Sickness absence in a re-employment program as a predictor of labor market attachment among long-term unemployed individuals: A 6-year cohort study in Finland

This paper investigates whether sickness absence during participation in a subsidized re-employment program in Finland influenced future employability of long-term unemployed people. Findings suggested an increased risk of poor labor market attachment among unemployed people with sickness absence, especially among young adults with >30 days of sickness absence. Sick unemployed people may require extra support in terms of healthcare and rehabilitation. Sickness absence in a re-employment program as a predictor of labor market attachment among long-term unemployed individuals: A 6-year cohort study in Finland. Objectives We examined whether sickness absence during participation in a state subsidized re-employment program among long-term unemployed people was associated with subsequent labor market attachment. Methods We linked 18 944 long-term unemployed participants (aged 18–60 years) of a six-month subsidized re-employment program in Finland to their records of sickness absence during the program and labor market status after the program. We used the latent class growth model to identify labor market attachment trajectories over a six-year follow-up period and multinomial logistic regression to investigate the association between sickness absence and labor market attachment trajectories. Results We identified four labor market attachment trajectories: "strengthening", (77%), "delayed" (6%), "leav-ers" (10%), and "non-attached" (7%). Sickness absence was associated with an increased risk of belonging to the leavers and non-attached trajectories. Having >30 days of sickness absence during the six-month re-employment program increased the risk for belonging to the future non-attached trajectory in all age groups, but in particular for those aged 30–44 [odds ratio (OR) 7.35, 95% confidence interval (CI) 4.85–11.14] and 18–29 years (OR 5.38, 95% CI 3.76–7.69). At these ages, having fewer than 30 days sickness absences was also associated with an elevated risk of belonging to the non-attached trajectory, while this risk was lower for those aged 45–60. Conclusions Sickness absence during participation in a subsidized re-employment program increased the risk for poor labor market attachment during the subsequent six years. The risk was particularly high among younger participants with >30 days of sickness absence.

There is a large body of research on the effects of unemployment on health. There is also a growing number of studies suggesting improved health status among reemployed individuals compared to those who remain unemployed (1)(2)(3)(4)(5). Much of the concern on how to improve the re-employment prospects of unemployed people is associated with concern about how to prevent health deterioration resulting from long-term unemployment (6,7). In attempting to address this concern, labor authorities in several European countries have instituted active labor market policy measures such as job training, subsidized re-employment programs, and re-education courses, but the effectiveness of these measures has remained controversial (8)(9)(10). Hence policy-makers need to review and develop the current measures and seek for means that would contribute to the realization of the goal of improving re-employment.
A better understanding of the factors that constitute barriers to re-employment is crucial when planning preventive interventions aimed at promoting re-employ-ment. Earlier studies have identified self-perceived general poor health (11)(12)(13)(14) and musculoskeletal pain (15) as factors that may reduce the likelihood of reemployment. Research also suggests that unemployed people with physician-diagnosed mental problems (6,16,17) are at greater risk of not re-entering the labor market than their counterparts with no such health problems. However, an aspect that has been less explored is the association between previous sickness absence and re-employment. Assessing sickness absence during unemployment is complex due to the lack of a concrete everyday job. Our access to the sickness absence records of unemployed people during their participation in active labor market policy measures (ALMP) affords us the possibility of addressing the knowledge gap on the role of previous sickness absence on re-employment outcomes.
Sickness absence, particularly medically certified absence, is a well-known risk marker for mortality (17,18) as well as many chronic health problems (19). Medically certified sickness absence has also been associated with increased risk of future sickness absences (20), job termination (21,22) and unemployment (22)(23)(24) among active employees. Therefore, sickness absence during participation in ALMP could serve as a strong proxy for poor health and as such presents an objective measure to investigate the sustainability of the re-employment intervention for persons with serious health problems.
In our earlier study (17), we identified employment trajectories of re-employed people after their participation in a state subsidized re-employment program. Our aim in the present study is to examine whether sickness absence during such programs is a predictor of subsequent labor market attachment. We hypothesize that sickness absence is associated with poor labor market attachment, and that the risk of poor attachment increases with increasing number of sickness absence days.

Design and study population
We derived our study data from the Finnish Public Sector (FPS) study, which is an ongoing prospective study of employees in ten towns and five hospital districts in Finland. The overall aim of the FPS study is to assess the work life of employees and the impact of work and work-related changes on employee health and well-being (25). The Ethics Committee of the Hospital District of Helsinki and Uusimaa approved the study.
We used the part of the FPS study that included data on originally long-term unemployed individu-als (N=23 213) who had their first period of subsidized re-employment in the service of the ten towns in 1994-2005. Subsidized re-employment schemes are an essential component of Finland's ALMP measures. They are designed for long-term unemployed people (as a rule those unemployed for ≥12 months) who have problems finding a job in the regular labor market. Participation in the program is voluntary and the selection of the participants is coordinated by the municipalities in cooperation with the local employment authorities. Individuals are selected if upon assessment they are considered capable of performing a full-time job. The employment contract lasts for six months. For the purposes of this study, the sample included the individuals (N=18 944) who completed the full six months' participation in the subsidized program and excluded those who were compelled to drop out due to poor health or who discontinued the program due to finding a job in the open labor market. All participants had to be 18-60 years old at the end of the scheme. Those who moved into retirement during the six-year follow-up were excluded.

Data on labor market attachment
Starting at the end of the subsidized re-employment period, subsequent employment of each individual was followed-up for six years, which were divided into 12 six-month periods to enable analysis of the labor market attachment trajectories. Labor market attachment here refers to the number of months (0-6) as an employee or entrepreneur during each of the 12 time periods. We obtained information on the employment histories from the register of the Finnish Centre for Pensions. This register includes monthly records of all employment contracts, as well as entrepreneurship which is eligible for inclusion in the formation of the statutory earningsrelated pension insurance.

Data on sickness absence
We retrieved information on sickness absence from the register of the Social Insurance Institution of Finland (KELA). This nation level register contains records of all sickness absence periods lasting >10 working days, while costs of shorter periods are covered by the employer (26). Like other employees, participants in the subsidized re-employment program were also entitled to an earnings-related sickness allowance. We categorized the total number of their sickness absence days across the six-month period as: 0-≤10, 11-29, and ≥30 sickness absence days.

Background variables
Information on gender and age (18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33)(34)(35)(36)(37)(38)(39)(40)(41)(42)(43)(44), and 45-60) was obtained from the employers' registers. Educational level (basic, vocational school, and college or university degree) was retrieved from Statistics Finland, while information on chronic diseases was retrieved from the records of the Social Insurance Institution on entitlements to special reimbursements for the costs of purchased drugs for severe and chronic diseases. The chronic disease' variable was a summed score of six common diseases (heart disease, rheumatoid arthritis, asthma or chronic obstructive pulmonary disease (COPD), chronic hypertension, and severe mental problems). We categorized participants into two groups based on whether or not they had at least one of the chronic diseases mentioned. Information on calendar year in subsidized re-employment was derived from the employer's register, and calendar time was categorized as: 1994-1997, 1998-2001, and 2002-2005 in order to control for variations in unemployment and subsidized re-employment rates. We used the information on the ten towns where the participants had worked as a proxy for dichotomizing the area of residence into small towns and big towns. The small towns included Raisio, Naantali, Nokia, Valkeakoski and Virrat, while the big towns were Tampere, Turku, Oulu Vantaa and Espoo.

Statistical analysis
We used latent class growth model with zero-inflated Poisson (LCGM-ZIP) (27,28) to identify subgroups within the population following a similar pattern of change in labor market attachment during the six-year follow-up period. We adopted an exploratory approach and estimated as many classes as possible (two through six latent class solutions) that yielded proper solutions, in the search for the optimum number of trajectory classes. We specified a quadratic growth term in each model (k number of classes), assuming that labor market attachment would decrease over time after an initial increase. We compared the models (ie, the k and k-1 models) using four selection criteria: (i) the Bayesian Information Criteria (BIC), where a model with lower BIC values indicated a well-fitting model (29); (ii) the Lo Mendell and Rubin Adjusted Likelihood Ratio test (LMR-LRT), where a significant P-value (P<0.05) indicates that the k class fits better than the k-1 class model (30); (iii) the average posterior probabilities of group membership for each class, where higher values (closer to 1) suggested that the trajectories correctly classified individuals with a similar pattern of labor market attachment and discriminated between individuals with dissimilar attachment patterns (31); (iv) the practical usefulness of the trajectories. To evaluate this, we examined both the distinctiveness and the sizes (proportions) of each of the trajectory groups (28). For trajectory groups to serve a useful substantive purpose, they should be distinguishable in terms of their shapes and other explanatory characteristics. They should also be of reasonable sizes (≥5%) to ensure precision (27,31).
Upon establishing the optimum number of trajectory classes, we then used multinomial logistic regression to investigate the association between sickness absence and labor market attachment trajectories. We ran both unadjusted and adjusted models, where the adjusted model included age, gender, educational level, calendar year in a subsidized re-employment program, chronic conditions, and size of town. We also examined whether age and gender acted as potential modifiers by entering an interaction term between sickness absence and each of the variables in the fully adjusted model. If the interaction term was significant (P<0.05), the analysis was stratified and the stratum-specific estimates calculated. We presented the results of the regression analyses as odds ratio (OR) with their 95% confidence intervals (95% CI). We used Mplus version 7 for LCGA-ZIP and IBM SPSS Statistics for Windows version 23.0 (IBM Corp, Armonk, NY, USA) for the multinomial logistic regression.

Results
Altogether, 1172 of the 18 944 study participants had a recorded sickness absence during the six month period they were enrolled in the re-employment program. The absence lasted 11-29 days among 708 (60%) individuals, while the rest had >30 days of sickness absence. Having >30 absence days was more common in the older age group, among those with basic educational qualifications only, and in those who participated in the program between 2002-2005. Participants with chronic diseases were also more likely to have >30 days of sickness absence than those without such diseases (table 1).
On the basis of the information criteria, the fourtrajectory solution discerned the different labor market attachment trajectories best over the six-year follow-up. figure 1 illustrates the four labor market attachment trajectories. The curves are constructed based on the mean proportion of months of the attachment at each half-year period of the individuals classified into each trajectory. The strengthening trajectory (N=14 577, 77%) represented those with a relatively stable attachment throughout the follow-up time. The delayed trajectory (N=1101, 6%) included those whose initially weak attachment steadily improved after 36 months. In the leavers trajectory (N=1970, 10%), attachment declined with time, while among the non-attached (N=1296, 7%) the trajectory assumed a very low level throughout the follow-up period. reduces the likelihood of regaining a paid job. Not only is our study in agreement with earlier studies (as indicated in the association between sickness absence and the non-attached trajectory), but our findings moreover show that poor health could also constitute a risk factor even among those who regained employment immediately after the subsidized re-employment period (as seen in the association between sickness absence and the leavers trajectory). This has an important implication in that the emphasis should not only be on getting the unemployed back to work, but also on ensuring that re-employed people maintain favorable labor market attachment over time.
Earlier studies have suggested that poor health can negatively influence the job-seeking behavior of unemployed people that in turn may result in reduced likelihood of finding paid jobs (13). It has also been suggested that employers may be reluctant to hire job seekers with a history of poor health. Given that long-term sickness absences are mostly associated with ill-health (32) mostly due to serious health conditions (19) that can interfere with everyday activities, it is plausible that these explanations may account for our findings of an association between sickness absence and poor labor market attachment among previously long-term unemployed people.
There is evidence that younger individuals tend to have fewer long-term sickness absences (33) and better chances of re-employment (11,12). Our findings revealed that younger long-term unemployed people with >30 days of sickness absence had a higher odds for poor labor market attachment than their older counterparts. This finding is similar to that of Virtanen et al (22) in which younger temporary employees with high sickness absence were reported to be at increased risk of subsequent unemployment, whereas among older temporary employees no such association was found. Knutsson & Goine (34) stratified sickness absence diagnosis by age and found that psychiatric diseases and allergies were more prevalent among younger individuals, while cardiovascular diseases were more common among older people. That study also reported that musculoskeletal diseases increased from ages 16-44 years, and thereafter leveled off. Virtanen et al (22) suggested that these differences in the diagnoses underlying sickness absence among younger and older people may explain the age-differences in the association between sickness absence and labor market outcomes.
Schuring et al (35) found that a poor self-rated health increased the likelihood of non-re-employment among men, but had less influence among women. In our study, gender did not modify the association between sickness absence and labor market attachment. It appears that the effect of gender on labor market outcomes may be sensitive to the health indicator measured.
It is noteworthy that the 94% of the study subjects were able to work the six-month period without needing long sick leaves. The high figure may be due to healthrelated selection in enrolling individuals on subsidized schemes or due to poor health among those who did not complete the period (and were therefore excluded from the study sample). But on the other hand, there may have been a high "sickness presence" either due to high motivation of the participants to demonstrate their work ability or due to the liberal attitude of the employer to the sometimes relatively poor productivity of these workers. In sum, those 6% who needed longer-term sick leaves evidently had relatively severe problems in health -notably in the areas that were not covered by the chronic conditions adjusted in the analyses.
It is also noteworthy that the trajectory analysis did  not produce a group with full, ie, six-month labor market attachment: at best, in those assuming the trajectory of "strengthening attachment", the employment rate remained at the level of four months. Thus, we can conclude that subsidized re-employment leads to permanent full employment relatively seldom even if the health and work ability of the participant are optimal. The strengths of this study include the prospective design with large sample size. The use of registry-based data both for sickness absence and labor market attachment is also a strength, as it eliminated problems relating to attrition, recall bias, and subjective interpretation of both health and employment status. A limitation of our study is the lack of data on the nature of the employment contracts of the participants during follow-up (ie, whether they had permanent or temporary jobs or whether the employment was full-or part-time), which may influence the relationship between sickness absence and employment outcomes (22). Another limitation relates to the trajectory classes, which is only an approximation of reality since individuals were assigned to their mostly likely classes according to their average posterior probability values. The inability to adjust for behavioral risk factors and other health variables aside from those covered in the reimbursement scheme could be another source of residual confounding, although, in Hultin et al's study (23), the association between sickness absence and unemployment persisted even after controlling for several health indicators. This study finding may not be extrapolated to unemployed population in general since our sample consisted of a selected group of long-term unemployed people ie, those participating in a state subsidized re-employment program. However, on average, they may be regarded as a representative sample of re-employed people in Finland given that they were pooled from ten towns in Finland with varying recruitment criteria.
In conclusion, sickness absence among long-term unemployed people participating in a state subsidized re-employment program predicted future poor employability ie, poor labor market attachment. The risk for poor labor market attachment was particularly marked for younger unemployed people. Sick unemployed people face a double burden by virtue of their health and labor market status, which is why it is important to provide them with adequate support, including health care and rehabilitation to enhance both their chances of re-employment and maintaining favorable labor market attachment in the long term.