Trends of Diagnosis-Specific Work Disability After Newly Diagnosed Diabetes: A 4-Year Nationwide Prospective Cohort Study

Type 2 diabetes is common among working-age populations, as its onset is typically in middle age (1). To date, the estimated number of 20- to 59-year-old adults with diabetes in developed countries is nearly 50 million (2). Although the association of diabetes with morbidity and mortality is well documented (3), little is known about work disability among people with diabetes (4–7). Previous research has reported that people with diabetes lose 2‒10 more workdays owing to illness per year compared with those without diabetes (4). In the U.S., diabetes accounts for 25 million days of productivity loss owing sickness absence and 113 million days owing to reduced performance at work (8).

There is little evidence to show which specific diseases contribute to excess work disability among those with diabetes. One study of an occupational cohort followed 506 participants after they were diagnosed with diabetes and reported a higher amount of sickness absences among them than among those without diabetes, owing to metabolic and circulatory causes but not owing to other causes (5). However, as diabetes increases the risk of, for example, depression (9–11), it is plausible to expect a higher prevalence of work disability owing to mental disorders among people with diabetes. Moreover, in order to manage the disease burden associated with diabetes, it is important to identify the groups most vulnerable to work disability, such as people with socioeconomic disadvantage.

In this study, we used a large nationwide register-based data set, which includes information on work disability for all working-age inhabitants of Sweden, in order to investigate trends of diagnosis-specific work disability (sickness absence and disability pension) among people with diabetes for 4 years directly after the recorded onset of diabetes. We compared work disability trends among people with diabetes with trends among those without diabetes. We also investigated whether significant sociodemographic factors differently shaped the risk of work disability among these groups and sought to identify the diagnostic and sociodemographic groups with the highest levels of work disability.

This prospective cohort study was based on data from the population-based Insurance Medicine All-Sweden research database, which includes annual microdata covering nationwide register linkage data for ∼12 million people over several decades (12–14). The present cohort was identified from Insurance Medicine All-Sweden as all individuals aged 25‒59 years on 31 December 2005 who had been living in Sweden since 31 December 2002. Individuals with incident diabetes were those with the first indication of recorded diabetes diagnosis between 1 January and 31 December 2006 (n = 14,098). We restricted the study population to working-aged individuals aged 25‒59 years to avoid selective loss to follow-up due to old-age pensioning (at the age of 65 years in Sweden). The comparisons to the general Swedish population were made on the basis of a 1% random sample of the cohort aged 25‒59 years on 31 December 2005 and living in Sweden on 31 December 2006 (n = 39,056). Inclusion criterion to the reference group was defined as having no indication of diabetes between 2003 and 2010. All participants were followed for 4 years (2007‒2010). The project was approved by the Regional Ethical Review Board, Stockholm, Sweden.

Data were obtained from the following nationwide Swedish registers and linked using the personal identity numbers that are unique to each resident of Sweden:

• 1. Statistics Sweden: Longitudinal Integration Database for Health Insurance and Labour Market Studies (LISA) on age, sex, education level, family situation, type of living area, and birth country; all derived on 31 December 2005. Year of emigration was elicited in order to exclude those who no longer lived in Sweden during the follow-up.

• 2. The National Board of Health and Welfare: a) ICD-10 diagnoses and dates for diagnosis-specific data on hospitalizations or specialized outpatient care from 1 January 2003 and 31 December 2010, b) Swedish Prescribed Drug Register: medication purchases from 1 July 2005 to 31 December 2010, and c) death register from 1 January 2006 to 31 December 2010 to exclude those who died during the follow-up.

  • 3. Swedish Social Insurance Agency: dates and ICD-10 diagnoses for sickness absences and disability pensions from 1 January 2007 to 31 December 2010.

Incident diabetes was defined as including at least one of the following: 1) purchase of insulin or other diabetes medication (Anatomical Therapeutic Chemical code A10) between 1 January and 31 December 2006 and none between 1 July and 31 December 2005 and 2) in- or outpatient specialized health care record with ICD-10 code E10, E11, E12, E13, or E14 between 1 January and 31 December 2006 and none between 1 January 2003 and 31 December 2005.

Work disability was defined as days of sickness absence or disability pension on a yearly basis, as indicated by sick leave and disability pension benefits from the Swedish Social Insurance Agency, from 1 January 2007 to 31 December 2010 (that is, for a period of 4 years after the diabetes diagnosis). In Sweden, all people with income from work or unemployment benefits who have reduced work capacity owing to a disease or injury can receive sick leave benefits. A social insurance officer decides whether a claimant fulfils the criteria for work disability benefit. The assessment for this is based on comprehensive evaluation, e.g., from the treating physician about the claimant’s medical and work situation. Two requisites are necessary: one is that the individual has a disease or injury and the other that disease or injury has impaired work capacity to the extent that he or she is unable to respond to the demands of the work (15,16). All individuals, including those with no previous income, can be granted disability pension if their work capacity is permanently reduced owing to a disease or injury. Employees usually receive sick pay from their employer for the first 14 days of a sick leave spell, and these days are not included in the register.

We examined the following ICD-10 diagnostic groups: certain infectious and parasitic diseases (A00–B99); neoplasms (C00–D48); diabetes (E10–E14); other endocrine, nutritional, and metabolic diseases (E00–E09 and E15–E90); mental disorders (F00–F99); diseases of the nervous system (G00–G99); diseases of the eye and adnexa (H00–H59); diseases of the circulatory system (I00–I99); diseases of the respiratory system (J00–J99); diseases of the digestive system (K00–K93); diseases of the skin and subcutaneous tissue (L00–L99); musculoskeletal disorders (M00–M99): injuries (S00–T35, T66–T78, and T79); poisonings (T36–T65); and other (diagnostic groups N, O, P, Q, R, V, Y, Z, U, and missing diagnosis).

Sociodemographic factors, all measured on 31 December 2005, were age, sex, education level (compulsory school [low], high school [medium], university [high]), type of living area (large city, medium-sized town, small municipality), family situation/living arrangements (married/cohabitant, single/divorced/separated/widowed and no children in the household, single/divorced/separated/widowed and children in the household), and birth country (Sweden, other).

Comparisons in the baseline descriptive characteristics between people with diabetes and those without diabetes were made using χ² test for categorical variables and t test for age. We then examined the 4-year work disability rates for all diagnostic groups among individuals with and without diabetes.

To assess the annual mean difference between the work disability of people with and without diabetes, we calculated the least square means of annual days of work disability for each diagnostic group and expressed the differences as rate ratios and their 95% CIs. The models were adjusted for sociodemographic factors (age, sex, education, type of living area, family situation, and birth country). We applied a repeated-measures negative binomial regression analysis using the generalized estimating equations method with an autoregressive correlation structure (17). The generalized estimating equations method takes into account the intraindividual correlation between measurements. The autoregressive correlation structure assumes correlations between time points to be greater the nearer the measurements are to each other. To obtain adjusted annual means, we calculated the exponential functions of the least square means. To assess change over time, we entered a “year” variable, and to examine change over time between individuals with diabetes and those without diabetes, we entered a “year × diabetes status” interaction term to the models in addition to the main effects.

Next, we examined the five most common reasons for work disability for further analyses: mental disorders (F00–F99), musculoskeletal diseases (M00–M99), diseases of the circulatory system (I00–I99), diseases of the nervous system (G00–G99), and injuries (S00–T35, T66–T78, and T79). We examined the association between sociodemographic factors and diagnosis-specific work disability among individuals with and without diabetes. For those with diabetes, we also analyzed the association between sociodemographic factors and work disability owing to diabetes (E10–E14). SAS, version 9.4 (Cary, NC), was used for all analyses.

Diabetes was associated with all sociodemographic factors in this study: older age, male sex, low level of education, living in smaller towns rather than large cities, not being married or cohabitant, having no children in household, and birth country other than Sweden (Table 1). As shown in Supplementary Table 1, the highest annual work disability rates across the 4-year follow-up among people with and without diabetes were in the diagnostic group of mental disorders (unadjusted mean 30.5 days/year among those with diabetes and 12.5 days/year among those without diabetes), followed by musculoskeletal diseases (29.4 and 11.4 days, respectively) and diseases of the circulatory system (9.3 and 1.3 days, respectively). Besides mental, musculoskeletal, and circulatory diseases, all other diagnostic groups, including those of diabetes diagnosis, made a relatively small contribution to the total burden of work disability among people with diabetes (mean annual compensated disability days <5). However, even in these minor diagnostic groups, people with diabetes had more work disability days than those without diabetes.

Over the 4-year study period, there was a slight but significant decrease in work disability owing to mental disorders, musculoskeletal diseases, injuries, diseases of the digestive system, and “other” diseases among people with diabetes (Table 2).

Development of multivariable adjusted difference in work disability over 4 years between people with and without diabetes in the five most common diagnostic groups is presented in Fig. 1. We found higher work disability risk among people with diabetes in all diagnostic groups (P < 0.001). The greatest discrepancy was found for mental disorders, followed by musculoskeletal diseases, circulatory diseases, diseases of the nervous system, and injuries. P values for trend over time comparing participants with and without diabetes were as follows: P = 0.037 for mental disorders, P = 0.234 for musculoskeletal diseases, P = 0.911 for circulatory diseases, P = 0.624 for diseases of the nervous system, and P = 0.203 for injuries. These interaction tests suggest that, with the exception of mental disorders, the overall trends of work disability in these diagnostic groups were similar among people with diabetes and those without diabetes. The adjusted annual mean days of work disability owing to mental disorders from 2007 to 2010 were 32.8, 33.4, 32.8, and 30.9 among people with diabetes and 13.7, 13.6, 13.2, and 12.1, respectively, among those without diabetes; thus, the disability rate for mental disorders among people with diabetes first increased and then began to decrease in the last 2 years of follow-up. These findings were replicated in a sensitivity analysis including occupational status (categorized as leaders, office workers, manual laborers, military personnel, unknown occupation, and nonemployed) as a covariate (data not shown).

In general, the association between sociodemographic characteristics and diagnosis-specific work disability among people with and without diabetes did not greatly differ (Supplementary Tables 2–4). Older age, female sex, low level of education, and birth country other than Sweden were associated with higher work disability rates. However, based on the observed nonoverlapping CIs, some of the associations were found to be different between individuals with diabetes and those without: there was a positive association between age and work disability owing to mental disorders among people without diabetes, while the association was inverse among those with diabetes (Supplementary Table 2). In addition, the association of female sex and low education with mental disorders and the association between older age and work disability owing to circulatory diseases were stronger among those without diabetes, and the association between female sex and work disability owing to injuries was seen only among those without diabetes (Supplementary Tables 2 and 3). Among people with diabetes (Supplementary Table 4), older age was the only significant predictor of work disability owing to diagnosis of diabetes.

To identify the 10 diagnostic and sociodemographic groups most vulnerable to work disability, we collected the groups with the highest mean annual days of work disability according to diabetes status, diagnosis of work disability, and sociodemographic factors, based on the data in Supplementary Tables 2–4. The summary of these is listed in Table 3. All 10 groups comprised people with diabetes. The highest mean number of work disability days was found among those who had work disability owing to mental disorders and lived alone, i.e., were not married or cohabitant and had no children in their household (mean annual days 45.3), followed by those with work disability owing to mental disorders and birth country other than Sweden (41.8 days), those with work disability owing to musculoskeletal diseases and birth country other than Sweden (38.9 days), and those with work disability owing to mental disorders who were not married or cohabitant but had children in the household (38.7 days).

In this 4-year prospective cohort study of the Swedish general population, we found that among people with newly diagnosed diabetes, the highest rate of subsequent work disability compared with those without diabetes was owing to diagnoses of mental, musculoskeletal, and circulatory diseases; diseases of the nervous system; and injuries. Diagnosis of diabetes as the cause of work disability was rare. Overall, the trends over 4 years were similar in people with diabetes and those without diabetes, showing the greatest decreasing trend in mental and musculoskeletal disorders toward the end of follow-up, although the difference in mental disorders first widened and then narrowed. The highest rate of work disability was found among those with diabetes who lived alone and had work disability owing to mental disorders, followed by those with diabetes whose birth country was other than Sweden and had mental or musculoskeletal disorders.

The present findings suggesting a particular role of comorbid conditions in diabetes-related work disability are in line with the widely reported high comorbidity in diabetes, which often are direct diabetes-related complications (e.g., coronary heart disease and stroke) (18,19). A previous study also noted that more than half of the excess risk of all-cause sickness absence in people with diabetes was attributable to comorbid conditions (6). Mental and musculoskeletal disorders were the most common causes of work disability in this cohort. This is in accordance with reports from the general populations of other European countries such as France, U.K. (20), and Finland (21).

Research on the association between diabetes and mental disorders has largely focused on depression, and meta-analyses suggest that the association between diabetes and depression might be bidirectional, that is, that depression predicts diabetes and vice versa. However, the strength of each association may be modest (9–11). Possible mechanisms explaining why depression may predict diabetes include biological pathways (such as hypothalamic-pituitary adrenal axis hyperactivation and the associated increased release of cortisol, which further stimulates glucose production), health risk behaviors, and poor adherence to the dietary and exercise regimes used to treat preclinical diabetes. Diabetes, as a predictor of depression, might in turn relate to direct biochemical mechanisms in hyperglycemia, crisis when confronted with diagnosis of a chronic disease, and the burden of dealing with diabetes complications (10).

Excess work disability owing to musculoskeletal diseases is plausible because diabetes has been associated with, for example, diabetic cheiroarthropathy (i.e., thickening of the skin and limited joint mobility of hands), Dupuytren contracture, flexor tenosynovitis (i.e., trigger finger), carpal tunnel syndrome, adhesive capsulitis of the shoulder (i.e., frozen shoulder), diffuse idiopathic skeletal hyperostosis, neuropathic osteoarthropathy, and gout, all of which are linked to diabetes-related conditions— hyperglycemia, micro- and macrovascular complications, and diabetic polyneuropathies (19).

Given the high comorbidity between diabetes and circulatory diseases, excess work disability owing to these diseases among people with diabetes is not surprising. A similar discrepancy in sickness absence has previously been reported (5). In the U.S. alone, one-third of medical expenditure on cardiovascular and peripheral vascular diseases is estimated to be associated with diabetes (8). Our other finding, showing that people with diabetes had excess work disability owing to diseases of the nervous system, supports the result of a study in which one-third of the expenditure for neurological diseases was attributed to diabetes (8).

Only a handful of studies have investigated the risk of occupational injuries among people with diabetes and suggest a higher risk (22,23), although not all studies confirm this finding (24): one study found an elevated risk only in a subgroup of employees with long diabetes duration (25). Explanations for the higher injury risk owing to diabetes include potential causal pathways from the disease itself, such as hypoglycemia (when treated with insulin) and diabetes complications, e.g., impaired vision and peripheral nerve sensory impairments. However, as several other chronic diseases (e.g., coronary heart disease, depression, and asthma) have similarly indicated an association with injuries, the pathway between chronic diseases and injuries might also be nonspecific, including disease-induced fatigue or obesity, which are both prevalent in many chronic conditions (23).

We sought to identify the diagnostic and sociodemographic subgroups that had the highest level of work disability and found that the top 10 exclusively comprised people with diabetes, which is plausible given the overall higher work disability rate in those with diabetes. The highest levels of work disability were found among those who lived alone and had work disability owing to mental disorders. High absence rates were also found among those whose birth country was not Sweden and who had work disability owing to mental or musculoskeletal disorders and those who were living with children without a partner and had work disability owing to mental disorders. In sum, the people most vulnerable to work disability are characterized by diabetes and mental or musculoskeletal disorders and might also live alone, be single parents, or immigrants. Possible overarching factors for these people are economic hardship and weak or no social ties and support from other people (26,27). Our study is in line with previous research that shows an association between living alone, single parenthood, and mood disorders (26,28,29). Nevertheless, our study is the first to suggest that mental disorders associated with living alone and single parenthood might be particularly prevalent among people with diabetes.

The higher prevalence of depression among ethnic minority groups and immigrants is well documented, and it has been suggested that depression among these groups is more persistent and disabling (27). We found a higher prevalence of diabetes among foreign-born people than among those born in Sweden and demonstrated that people with diabetes who were born outside Sweden had high levels of work disability owing to mental disorders. Mental disorders among immigrants may reflect, for example, economic and social stressors of resettlement and exposure to racism, which may be compounded by poor health and problems with cultural competence, i.e., accessing health care and communicating with health care professionals (27,30).

The major strengths of this study include a large nationwide database that covers virtually all working-age residents of Sweden and has no loss to follow-up, the prospective cohort study design, and data from high-quality independent administrative registers (31). We were able to detect new cases of diabetes based on the data regarding prescribed diabetes medication, hospitalization, or visits to specialized health care owing to diabetes within a given year and excluded those who already had an indication of diabetes before that year. To our knowledge, this was the first large-scale population-wide study to use diagnosis-specific work disability outcomes. Our large data set enabled us to examine rare diagnostic groups and analyze subgroups with the highest disability rates.

However, one important limitation of this study is that undiagnosed diabetes cases as well as those treated by lifestyle intervention only were not included. Prediabetes and undiagnosed diabetes have been associated with a higher risk of cardiovascular diseases (3); thus, inclusion of these cases in the reference group might have attenuated the association between diabetes and work disability owing to circulatory diseases. However, prediabetes and undiagnosed diabetes have not been associated with depression (32). Those with lifestyle intervention only have milder diabetes. Treatment with lifestyle intervention only decreased between 1996 and 2006 because of an almost fourfold increase in the use of oral antidiabetes medication as first-line medication, which was not explained by an increase in diabetes incidence (33). In addition, as we only had data on prescribed drug purchases from July 2005 onward, we might have missed the subjects with diabetes who had medication before this and had for some reason had a break in their diabetes medication between July 2005 and January 2006. However, in Sweden, prescribed medication can only be bought for 3 months at a time, which is why most should have been included. The records of outpatient visits were from specialized health care, but a large part of diabetes care is managed in primary health care. However, we considered hospitalization and outpatient specialized health care visits a secondary data source and diabetes medication a primary data source in the detection of diabetes cases.

Another limitation is that we had no data on work disability episodes that lasted <14 days among those who were employed. In Sweden, a physician’s certificate is usually only needed after the seventh day of a sick leave period. Thus, the validity of sick leave diagnoses in our data can be considered high because two persons, one being a medical professional, have agreed on them. Studies reporting the reasons behind self-certified sickness absences are very scarce. In the Whitehall II study of British civil servants, the main reasons for short-term (<7 days) sickness absence were respiratory diseases, gastroenteritis, and headache/migraine (34). We are not aware of any studies that compare the short-term sickness absences of people with and without diabetes. Although diabetes is associated with serious bacterial infections such as pneumonia, sepsis, and tuberculosis, as well as skin and soft-tissue infections (35–37), it is unknown whether people with diabetes are more susceptible to acute viral infections such as “the common cold” and gastroenteritis or migraine and headache (35–39). However, in our study, individuals with diabetes seldom had ≥14 work disability days owing to infectious diseases, respiratory diseases, diseases of the digestive system, or skin diseases.

We had no data on the length of stay of immigrants in Sweden, and our data did not include information on second-generation immigrant status. However, people in our data had been living in Sweden at least 4 years prior to the beginning of follow-up. In addition, all covariates were measured at the study baseline, which may have introduced some inaccuracy to the associations if, for example, the family situation changed during the 4-year follow-up.

In summary, this study demonstrates that diabetes is associated with higher levels of work disability owing to comorbid diseases, particularly mental disorders, musculoskeletal diseases, and circulatory diseases. The most vulnerable groups were found among individuals who had mental or musculoskeletal disorders and lived alone, were immigrants, or were single parents. For the management of diabetes and the prevention of work disability among people with diabetes, it is highly important to monitor comorbid conditions and pay attention to the most vulnerable socioeconomic groups.

Funding. M.V. and J.E. are supported by the Academy of Finland (258598 and 265174), and J.E. is also supported by the Finnish Work Environment Fund (114260). E.M.-R. was supported by the Swedish Research Council. P.T., L.K., and K.A. had funding from the Swedish Research Council for Health, Working Life and Welfare. T.L. was supported by the Association for Promotion of Occupational Health.

The study sponsors had no role in the study design, analysis, or interpretation of data or the preparation, review, or approval of the manuscript.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. M.V. wrote the manuscript and researched data. J.E. performed statistical analyses, wrote the manuscript, and reviewed and edited the manuscript. E.M.-R., P.T., L.K., and K.A. obtained and managed the data, helped with study design and statistical analysis, and reviewed and edited the manuscript. T.L. helped with study design and reviewed and edited the manuscript. J.P. helped with study design and statistical analyses and reviewed and edited the manuscript. M.V. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.