Objectives To evaluate the cost effectiveness, cost utility and cost benefit of a workplace intervention compared with usual care for sick-listed employees with distress.
Methods An economic evaluation was conducted alongside a randomised controlled trial. Employees with distress and who were sick-listed for 2–8 weeks were randomised to a workplace intervention (n=73) or to usual care (n=72). The workplace intervention is a stepwise process involving the sick-listed employee and their supervisor, aimed at formulating a consensus-based plan for return to work (RTW). The effect outcomes were lasting RTW and quality-adjusted life years (QALYs). Healthcare utilisation was measured over 12 months. Cost effectiveness analyses (CEA) and cost utility analyses (CUA) were conducted from the societal perspective and cost benefit analyses (CBA) from the employer perspective. Bootstrapping techniques were used to estimate cost and effect differences, related CIs, and cost effectiveness and cost utility ratios. Cost effectiveness planes were presented and subgroup analyses were performed.
Results CEA and CUA revealed no statistically significant differences in lasting RTW, QALYs or costs. The CBA indicated a statistically significant higher cost of occupational health services in the workplace intervention group. The workplace intervention was not cost effective according to the CEA, CUA and CBA.
Conclusions Widespread implementation of the workplace intervention for sick-listed employees with distress is not recommended because there was no economic benefit compared with usual care. Future trials should confirm if the workplace intervention is cost effective for the subgroup employees who intended to return to work despite symptoms.
This trial has been registered at the Dutch National Trial Register ISRCTN92307123.
- Participatory workplace intervention
- return to work
- economic evaluation
- mental health
- occupational health practice
- intervention studies
Statistics from Altmetric.com
- Participatory workplace intervention
- return to work
- economic evaluation
- mental health
- occupational health practice
- intervention studies
What this paper adds
Mental health problems are highly prevalent and increasingly related to sick leave and productivity loss.
There is a need for interventions aimed at facilitating return to work for employees with mental health problems.
Overall, the workplace intervention had no economic benefit compared with usual care.
However, the workplace intervention was cost effective for the subgroup of employees with baseline intentions to return to work despite symptoms.
Future studies are needed to confirm the findings for employees with distress and baseline intentions to return to work despite symptoms.
Mental health problems are highly prevalent and increasingly related to sick leave and productivity loss.1–3 In the Netherlands, about one third of all disability benefits are paid to workers with mental health problems4 and the majority (approximately 80%) of these mental health problems are stress-related.5 6 The economic impact of stress-related disorders in the working population is enormous due to the cost of productivity loss and treatment. Furthermore, several studies have shown that the cost of lost productivity is higher than the cost of healthcare consumption.7–9 Indeed, productivity losses due to mental health problems are estimated to account for over 80% of the total costs of mental health illness in workers.2 Therefore, reducing sick leave duration is essential if the economic burden for employers and society is to be lowered. Nevertheless, most interventions for mental health problems are aimed at improving mental health states and not at facilitating return to work (RTW), although it is known that better health states do not necessarily lead to a RTW.10 Thus, there is a need for interventions to facilitate RTW.
Two cost effectiveness studies on interventions for stress-related disorders, with a focus on RTW, were conducted in primary care in the Netherlands.9 11 The intervention studied by Uegaki et al took an active approach to RTW and was applied in a general practice setting. This intervention was not cost effective compared with usual care.9 Brouwers et al also studied the cost effectiveness of an active approach applied by social workers in general practice. The same intervention was earlier found to be effective as regards RTW in an occupational setting.10 Again, the authors concluded that the active intervention was not cost effective compared with usual primary care. Since these interventions seemed to lack power to influence important players in the workplace, the authors recommended that future interventions should be directed more explicitly at work and should involve important stakeholders from the workplace.11 12
A cost effective participatory workplace intervention for the RTW of workers on sick leave due to low back pain13 was adjusted for sick-listed employees with mental health problems.14 This intervention is a structured procedure involving the sick-listed employee and their supervisor, guided by a RTW coordinator. It requires the active participation of both employee and supervisor, and is aimed at formulating a consensus-based RTW plan through the identification of barriers and solutions for RTW.
The objective of this study was to conduct an economic evaluation of a workplace intervention compared with usual care. Cost effectiveness and cost utility were determined from the societal perspective and cost benefit from the employer perspective. Most economic evaluations have been performed from the societal perspective but since Dutch employers continue to pay wages for sick-listed employees for the first 2 years of sick leave and pay for interventions conducted by occupational health services, it is important to consider the employer perspective as well.
An economic evaluation from a societal and employer perspective was conducted alongside a randomised controlled trial (RCT). The study design, protocol and procedures were approved by the Medical Ethics Committee of the VU University Medical Center. The study design has been reported in detail elsewhere.15
The study population consisted of employees of three large Dutch organisations: the VU University, the VU University Medical Center and Corus (a steel company). Between April 2006 and May 2008, all eligible sick-listed employees were screened using a short questionnaire.16 Employees with distress and who had been on sick leave for 2–8 weeks were recruited by the research assistant. Exclusion criteria were: (1) a conflict at work with legal involvement; (2) working less than 12 h a week; (3) pregnancy; (4) any other episode of sick leave within 1 month before the current episode; and (5) inability to complete questionnaires written in the Dutch language. Employees with severe psychiatric disorders (mania, psychosis or severe risk of suicide) and employees with a terminal illness were not referred to the workplace intervention by the occupational physician. Baseline measurement included data regarding the personal and work characteristics of the study population and stress-related symptoms measured by the Four-Dimensional Symptom Questionnaire (4DSQ).17
Randomisation and blinding
An independent statistician prepared the randomisation scheme by using a computer-generated list of random numbers. Employees were pre-stratified by company (VU University, VU University Medical Center, Corus) and by full or part time sick leave. This resulted in six strata. Block randomisation (with blocks of four) was applied to ensure equal group sizes within each stratum. Of the four participants in each stratum, two were assigned to the workplace intervention and two were assigned to usual care. Treatment allocation was concealed, as based on the randomisation scheme, sealed opaque envelopes were prepared before the start of the study containing a referral either to workplace intervention or to usual care. After completing the baseline questionnaire each employee opened an envelope provided by the research assistant.
The participants and occupational health professionals were not blinded to the intervention. Sick leave data were blinded because they were extracted from the computerised registrations of the occupational health services. All self-reported data were entered into the computer by a research assistant, and analysis of the data by the researcher was blinded.
Usual care consisted of treatment by the occupational physician according to the evidence-based guideline of the Dutch Association of Occupational Physicians (NVAB) published in 2000 and updated in 2007.18 This guideline aims to facilitate the optimal functioning of employees with mental health problems and to prevent long-term sick leave and frequent recurrences. An early start to the treatment by occupational physicians is recommended. Occupational physicians act as motivating counsellors using cognitive behavioural elements to enhance the problem-solving capacity of employees. In addition, the Improved Gatekeeper Act requires that both the employer and employee take responsibility for a RTW plan.
The employees allocated to the workplace intervention received care from their occupational physician according to the guideline and were additionally referred to a RTW coordinator (eg, company social worker) for the workplace intervention. The participatory workplace intervention consisted of a stepwise communication process to identify and solve obstacles to RTW. This intervention is based on consensus between the sick-listed employee and their supervisor. Three meetings were planned to take place within 2 weeks. The purpose of the first meeting between the sick-listed employee and the RTW coordinator was to identify obstacles for RTW from the perspective of the employee. The second meeting was between the supervisor and the RTW coordinator, where obstacles to the employee's RTW were identified from the perspective of the supervisor. In the third meeting, which was generally the longest, the employee, supervisor and RTW coordinator discussed solutions and formulated a consensus-based plan for their implementation. The mean total duration of the three meetings was 3 h and 45 min.19 The complete workplace intervention took the RTW coordinator 7 h, including the time needed for organisation and administration.
Measures of effect were lasting RTW and quality-adjusted life years (QALYs). Lasting RTW was defined as the duration of distress-related sick leave in calendar days from the day of randomisation until full RTW to the employee's previous or another position with equal earnings, for at least 4 weeks without (partial or full) recurrence of sick leave. Health-related quality of life was measured using the EuroQol-5D at baseline and at 3, 6 and 12-month follow-up.20 QALYs gained at 12-month follow-up were calculated by multiplying the time an employee spent in a particular health state while working for the utility based on EuroQol scores using Dutch values.21 Transitions between health states were linearly interpolated.
As mental health problems often co-exist with other health problems, all healthcare-related costs were considered. Healthcare utilisation was measured using the Trimbos/iMTA questionnaire for Costs associated with Psychiatric Illness (TiC-P) over a 4-week period at baseline and at 3, 6, 9 and 12-month follow-up.22 These data were linearly interpolated over 12 months. Data on the use of occupational health services (occupational physician, company social worker and psychologist) were extracted from the computerised medical records of the occupational health services. The cost prices used for valuing resource utilisation are presented in table 1. The index year for the study was 2008. We used standard cost prices according to the Dutch Manual for Costing,23 the Dutch Central Organization for Health Care Charges or (if not available) calculated mean cost prices according to providers. Medication costs were valued using the prices of the Royal Dutch Society for Pharmacy.24
The costs of the workplace intervention are listed in table 4 and were calculated according to the bottom-up approach based on the time invested by a RTW coordinator in the workplace intervention for one employee and the costs for training occupational physicians and RTW coordinators. The median time invested by RTW coordinators was 7 h19 and this was multiplied by the fee for 1 h of a company social worker. The costs of the work adaptations were not registered because most work adaptations were organisational changes in the workplace, the costs of which are difficult to estimate.
The costs of the occupational health services were paid by the employer. Therefore, we calculated the costs of occupational health services based on real prices for the analysis from the employer's perspective.
Productivity loss from paid work was quantified in terms of net cumulative number of days of sick leave over a period of 12 months. In cases of partial sick leave, we assumed that subjects were 100% productive during the hours of partial work resumption. The cumulative number of calendar days of sick leave was converted into work-hour equivalents based on a Dutch average of 1540 work hours per year.23 The costs of production losses were calculated by multiplying the number of sick leave hours by the estimated cost of production loss for an employee per hour of sick leave based on age and sex. The costs of production losses were calculated using the human capital approach (HCA) and the friction cost approach (FCA). In the FCA, the assumption is made that every worker in the production process can be replaced and production losses cease to exist after a certain friction period and that the decrease in productivity is less than 100% of the time lost at work (ie, elasticity). A friction period of 154 days and an elasticity of 0.8 were applied in the FCA.23 25
The economic evaluation was performed according to the intention-to-treat principle. Only 1.8% of the items of the TiC-P and EuroQol were missing. A data file for complete case analysis was acquired by item-based imputation for the EuroQol data26 and hot deck imputation for the TiC-P and EuroQol data.27 Discounting of costs was not applied because the follow-up was 1 year.28
For the cost effectiveness analyses (CEA), the incremental cost effectiveness ratio (ICER) was calculated by the differences in costs of all healthcare utilisation (including the intervention costs) between the workplace intervention and usual care divided by the difference in days of the duration of sick leave until lasting RTW between workplace intervention and usual care. The ICER indicates the additional investments needed to gain one extra unit of effect. For the cost utility analyses (CUA), total costs were calculated by the sum of the healthcare costs (including the intervention costs) and the costs of productivity loss between workplace intervention and usual care. The incremental cost utility ratio (ICUR) was calculated by the difference in total costs divided by the difference in QALYs. The cost benefit analysis (CBA) from the perspective of the employer calculated a net monetary benefit by subtracting the difference in costs of occupational health services (including the intervention costs) between workplace intervention and usual care from the difference in costs of productivity loss between workplace intervention and usual care. The CUA and CBA were performed using both the HCA and FCA.
The 95% CIs around the mean cost differences were obtained by a bias corrected and accelerated bootstrapping procedure with 1000 replications.29 Bootstrapped cost effect pairs were plotted on a cost effectiveness plane. Cost effectiveness acceptability curves were generated if the ICER or ICUR was located in the north-east or south-west quadrant.30 A sensitivity analysis for the CEA was conducted to assess the effect of one extreme outlier whose healthcare costs were approximately eight times higher than the upper limit of the 95% CI and for whom the costs were mainly caused by hospitalisation for cancer (67%). Data processing was performed in SPSS 14.0. Calculation of CIs, and CEA and CUA analyses were conducted in R.31
In the effectiveness analyses, we found a significant interaction effect between intervention and baseline intention to return to work despite symptoms. The question about the intention to return to work was formulated thus: ‘Do you intend to return to work when still experiencing symptoms?’. The workplace intervention appeared to be effective in the subgroup of employees with an intention to return to work despite symptoms.32 Therefore, we conducted, on an explorative basis, CEA, CUA and CBA subgroup analyses in this subgroup.
The screening questionnaire initially identified 686 employees eligible for participation. Of these, 145 fulfilled all inclusion criteria, signed informed consent and were randomised to the workplace intervention (n=73) or usual care (n=72). The flow of participants has been reported elsewhere in detail.32 Table 2 shows the baseline characteristics in the workplace intervention and usual care groups. There were no relevant differences between the two groups with regard to the demographic characteristics, prognostic variables or stress-related symptoms. Two employees in the usual care group withdrew from the study and so no follow-up data regarding the self-reported outcomes of these two employees were available. Administrative sick leave data were available for all employees for the entire 12-month follow-up period.
Effects on sick leave and QALYs
The mean duration of sick leave until lasting RTW was 133 (SD 109) days in the workplace intervention group and 134 (SD 108) days in the usual care group. For the workplace intervention group, the mean utilities at baseline and at 3, 6 and 12-month follow-up were 0.59 (SD 0.28), 0.76 (SD 0.25), 0.80 (SD 0.23) and 0.83 (SD 0.18), respectively. For the usual care group, the respective utilities were 0.65 (SD 0.21), 0.77 (SD 0.20), 0.79 (SD 0.21) and 0.83 (SD 0.15). The mean QALY in the workplace intervention group was 0.77 (SD 0.17) and in the usual care group 0.78 (SD 0.15). The differences in effects for lasting RTW and QALYs are presented in table 3. No significant differences were found in the effects between workplace intervention and usual care.
With one exception all employees visited their occupational physician during the 12-month follow-up. Seventy two of the 145 employees (50%) visited a specialised mental healthcare provider with a frequency of 7.3 visits in the workplace intervention group and 4.5 visits in the usual care group. Medical specialists were frequently consulted: 77 of 145 employees (53%) consulted a medical specialist with a frequency of 3.7 consultations in the workplace intervention group and 6.8 consultations in the usual care group.
The costs of the workplace intervention were €763 per employee (table 4). The mean costs for healthcare utilisation and productivity loss are presented in table 1. For both groups, mean productivity loss costs represented approximately 85% of mean total costs.
Cost effectiveness analyses
Table 3 shows that there were no differences in effects between workplace intervention and usual care. An ICER of 627 was found, meaning that an additional €627 are needed in the workplace intervention group to achieve a 1-day reduction in sick leave compared with usual care. However, the cost effectiveness plane (figure 1) shows that this estimate is not robust. The acceptability curve showed that regardless of the amount one is willing to pay per day of sick leave, the probability that the intervention was cost effective did not exceed 50%. Sensitivity analysis with exclusion of one outlier did not change the direction of the CEA results.
Cost utility analyses
As a result of a marginally lower mean QALY in the workplace intervention compared with the usual care group and higher total costs according to the HCA (table 3), a negative ICUR of −184 562 was found. Figure 1 shows that the estimate of the ICUR was not robust. Application of the FCA resulted in an ICUR of −155 850 and did not change the results on the cost effectiveness plane (table 3).
Cost benefit analyses
The CBA showed no net monetary benefit of the workplace intervention compared with usual care as regards the costs of occupational health services and the costs of productivity loss. The costs of occupational health services were €584 (95% CI 321 to 820) higher in the workplace intervention group than in the usual care group (p<0.01).
The costs of productivity loss according to the HCA were €1403 (95% CI −3244 to 6329) higher in the workplace intervention group, and according to the FCA €1116 (−2196 to 4591) higher in the workplace intervention group (table 3). The workplace intervention resulted in extra costs for the employer because the costs of both the occupational health services and productivity loss were higher with workplace intervention than usual care.
The cost effectiveness planes of the subgroup analysis are shown in the lower half of figure 1. The CEA for the subgroup of employees who at baseline intended to return to work despite symptoms showed that workplace intervention was significantly more effective and associated with less costs (not statistically significant) compared with usual care (table 3). An ICER of −10 was found, indicating that the workplace intervention was more effective and less costly than usual care for this subgroup. Figure 1 shows that 88% of the bootstrapped cost effect pairs were located in the south-east quadrant. The mean ICUR of this subgroup of −124 044 indicated that workplace intervention was more effective and less costly than usual care for this subgroup. The CBA analysis for this subgroup showed a mean of €322 higher healthcare costs for the employer, while the costs of productivity loss were on average €6565 lower as a result of the workplace intervention compared with usual care. The monetary benefit of the workplace intervention for the employer was €6243.
Overall, the workplace intervention had no economic benefit compared with usual care. No differences were found in effects on lasting RTW or QALYs between the workplace intervention and usual care over the 12-month follow-up. From the societal perspective, the workplace intervention and usual care groups did not differ significantly as regards the costs of either healthcare utilisation or productivity loss. However, for the employer the costs of healthcare utilisation were significantly higher in the workplace intervention group. For the subgroup of employees with baseline intentions to return to work despite their symptoms, the workplace intervention was significantly more effective and less costly, and a net monetary benefit of €6243 was found.
Strengths and limitations
The main strengths of our study are the pragmatic RCT design and minimal loss to follow-up. The primary outcome was retrieved from the continuous sick leave registration systems of the occupational health services and was available for all employees and measured blindly. Furthermore, as only two participants in the usual care group were lost to follow-up, a minimal proportion of the self-reported outcomes was imputed.
The study also had some methodological limitations. First of all, the costs of work adaptations were not registered, and therefore the cost of the workplace intervention may have been underestimated. The feasibility study showed that most work adaptations that were realised comprised changes in tasks, arrangements for better communication and training.19 The costs of these work adaptations are difficult to estimate. Costs of changes in workplace design or equipment are less difficult to determine, but these were not frequent for employees with distress. Adding the costs of work adaptations to the total healthcare costs would have increased the difference between the groups. However, work adaptations were (to a lesser extent) part of usual care as well because of the legal obligation for employers to provide work adaptations under the Improved Gatekeeper Act.
Second, net cumulative days of sick leave were used as proxy for productivity loss. Productivity loss caused by reduced productivity when an employee is at work,33 the recently introduced concept of presenteeism, was not measured in this study. Presenteeism is an important cause of productivity loss for employees with mental health problems.33 34 Despite the fact that in our study all employees were sick-listed at baseline, productivity losses due to presenteeism may have occurred when they returned to work. Therefore, an underestimation of the costs of productivity loss in both groups is likely. A future study should evaluate if workplace intervention will reduce presenteeism compared with usual care.
Third, confidence intervals for cost differences were very wide. This is a common problem in economic evaluations alongside RCTs, caused by the small sample sizes that are based on detecting relevant differences in effects. Because the distribution of cost data typically is heavily skewed, large study populations are needed.
Lastly, 20 out of 73 participants did not receive the workplace intervention. For seven employees this was due to a full RTW before the occupational physician could refer the employee to the intervention. In some cases the occupational physician decided not to refer because of problems in the employee's personal situation only or for medical reasons. Also, some supervisors refused to participate in the intervention. The large proportion of employees who did not receive the workplace intervention may have reduced the contrast between the groups.
Comparison with the literature
Economic evaluations of RTW interventions for mental health problems are scarce. As described in the introduction, two RTW interventions applied by primary care treatment providers were not cost effective.9 11 Schene et al found that a 28-week intensive intervention consisting of group and individual sessions of occupational therapy for workers with major depression, directed at RTW, had a probability of 76% of being more cost effective than usual care alone.35 Two other cost effective RTW interventions were found for employees with mental health problems.8 36 However, the content of these interventions was very different and none can be classified as a workplace intervention.
Conversely, the participatory workplace intervention for sick-listed employees with low back pain was found to be cost effective.13 37 Loisel et al found a reduction in sick leave and lower costs after the workplace intervention. Comparison of our results to those of Loisel et al is difficult because the insurer perspective was used in their study.37 Steenstra et al found that workplace intervention was more effective than usual care as regards RTW at slightly higher costs and was equally effective as usual care at equal costs on other outcomes. An investment of €19 in the workplace intervention resulted in a sick leave reduction of 1 day compared with usual care.13 Our study results showed no benefit of the workplace intervention for employees with distress. This indicates that the workplace intervention may affect RTW differently for employees with physical and mental health problems.
The lack of effectiveness of the workplace intervention may be attributed to either the intervention or the study population. The guideline for occupational physicians for treating workers with mental health problems, which was applied in both groups, recommends workplace accommodations. Also, employees and supervisors are legally obliged to make a RTW plan under the Improved Gatekeeper Act. These elements of usual care may have reduced the contrast between the groups in our study. However, we believe that the contacts between the employee and their supervisor in the workplace intervention were certainly more intensive and more structured than in usual care. For future studies, it is recommended that employee and supervisor contacts in both groups should be monitored.
With regard to the study population, we found indications for a different effect of the workplace intervention for employees who at baseline intended to return to work despite symptoms and employees without baseline intentions to return to work despite symptoms. The subgroup analyses indicate an economic benefit of the workplace intervention for employees with baseline intentions to return to work despite symptoms, in the cost effectiveness, cost utility and cost benefit analyses. For employees without baseline intentions to return to work, no difference was found on lasting RTW. Gender, company and work characteristics such as decision latitude and job demands did not influence these findings.32 Despite the significant interaction term upon which we based the subgroup analyses, we interpret the findings of the subgroup analyses primarily as hypothesis generating. The findings should be confirmed in a future RCT comparing the workplace intervention to usual care with identification in advance of individuals intending to return to work despite symptoms. Employees without a baseline intention to return to work despite symptoms require a different treatment approach. Cognitive behavioural interventions are recommended to change the motivation for RTW despite symptoms.32
Overall, implementation of the workplace intervention for sick-listed employees with distress cannot be recommended on the basis of our study results. Future studies are needed to confirm the findings that the workplace intervention is cost effective for the subgroup of employees intending to return to work despite symptoms.
We would like to thank all participants, occupational physicians and return to work coordinators for their participation. Furthermore, we would like to thank the occupational health services for providing us with data. We also wish to thank Sjennie Daelmans for her contributions to the recruitment of participants and data entry, and Judith Bosmans for her advice with regard to the processing of the medication data.
Linked articles 050849.
Funding This study is financially supported by the Dutch Ministry of Social Affairs and Employment and the participating occupational health services.
Competing interests None.
Ethics approval This study was conducted with the approval of the Medical Ethics Committee of the VU University Medical Center.
Patient consent Obtained.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.