Article Text

Download PDFPDF

Educational differences in disability pension among Swedish middle-aged men: role of factors in late adolescence and work characteristics in adulthood
  1. Elin Johansson,
  2. Ola Leijon,
  3. Daniel Falkstedt,
  4. Ahmed Farah,
  5. Tomas Hemmingsson
  1. Division of Occupational and Environmental Medicine, Department of Public Health Sciences, Karolinska Institutet, Stockholm, Sweden
  1. Correspondence to Dr Tomas Hemmingsson, Division of Occupational and Environmental Medicine, Department of Public Health Sciences, Karolinska Institutet, Norrbacka 4th floor, SE 171 76 Stockholm, Sweden; tomas.hemmingsson{at}ki.se

Abstract

Background The association between level of education and disability pension (DP) is well known. Earlier studies have investigated the importance of early life factors and work characteristics but not in combination. The aim of this study was to investigate the association between level of education and DP among Swedish middle-aged working men and to what extent such an association can be explained by factors measured in late adolescence and work characteristics in adulthood.

Methods Information about IQ, health-related lifestyle factors, psychiatric and musculoskeletal diagnoses was obtained from the 1969 conscription cohort, consisting of 49 321 Swedish men. Data collected when subjects were 18–20 years of age were combined with national register-based information about level of education, job control and physical strain at work in adulthood, and information about DP between 1991 and 2002.

Results There was a strong graded association between level of education and DP. Those with the lowest level of education had a four times greater probability of having DP as compared with those with the highest level. In multivariable analyses, factors measured in late adolescence, IQ in particular, attenuated the association more than work-related characteristics in adulthood.

Conclusions The authors found an association between level of education and DP among Swedish middle-aged working men. A large part of the association was explained by factors measured in late adolescence, IQ in particular, and somewhat less by work characteristics measured in adulthood. Level of education remained as a significant predictor of DP in middle age after full adjustment.

  • Education
  • disability pension
  • adolescents
  • cognitive ability
  • work characteristics
  • child health
  • exercise
  • health behaviour
  • lifestyle
  • obesity
  • occupational health
  • health promotion
  • low back pain
  • physical activity
  • epidemiology
  • inequalities
  • ischaemic heart disease
  • mental health

Statistics from Altmetric.com

Introduction

About 8% of the Swedish population, aged 16–64 years, received full time or part time disability pension (DP) in March 2010.1 ,2 The diagnoses most often applied for are musculoskeletal and psychiatric disorders.3 Financing DP is a considerable economic burden on society, not only in Sweden.

Several studies have concluded that less years of education increase the probability of receiving DP in middle age.4–7 Risk factors for DP may be found early8 as well as later in life.9 An association between level of education and DP may be confounded by early life risk factors or mediated by adult life circumstances (figure 1).

Figure 1

Graphical representation of potential contributions of factors in late adolescence and in adulthood (work conditions) in explaining the association between level of education and disability pension in early middle age.

It has been shown that work characteristics may play an important role as mediator in the relationships between educational level and health outcomes.10 Certainly, the relationship between physical aspects as well as psychosocial aspects of work and DP in middle age is well established.4 ,9 ,11 In a Norwegian study, Krokstad et al4 found that occupational factors attenuated the association between level of education and DP to a larger extent than medical factors.

However, the association may be explained by different factors throughout the life course, some of which may be confounders in the association between educational level and DP. Among them are health-related lifestyle factors, often established in adolescence.12 Associations between overweight, smoking, risky use of alcohol and DP have been found in many studies, although these health-related lifestyle factors are often measured in adulthood.13–15 Psychological disabilities, musculoskeletal disorders and IQ in young age also seem to be risk factors for DP and may thus contribute to the association between level of education and DP in terms of confounding.14 ,16 ,17 IQ may be determined by level of education and to some extent act as a mediator.

The aim of the present study was to investigate the association between level of education and DP among Swedish middle-aged working men and to examine to what extent this association can be confounded or mediated by factors in late adolescence and work characteristics in adulthood. No earlier study on the association has been conducted where factors measured in late adolescence (prior to labour market entry) as well as later work characteristics have been included as possible explanatory risk factors.

Methods

Study population

The study population consisted of 49 321 Swedish men, born 1949–1951, who conscripted for compulsory military service in 1969–1970. A large majority (97%–98%) of all Swedish men aged 18–20 years went through conscription those years. Men with severe handicaps or congenital disorders were exempted. At conscription, all young men went through tests, interviews and examinations at any of seven conscription centres. The cohort and potential explanatory factors measured at conscription have been described more comprehensively elsewhere.13 ,16 The cohort was followed up 1991–2002 when information on DP was obtained.

Only subjects in the labour force and under risk of becoming disability pensioners were included in the study. Subjects alive and not receiving DP until 1990 and who had reported a job title in the National Population and Housing Censuses 1990 were considered as being part of the labour force (table 1). Subjects with missing data on any variable were excluded.

Table 1

Description of the cohort (full cohort n=49 321), exclusion criteria, number of excluded subjects and number of subjects remaining for follow-up

Level of education

Information about highest level of education was collected at age 39–41 years. This information was not available before 1990 but most subjects had completed final level of education before this age. Data were obtained from the Longitudinal Database of Education, Income and Occupation (LOUISE) administered by Statistics Sweden.14 LOUISE is a longitudinal database containing information about education, income and employment and includes all Swedish citizens from 16 years of age. The subjects were divided into five categories according to highest education obtained: ≤9, 10–11, 12–13, 14 and ≥15 years. The five categories are based on SUN (Swedish education nomenclature, Statistics Sweden) where education is coded according to Swedish standards and correlate with elementary school, upper secondary school, university education <2 years, university education ≥2 years and at least 3 years of university education. Fifteen years of education or more (at least 3 years of university education) was set as the reference category.

Disability pension

The outcome was received DP from 1 January 1991, when the subjects were 39–41 years of age, until the end of 2002. Information on income from DP was obtained from the LOUISE database,14 described above.

The number of new disability pensioners increased each year during the follow-up. In 1991, in the beginning of the follow-up, 118 subjects received DP compared with 340 subjects in 2002 (not shown in table).

Potential explanatory factors

On a theoretical basis factors that earlier studies have reported as associated to DP were selected and, thus, could potentially explain the association between level of education and DP.4 ,5 ,7–9 ,11 ,18 ,19

Factors measured in late adolescents

From the examination at conscription, we used six potential explanatory factors for this study: IQ, smoking, alcohol consumption, overweight, psychiatric and musculoskeletal diagnoses.

The IQ tests performed included tests of logic/general intelligence, verbal tests of synonym detection and further tests of visuospatial/geometric perception and technical/mechanical skills on the basis of mathematics/physics problems. All the tests were progressive, starting with relatively simple questions that then became more difficult. The logic/general intelligence test consisted of 40 items and took 12 min. Subjects were presented with an array of shapes and letters in combination and were asked to ‘strike through the square under the longest word’ (out of five). The verbal IQ tests lasted for 7 min and consisted of 40 questions. The items consisted in rows of five words, and subjects were instructed to underline the odd one out of, say, ‘teapot, sandwich, milk, egg, meat’, or—with greater difficultly—‘insist, request, question, believe, admire’. The visuospatial test took 40 min and involved the presentation of probe geometric shapes (eg, an isosceles triangle bisected by a vertical line) followed by four items consisting of different pairs of triangles of different sizes and with different orientations. Subjects were asked to state which options they had to make up the probe item. The outcome of each test was ranked 1–9. The standard-nine values were transformed into a composite standard-nine scale to measure general ability, corresponding to approximate IQ bands of <74, 74–81, 82–89, 90–95, 96–104, 105–110, 111–118, 119–126 and >126.15

Overweight, smoking and risky use of alcohol were potential explanatory factors used as indicators of unfavourable health-related lifestyle factors established in adolescence. Measurements of body height and weight of each conscript were used to calculate body mass index (BMI) by using bodyweight (kilograms) divided by height (metres) squared. Conscripts with a BMI ≥25 were considered overweight. All conscripts were asked to fill in two questionnaires concerning items on social background, health-related life style factors and adjustment, psychological factors and health, and use of substances such as alcohol consumption and tobacco smoking. Men who reported smoking 5 cigarettes/day or more was categorised as smokers. Men who reported at least one of the following indicators of problem drinking were categorised as risky users of alcohol: consumption of at least 250 g of 100% alcohol per week, to have taken a ‘hair-of-the-dog’ during a hangover (ie, to drink alcohol in the morning in order to ease hangover), to have been apprehended for drunkenness or to have often been drunk (the alternatives given were ‘often’, ‘rather often’, ‘sometimes’ and ‘never’).

All men were examined by a physician and information about any musculoskeletal diagnose was obtained according to ICD-8 (710–738). If the physician suspected psychological disorders, the conscript was seen by a psychiatrist. The conscript was then, if considered appropriate, given a psychiatric diagnose according to ICD-8 (290–315).

Work characteristics in adulthood

The National Population and Housing Censuses 1990 provided information on occupation, when subjects were 39–41 years old. All jobs were coded according to Nordic Occupation Classification (NYK),17 which corresponds to the International Classification (ISCO-88).20 Thereafter, work characteristics were determined using two different exposure classification systems to assess the average level of job control and physical strain at work, respectively, in different occupations.

In order to classify the subjects according to job control, we used an exposure classification system based on questions from the Swedish Work Environment Surveys 1989–1997. These surveys included data on almost 49 000 Swedes. The questions concerned different aspects of the working conditions, for example, job control, formed by measures of decision authority and skill discretion. Three hundred and twenty different job titles were classified according to the exposure classification system.21 It was then applied to rate all job titles on a 10 digit scale concerning job control which made it possible to classify subjects into one of four equally large groups: high, medium–high, medium–low or low job control.

The exposure classification system referring to physical workload was based on the Swedish Annual Level-of-Living Surveys (1979–1981). The exposure classification was founded on an index regarding eight physical risk factors: heavy lifts daily, repetitive and one-sided work movements, awkward work postures, heavy shaking or vibrations, daily perspiration from physical exertion, contact with dirt, deafening noise, and risk for exposure to accidents.22 ,23 Job titles were then classified into one of the four groups, ranked by the proportion of employees reporting different strains at work: low, medium–low, medium–high or high physical strain at work.

Data analysis

Prevalence of DP and potential explanatory factors was calculated for each level of education. All factors were kept throughout the analysis.

In order to examine the association between level of education and DP, HRs with 95% CIs were estimated by Cox proportional hazards model. Analyses of crude HR and 95% CI for receiving DP were conducted for all potential explanatory factors.

Those who died or emigrated were censored on the day of death or emigration and those received DP were censured on July 2nd the year of receiving DP. Since data on DP were based on annual income from DP, the exact date of first received DP is unknown. The date was set as close to the middle of the year as possible, at July 2nd, of the year when income from DP was first registered.

Adjustments were made for potential explanatory factors measured in late adolescence (age 18–20) as well as work characteristics measured later in life (age 39–41). In the first step, no adjustments were conducted. In the second step, all potential explanatory factors measured in late adolescents, that is, IQ, health-related lifestyle factors, psychiatric and musculoskeletal diagnosis, were adjusted for, one by one, and simultaneously. In the third step, work characteristics measured later in life, that is, job control and physical strain at work, were adjusted for, one by one, and simultaneously. In the fourth and final step, an analysis with adjustments for all potential explanatory factors together was conducted. All the statistical analysis was conducted in SAS V 9.1® statistical package software. The assumption of proportional hazards was checked with the ‘Supremum test for proportional hazards assumption’ in SAS. The Stockholm Regional Ethical Review Board approved the study.

Results

The full cohort consisted of 49 321 male subjects. After exclusion of 9419 subjects who were either dead, receiving DP before 1990 or had missing data on any variable, 39 902 subjects remained for follow-up (table 1).

All factors in late adolescence (low IQ, unfavourable health-related lifestyle factors, psychiatric and musculoskeletal diagnosis) and work characteristics in adulthood (low job control and high physical strain at work in 1990) were associated with a higher probability of receiving DP (table 2).

Table 2

Crude analyses, unadjusted HRs and 95% CIs, on the association between intelligence (IQ), body mass index (BMI), smoking, risky use of alcohol, psychiatric diagnosis, musculoskeletal diagnosis at conscription in 1969–1970, job control, physical strain at work in 1990 and the risk of disability pension in 1991–2002

During the follow-up period, 1977 of the 39 902 subjects were receiving DP, that is, 5%. The number of disability pensioners increased with lower level of education (table 3) and was 7.2% among those with the lowest level of education compared with 1.9% among those with the highest level.

Table 3

Number and percentage of disability pensioners in 1991–2002 in each level of education and prevalence (%) of different risk indicators by level of education

In the group with the highest level of education, 69.3% had a high IQ at 18–20 years of age compared with 9% in the group with the lowest level of education (table 3). Overweight, smoking and risky use of alcohol in late adolescence were clearly more common with lower levels of education. The same trend was seen for psychiatric diagnoses in late adolescence but not for musculoskeletal diagnoses (table 3).

Those with higher level of education much more often had high-control jobs than individuals with less education: 57.6% in the group with highest level of education compared with 7.8% in the group with lowest level (table 3). Conversely, subjects with lower level of education more often had jobs that involved high physical strain: 45.2% in the group with the lowest level of education compared with 1.5% in the group with the highest level.

In table 4, we present RRs (HRs and 95% CI) for receiving DP by level of education, with adjustments for IQ, health-related lifestyle factors, psychiatric diagnosis, musculoskeletal diagnosis, job control and physical strain at work. Risk reductions were not equally large for each level of education. For all potential explanatory factors presented, the lower the level of education, the larger the attenuation of the association between level of education and DP.

Table 4

HRs and 95% CIs for the risk of disability pension 1991–2002 by level of education in 1990: adjusted for single and combined factors in late adolescence (ie, IQ, health-related lifestyle factors, psychiatric diagnosis and musculoskeletal diagnosis) and work characteristics (ie, job control and physical strain at work) in adulthood

In the first step, no adjustments were done. A clear gradient in increasing RR of receiving DP with lower level of education was demonstrated. The risk for receiving DP was four times higher in the group with lowest level of education compared with the group with highest level of education. In the second step, adjustments for factors in late adolescence (IQ, health-related lifestyle factors, psychiatric and musculoskeletal diagnosis) were done, for each potential explanatory factor separately and for all factors simultaneously (table 4). Altogether, these factors attenuated the association between level of education and DP by 33%–63%. IQ was the single factor that attenuated the association most. Health-related lifestyle factors attenuated the association to some extent. The attenuation from psychiatric or musculoskeletal diagnosis was negligible. In the third step, adjustments for work characteristics in adulthood (job control and physical strain at work) were done, both separately and simultaneously (table 4). Together, work characteristics attenuated the association between level of education and DP by 22%–47%. In the fourth and final step, adjustments for all potential explanatory factors were done simultaneously. Together, factors in late adolescence and work characteristics in adulthood attenuated the association between level of education and DP by 44%–78%.

Discussion

In summary, we found a strong association between level of education and DP in Swedish middle-aged working men. Potential explanatory factors measured in late adolescence and work characteristics measured in adulthood explained a large part of the association. Factors measured in late adolescence, IQ in particular, explained the larger part.

Comparison with previous studies

In concordance with earlier studies, we found that the prevalence of disability pensioners was increasing with lower level of education.4 ,9 ,11 A few previous studies have investigated potential explanatory factors for the association between level of education and DP and long-term sickness absence.4 ,24 In those studies, all potential explanatory factors were measured at one occasion in adulthood only. Moreover, IQ was not included and occupational factors were based on self-reports. In contrast to the present study, those studies found that work-related factors attenuated the association more than health-related lifestyle factors. In the present study, we found factors measured in late adolescence to be more important. Work characteristics in adulthood explained only about 15% after adjustments had been made for factors measured in late adolescence. Our findings address the importance of applying a life course perspective when investigating the association between level of education and DP. Health-related lifestyle factors are often established in adolescence and thus affect health in adulthood. Childhood adversities8 and musculoskeletal disorders5 are also known risk factors for DP. Lower IQ in young age have also previously been associated with DP in early adult life.12

Gravseth et al25 explored the association and interrelation of level of education, IQ, mental function, BMI and height for the risk of receiving DP between 24 and 36 years of age.25 They found, like in the present study, that both low level of education and low IQ were independently associated with increased risk for DP. Because of the short time of exposure to bad working conditions at such early age, one could assume that work characteristics would not have a major influence on the association. In the present study, subjects were over 40 years of age and had been exposed to occupational risk factors for several years, but despite this, we found factors measured in late adolescence and IQ in particular to be more important than work characteristics.

Interpretation of the results

Factors prior to labour market entry, IQ in particular, are likely to be important predictors of future occupation. Individuals with lower IQ more often end up in jobs that are more physically and psychosocially demanding.26 Such occupations have been recognised as predictors for poor health and DP.4 ,7 ,9 ,11 However, during the last decade, a large number of studies have shown that a lower IQ is a risk factor for mortality and morbidity in its own.27 ,28 This might explain the somewhat surprising result that IQ explained a larger part of the association between level of education and DP than medical diagnoses and health-related lifestyle factors, more obviously connected to health.

Even though a large part of the association between level of education and DP was explained by factors measured in late adolescence and work characteristics in adulthood, a significant association remained. Lower level of education may contribute to an increased risk of DP. For instance, it is suggested that more years of schooling is associated with higher capabilities to manage health problems and with reduced poor health-related behaviour.29 ,30 It is, of course, also possible that our analysis did not include all factors explaining the association between level of education and DP. However, earlier studies have shown that, for instance, family-related factors such as marital status and number of children had no substantial effect on the association between level of education and DP.7 ,31

Methodological considerations

The main strength of this study was the possibility to combine data on IQ, health-related lifestyle factors, psychiatric and musculoskeletal diagnoses at labour market entry with register-based prospective information, with minimal loss, on work characteristics, level of education and DP. The cohort includes a large study population, representative of Swedish males born around 1950. DP is an unspecified health conditions in most cases. In this study, we did not have information on specific medical causes of DP. This must be considered as a limitation.

However, there is some misclassification in the registers containing information on educational attainment, especially among those classified as having a low level of education.14 Since the group classified with low education actually contained subjects with more education, some underestimation of the association between level of education and DP probably occurred.

Potential explanatory factors measured in late adolescence were measured at age 18–20. Thus, bias that could occur when retrospective information is used was therefore minimised. Information about overweight, smoking and risky use of alcohol must, to some extent, be misclassified as indicators of health-related lifestyle factors at the beginning of the follow-up 20 years later. Overweight must have increased and smoking decreased considerably among the men in middle age. However, social disparities persist.32

In our analyses, we have included all variables as categorical variables, that is, using two to five categories (see tables). In additional analyses, we used a more refined categorisation or continuous variables. Using a more refined categorisation or continuous variables did not change the results. In our analyses, we adjust for, for example, job control in four categories (low, medium–low, medium–high and high control). The information on job control in its original form consists of values on a scale from 1.00 to 9.99. In the additional analyses, we adjusted for job control as a continuous variable. The attenuating effect from adjusting for job control in the form of a continuous variable or in four categories in the association between educational level and DP was very similar in those analyses.

The differences in IQ between different educational levels may partly be a result of schooling already at age 18 for some part of the cohort, that is, between those who did not continue at school after age 16 and all other groups (in the group classified as having had 10–11 years of schooling, 11 years was most common) and partly act as a mediator. In a recent Swedish study, it was estimated, in agreement with several previous studies from other countries, that each extra year of education may correspond to an increase in IQ of approximately 2 points.33

Work characteristics were classified according to job-exposure classification systems based on job title. Such an approach has been proposed as a more objective way to collect information than self-reports, though inferred information is not sensitive to individuals' claims and expectations.21 A disadvantage with using inferred information is that some of the true variations are eliminated which may lead to underestimation of true associations. It is often argued that the inferred information based on job title is preferable since such information is not sensitive to the influence of individual characteristics such as different kinds of health problems. However, in the Swedish SHEEP study and in the British Whitehall study, associations between job control inferred from job titles and coronary heart disease and self-reported job control and coronary heart disease were very similar.34 ,35 Therefore, it may be argued that inferred data on work characteristics may be considered a valid source of information on, at least, level of job control. Data on work-related exposures were measured only once, that is, in 1990, but could have changed over time. We conducted additional analyses where we restricted the data set to those stable in one of the four groups of job control (low, medium–low, medium–high and high control) between the years 1985 and 1990. In these analyses, it was shown that the crude association between level of education and DP was very similar. The attenuating effect from adjusting for IQ, which is likely to be more stable over time, was very similar and the effect from adjusting for job control was also very similar. The differences in attenuating effect between IQ and work characteristics may not to a large extent be explained by differences in measurement precision.

These job-exposure classification systems have been used in previous studies.23 ,36

An obvious limitation of the present study was that it only included men.

Conclusions

We found an association between level of education and DP among Swedish middle-aged working men. Although a large part of the association was suggested to be explained by factors measured in late adolescence, IQ in particular, and somewhat less by work characteristics measured in adulthood, level of education remained as a significant predictor of DP in middle age after full adjustment.

What is already known on this subject

  • The association between level of education and disability pension is well known.

  • Earlier studies have investigated the importance of early life factors and work characteristics but not in combination.

What this study adds

  • This study showed that a large part of the association between level of education and disability pension was explained by factors measured in late adolescence, IQ in particular, and somewhat less by work characteristics measured in adulthood.

  • In order to lower the risk of disability pension in middle age factors from early life as well as work-related factors must be considered.

References

Footnotes

  • Funding This study was supported by the Swedish Council for Working Life and Social Research (Project No. 2008-0907).

  • Competing interests None.

  • Ethics approval Karolinska Institutet.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.