Using statistical tools for multistate models in the analysis of official register data on different sick leave benefits made it possible to describe the composite process of RTW in detail. During a 4-year follow-up after a group of patients had finished a rehabilitation program, the average number of transitions between different benefits and work was 3.7 times. The maximum number of transitions for one individual was 18. This shows that the RTW process for people on long-term sick leave benefits, with non-specific musculoskeletal and mental health problems, may be long and complex. In addition to the high number of transitions between different benefits, the proportion of individuals that returned to full-time work, i.e. no registered benefit, increased from about 10% at departure from the rehabilitation clinic to above 50% throughout the follow-up period. It seem to be a trend that those being on partial sick leave at departure from the rehabilitation clinic had higher probability for working full-time during the first years of follow-up, than those being on full-time sick leave.
Our findings are in line with other studies showing that RTW can be a long and resource consuming process [19, 25]. It also emphasises the need for analyses of multiple, longitudinal and repeated events for a better understanding of the RTW process [3, 5, 26]. RTW is not an absorbing state and cannot be measured at a single point of time [25, 26]. Multi-state models, as used in our study, can be a useful method to analyze longitudinal RTW data. However there is no clear agreement about how long of a follow-up period is needed to get the best measurement of the effect on work and benefits, after sick leave and work-related interventions [27, 28]. Our data indicate that several years are needed to get an adequate picture of the RTW outcome. Differences in design, populations and follow-up period in RTW studies make it difficult to compare results between studies [1, 27, 28].
Our data also demonstrates the importance of including information on partial sick leave benefits and part-time work, when evaluating RTW after rehabilitation programmes . This is in accordance with recommendations in research on sick leave [1, 5, 19]. For individuals on long-term sick leave, with limited ability to function, working part-time may be a sufficient goal after rehabilitation, especially when the alternative would be full-time disability pension. Since work ability is not a fixed either/or category, but changes over time, and with varying circumstances, working part-time may also play a role in transitions to full-time employment . It may therefore be important to have a flexible working life, giving opportunities for part-time work for workers with temporary low work ability .
We found an increased probability for working after receiving either medical or vocational rehabilitation allowances, at the time of departure from a rehabilitation clinic. This finding is in accordance with the idea behind the legal regulations for sick leave benefits in Norway. Medical rehabilitation allowances should only be granted to individuals with treatable medical conditions, whereas vocational rehabilitation allowance is restricted to individuals with an expected likelihood for RTW after e.g. work training or professional re-education. After the first year on sick leave benefits, medical and vocational rehabilitation allowances constitute a secure source of income for long-term sick workers in Norway. These allowances are equivalent to the sickness benefits system in Sweden before 2010, where sick leave benefits could be expanded beyond the first year in certain medical cases .
However receiving such benefits has been claimed to be a step toward disability pension, indicating that RTW is difficult to achieve after more than one year out of work. In Sweden the sick leave track record was found to be the most important predictor of the probability of being granted a disability pension [31, 32]. The probability of being granted disability pension following sick leave often begins with short term sick leave periods followed by episodes increasing in length, and with shorter and shorter intervals between them . Few studies have explored outcomes after such allowances in Norway, but Landstad and co-authors  found that only 27% returned to work after receiving medical rehabilitation allowance. RTW was, among other factors, associated with prior work life affiliation, contact and satisfaction with the work place during rehabilitation, and experiencing a high degree of influence over their own rehabilitation process . Even though this was not an intervention study, the findings may have relevance in the interpretation of the optimistic RTW outcomes in our study. The rehabilitation program emphasized client participation and contact with the workplace, and the final part of the intervention was to agree on a RTW plan. Workplace engagement, in terms of coordination and cooperation between client, employer and different stakeholders involved in the RTW process is stated to be crucial for success in re-integrating individuals on sick leave in working life [3, 20, 34–36]. Offering adequate rehabilitation efforts may alter a negative trend out of working life when it is given to the right type of patients.
Furthermore we found an increase in medical and vocational rehabilitation allowance and a simultaneously decrease in sick leave benefit during the first months after the rehabilitation program. This may be interpreted as an effect of the social security legislation on time limits for sick leave benefit to maximum one year continuously. As the participants on average already had been out of work for 9 months before the rehabilitation stay, some were about to cross the time limit at departure from the clinic. These legislations, and the decrease in social security disbursement to 66% of the salary after one year, may also be an important incentive to RTW.
The current study adds to previous literature by a detailed analysis exploring the RTW process and pathways back to and out of work after work-related rehabilitation. To our knowledge, a multistate model on sick leave data has not earlier included this amount of possible outcomes. Lie et al. (2008) restricted their analysis to three different states, in their follow-up of low back pain patients in Norway . Analysis on register data in the Nordic countries has been restricted to sick leave benefits and disability pension  and to unemployment benefits .
The use of social security benefits and RTW is a frequent issue in public health research, irrespective of nationality [31, 37]. Comparison between studies may be difficult due to the large variation between countries in the regulation of sick leave compensations and the granting of disability pension . Within the Nordic countries the social security system, representing The Nordic welfare model, is more similar, and makes comparison between the Nordic countries more suitable [5, 27, 37, 38]. These countries are facing the same challenges towards undesirable high rates of sick leave and disability pension . The countries are also somewhat similar in the access to official registers, making it possible to follow individuals on sick leave through the social security system . Our findings may therefore be generalized to the Nordic countries.
Despite the strength of using register data in the long term follow-up of after rehabilitation, there are some limitations in this study. Register data from The National Labour and Welfare in Norway contains spare information on whether a person is actually working or not. Hence we assumed that individuals were working in the time gaps between dates of benefits in the register. This is believed to be a correct interpretation of such registers since people that support themselves without a job usually have no legal rights to receive sick leave benefits . Based on our analysis, we believe that the definition of work, being the time gaps in the register files, is valid. It was unusually many transitions from work directly to disability pension. This may be explained by the delays in the case procedures within The National Labour and Welfare system. Disability pension claimants may be without registered benefits in a short period while waiting for a decision. Data on unemployment benefits would have strengthened the study further. However the number of people on unemployment benefits in Norway is low.