To the best of our knowledge, this is the first study to examine both mental health problems and work capacity in relation to RTW in a general population-based cohort of employees off sick due to all-cause sickness absence. Low mental well-being was a strong determinant with a near three-fold increase in odds of late RTW in the adjusted model. Similar odds were found in a separate model assessing the relationship between self-reported persistent mental illness and RTW. Another important finding was that each of the four work capacity dimensions (knowledge, mental, collaborative and physical) predicted late RTW also in models adjusted for age, gender, persistent mental illness and low mental well-being, respectively. Somewhat unexpectedly, low mental well-being remained a significant determinant of late RTW after adjustment for all four work capacity dimensions.
Determinants of RTW
We found that individuals who reported persistent mental illness had a higher likelihood of late RTW. Our study lacks specific diagnostic information, but it can be assumed that many of those with self-reported persistent mental illness had anxiety and/or depressive disorders. Individuals in this study were initially off sick and the disease that led to the sick-leave episode might have triggered a relapse or worsening of mental illness. Furthermore, individuals with persistent mental illness might be more vulnerable to strain associated with the sick-leave episode, be it health or work related. Conditions at work might influence RTW and several studies have shown that mental illness is associated with stigma [
34]. It is possible that individuals with mental illness refrain from disclosing their problems and this might reduce the access to vocational support needed in the RTW process [
34]. It is well known from previous studies that individuals with mental health problems to a large extent go unidentified at both primary health care centres and specialized medical clinics [
35,
36]. From the current study, it is not clear whether persistent mental illness was detected or not. A medical consultation is a prerequisite for sickness benefit after 7 days, and all individuals in this study were thus seen by a doctor. However, the extent to which symptoms of CMD were identified and treated could not be determined in our study. In one of the few studies estimating undetected CMD in a national register of all-cause long-term sickness absence (exceeding 8 weeks), Soegaard [
36] found that, among individuals without any psychiatric sick-leave diagnosis, 19.9% had undetected depression and 6.4% had undetected anxiety symptoms [
36].
A more unexpected finding was that even after adjustment for all dimensions of capacity to work, low mental well-being was associated with late RTW. A Swedish population-based study concluded that very mild psychological distress at baseline, measured by the General Health Questionnaire (GHQ 12), had a hazard ratio of 1.7 for a disability pension for a somatic reason and a hazard ratio of 2.2 for a disability pension with a psychiatric reason 5 years later [
37]. Our study has a shorter follow-up period, but the findings are in the same direction. Thus, it seems important to identify low mental well-being to avoid lingering absence and, in the longer-term perspective, disability pensions. In psychiatric research, it has been suggested that well-being is an important complement to diagnostic procedures of symptoms because it has higher relevance to an individual’s quality of life, and it may better capture recovery or subclinical symptoms [
35,
38,
39].
Our study sample consisted of individuals off sick with all-cause sickness absence and low mental well-being might be secondary to severe acute or chronic disorders such as injuries, circulatory diseases or neoplasms. Psychiatric disorders constituted the most common diagnostic group among sick-listed women in Sweden in 2008, accounting for 33% of all cases of sickness absence. In men, psychiatric disorders were the second most common diagnostic group, and they accounted for 23% of all cases [
16]. The diagnostic panorama changes as the sick-leave periods become longer because most individuals with infectious diseases and injuries return to work at a faster pace. Thus, the study population is mixed over all diagnostic groups and this needs to be taken into account when interpreting the findings. It is likely that physicians identify low mental well-being more easily in individuals who primarily seek care for mental symptoms, and less easily in individuals who primarily seek care for other causes. Work capacity regarding knowledge, mental, collaborative and physical demands predicted RTW in our study. There are no previous studies that can provide data for direct comparison.
In a clinical population from primary and occupational health care, Reiso et al. [
17] used a single question: “to what degree is your ability to perform your ordinary, remunerative work reduced today?”. They found that individuals who reported that their ability to perform work was very much reduced had significantly prolonged sickness absence at follow-up at 1 and 3 months. The single question “current work ability compared with life-time best” was used in a study investigating RTW in a cohort of newly sick-listed patients in primary and occupational health care [
18]. Patients with musculoskeletal disorders and those with mental disorders improved their work ability assessments but significant association with RTW was found only for those with musculoskeletal disorders [
18]. This is in line with our findings that low mental well-being was a strong determinant for late RTW. On the other hand, it might be that work capacity assessments are more complex both in the clinical setting and as self-assessments. A recent qualitative study explored the capacity to work while depressed and anxious and identified a dynamic and varying capacity related both to fluctuations in symptoms and to work tasks and work environment. Even the social environment and life outside work were highlighted as important for the capacity to work [
40]. New instruments or items need to be developed to better capture the complexity of the phenomenon in a population-based study but probably also in the clinical setting. Niewenhuijsen et al. [
41] concluded in a review that there was a lack of functioning instruments detecting the very specific deficits in mental disorders.
Practice implications
The clinical implications from this study should be regarded from a sickness absence perspective. It might be important to develop easily administered screening questions to identify individuals in need of more intense interventions with the aim of reducing time to RTW. Several future studies are needed but, from this study, it seems more important to screen for mental well-being than for work capacity assessment. The length of sickness absence should be adapted to the individual resources and work demands and, in some cases, a longer period of absence is needed, relevant and well-motivated. However, to be sick-listed might lead to negative and unwanted consequences such as reduced work motivation, social isolation, stigmatization, changed self-image, economic strain and secondary health problems. National sickness insurance regulations must also be considered and the way they might affect individuals on prolonged sick leave. In Sweden, for example, a new and stricter sickness insurance regulation means that after 3 months sick leave, a work place transfer to another job at the same employer but with lesser demands must take place. After 6 months of sick leave, if the person cannot return to work, a new job on the open labour market is the alternative. Regulations like this increase the pressure on the individual and the employer to make work possible and there is a large responsibility also for the health care to be efficient and professional in support of sick-listed individuals. Early identification of mental health problems is an important component in this process.
Methodological considerations
The strengths of this study are the selection of a general population-based consecutive sample of newly sick-listed individuals obtained from national registers, the use of two different questions to capture mental health problems, and the inclusion of different dimensions of work capacity. The outcome measure, RTW, was prospective and the register data were of high quality.
There are some limitations to discuss. Different types of bias can occur from self-assessed measurements including social desirability, recall bias and responses biased by symptoms associated with mental health problems (e.g. depression). Validated and recommended instruments have been used for the measurement of health-related variables [
20,
24]. The validity of the WHO (Ten) Mental Well-Being Scale was tested in the PART project; a population-based study from Sweden found that the WHO (Ten) Mental Well-Being Index was more likely to identify individuals in need of care and psychiatric diagnosis compared with diagnostic questionnaires on CMD, which were seen as too inclusive [
24]. Mental and physical capacity to work was measured by validated instruments [
25,
42]. However, knowledge and collaborative work capacity items were developed within the project and have not been tested for validity. Even though the associations were similar to those found for mental and physical work capacity, we cannot claim that these new constructs are equally valid. It can be argued that social desirability played a role in the assessments of collaborative work capacity; the second lowest proportion reporting low work capacity was found for this dimension. It is likely that an overestimation of one’s own capacity to collaborate is more common than an underestimation. Regarding memory and symptom bias, it can be argued that individuals with CMD might undervalue their capacities to a greater extent than the mentally healthy. However, the findings in this study were quite strong and remained significant even after adjustments, so even though there might be a slight overestimation of a specific OR, the direction would be the same with some changes in the level of estimations of, for example, capacity to work.
There was some dropout in the first phase of inclusion in this study. Not all individuals who actually became sick-listed between February 18 and April 15, 2008, were registered as off sick during that period. In our study, 49% were included, which is a large proportion of the target population. Multivariable analyses controlling for possible confounders such as gender, educational level, SES, and total number of sick-leave days during the year before to the index episode were done as a way to manage any bias related to overrepresentation of certain groups among those registered after April 15.
There was also a systematic dropout in the questionnaire study and the estimates need to be interpreted with this in mind. The dropout was higher among younger persons. From a sick-leave perspective, this means that a group with low risk for long-term sick leave had a higher dropout rate. The proportion for early RTW might have been higher if the age-related dropout had not occurred. On the other hand, young individuals with mental health problems might have more problems related to the capacity to work than older individuals due to their lack of experience of work life. However, this would strengthen the overall association between RTW and persistent mental illness and low mental well-being rather than attenuate it. The lower representation of immigrants most likely led to an underestimation of associations because mental health problems can be expected to be higher in that group compared with nonmigrants.
The internal dropouts in the WHO (Ten) Mental Well-Being Index were replaced by imputation and the variable was analysed in several ways. Although there was no systemic dropout regarding any of the specific items, missing data were more prevalent for the WHO (Ten) Mental Well-Being Scale compared with questions using the Likert scale that was placed both immediately before and after the well-being index. This indicates that questions regarding mental health are sensitive and the missing data were not missing at random [
43]. Through imputation, the sample size was increased. The proportions reporting high and low mental well-being before and after imputation did not change but a bias might have been introduced due to possible misclassification.
We do not have any information on dropout regarding work capacity but it can be expected that individuals with low mental work capacity refrain to a higher degree from participation in a questionnaire study with a large number of cognitively demanding questions [
44]. If this is the case, the proportions reporting low work capacity might be low in comparison with the true distribution.
The choice of confounders in this study was mainly based on earlier research on CMD and sickness absence. It is well known that both vary with age, gender, education, socioeconomic position, marital status; and for sickness absence, also with hours worked and earlier sickness absence. An initial bivariate analysis of the associations between these variables and RTW was done and in most cases the non significant variables were excluded from further analyses. However, the basic covariate and gender were included even though the associations were non-significant in the bivariate analyses to keep these factors under control in different sub analyses.