Article Text
PDF
Abstract
Objectives To evaluate the cost-effectiveness of a work style (WS) intervention and a work style plus physical activity (WSPA) intervention in computer workers with neck and upper limb symptoms compared with usual care.
Methods An economic evaluation was conducted from an employer's perspective and alongside a randomised controlled trial in which 466 computer workers with neck and upper limb symptoms were randomised to a WS group (N=152), a WSPA group (N=156) or a usual care group (N=158). Total costs were compared to the effects on recovery and pain intensity. In the primary analyses, missing effect data were imputed using multiple imputation techniques.
Results Total costs during the 12-month intervention and follow-up period were €1907 (WS), €2811 (WSPA) and €2310 (usual care). Differences between groups were not statistically significant. Neither intervention was more effective than usual care in improving overall recovery. The WS intervention was more effective than usual care in reducing current pain, average pain and worst pain in the past 4 weeks, but the WSPA intervention was not. The acceptability curve showed that when a company is willing to pay approximately €900 for a 1-point reduction in average pain (scale from 0 to 10), the probability of cost-effectiveness compared to usual care is 95%. Similar results were observed for current and worst pain.
Conclusions This study shows that the WS intervention was not cost-effective for improving recovery but was cost-effective for reducing pain intensity, although this reduction was not clinically significant. The WSPA intervention was not cost-effective compared with usual care.
Trial registration number ISRCTN87019406.
- Ergonomics
- occupational health practice
- psychology
- repetitive strain injury
- health promotion
Statistics from Altmetric.com
What this paper adds
A recent randomised controlled trial showed that a work style intervention was effective in reducing pain but a work style plus physical activity intervention was not.
However, knowledge about the cost-effectiveness of interventions to manage neck and upper limb symptoms is missing.
A work style intervention does not seem to be cost-effective for improving recovery from neck and upper limb symptoms but does seem to be cost-effective for reducing pain intensity.
Combining a work style intervention with a lifestyle physical activity intervention does not seem to be cost-effective.
Before companies can be advised to implement a work style intervention, research is needed to investigate whether or not the observed pain reductions are clinically relevant and which specific risk groups may benefit most from a work style intervention.
Neck and upper limb symptoms are frequently reported by computer workers. In 2002, 28% of the general Dutch working population had experienced pain or stiffness in the neck, shoulder, arms, hands or wrists in the previous 12 months.1 In Europe, the prevalence of work-related neck/shoulder pain was 25% and 15% for work-related arm pain.2 In the Netherlands, the total annual costs of neck and upper limb symptoms due to decreased productivity (sick leave and chronic disability) and medical costs were recently estimated at €2.1 billion.3 Recent conservative estimates of the annual costs of all musculoskeletal disorders in the USA ranged from $45 to $54 billion.4 So, work-related neck and upper limb symptoms are not only a medical but also a huge socio-economic problem worldwide.
Neck and upper limb symptoms have a multi-factorial aetiology. Previous studies have shown that individual factors such as gender5 and comorbidity,6 physical factors such as self-reported duration of mouse use,7 and organisational factors such as high work demands and little control,8 are associated with neck and upper limb symptoms. Despite this multi-factorial origin, most intervention studies aiming to reduce neck and upper limb symptoms focused only on the physical components of the workplace.9–12 In recent years attention has shifted from single component interventions to multi-component interventions or ‘integrative’ interventions that include both physical and psychosocial components.13 Also, more attention has been paid to the concept of work style, which influences individual responses to high work demands.14 In the present study these responses were conceptualised as changes in body posture, appropriate workplace adjustment, number and quality of breaks, and work stress. Work style is predictive of future pain and functional limitations in office workers with neck and upper limb symptoms.15 In theory, increasing physical activity may be effective in reducing neck and upper limb symptoms,16–18 but the results of studies investigating the effectiveness of exercise programs on neck and upper limb symptoms are inconsistent.19 This inconsistency might be due to lack of adherence to exercise programs. However, lifestyle physical activity interventions that aim to increase (daily) lifestyle physical activity are promising regarding the maintenance of physical activity.20 A recent randomised controlled trial21 showed that a work style intervention was effective in reducing pain but a work style plus physical activity intervention was not. Currently, knowledge about the cost-effectiveness of interventions to manage neck and upper limb symptoms is still lacking.22 Information about the cost-effectiveness of interventions is relevant for employers because it provides insight into the return on investment and thus allows better decisions concerning interventions. In the present study we evaluated the cost-effectiveness of these interventions compared with usual care from an employer's perspective.
Methods
Design
This cost-effectiveness study was conducted alongside a randomised controlled trial with two intervention groups and a usual care group. The intervention period lasted 6 months and measurements were conducted at baseline, and at 6 and 12 months follow-up. The design of the randomised controlled trial has been described extensively in a previous paper.23 The study was approved by the Medical Ethics Committee of the VU University Medical Center.
Study population
The source population for this study consisted of approximately 8000 computer workers with neck and upper limb symptoms from the head offices of various branches (eg, insurance, science, energy, transportation policy and taxes) of seven Dutch companies. Workers with chronic and with recurrent symptoms in the past 6 months as well as workers with recent symptoms were included. A short questionnaire was used to select workers who were eligible for participation in the study. The inclusion criteria were as follows:
Frequent or long-term pain, stiffness or tingling in the neck, shoulders, arms, wrists and/or hands in the preceding 6 months
Computer work for at least 3 days per week and 3 hours per day
Employment contract lasting at least until the last follow-up measurement
Currently not under treatment for neck and upper limb symptoms by a medical professional
No (self-reported) non-work related somatic disease
Currently no sickness absence for more than 50% of working hours.
All workers who gave informed consent and completed the baseline questionnaire were randomised to (1) a work style intervention (WS group), (2) a work style and physical activity intervention (WSPA group) or (3) usual care.
Interventions
The interventions were based on two theoretical models.24 25 The goal of the work style intervention was behavioural change with regard to body posture, workplace adjustment, breaks and coping with risk factors for work stress. The goal of the lifestyle physical activity intervention was increased engagement in moderate to high intensity physical activity following the provision of group counselling but not supervised exercise programs. The two intervention groups (WS and WSPA) participated in six interactive group meetings (with a maximum of 10 participants) in a 6-month period. The duration of the group meetings was maximally 90 min in the WSPA group and 60 min in the WS group. The additional time in the WSPA group was used for the lifestyle physical activity intervention. The usual care group did not attend group meetings but did receive ‘breaks and exercise reminder software’ as in the WS and WSPA groups. A detailed description of each group meeting has been published elsewhere.23
Outcome measures
Recovery from neck and upper limb symptoms was assessed at 6 and 12 months after randomisation using a 7-point scale ranging from ‘much worse’ to ‘completely recovered’ compared to baseline. Recovery was assessed separately for the neck/shoulder region and the arm/wrist/hand region. Participants were considered to have recovered when they indicated they had completely or to a large extent recovered in both body regions.
Pain intensity (current pain, average pain and worst pain in the past 4 weeks) was assessed at baseline, and at 6 and 12 months after randomisation using an 11-point numerical rating scale ranging from 0 ‘no pain’ to 10 ‘worst pain ever’.26 Pain intensity was assessed separately for the neck/shoulder region and the arm/wrist/hand region. An overall pain measure was constructed using the highest score in either the neck/shoulder region or the arm/wrist/hand region.
Cost data
The economic evaluation was conducted from the employer's perspective. Consequently, only costs relevant to the employer were included, that is the costs of the intervention and the costs of production losses. The intervention costs were estimated using a bottom-up approach and included the costs of the counsellors, that is training of the counsellors (€12.5 per participant (pp)), time spent on group meetings (WS group: €79.5 pp; WSPA group: €109.5 pp), the costs of elastic bands used in the WSPA group (€3.2 pp) and the costs of ‘breaks and exercise reminder software’, which were dependent on the number of licences that each company held (€4.5 pp, €25.0 pp). The elastic band was used as an extra stimulus to increase daily physical activity.
Costs of production losses were defined as costs due to absenteeism from paid work because of illness and costs due to absenteeism because of time spent at group meetings. Work absenteeism was estimated using company records. Data on absenteeism were collected for the year prior to randomisation and the year after randomisation. Absenteeism due to any reason was included because companies did not register the causes of absenteeism. In the present study we only included 100% sickness absence in the cost-effectiveness analyses, meaning that part time absenteeism was counted as 100% absenteeism. In the Netherlands, workers can be on partial sick leave as part of a return to work program.
The human capital approach was used for estimating the costs of work absenteeism due to illness using the mean income of the Dutch population according to age and gender,27 because data on individual income were not available for every participant. We used elasticities to account for the relationship between sickness absence and production losses. For the Netherlands, the elasticity between labour time and labour production was estimated to be 0.8, which means that production losses were estimated to be 80% in workers with 100% sickness absence.28 This also means that 20% of the production losses will be compensated for by the sick worker or his colleagues. In the primary analyses, costs of work absenteeism were estimated using an elasticity of 0.8. In order to estimate the costs of attending group meetings, the counsellors noted those present at these meetings. The costs of attending group meetings were estimated using the method of Oostenbrink et al27 and by using an elasticity value of 0.3. The elasticity value of 0.3 is based on the study of Jacob-Tacken et al29 who found that shorter absenteeism was associated with lower productivity costs because lost work was more often compensated for during normal working hours.
Data analysis
Cost data were complete for 93% (433/466) of the participants. Participants with missing cost data (N=33) did not give permission for analysis of their absenteeism data. All analyses were based on the data of the 433 participants with complete cost data. Effect data were missing for 143 (33%) participants with regard to recovery and for 136 (31%) participants with regard to pain. In the primary analyses, missing effect data were supplied by applying multiple imputation using the multiple imputation by chained equation procedure.30 This is a flexible regression based imputation method, which allows one to specify the multivariate structure in the data and to use all the observed data of other variables to estimate the missing values. We constructed five imputed datasets.
Differences (and 95% CIs) in recovery and pain reduction between the groups at follow-up were adjusted for pain at baseline and calculated using linear regression analysis. For recovery, we used a logistic regression model to calculate the log odds of recovery for each patient and converted these to percentage scores. Use of these percentage scores allowed differences between groups to be analysed by linear regression. ANOVA (p<0.05) was used to analyse differences in absenteeism parameters between year 1 (prior to randomisation) and year 2 (during the intervention and follow-up). Described absenteeism parameters were: average number of absent days, frequency of absent periods and 12-month prevalence of: absenteeism in general (ie, ≥1 day), short-term absenteeism (ie, <2 weeks), long-term absenteeism (ie, ≥2 weeks; <3 months) and very long-term absenteeism (≥3 months). Mean differences in costs were estimated by using linear regression analyses. Differences in costs and effects were estimated in each imputed dataset and results were combined using Rubin's rules.31 Bootstrapping with imputed datasets was used to estimate the 95% CIs around these differences and results were combined as previously described.32 All cost estimates were adjusted for the costs expended in the previous year.
The bootstrapped cost–effect pairs estimated from each imputed dataset were plotted on a cost-effectiveness plane. Acceptability curves were calculated to show the probability that the interventions were cost-effective at a certain ceiling ratio (ie, the amount of money that the employer was willing to pay for one recovered worker or one unit pain reduction). Acceptability curves were based on the net monetary benefit approach.33 Net monetary benefits were estimated in each imputed dataset and subsequently averaged.32 The following pair-wise comparisons were made: WS versus usual care, WSPA versus usual care, and WSPA versus WS. All analyses were conducted for overall pain and recovery and for pain and recovery in the neck/shoulder region and the arm/wrist/hand region.
Sensitivity analyses
To assess the robustness of the results two sensitivity analyses were conducted.34 The first sensitivity analysis consisted of three sub-analyses in which elasticity values were used other than the 0.8 that was employed in the primary analyses. Jacob-Tacken et al29 found that lost work due to absenteeism is often compensated for without incurring productivity costs. Therefore, conventional methods multiplying the duration of absence by the daily wage often overestimate production losses. The findings of Jacob-Tacken et al29 indicate that only 30–55% of the productivity costs calculated with conventional methods need to be taken into account. Based on the study of Jacob-Tacken et al,29 we conducted three sub-analyses with elasticity values of 0.3, 0.55 and 1.0. In the second sensitivity analysis, only data of participants with complete cost and effect data were taken into account (ie, complete cases analysis). So that the second sensitivity analyses could be compared with the primary analyses, an elasticity of 0.8 was also used in the complete case analysis.
Results
Figure 1 shows that of the approximately 8000 workers who were invited to participate in the study, only 466 were actually recruited and randomised into one of the three groups: 152 into the WS group, 156 into the WSPA group and 158 into the usual care group. All participants reported chronic or recurrent symptoms for a period of at least 6 months. None of the participants reported symptoms only in the past 2 weeks. Complete cost data were available for 93% (N=433) of the participants. Complete cost and effect data were available for 62% (N=290) of the participants. Effect data were missing and therefore imputed for 143 participants with complete cost data (WS: N=45; WSPA: N=43; usual care: N=55). Participants with imputed data were slightly younger on average (43.1 vs 44.8 years) and more likely to be female (51.0% vs 38.3%) compared to participants without imputed data (N=290). No differences with regard to pain intensity or absenteeism at baseline were observed between these groups. Absenteeism during intervention and follow-up was significantly higher in participants with imputed data (25.7 days per year) compared to participants without imputed data (13.3 days per year).
Table 1 presents the baseline characteristics of all 433 participants with complete cost data. There were no differences in gender, age or pain between the study groups.
Baseline characteristics of participants with complete cost data (N=433)
Compliance
The number of participants who attended at least three out of six group meetings was very similar between the intervention groups, being 124 (82%) in the WS group and 122 (78%) in the WSPA group.
Absenteeism
Table 2 presents absenteeism parameters in the year prior to randomisation (year 1) and the year during the intervention and follow-up period (year 2). The average number of absence days was low for most of the participants and very high for a few participants in all three study groups. In year 1, the average number of absence days was highest in the WSPA group and lowest in the usual care group. This difference reduced slightly during the intervention year because the average number of absence days increased by 4.7 days in the usual care group and reduced by 0.9 days in the WS and WSPA groups. None of the observed changes in absenteeism parameters between the years were statistically significant. The number of participants with very long absenteeism in year 1 was somewhat higher in the WSPA group compared to the other groups. However, this difference remained the same in the year during the intervention.
Absenteeism parameters at baseline (in the year before the intervention) and in the year of the intervention and follow-up of participants with complete follow-up data (N=433)
Effectiveness of the intervention
Overall recovery and pain
Table 3 shows the effectiveness of the interventions on overall recovery and pain and on recovery and pain in the neck/shoulder region and arm/wrist/hand region between baseline and the 12 months follow-up. Neither intervention was more effective than usual care in improving overall recovery. The WS intervention was significantly more effective than usual care in reducing current pain, average pain and worst pain in the past 4 weeks, but the combined WSPA intervention was not.
Differences in recovery and pain between study groups (multiply imputed data) at 12 months follow-up (and 95% CIs estimated with bootstrapping)
Neck/shoulder and arm/wrist/hand region
Neither intervention was more effective than usual care in improving recovery in either of the body regions. The WS intervention and the WSPA intervention were both significantly more effective than usual care in reducing current, average and worst neck/shoulder pain in the past 4 weeks but not in reducing pain in the arm/wrist/hand region.
Total costs
Table 4 presents the mean costs (and standard deviations) of all three study groups during the 12 months follow-up. No statistically significant differences in costs were observed between study groups after adjustment for the costs of sickness absence in the year prior to the intervention.
Mean total costs (and SD) for each study group during the intervention and follow-up period of 12 months
Cost-effectiveness
Overall recovery and pain
Neither intervention was cost-effective in improving overall recovery between baseline and 12 months follow-up compared to usual care.
Figure 2 shows the cost-effectiveness plane for the WS intervention versus usual care for reducing average pain in the past 4 weeks between baseline and 12 months follow-up. The cost-effectiveness plane shows that, on the whole, the WS intervention was more effective than usual care as 98.2% of the cost–effect pairs are located on the right side of the y-axis (more effective). It also shows that the costs of the WS intervention were lower than the costs of usual care because 83.7% of the cost–effect pairs are located below the x-axis (less expensive). These cost differences were not statistically significant. The acceptability curve in figure 1 shows that when a company is willing to pay approximately €900 for a 1-point reduction in average pain (on a scale from 0 to 10) the probability of cost-effectiveness compared to usual care is 95%. Based on these results, the WS intervention was considered to be cost-effective in reducing average pain. Similar results were observed for current and worst pain. In contrast to the WS intervention, the WSPA intervention was not cost-effective in reducing average pain compared to usual care. The observed differences in costs and effects between the WSPA intervention and usual care were not statistically significant. For average pain, 92.5% of the cost–effect pairs were located on the right side of the y-axis (more effective) and 64% were above the x-axis (more expensive). The cost-effectiveness planes for current pain and worst pain reduction showed similar results.
Cost-effectiveness plane for the work style intervention versus usual care for average overall pain reduction at 12 months follow-up (upper figure). Acceptability curve of work style intervention versus usual care for average overall pain reduction (lower figure).
Neck/shoulder and arm/wrist/hand region
The WS intervention was cost-effective in improving recovery in the neck/shoulder region but not in improving recovery in the arm/wrist/hand region. When a company is willing to pay approximately €3000 for one recovered worker (neck/shoulder region), the probability that the work style intervention is cost-effective compared to usual care is 95%. The WSPA intervention was not cost-effective in improving recovery in either of the two body regions.
The WS intervention was cost-effective in reducing average, current and worst neck/shoulder pain between baseline and 12 months follow-up. The WS intervention was more effective than usual care, although no statistically significant differences in costs were observed. The acceptability curves showed that the probability of cost-effectiveness is 95% if companies are willing to pay approximately €600 for a 1-point reduction in average neck/shoulder pain (on a scale from 0 to 10). The WSPA intervention was more effective than usual care in reducing all pain outcomes in the neck/shoulder region, but the associated costs were slightly, but not significantly, higher. The probability of cost-effectiveness is 95% if companies are willing to pay approximately €2200 for a 1-point reduction in average neck/shoulder pain. Neither interventions was cost-effective in reducing any of the pain outcomes in the arm/wrist/hand region compared to usual care.
Sensitivity analyses
In the first sensitivity analysis, total costs were calculated using elasticity values of 0.3, 0.55 and 1.0 for absenteeism due to illness. When these elasticity values were used, no differences in costs were observed between groups, similar to the primary analyses.
The second sensitivity analysis consisted of a complete case analysis (with an elasticity value of 0.8), meaning that only the data of participants with complete cost and effect data were taken into account. Based on this complete case analysis, total costs were lower than in the primary analyses (with data imputation), being €1459 (SD €1819) in the WS group, €2789 (SD €6351) in the WSPA group, and €1544 (SD €2540) in the usual care group. Despite these lower costs in all study groups, total costs did not differ significantly between study groups. The effect on recovery was similar, as in the primary analyses. However, in contrast to the primary analyses, the effects on pain were smaller and no longer significant in the WS group compared with usual care (−0.44 for average pain; 95% CI −1.04 to 0.17).
Discussion
In the present study we evaluated the cost-effectiveness of the WS intervention and the WSPA intervention compared with usual care in computer workers with neck and upper limb symptoms from the employer's perspective. Neither intervention was cost-effective for improving recovery from neck and upper limb symptoms. In general, the WS intervention was more effective in reducing pain than usual care while costs were lower, although this difference in costs was not statistically significant and pain reductions were small. Despite these small changes in pain, the WS intervention was considered cost-effective in reducing pain while the WSPA intervention was not. The WSPA intervention was not more effective in reducing overall pain compared to usual care and cost differences were not statistically significant.
Interpretation of the results
The WS intervention was cost-effective for pain reduction, but a few issues need to be addressed concerning whether or not this intervention should be recommended for implementation. First of all, no significant effects on recovery were found and the observed pain reductions in the WS group compared with usual care were significant but small. According to the guidelines for minimal important change in pain and functional status for low back pain,35 observed pain reductions in the present study were less than 30% and therefore not clinically relevant. Based on these results, implementation of the WS intervention in its current form cannot be recommended. The results of two longitudinal studies36 37 might explain the small effect of our intervention. Both studies found no association between objectively recorded computer use and (the onset of) chronic neck and upper limb symptoms and identified only a few factors causing neck and upper limb symptoms. Lack of knowledge of the aetiology of neck and upper limb symptoms makes it difficult to develop (cost-)effective interventions for workers with neck and upper limb symptoms. Second, the pain reductions observed in the WSPA group were smaller than in the WS group, although both groups received the same work style intervention. The combination of improving work style and increasing physical activity behaviour may have caused a lack of focus resulting in smaller behavioural changes. However, this was not confirmed by our finding (published elsewhere) that similar effects on work style behaviour were established in the two groups.38 Pain reductions were observed together with positive changes in some of these work style parameters, which supports the proposed mechanism of pain reduction through work style improvement. Third, the lower costs in the WS group are primarily the result of a rise in the number of absence days (ie, 4.7 days) in the usual care group. This observed rise is remarkable because the usual care group received no intervention. It suggests that the observed reduction in absenteeism is much smaller than the normal variation in the usual care group. Finally, some absenteeism frequency parameters increased in the WS and WSPA groups. The causes are unknown but might be due to reorganisation in one of the companies.
Strength of this study
This is the first study to evaluate the cost-effectiveness of a work style intervention and a work style plus physical activity intervention to manage neck and upper limb symptoms in a working population. The strength of this economic evaluation is that it was conducted alongside a large randomised controlled trial with a follow-up of 1 year. Within a randomised controlled trial, data are collected prospectively and the risk of bias is limited. Another strength of the study is the fact that we collected cost data in the year prior to randomisation in order to adjust for possible baseline cost differences. In most cost-effectiveness studies only cost data in the year of intervention are collected.
Limitations of the study
The main limitation of the study is the high number of participants with missing effect data. Since the participants with missing data were on sick leave for twice as long as the participants without missing data, data imputation was necessary. Although data imputation causes uncertainty, multiple imputation accounts for this uncertainty. This method has been proven to generate valid statistical inferences.31 Second, absenteeism data were highly skewed resulting in large standard deviations. Since a few participants with numerous absence days were observed in all three study groups, we do not expect that this skewness has affected the outcomes of our study. A third limitation of the study arises from the fact that income data could not be provided from company records. Therefore, income data were estimated based on the mean income of the Dutch population according to age and gender. In economic evaluations from the employer's perspective, this method is not customary and often not desirable because the main goal is to obtain company specific results, and estimated costs based on mean income data are not equal to real costs. However, the goal of the present economic evaluation was to study the general cost-effectiveness of the WS and WSPA interventions. By studying the cost-effectiveness of our interventions in seven heterogeneous companies and by using mean income levels instead of company specific income levels, the results of our study are more easily translated to other companies. Finally, the subjective measures for recovery may have been affected by psychological factors. More objective instruments such as the Northwick Park Neck Pain questionnaire and the Neck Disability Index could not be used because these instruments have been developed specifically for patients with neck pain and not for computer workers with relatively mild neck and upper limb symptoms.
Recommendations for future research
Future research should investigate the determinants of pain reduction and recovery from neck and upper limb symptoms in order to develop (cost-)effective interventions. In addition, more high quality studies are needed to investigate the (cost-)effectiveness of these interventions. Cost-effectiveness studies should collect baseline cost data and adjust the cost-effectiveness analyses for baseline differences if necessary. Furthermore, future studies should examine the association between pain reduction and estimates of productivity (eg, work performance) in order to make cost-effectiveness studies easier to interpret. Finally, subgroup analyses could be conducted in order to investigate whether or not interventions should be targeted towards specific (risk) groups.
Conclusion
The WS intervention was not more effective than usual care in improving recovery. In general, the WS intervention was more effective in reducing pain than usual care and was associated with lower costs, although these cost differences were not statistically significant and pain reductions were not clinically relevant. The WS intervention was considered cost-effective if a company is willing to pay approximately €900 for a 1-point reduction in average pain. In general, the WSPA intervention was not more effective than usual care in improving recovery and reducing pain and was associated with slightly higher costs (not statistically significant).
References
Footnotes
Funding This study was funded by Body@Work Research Center on Physical Activity, Work and Health, TNO-VUmc, Amsterdam, The Netherlands.
Competing interests None.
Patient consent obtained.
Ethics approval This study was conducted with the approval of the VU University Medical Center, Amsterdam, The Netherlands.
Provenance and peer review Not commissioned; externally peer reviewed.