*3.2. Logistic Regression Models*

The adjusted binary logistic regression analysis showed that the MDT group had an OR of 4.62 (95% CI 2.27–9.41) for RTW or increased employability. The corresponding number for the ACT group was OR 2.35 (95% CI 1.07–5.19), as shown in Table 3.


**Table 3.** Binary logistic regressions presenting odds ratios (OR) of reporting return to work or increased employability vs. negative or unchanged outcomes.

Odds ratio (OR), 95% CI: 95% confidence interval. \* *p* < 0.05, \*\* *p* < 0.01. aMultidisciplinary treatment. bAcceptance and Commitment Therapy. cHospital Anxiety and Depression Scale, ranging from 0 to 21. Model 1 = Intervention group + age + education level, Model 2 = Model 1 + HADS + Self-efficacy + Employment contract + Extent of sick leave + Years with income replacement.

#### *3.3. Multinomial Regression Models*

The adjusted multinomial logistic regression analysis showed that the multidisciplinary team group had an odds ratio (OR) of 3.31 (95% CI 1.39–7.87) for RTW, an OR of 4.24 (95% CI 1.60–11.26) for increased employability, and an OR of 0.19 (95% CI 0.05–0.72) on negative outcome. The adjusted multinomial logistic regression analysis showed that the ACT intervention group had an OR of 3.22 (95% CI 1.13–9.15) on increased employability, but no significant effect on RTW. See Table 4.

Due to a large number of internal missing values in the outcome measure (34.0%), a multinomial logistic regression analysis was also performed, in which non-responding participants were assumed to have made no change (reference category) in outcome. Similar results, despite somewhat lowered effects, showed that the multidisciplinary team group had an adjusted OR of 2.57 (95% CI 1.19–5.58) on RTW; an adjusted OR of 2.80 (95% CI 1.20–6.52) on increased employability; and an adjusted OR of 0.18 (95% CI 0.05–0.62) on negative outcome. The ACT intervention group had an adjusted OR of 2.79 (95% CI 1.12–6.97) on increased employability.


**Table 4.** Results of multinomial logistic regression of intervention group's effect on negative outcome, return to work, or increased employability.

Reference outcome category: No change. For the adjusted model, Nagelkerke r<sup>2</sup> = 37.8%. \* *p* < 0.05, \*\* *p* < 0.01. aMultidisciplinary treatment. bAcceptance and Commitment Therapy. cHospital Anxiety and Depression Scale, ranging from 0 to 21. Odds ratio (OR), significant level and confidence interval (CI) for having made RTW or (positive) system position change or negative change.

#### **4. Discussion**

This study aimed to investigate the effects on RTW and increased employability of two vocational rehabilitation interventions in patients on long-term sick leave due to mental illness and/or chronic pain. In this study, a non-dichotomous outcome was used to capture not only RTW, but also changes in increased or decreased employability. The results indicate that multidisciplinary interventions and individualized treatments may increase RTW and employability in patients on long-term sick leave due to common mental illness and/or chronic pain. The results in this study are similar to results presented in other studies regarding RTW [18–23]. Also, in review studies, this type of multi-domain intervention seems to be the most effective intervention for RTW outcome when managing musculoskeletal and pain-related conditions [10]. The results also suggest that multidisciplinary team and ACT interventions may increase employability, i.e., transferring individuals in the insurance system to increased availability for employment and work.

One reason for removing the separate ACT group in the second phase of the project was that the intervention providers discovered that some people who were randomized to ACT were not keen on receiving psychological therapy [24]; this might also explain the somewhat weaker outcome of the ACT group.

At the time when the study was conducted, the participants were about to lose their sickness insurance benefits due to a major change in the social insurance system. These circumstances may have led to the inclusion of participants who would not otherwise have been motivated to participate in vocational rehabilitation, thus creating a study population with relatively low motivation and belief in their own ability to RTW. Since longer periods of sick leave and absence from the labor market are known risk factors for RTW failure [6,7], this project was believed to target a difficult

group. Taking all of the factors into account, including the interventions, the multinomial regression explained 37.8% (r2) of the outcome, which leaves a 62.2% unexplained variance that is dependent on factors that were not assessed in the model. This indicates that there are other factors outside the intervention program that affect the outcomes. For instance, this study did not include any work-directed modifications or interventions, which is seen as a factor to increase the chances of RTW goals [25]. The causal chain between sick leave, mental illness, and chronic pain is not fully explored. Sick leave is warranted by an inability to work, but sick leave may in itself also contribute to depression, unhealthy living, and stress [26]. Previous research has identified several important factors for RTW [8]. Current demands in the labor market are important for re-entry to the workforce, including factors on a systemic/organizational level [27]. Many of these factors are not affected by health and rehabilitation interventions such as this project. Instead, several of these factors are found "upstream" in the analogy of a "river", which is used to describe how previous social, financial, environmental, and historic factors ultimately go on to profoundly influence present outcomes [28]. In the present study, this implies that even if a participant in the project reduces their ill health and increases employability, these results might not be captured in a sole RTW outcome; thus, RTW is (also) dependent on upstream factors that are determined outside the reach of the project. Especially when considering the long-term sick leave period among the project participants (on average 7.7 years), it is reasonable to assume that the labor market has changed in several aspects during their absence, such as work content and demands etc. Therefore, an outcome measure was constructed that also captures changes in different transition stages made by program participants, such as increased employability. This does not apply to all of the participants; several had part-time jobs.

In Sweden, as well as in many other countries, concern and efforts provided by society for people on sick leave are regulated by different laws and principals and divided among different authorities. There have been doubts about whether a multi-actor model is optimal, and limitations have been found in the system. This project included several authorities and professions with one collaborative goal: to support the participants in their vocational rehabilitation and increase their likelihood for RTW. This project was also a collaborative challenge, as different professions in the MDT worked closely with one goal. It should be noted that the MDT and ACT groups received about the same total number of sessions in the project. The ACT group only received sessions with an ACT psychologist; the MDT group received on average about half of the sessions from an ACT psychologist and half from the other professions.

The major findings in this study imply that a multidisciplinary team-based intervention directed at people on long-term sick leave, including ACT counseling, seems to help people, even those on long-term sick leave, to RTW. The findings add to the evidence that multidisciplinary interventions such as vocational rehabilitation may increase RTW among patients with mental illness and/or chronic pain [29].

There is a need to further investigate multidisciplinary RTW interventions, determine their core components, and consider combining them with workplace collaboration, including the employer and interventions in the work environment.

#### *Strengths and Limitations*

The strengths of this study include the randomized prospective controlled design and the experimental design, which suggest that the effects on RTW are in fact effects of the interventions.

This study also has some limitations. Reliable evidence for RTW should be assessed using information on validated employment and work activity. In this study, the outcome was based on self-reported data and actual RTW was not measured; instead, a changed proportion of income source was used as an indicator for changes in working status and employability. However, similar studies often use proxy variables for measuring RTW [23]. Since the data—both exposures and outcomes—are self-reported, there is a risk of recall bias. Another limitation of this study is that education level, HADS, and self-efficacy were measured after randomization, so the scores could have already been

influenced by the knowledge of intervention group affiliation. The results were not adjusted by type of disease (mental illness or chronic pain).

Many participants did not show up, dropped out, or did not answer the follow-up questionnaires, which weakens the assessment of the outcomes. There could be different reasons for the high attrition: perhaps the participants did not actually want to be part of the project, were too sick to participate, or felt they had a problem that would not benefit from the project.

Since participants in the treatment groups received individual treatment as well as organizational collaboration, this raises the question of which of these interventions mediated the effect, if either. The reason to include organizational collaboration was to set and mutually agree on each individual's RTW goal. Another potential problem with this study is that the two types of multidisciplinary team interventions that were combined were not identical, but they were similar.
