities. In addition, the differences between WMR and common medical rehabilitation in the scores which we used to describe the dose received (contents, consistency, and goal attainment) were much less pronounced than the differences in the dose delivered. Although the documented treatment dose was achieved in many centres, the strong interdisciplinary approach that characterised the programmes that were tested in the randomised clinical trials might still be a challenge to be accomplished in several departments. A similar observation was made in the evaluation of the Danish return-to-work programme.34 Team fidelity strongly varied between the municipalities in that study, indicating that establishing an interdisciplinary return-to-work team and working successfully in such a team is challenging.34
When interpreting our findings, the following limitations must be considered. First, although propensity score matching is a powerful tool to reduce bias, a propensity score matched analysis is still based on observational data only.32 It is not a randomised controlled trial. Consequently, the results are potentially biased by unobserved differences between patients treated by WMR and by common medical rehabilitation. Second, the final sample size was smaller than that outlined in our study protocol,16 and selection bias due to selective non-participation and selective follow-up non-response might have biassed our estimates.
The limitations of the study are balanced by the following strengths. First, we used a rigorous approach to check guideline fidelity and used these findings for a per-protocol analysis. This gave us valuable insights into how implementation can be improved. Second, our pragmatic effectiveness trial is one of the largest trials to test the effects of a new rehabilitation approach in real-world care environments. We believe that implementation research should document fidelity, and additionally check whether the anticipated goals are also achieved in real-world care settings and assess the degree to which findings from efficacy studies can be replicated.
To conclude, the nationwide implementation of WMR improved work participation; however, the beneficial effects of WMR were small. Better guideline fidelity, for example, reaching the appropriate patients and accomplishing the treatment dose in line with the guideline’s recommendations, will probably improve the outcomes in routine care.
References
Footnotes
Correction notice This article has been corrected since it published Online First.
Contributors MB, MS and MS planned and designed the study. MM did the data management. MB and MM analysed the data. All authors interpreted the data. MB drafted the first version of the manuscript. All authors read and approved the final manuscript.
Funding This research was funded by the Federal German Pension Insurance, and this institution paid money to the University of Lübeck and the University of Würzburg.
Competing interests None declared.
Patient consent for publication Not required.
Ethics approval All procedures performed in the studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee, and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. The study protocol was approved by the Ethics Committee of the University of Lübeck (15-356).
Provenance and peer review Not commissioned; externally peer reviewed.
Data availability statement All data relevant to the study are included in the article or uploaded as supplementary information.