Having high STSL in 2002 increased the odds of future long-term sick leave in 2007. The increased odds of long-term sick leave remained after adjustment for socio-demographic factors and self-reported health status and register data on in-patient care from 2002. The results did not support STSL (high or low) to be a substantial risk factor for adverse labor market position. Exposure to STSL implied no statistically significant increased risks of unemployment or disability pension after adjustment for health status and socio-demographic factors.
Studies on the consequences of short-term sickness absence are scarce; however, our results are in line with those published [11, 18]. These studies showed that frequent short-term absence was a predictor for later long-term sickness absence  and that sickness absence of more than 15 days per year was a risk factor for terminated employment, mainly resulting in unemployment [11, 29]. In our study, we found an increased risk of long-term sick leave among individuals with STSL. The OR of short-term unemployment was above one but not statistically significant after adjustment for all measures of health status. Few of these previous studies had the possibility to control for health status. The differences between our study and previous studies, regarding the risk of unemployment, could be due to the differences in exposure definitions or due to the fact that the association previously found between sick leave and unemployment mainly is due to the ill health that sick leave reflects.
Another study showed that the duration of sick-leave spells and the total number of sick-leave days in a year were the strongest predictors of disability pension, which led the researchers to conclude that the pathway to disability pension starts with short term sick-leave periods, then increasing in length until the disability pension . Our findings show that STSL increased the odds of long-term sick leave, but not of disability pension. However, it is important to note that our study investigated an initially healthy cohort, and it is likely that it may take longer than the five-year follow up to transition from short-term sickness absentee, to long-term sickness absentee to disability pensioner. The time period from short-term to long-term sick leave and later exclusion from the labor market may depend on severity of disease and type of work load but we have not found any studies regarding this issue.
Strengths and limitations
The strength of this study is firstly the longitudinal design and the large number of participants, and secondly the opportunity to combine survey data and register data, available from 2001 until 2007. This allowed us to limit the sample by excluding persons with long-term sick leave and in-patient care in 2001 and thereby avoid short-term STSL in 2002 being a continuation of previous severe illness .
In order examine the consequences of sick leave when controlling for health status, we used our access to self-reported health assessments in the survey from 2002. This captured different aspects of general health measured by SRH, GHQ-12, and LLSI. The GHQ-12 questionnaire is a well-established and validated instrument, measuring mental well-being . Further, the questions covering long-term limiting illness and musculoskeletal pain capture the most common somatic diagnosis among sickness absent, e.g. musculoskeletal diseases. The measure somatic disease covered severe diagnoses such as myocardial infarction or diabetes, but is not an all-encompassing measure of somatic diagnoses, however the access to registered in-patient care made it possible to control for those with ill health that required in-patient care for shorter or longer periods during 2002, like tumors, psychiatric diseases, and coronary heart diseases. Hence, the analyses were adjusted for both self-reported and register data on diseases and health conditions. Nevertheless, it is not possible to fully discriminate between the consequences of sick leave and the ill-health as we cannot link the health status with each specific sick-leave spell. Despite this shortcoming, we believe that this study has a major advantage compared to several previous studies of the consequences of sick leave [32, 33], in its adjustments for both self-reported and registered ill health.
Previous studies indicate that adverse work conditions and smoking are risk factors for disability pension [25–28]. Since these risk factors are likely to have their effect on disability pension through illness and sick leave, we have not considered them as confounders in our study.
One weakness in our study is the lack of register data on exposure, which implies an inexact exposure measurement and may imply recall bias and risk of response bias due to social desirability. The registers of sick leave in Sweden do not have valid data on the first 14 days of sick leave, since the 14 first days of sick leave are compensated by the employer and hence not included in the register held by the Swedish National Social Insurance Agency. Data on exposure to STSL were thus based on self-reported survey data. However, studies on the validity and reliability of register data and self-reported data on sick leave have not shown any significant differences between these two types of information [42–44]. The lack of register data also implies that the outcome long-term sick leave will only include individuals with sick-leave spells of at least 15 days, since these are the ones registered by the Swedish National Social Insurance Agency. Unfortunately, no other register data on the outcome long-term sick leave is available. If differential, this misclassification is likely to be more common among the exposed group, which would underestimate the odds ratios of long-term sick leave.
STSL can been seen as a coping strategy to prevent later long-term sick leave , or as an indicator of an underlying severe disease . In this study, those with chronic or severe disease, recognized by long-term sick leave, in-patient care, or disability pension the year before the short-term absence, were excluded from the study group. Hence, the study population was a “healthy population”, and most likely their subsequent ill health was either of short-lasting nature, or the beginning of a potential chronic, but not yet medically diagnosed, health condition. However, as mentioned above, to successfully investigate if a pathway exists from STSL, via long-term sick leave, to long-term labor-market exclusion, a longer follow up period than five years is likely to be needed. From our results we cannot determine through which mechanisms STSL affect long-term sick leave. Possible mediators between short-term and long-term sick leave remain for future studies to explore.