Morbidity and suicide mortality following sick leave in relation to changes of social insurance regulations in Sweden

Backgrounds: Stricter regulations including time limits for sick leave have been introduced in Sweden in 2008, which might have resulted in higher morbidity in those on longer sick-leave spells after the introduction. This study aimed to examine (i) the association between all-cause and diagnosis-specific sickness absence and sick-leave duration with subsequent morbidity and suicide mortality and (ii) differences in socio-demographics and morbidity in individuals on sickness absence regarding changes of social insurance regulations. Methods: A population-based prospective study was conducted of two cohorts of individuals who lived in Sweden, aged between 20 and 64 years at 31 December 2005 (n = 4 477 678) and at 31 December 2008 (n = 4 500 400), respectively. Each of the cohorts was followed regarding inpatient healthcare and suicide. Hazard ratios (HR) and 95% confidence intervals were estimated by Cox regression models. Results: In the multivariate analyses, all-cause and diagnosis-specific sickness absence and sick-leave duration showed higher HRs for inpatient care and suicide in both cohorts (range of HR:1.10–2.59). HRs of inpatient care and suicide among individuals with mental sickness absence 2009 were reduced more after controlling for morbidity-related covariates, than such sickness absence in 2006. Individuals with mental and somatic sickness absence and sickness absence > 180 days in 2009 had higher HRs of somatic inpatient care than those on sickness absence in 2006. Conclusions: Diagnosis-specific sickness absence and long-term sickness absence in 2009 might be associated with more severe morbidity or work incapacity than in 2006 due to the stricter regulations.

Introduction

Sickness absence represents a substantial public health problem in many European countries.^1–3 Mental and musculoskeletal disorders comprise the most common diagnostic groups behind sickness absence.^3,4 The proportion of depressive, anxiety and stress-related disorders summarized as common mental disorders (CMD), account for large societal costs of sickness absence^3,5 and recurrence of such sickness absence is reported to be high.⁶ In Sweden in 2012, 93% of the mental sickness absence was with CMD.^7,8

Several recent prospective cohort studies suggest that the risk of morbidity, e.g. depression and suicidal behaviour, is higher in individuals with all-cause and cause-specific sickness absence, including CMDs, other mental as well as somatic diagnoses.^9–15 Higher risks of morbidity and suicidal behaviour in long-term sickness absentees have also been reported.^9,10,15

Sick-leave rates may vary under the influence of changes of national social insurance regulations.¹⁶ In 2008, the Swedish government launched stricter rules for sickness benefits in order to facilitate early return to work and reduce the high level of benefit dependency. The new regulations included set time for careful assessments of work capacity, of allowed time for sickness benefits, and stricter assessments related to eligibility for disability pension.^3–5 Recent studies suggest that these regulation changes are related to the shorter duration of sick-leave spells and to fewer transitions to disability pension.^5,17–19 It is possible that those on sickness absence for longer time periods (i.e. the time periods related to assessments according to the new regulations) after the stricter rules were introduced had higher medical severity or work incapacity than before these regulations were implemented. This means that both morbidity and mortality, e.g. in terms of suicide, might differ between those on long-term sickness absence before and after the new rules—something that has not been studied before.

Aims

The aim of this study was to investigate the association between all-cause and cause-specific sickness absence as well as sick-leave duration with subsequent risk of inpatient care due to mental and somatic disorders as well as of suicidal behaviour (suicide attempt and suicide) before and after the implementation of the new regulations in Sweden in 2008. Additionally, the study aimed to examine differences in socio-demographic and previous and current morbidity in individuals on sickness absence before and after the new regulations.

Methods

Study population

We conducted a population-based prospective study, comparing two cohorts. The study population comprised individuals resident in Sweden, aged between 20 and 64 years at 31 December 2005 and alive and living in Sweden at 31 December 2006 (excluding 581 619 individuals with old-age or disability pension in 2006) (baseline). An additional similar study population included individuals resident in Sweden, aged between 20 and 64 years at 31 December 2008, alive and living in Sweden at 31 December 2009 (excluding 583 144 individuals with old-age or disability pension in 2009) (baseline). These two cohorts are hereafter referred to as ‘cohort 2006’ and ‘cohort 2009’, respectively. Individuals with an ongoing sick-leave spell at the turn of 2005/06 and 2008/09 in the respective cohort were excluded (N = 231 627 vs. N = 142 111). This left 4 477 678 individuals in cohort 2006 and 4 500 400 individuals in cohort 2009. We used the IMAS (Insurance Medicine All Sweden) study including nationwide register data obtained from Swedish authorities and merged for each individual by means of the individuals’ personal numbers attributed to all Swedish inhabitants.

Exposure

Exposure was measured as having at least one new (incident) sick-leave spell during 2006 or 2009, respectively. Information on all new sick-leave spells with sickness benefits regarding diagnosis and duration initiated during 2006 and 2009 was obtained from the Social Insurance Agency (SIA). The main diagnosis in the first new sick-leave spell was reported by the sickness certifying physicians and coded by using the corresponding codes of the International Classification of Diseases (ICD) version 10.²⁰ The main sick-leave diagnosis in the first sick-leave spell was in this study categorized into the following five groups: sickness absence due to CMD included ‘depressive disorders’ (F32–F33), ‘anxiety disorders’ (F40–42) and ‘reaction to severe stress, and adjustment disorders’ (F43).⁶ All other mental diagnoses (F00–F31, F34–F39, F44–F99) were combined into one group. Sickness absence due to musculoskeletal diagnoses comprised ICD-10 codes M00-99. The remaining sick-leave diagnoses were grouped as other somatic diagnoses. Missing sick-leave diagnoses (≈13%) were coded as an own group. Sick-leave duration was defined as the total number of gross days of the first new sick-leave spells i.e. a continuous, not interrupted period with benefits from the SIA. Information on sick-leave duration in 2006 or 2009, was grouped into six categories: no sickness absence, 1–4 days, 15–90 days, 91–180 days, 181–365 days, >365 days, in accordance with previous studies.^9,10 That is, the sick-leave spell was followed till its end, even if that was in a later year.

Covariates

Information on sex, age, country of birth, type of area of residence, educational level and family situation at baseline was obtained from Statistics Sweden. Information on missing values on any variable was categorized in an own category. The variables were coded as indicated in table 1.

Other variables of interest included previous and ongoing mental and somatic healthcare from inpatient and specialised outpatient care, previous suicide attempt from inpatient care, as well as dispensed prescription of antidepressants. This information was obtained from the National Patient Register and the Prescribed Drug Register, respectively. Categorisation of previous and ongoing healthcare was based on the diagnosis-specific median length of inpatient care in 2000–06 or in 2003–09 (no inpatient care; ≤median length; >median length) and total number of outpatient care visits, 2001–06/2004–09 (no visits; ≤median visits; >median visits). As suicide attempt can be underreported or reported as undetermined causes,²¹ previous inpatient care for suicide attempt was coded according to ICD-10: X60–X84, Y10- Y34 during 2000–05/2003–08 and this variable was dichotomised into yes/no. Additionally, prescribed antidepressants dispensed during 2006 and 2009, respectively, based on the codes in the Anatomical Therapeutic Chemical Classification System (ATC codes N06A) were used, also coded as yes/no (table 1).

Outcome measures

Information on inpatient care due to mental disorders, somatic disorders and suicide attempt and on completed suicide was obtained from the National Patient Register and the Cause of Death Register kept by the National Board of Health and Welfare. These outcome measures were coded according to ICD-10: mental disorders (F00–F99), somatic disorders (A00–E90, G00– Z99), suicide attempt/suicide (X60–X84, Y10–Y34). ICD codes Y10–Y34 (undetermined intent) was included to limit underreporting of suicide attempt/suicide.^22,23 Sensitivity analyses without such undetermined cases showed similar results.

Statistical analyses

All analyses were performed separately for the two cohorts. Potential differences between the two cohorts among individuals with sickness absence were tested with chi-squared tests regarding all socio-demographic and morbidity-related factors. Crude and adjusted hazard ratios (HR) and 95% confidence intervals (CI) for inpatient healthcare due to mental, somatic disorders and suicide attempt as well as completed suicide in relation to the exposure variables were estimated by Cox proportional hazard regression models after testing that the proportional hazard assumption was met. The two cohorts were separately followed from 1 January 2007/1 January 2010 until the respective event (inpatient healthcare due to mental, somatic disorders and suicide attempt or death due to suicide), or end of follow up after 2 years (31 December 2008/31 December 2011) whichever came first. Censoring was due to death (due to other reasons than suicide) and emigration. Participants with no new sick-leave spells during 2006/09 were used as the respective reference group. Besides the crude models, we adjusted HRs for socio-demographic factors in the first model. The final model was additionally adjusted for previous and ongoing mental and somatic healthcare from inpatient and specialised outpatient care, previous suicide attempt from inpatient care, and current antidepressant prescription. The partial likelihood ratio test was performed in relation to the interaction regarding the association between sick-leave measures and the different outcome measures in the two cohorts. Data processing was performed using SPSS for Windows V.22.0.

Social insurance regulations

All people in Sweden with income from work or unemployment benefits can from the age of 16 years receive sickness benefits if unable to work due to a disease or injury from SIA. Employees receive sick pay for the first 2 weeks of the sick-leave spell from the employer. Afterwards, employees can receive sickness benefits from SIA. Unemployed individuals and those with chronic diseases can be granted sickness benefit from SIA from the second day of a sick-leave spell. All have one qualifying day, with the exception of self-employed, who can have more qualifying days according to the insurance they have chosen. Before 2008, there was no limit to the duration of a sick-leave spell. Since July 2008, stricter assessments related to eligibility for sickness absence and disability pension were introduced and provision of sickness benefit was limited to 364 days during an 18-month period. After that, sickness benefit could be extended to another 550 days for individuals whose working capacity was still reduced after 1 year due to specific reasons including a high medical severity of the underlying disease.^3,5 After 90 sick-leave days the absentee’s work capacity was recommended to be assessed by the officer at the SIA, who based this assessment on information from the absentee, the certifying physician and sometimes also from the employer for other jobs at the work place, after 180 days for all jobs at the labour market.

Results

Table 1 shows descriptive statistics for the two cohorts regarding exposure, covariates and outcome characteristics. Chi-squared tests showed that individuals with incident sickness absence in the cohort 2006 and 2009 differed significantly regarding all measured characteristics and outcome measures with the exception of sex and suicide. The proportion of individuals on sickness absence due to all other somatic diagnoses than musculoskeletal diagnoses was higher in 2009 (50.2%) than in 2006 (46.5%). There was a higher proportion of individuals whose first sick-leave spell lasted for more than 365 days in cohort 2006 (9.2%) than in cohort 2009 (4.7%). Among those with sickness absence, a higher proportion of the individuals in the cohort 2006 (17.0%) had only a compulsory educational level than those in the cohort 2009 (14.8%). With regard to healthcare characteristics, more individuals on sickness absence in 2009 had longer previous and ongoing healthcare visits, previous suicide attempt, as well as subsequent inpatient care and suicide attempt than those on sick leave in 2006 (table 1).

Inpatient care due to mental and somatic disorders

In the univariate model, all-cause and diagnosis-specific sickness absence as well as sick-leave duration were associated with higher HRs of mental inpatient healthcare in both cohorts. After controlling for socio-demographic factors, previous and ongoing healthcare, previous suicide attempt and antidepressant prescription, all sick-leave variables were still associated with a higher risk of mental inpatient care compared with those without sickness absence in the respective cohorts. The associations between sickness absence due to CMD and other mental diagnoses as well as sickness absence >365 days with future mental inpatient healthcare were attenuated more strongly among those on sickness absence in the cohort 2009 than those on sickness absence in the cohort 2006, after controlling for all covariates (table 2).

Table 3 shows higher HRs of somatic inpatient healthcare among individuals with all-cause sickness absence, sickness absence due to mental and somatic diagnoses, and with different sick-leave durations in the respective cohort in the uni- and multi-variate analyses. In the univariate model, individuals with all-cause sickness absence, sickness absence due to CMD, musculoskeletal diagnoses, and other somatic diagnoses and sickness absence >14 days in 2009 had higher risk estimates of somatic inpatient care than individuals with such sickness absence in 2006. All-cause sickness absence, sickness absence due to CMD and musculoskeletal diagnoses as well as sick-leave duration >180 days in 2009 remained to be associated with a higher HR of somatic inpatient healthcare compared with the sickness absence in 2006 after controlling for all covariates.

Suicide attempt and suicide

We observed that all-cause sickness absence, sick-leave diagnoses and sick-leave duration were associated with higher HRs of suicide attempt in both cohorts in the crude analyses. The high risks of future suicide attempt were strongly reduced in mental sickness absence, particularly for those with such sickness absence in 2009 in the multivariate model. Still, the high risk estimates remained for most of the sick-leave variables after adjustment for all covariates in 2006 and 2009. The HRs were markedly higher for those whose sick-leave spell exceeded 365 days in 2006 and 2009 compared with those without sickness absence in the final models (HR 2.14; 95% CI 1.84–2.50 vs. HR 1.82; 95% CI 1.46–2.27) (table 4).

Moreover, associations of different sickness absence measures before and after the regulatory changes and subsequent suicide have been investigated (data not shown). In the univariate models, individuals with all-cause and diagnosis-specific sickness absence and sick-leave duration in 2006 and in 2009 had higher HRs of suicide. In the multivariate models, the HRs of suicide were strongly reduced, but still showed higher risks among those with all-cause sickness absence as well as among those with sickness absence due to CMD and other mental diagnoses in 2006. Sick-leave duration between 91 and 365 days in 2006 showed higher risks of suicide. Individuals with all-cause sickness absence and sickness absence due to CMD in 2009 had higher risks of suicide in the final model. Moreover, 15–365 days of sickness absence in 2009 were associated with higher HRs of suicide after adjusting for all covariates (data not shown).

Discussions

Main findings

This study comparing two population-based cohorts of individuals from before and after the implementation of stricter sickness insurance regulations in 2008, shows that individuals on sickness absence in 2006 and 2009 differed significantly regarding socio-demographic and morbidity-related factors. A lower proportion of individuals was on sickness absence in 2009 compared with in 2006, and those in 2009 to a higher extent had previous or current morbidity. All-cause and diagnosis-specific sickness absence as well as sick-leave duration were associated with higher HRs of subsequent diagnosis-specific inpatient healthcare and suicidal behaviour in both the cohorts. When contrasting the two cohorts, individuals with all-cause sickness absence, sickness absence due to CMD and musculoskeletal diagnoses and on sickness absence >180 days in 2009 had higher HRs of subsequent somatic inpatient care than corresponding groups in 2006.

Methodological considerations

To the best of our knowledge, this is the first study with a prospective design scrutinizing whether the stricter social insurance regulations introduced in Sweden in 2008 showed any associations with future morbidity and suicidal behaviour in sickness absentees. Strengths are that the study was based on high quality register data with national coverage,^24,25 no drop outs and the large sample size that allowed for investigations also of subgroups and of rare outcomes such as suicide attempt and suicide. Moreover, we could control for a number of confounders, including previous and ongoing inpatient and specialised outpatient care due to mental and somatic disorders, previous suicide attempt leading to inpatient healthcare, and ongoing antidepressant prescription.

The present study has some limitations. The follow-up time was relatively short, which means that for the detailed analyses regarding specific sick-leave diagnoses and duration, the number of suicides was low. The validity of the sick-leave diagnoses are sometimes questioned, however, seldom studied. Still, the one study we found on this, reported acceptable validity of sick-leave diagnoses.²⁶ There might be misclassifications and underreporting of mental diagnoses in the sick-leave data, though. Some information on sick-leave diagnoses was missing due to the fact that the information was not coded into ICD-codes by the physicians or impossible to read or scan by the personnel at the SIA.²⁷ As missing information on sick-leave diagnoses appeared to be randomly distributed across the different outcome measures, this misclassification is assumed to be non-differential.

Also, we included the first new sick-leave spell of the people in the cohorts, which means that a next spell might have been long or due to another diagnosis. However, the majority of the study populations had only one new sick-leave spell during the exposure year and those rates did not differ between the two cohorts (85.1% in cohort 2006 vs. 85.7% in cohort 2009). In addition, most of the shorter sick-leave spells (<14 days) were not included for employees, but for unemployed individuals. Therefore, estimates related to sick-leave spells below 14 days might be overestimated. In this study, we were able to control for a number of healthcare and medication factors. Still, residual confounding is possible, particularly with respect to somatic morbidity. Control for mental morbidity included information on antidepressant treatment. We did not have an equivalent measure for somatic disorders. When it comes to comparison of the two cohorts, it is also important to keep in mind that we were not able to fully distinguish between a potential period effect by the two cohorts and the potential effect of the new sickness insurance regulations on the studied outcomes.²⁸ Another potential period effect could be driven by the recession. Still unemployment rates were considerably lower in the cohort 2009 (9%) compared with the cohort in 2006 (15%). It is therefore not specifically likely that the recession had a major effect on the changes observed in this study. Moreover, as there only are 2 years between the cohorts and no other major changes occurred, we assume no other important period effects were present.

Morbidity and suicide mortality

We found that all-cause sickness absence in 2006 and 2009 implied an ∼1.5-fold higher risk of inpatient care due to mental and somatic disorders, suicide attempt and suicide, after controlling for socio-demographic factors, healthcare factors and prescription of antidepressants. Our finding is in this respect comparable to other prospective cohort studies.^{10,12,14,15,29–32}

We also investigated different sick-leave diagnoses regarding future inpatient care and suicidal behaviour. Previous such studies report higher risk of suicidal behaviour among individuals with sickness absence due to specific diagnoses, especially mental diagnoses.^9–13,33,34 Besides this pattern, we here also observed that those on sickness absence due to musculoskeletal diagnoses and other somatic disorders had a higher risk of suicide attempt. Furthermore, sickness absence due to somatic diagnoses showed a higher risk of inpatient care due to mental disorders, whereas mental sickness absence was associated with a higher risk of somatic inpatient care. It is common that somatic health problems coexist with mental disorders such as depressive disorders.^35–37 Such associations indicate comorbidity among the sickness absentees.^14,15,38 Moreover, those on sickness absence due to a CMD in 2009 had a higher risk of somatic inpatient care than those on sickness absence due to a CMD in 2006, even after controlling for a number of covariates. This might reflect that individuals on sickness absence due to a CMD in 2009 were more likely to have comorbid somatic disorders than those with such sickness absence in 2006.

In addition, the high future risks of inpatient care due to mental disorders, suicide attempt and suicide among sickness absentees in 2009 were more reduced among those with mental sickness absence after additionally adjusting for healthcare factors and antidepressant prescription than among those on mental sickness absence in 2006. This might be due to a stronger influence of morbidity, in terms of healthcare and medication, on mental sickness absentees in 2009 than on those with mental sickness absence in 2006. The descriptive statistics showed that more individuals on sickness absence in 2009 had antidepressants prescription, longer inpatient care, more frequent outpatient care, as well as previous suicide attempt than those on sickness absence in 2006. Also, we found a lower proportion of individuals with sickness absence in 2009 who had lower educational levels compared with those in 2006. This is in line with a recent report regarding the increasing proportions of higher educated individuals with sickness absence.³⁹ A probable explanation for this finding includes the fact that regulations in 2008 implied stricter rules for granting disability pension, which might have resulted in sickness absence with higher medical severity in 2009 than 2006.

Also, a longer duration of the first new sick-leave spell was associated with a higher risk of inpatient care and suicidal behaviour. This is consistent with previous studies, reporting a higher risk of adverse health outcomes and suicidal behaviour in individuals with long-term sickness absence.^{2,9,10,15,33,38,40} A higher risk of inpatient care due to somatic disorders was observed among individuals with long sick-leave duration (>180 days) in 2009 than those sickness absent in 2006. The new regulations in 2008 meant that work capacity was assessed more often and rehabilitation measures introduced earlier in the process.^3,5 Therefore, long sickness absence in 2009 may represent a less heterogeneous group which consists of individuals with more severe morbidity and/or reduced work capacity than among those in 2006, which in turn can be the reason for their a higher risk of subsequent hospitalization.

Our findings point at the importance of being particularly aware of long-term sickness absence after the introduction of the new social insurance regulations in 2008 in Sweden. Further knowledge is warranted on possible selection mechanisms into sickness absence as well as on possible effects of such regulatory changes, not only regarding the outcomes here studied, but also for e.g. sustainable return to work, risk of presenteeism, lower work capacity/productivity and higher risk for subsequent sickness absence or disability pension.

Conflicts of interest: None declared.

Funding

This work was supported by the Swedish Research Council (grant number 522-2010-2683), Karolinska Institutet funds for doctoral education, and the Swedish Research Council for Health, Working Life and Welfare.

References