Sickness absence and early retirement cause a considerable deficit in the workforce and pose a major threat to productivity in all developed countries with increasingly ageing populations.1^–3 Concern over this threat has stimulated particular interest in finding new ways to reduce work disability. Traditionally, the measures taken have been targeted at people with diseases known to be leading causes of work disability, that is, musculoskeletal and mental disorders.4 5 Recently, emphasis has also been placed on identifying new target groups for the prevention of work disability.6^–10 More specifically, interventions have been directed at employees at risk of future work disability in addition to those who already have a disabling disease. The “employees at risk” group refers to people with no history of long-term sickness absences or severe illnesses but with an increased risk of future work disability due to harmful symptoms, detrimental health risk behaviours and shorter periods of work disability. When screening these employees occupational health care professionals use questionnaires and interviews to examine predetermined risk indicators, for example impairments due to musculoskeletal problems, symptoms of distress and mood disturbance, fatigue, excess alcohol consumption and weight problems. Repetitive short sick leaves are also considered a risk marker as an unhealthy lifestyle may be associated with short sick leaves.11

Preventive interventions for employees at risk are typically implemented in the workplace within the health care service and, in addition, at rehabilitation institutions.6 7 9 10 In Finland, the most important preventive intervention is the vocationally oriented multidisciplinary rehabilitation programme known as ASLAK, with approximately 15 000 participants every year.12 Although mainly preventive in nature, the intervention is implemented within the framework of the nationwide rehabilitation system. Participation is funded by the Social Insurance Institution (SII), the main provider of rehabilitation measures for the working population in Finland. Thus, the concept of rehabilitation is widened towards prevention8 13 and the rehabilitation institutions are accessible to new target groups.

Preventive interventions specifically targeted at employees at risk are considered a possible way of sustaining their ability to work and reducing the risk of sickness absence, and ultimately, disability pension among the working-age population. However, the existing evidence of the effectiveness of preventive interventions on temporary disability is scarce and inconsistent.6 9 10 Most of the studies of preventive interventions have focused on the health-related consequences of neck and back pain and mental disorders, and only limited evidence supports the effectiveness of such interventions in terms of preventing sickness absences.13^–22 Interventions are difficult to compare due to variations in their settings, target groups and contents. Furthermore, previous studies have been based on relatively small samples and, to date, no study has evaluated the effect of a preventive intervention on permanent disability.

To assess the effectiveness of an early intervention on work disability in a real life setting, we conducted a register-based observational study in a large cohort of local government employees. We examined the possible association of a vocationally oriented multidisciplinary intervention with subsequent sickness absences and the risk of permanent disability over an 8-year period. We compared the participants with matched controls from the same work organisations.

METHODS

Participants

This observational study is part of the ongoing 10-Town Study examining work-related determinants of health among full-time local government employees working in ten Finnish towns.23 For this study, we identified the eligible population from employers’ records, that is 67 106 full-time local government employees who had been employed for at least 10 months in any year between 1994 and 2002. Using personal identification codes (a unique number that all Finns receive at birth and is used for all contacts with the social welfare and health care systems) we linked employee data to records in national rehabilitation registers. According to data from the rehabilitation register kept by the SII, a total of 2409 employees from the eligible population participated in a vocationally oriented multidisciplinary intervention provided by the SII between 1994 and 2002. As in our earlier study,24 we excluded those who worked for <3 months during the time period from 3 years before intervention to 3 years after intervention (n = 34). We also excluded employees who underwent any other rehabilitation provided by the SII between 1994 and 2002 (n = 139). From the permanent employees not participating in any rehabilitation provided by the SII, we randomly selected four controls for each participant matched for sex, age (<40, 40–50, >50 years), occupational position (higher grade non-manual, lower grade non-manual or manual) and employer (Helsinki area, south-west Finland, middle Finland, north Finland). The controls were drawn from all employees and very few of them were likely to fulfil the criteria for participation in vocationally oriented multidisciplinary intervention. Thus, the participants, who all fulfilled these criteria, were more likely to represent a population at an elevated risk of future disability than the controls. The final cohort consisted of 2236 employees participating in the vocationally oriented multidisciplinary intervention, and 8944 controls.

The Ethics Committee of the Finnish Institute of Occupational Health approved the study.

Vocationally oriented multidisciplinary intervention

Vocationally oriented multidisciplinary intervention provided by the SII is targeted at middle-aged or younger employees at risk of future work disability due to mild symptoms, an adverse lifestyle or short-term sickness absences. The threshold for eligibility to participate in this intervention is kept low and only conditions such as a history of long-term sickness absence, current severe illness or work disability are categorical exclusion criteria. In this study, the main indications for participation were low back pain (40%), neck pain (29%), psychiatric symptoms and disorders (6%), joint symptoms (6%) and fibromyalgia (4%). The participants were selected in groups of 8–10 employees with similar occupations and often the same employer, for example high school teachers working in the service of the same municipality. The relevant rehabilitation institutions are required to collaborate with occupational health care personnel in order to assess work-related health risks for each group during the intervention and also on a special collaboration day.

According to the SII rehabilitation guidelines, the aim of the vocationally oriented intervention is “by means of a learning process to create a prerequisite for maintaining and promoting employees’ long term working ability”. The multidisciplinary intervention team includes a physician, a psychologist, a physiotherapist and a social worker. In some instances, the team may additionally include a nurse, a work instructor or a physical training instructor. The 1-year process includes three inpatient periods at a rehabilitation centre totalling 15–22 days.

During their stay at the rehabilitation centre the participants are supported towards adopting a healthier lifestyle. As most of the participants have musculoskeletal symptoms, as in the cohort of this study, personal guidance to improve aerobic capacity and muscle endurance is provided. They also participate in group meetings and lessons dealing with the self-management of stress, adopting healthy dietary habits and controlling physical strain at work. The content of the intervention is work site specific and the rehabilitation centres collaborate with occupational health personnel in planning the programme. Given the differences in work related strains between workplaces and jobs, the emphasis on work related issues is variable. For instance, in the intervention for high school teachers the emphasis is on psychological education, while an intervention for construction workers emphasises physical fitness and ergonomics. During the intervention the purpose of information is to help the participants realise that they are at risk of health problems and deteriorating capabilities. Guidance and training aim at adopting active self-care. The intervention has a busy atmosphere with a scheduled 6–8 h programme each day.

The selection of participants is based on guidelines provided by the SII. Occupational health personnel are advised to use questionnaires and interviews during health checks to determine adverse lifestyles, symptoms and impairments related to the risk of future disability, as well as information on recurring short sick leaves. When the criteria for participation in a vocationally oriented multidisciplinary intervention are met, final acceptance takes place at the local SII offices and is based on a medical statement from occupational health care staff.

Assessment of sickness absence

We obtained data on all sickness absences in 1991–2006 from records kept by employers. The procedure for recording sick leaves within the public sector in Finland is considered reliable.25 A medical certificate is required for any absences longer than 3 days. The indices of sick leaves used in this study included the annual rates of short-term (⩽3 days) and long-term (>3 days) absences. We also examined sick leaves lasting >21 days, because long-standing illnesses show a stronger association with longer durations of sickness absence.26 To assess the rate of sickness absence, we divided the annual number of sickness absence episodes by days at risk in a given year. The number of contracted days represented the days at risk, from which the number of days absent from work for reasons other than sickness was subtracted (sick leaves are not recorded during these subtracted periods). The mean duration of days at risk was 7.1 years (SD 1.4) for participants and 6.7 years (SD 1.6) for controls.

We determined all sickness absences during an 8-year period, covering 2 years prior to the intervention (pre-years), the intervention year (actually the year when the intervention process started), and 5 subsequent years (post-years). For the controls, sickness absences were linked to the data in the same manner as for the participants, with pre- and post-years based on a randomly assigned year. Because data on absence were obtained from employers’ records, we were able to measure sickness absence only for those years the participant was in the service of the participating organisations. The proportion of actualised follow-up from all potential follow-up of sickness absences was 93.5% (10 450/11 180 years) for participants and 90.0% (40 261/44 720 years) for controls.

Assessment of disability pension

We collected disability pension data from the register maintained by the Finnish Centre of Pensions which provides virtually complete population pension data. The dates and causes (from medical certificates) of early retirement due to disability pension were obtained for all those participants and controls who had been granted a permanent (full-time or partial) or fixed-term disability pension between 1 January 1994 and 31 December 2006. In Finland a disability pension may be granted after 300 reimbursed sickness absence days (Sundays excluded).

Assessment of baseline health status

Measures of baseline health status included the prevalence of severe and/or chronic diseases and the purchases of prescribed drugs in the intervention year. We obtained data on eligibility for special medication reimbursement due to chronic or severe disease from the SII prescription register. In Finland, the national sickness insurance scheme provides special reimbursement of up to 72% or 100% for medication required for many chronic and severe diseases, such as hypertension, asthma, coronary heart disease, cancer, diabetes and epilepsy. We also collected data on lifetime history of cancer from the Finnish Cancer Registry, which provides data on all cancer cases diagnosed in Finland since 1953.27

To assess the consumption of analgesics and antidepressants, we obtained data from the SII prescription register. We selected these medications because they are used to treat musculoskeletal and mental disorders which are the leading causes of long-term sickness absences, and because studies show an association between prescribed medication and perceived severity of symptoms.28^–30 The national sickness insurance scheme covers the entire population and provides reimbursement for all filled outpatient prescriptions classified in the register according to the World Health Organization Anatomical Therapeutic Chemical (ATC) classification code.31 The prescription register contains data on the exact dates and defined daily doses (DDDs) of every purchase of prescription drugs, which can therefore be used to estimate annual drug consumption. The DDD is the assumed average maintenance dose per day for a drug used for its main indication in adults. From the prescription register we measured the DDDs of the purchases of prescribed analgesics (ATC codes N02 and M01A) and antidepressants (ATC code N06A) for participants and controls in the intervention year. The annual purchases for these drugs were divided into three categories: none, <100 DDDs and ⩾100 DDDs.

Statistical analysis

Baseline differences between the participants and controls in prevalent diseases and purchases of prescribed analgesics and antidepressants were studied with the χ² test. As in our earlier study,24 we examined the association between the intervention and sickness absence using information on repeated observations of sickness absence, and analysed these data with repeated measures Poisson regression analysis with the generalised estimating equations (GEE) method.32 The regression models for sickness absence spells were based on Poisson distribution and those for sickness absence days on negative binomial distribution. We calculated the numbers of annual sick leave days and the rates of annual sickness absences of varying duration (short <3 days, all long >3 days and very long >21 days) and their 95% confidence intervals (95% CI) for each of the eight time periods (2 years before the intervention, the intervention year and 5 years after the intervention). We tested curvilinear trends in sickness absence between the participants and controls by applying a Poisson model with an interaction term (time²×intervention) and treating time as a continuous variable in the analyses.

We used Cox proportional hazard models to examine the associations of the baseline characteristics and the intervention with disability pensions granted, and calculated the corresponding hazard ratios (HR) with 95% CI. The follow-up period for disability pensioning was calculated from 1 January following the year of intervention and the participants were censored on the date they were granted a disability pension, reached the age of 63 years (official retirement age) or died (23 participants and 132 controls, p = 0.106), or on 31 December of the 6th year of follow-up or on 31 December 2006, whichever came first.

Regarding the association between the intervention and risk of disability pension we found evidence against the validity of the proportional hazard assumption. The time-dependent interaction term between participation in intervention and the follow-up period was significant (p<0.001), confirming that the proportional hazard assumption was not justified. We calculated a Kaplan–Meier hazard curve to illustrate the association of disability pensioning for participants and controls and estimated the difference in the cumulative disability years between the two groups over a period of 7 years after the intervention. Because of non-proportionality of the hazard, we illustrated the time-dependent association by estimating the hazard ratio (HR with 95% CI) of disability pension of participants compared with controls for the first half of the follow-up.

Finally, we studied the differences between those participants and controls who were lost to follow-up, by comparing the total sick leave days during the first 4 years of the follow-up and the numbers of deaths between the two groups.

All analyses were performed with SAS 9.1.3 statistical software (SAS Institute, Cary, North Carolina).

RESULTS

The mean age of participants was 47.3 (SD 5.7, range 28–59) years at baseline, 74% were women, and 33% worked in higher grade non-manual occupations, 38% in lower grade non-manual occupations and 29% in manual occupations. As shown in table 1, baseline health status showed only minor differences between participants and controls. Nearly 18% of both participants and controls had a chronic disease. No differences were observed between the two groups in the prevalence of diabetes, hypertension, asthma or other severe disease as indicated by eligibility for special medication reimbursement, or in the lifetime history of cancer. However, in relative terms the prevalence of coronary heart disease was higher among controls (1.0%) than participants (0.5%) (p = 0.02); in absolute terms, however, the difference between the groups was marginal (0.5%). The main differences between participants and controls at baseline were in purchases of prescribed analgesics and antidepressants, with the participants having more purchases than the controls (p<0.001). In the intervention year, the mean DDDs of purchased analgesics in these two groups were 13.4 (95% CI 12.0 to 14.8) and 10.8 (10.0 to 11.6), respectively. For antidepressants, the corresponding figures were 8.7 (95% CI 7.0 to 10.8) and 7.7 (95% CI 6.6 to 8.9).

Table 1 also shows, as expected, that all of the baseline health measures were associated with an increased risk of disability pension. For example, manual workers were 2.4 (95% CI 1.9 to 3.1) times more likely to be granted a disability pension than higher grade non-manual workers, with no evidence of gender differences in this association (test for gender interaction p = 0.43). The hazard of permanent work disability for any severe or chronic disease with eligibility for special reimbursement was 3.2-fold (2.7 to 3.8) compared to those with no such reimbursement, and correspondingly, the hazards for purchases of prescribed analgesics and antidepressants in ⩾100 annual DDDs were 4.3- and 4.7-fold (3.2 to 6.1), respectively.

Sickness absence

As shown in table 2, there was a U-shaped association in the annual number of sickness absence days among the participants in intervention compared to controls over the 8-year period studied (p for curvilinear trend p<0.001). A similar pattern was observed for the number of very long (>21 days) spells of sickness absence (p = 0.03) but not for all long spells or for short spells. In the years preceding the intervention, the rates of absence days and very long spells were 1.2 (95% CI 1.0 to 1.4) times higher among participants than among matched controls. However, in the year of intervention and the 3 subsequent years the rates of sickness absence days and very long spells reduced to the same level as observed among controls. Later on, in the 4th and 5th years of the follow-up period, these rates again increased to the pre-intervention level. In absolute terms, the temporary decline in sickness absence totalled 6.7 absence days less per participant than would be expected assuming that the rate ratio between participants and controls was constant over the whole study period. Regarding all long sickness absence spells, the rate among participants compared with controls was 1.2–1.3 times higher (95% CI 1.1 to 1.3) in all years of follow-up, with no indication of a curvilinear trend (p = 0.43). For short spells, a constantly higher rate was similarly observed among participants, the rate ratio being 1.2–1.3 (95% CI 1.1 to 1.4).

Disability pensions

We identified 554 cases of new disability pension, 113 (5%) among participants of the intervention and 441 (5%) among controls, after the year of intervention. For the disability pensions granted to participants, the main diagnoses were musculoskeletal disorders (42% of granted pensions), psychiatric disorders (33%) and cardiovascular diseases (8%). The figures were almost the same for the controls (41%, 37% and 7%, respectively). As can be seen from the Kaplan–Meier cumulative hazard curve, the association of disability pension incidence for participants was non-proportional (fig 1). Compared with matched controls, the hazard among participants was lower in the first half of the follow-up period, but during the latter half it gradually increased to the same level. The number of cumulative disability pension years among participants was 56.4 years lower than among controls per 1000 employees during the 7 post-intervention years. As shown in table 3, in the first half of the 7-year follow-up period, the hazard of work disability among participants was only half of that among matched controls after adjusting for demographics, cancer history, special medication reimbursements for any chronic or severe disease, and purchases of prescribed analgesics and antidepressants.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Cumulative hazard function for disability pension among participants in the intervention and matched controls in the 7 years after the intervention.

DISCUSSION

This 8-year follow-up study of 2236 participants in an early intervention and 8944 matched controls suggests that the risk of work disability can be reduced in employees at risk of future work disability by means of preventive vocationally oriented multidisciplinary intervention. A positive effect of the intervention was consistently observed on several outcomes: days lost due to sickness absence, rate of very long sick leaves, and years lost due to disability pensioning. However, the effect of intervention was restricted to the 3–4 years immediately following the intervention.

To our knowledge, this is the first controlled large-scale study on the effectiveness of a vocationally oriented multidisciplinary intervention carried out in a real life situation. The strengths of our study are the use of register-based reliable and pertinent data on a large cohort and the long follow-up period. However, the use of register-based data is also a limitation because information on only a limited set of known prognostic factors was available for the intervention and control groups. In an observational study such as ours, important prognostic factors could be unequally distributed and only partially explain our findings.33 The main limitation is the lack of a randomisation mechanism in the allocation of groups to intervention and reference groups leaving our design open to confounding and bias. For the purpose of assessing the effectiveness of an intervention, the randomised clinical trial is the gold standard. Although we were able to carefully match participants with controls by demographics, health selection itself might be a bias. If the probability of being allocated to the intervention group or to the control group is potentially affected by health-related factors, the observed differences in relation to the outcome variables may result from selection bias. In our study, this would mean that the observed reduction in work disability might be due to the fact that the participants were healthier than the controls. For several reasons, however, we feel that such a bias is an unlikely explanation of our findings.

First, we used highly reliable national registers to measure the prevalence of severe and/or chronic diseases in the year of rehabilitation. The Finnish sickness insurance scheme provides special reimbursement for many longstanding and severe diseases requiring constant medication. For each disease category, strict medical criteria are set for the granting of special reimbursement. If selection bias were operative, one would expect a lower prevalence of reimbursements among the participants than the controls. However, the proportion of individuals eligible for special medication reimbursements was the same in both groups. No significant differences were observed for the most common chronic diseases, such as hypertension, diabetes or asthma, nor for the lifetime history of diagnosed cancer. The only disease category more common among controls than participants was coronary heart disease. However, when we excluded all cases of coronary heart disease from the analyses, the results remained unchanged (ie, the hazard ratio for disability pension among participants compared to controls was 0.55, 95% CI 0.37 to −0.81).

Second, at baseline participants purchased more prescribed analgesics and antidepressants than controls, although exactly the opposite would be expected if selection bias were to explain our findings. These purchases not only reflect the level of common symptoms, such as pain and mood disturbances, but may also be a marker of more severe underlying disease. Interestingly, the purchases of these prescribed drugs were strong predictors of disability pension. This finding is in line with earlier studies reporting an association between prescribed medication and perceived severity of symptoms.28^–30

Third, earlier studies have shown medically certified sickness absence to be a more powerful predictor of all-cause mortality in comparison with established self-reported health measures and available objective measures of specific physical illnesses and medical conditions.34 Furthermore, medically certified sickness absence is shown to strongly predict specific causes of death, such as cardiovascular disease, cancer, alcohol-related causes and suicide.25 Sickness absence is also a risk marker for future disability retirement.35 36 In our study, the rate of such absences was higher among participants than controls in all years of the follow-up period. This provides further evidence against the assumption that participants would have been healthier than controls.

Fourth, the time-dependent effect pattern supports a causal interpretation of the association between the intervention and work disability. If health selection had operated in such a way that participants were healthier, one would expect to see a constantly lower proportional hazard in the risk of disability pensioning. However, the hazard proved to be non-proportional, with a better prognosis concerning work disability being restricted to only the first 3–4 years following the intervention.

Fifth, differences in lost to follow-up or job changes between participants and controls during follow-up is a potential source of bias. Although data on disability pensions were collected from national registers providing virtually complete population pension data, data on sickness absence were derived from employers’ registers. However, the proportion of non-missing years of all years of potential follow-up of sickness absences was 93.5% (730/11 180 years missing) among participants and 90.0% (4459/44 720 years missing) among controls. A comparison of sick leave days showed no difference between participants and controls (RR 0.90; 95% CI 0.77 to 1.06) in the years −2 to 1 in relation to the year of intervention among the employees lost to follow-up after year 1. Moreover, compared to their baseline job, 2.4% of participants changed to a lower level job and 3.7% to a higher level job during follow-up, while 93.9% remained in the same job category. For the controls these proportions were almost the same (2.8%, 4.1% and 93.1%, respectively, χ² test p = 0.40). Deaths occurred among 1.0% of participants and 1.5% of controls. This suggests that employees who were lost to follow-up after the intervention did not severely bias the results obtained.

Taken together, the employees selected to participate in the preventive intervention seemed to be “at risk” according to the indication for vocationally oriented multidisciplinary intervention provided by the SII. Compared with controls, the participants had no excess in severe illnesses but were likely to have higher levels of symptoms related to, for example, musculoskeletal problems and mood disturbances, as indicated by a greater number of prescribed medication purchases. Furthermore, we found an increased risk of disability pension among relatively healthy employees who were prescribed analgesics or antidepressants, suggesting that the intervention group was indeed at a higher risk of work disability in the future than the controls.

Comparison with previous studies

Earlier Finnish studies have provided support for the effectiveness of vocationally oriented multidisciplinary intervention in various occupational groups in terms of favourable changes in work techniques.37^–40 Also, a short-term improvement in perceived health and work ability has been observed,38^–41 but the effects on sickness absence have been inconsistent. Two non-controlled studies found no effects39 41 and one study with a control group reported a decrease of 2 days in sickness absence among the participants in the intervention during the 1.5-year follow-up. Our findings are in accordance with the latter investigation, with the exception that our follow-up period was much longer. International comparisons would be preferable, but they are not easy to perform due to variations in the content and implementation of interventions in different countries. A Dutch trial has evaluated the effectiveness of an occupational health intervention programme among workers at risk for early retirement.42 The participants were older than 50 years and assumed that they would not be able to work up to their retirement. The intervention programme lasted 6 months, was carried out by an occupational physician and comprised at least three consultations. As a result, fewer employees retired early in the intervention group than in the control group, and the total average number of sick leave days within 2 years was lower in the intervention group than in the control group. In agreement with our results, no differences between the two groups were found 2 years after the randomised intervention.

In the majority of previous studies the preventive interventions have been single-modal programmes. Positive results as regards sickness absences have been reported in interventions implemented with a multidisciplinary programme for employees suffering from sub-acute low back pain.43

The estimated number of 6.7 saved sick leave days per employee in 4 years among the participants compared to the controls in our study was substantially less than the result in a Norwegian randomised clinical trial, where the difference in sick leave days was 31.4 in favour of the intervention participants during a follow-up of 3 years.19 However, the results of the studies may not be comparable, since the participants in the Norwegian study were on sick leave at the time of the intervention. A recent Finnish randomised controlled study on a pragmatic occupational health care personnel intervention targeted at employees at risk of future work disability suggested an effect of 11 sickness absence days less per employee per year when compared with controls receiving usual care.44

An interesting result in our study was the decreased rate of very long sickness absences. This is in line with a Dutch controlled study which explored the effects of a back pain intervention (individually tailored back school) on sick leaves during a follow-up period of 2 years.45 Indeed, lengthening of sickness absences caused by musculoskeletal disorders is known to lead, within months, to a high risk of permanent work disability.46

The effects of the intervention disappeared in 4 years. This result is in line with previous studies. Interventions that run longer than 1 year or employ new methods might be more effective in preventing work disability among employees at risk, but strong evidence to support this is still lacking.

Economic implications

Although our aim was not to explore the cost-effectiveness of the vocationally oriented multidisciplinary intervention, it may be relevant to briefly address the issue in light of the high costs of this particular intervention regime for the SII in Finland (€30 million in 2005).12 Previously, the societal costs of this intervention have been estimated at €3754 per participant in 1998.40 In Finland, the societal cost of 1 day of sickness absence is estimated to average €200 and the annual costs of an average disability pension €12 000. Given the estimated effect of 6.7 sickness days and 0.056 disability pension years saved per participant, as calculated in this study, the decrease in sickness absence and pension costs alone would cover about 54% of the cost of the vocationally oriented multidisciplinary intervention.

Previous studies have supported the cost-effectiveness of interventions for sub-acute low back pain in an outpatient setting.19 21 46 47 For example, in a Norwegian study19 the cost of an intervention was $303 (€220) per participant, and in a British study48 it was £917 (€1356), that is, considerably less than in our study. In Norway, the intervention was implemented by a physician and a physiotherapist, and in Britain by a physiotherapist in conjunction with work personnel. The main causes of the higher costs for the Finnish intervention are apparently attributed to the inpatient regime and multidisciplinary content. Nevertheless, a multidisciplinary approach in preventive measures has been called for7 and supported.49 In a Swedish randomised controlled trial, the addition of cognitive-behavioural content alone or cognitive-behavioural content and preventive physiotherapy together to a purely medical examination resulted in a 5-fold decrease in sickness absence longer than 15 days among employees with back and neck pain during a follow-up period of 1 year.49

CONCLUSIONS

Our results suggest that a vocationally oriented multidisciplinary intervention targeted at middle-aged employees at risk of future work disability reduces sickness absence and the likelihood of disability pensions over a period of 3–4 years. The fact that our cohort was from the public sector means that the findings may not be generalisable to the Finnish working population or to the private sector. Future research with a randomised controlled study design and comparison with a less expensive outpatient programme is needed to further explore the long-term benefits and cost-effectiveness of the vocationally oriented multidisciplinary intervention.