Article Text
PDF
Abstract
Background Low IQ has been shown to be an important risk factor for disability pension (DP) but whether the importance has changed over time remains unclear. It can be hypothesised that IQ has become more important for DP over time in parallel with a more demanding working life. The aim of this study was to investigate the relative risk of low IQ on the risk of DP before age 30 between 1971 and 2006.
Methods This study covered the entire Swedish male population born between 1951 and 1976, eligible for military conscription. Information about the study subjects was obtained by linkage of national registers. Associations between IQ and DP over time were analysed by descriptive measures (mean values, proportions, etc) and by Cox proportional hazards regressions. Analyses were adjusted for educational level.
Results The cohort consisted of 1 229 346 men. The proportion that received DP before the age of 30 increased over time, from 0.68% in the cohort born between 1951 and 1955 to 0.95% in the cohort born between 1971 and 1976. The relative risk of low IQ (adjusted for education) in relation to high IQ decreased from 5.68 (95% CI 4.71 to 6.85) in the cohort born between 1951 and 1955 to 2.62 (95% CI 2.25 to 3.05) in the cohort born between 1971 and 1976.
Conclusions Our results gave no support to the idea that the importance of low IQ for the risk of DP has increased in parallel with increasing demands in working life. In fact, low IQ has become less important as a risk factor for DP compared with high IQ between the early 1970s and 1990s. An increased educational level over the same time period is likely to be part of the explanation.
- COGNITION
- Cohort studies
- Life course epidemiology
Statistics from Altmetric.com
Introduction
The increased prevalence of disability pension (DP) among young adults (below age 30) in Sweden, as well as in many other European countries, over the past two decades1 calls for studies exploring the underlying factors of the increase. Intellectual performance (denoted here as IQ) and educational level have both been shown to be of major importance for receiving DP in early age.2 ,3 An inverse dose response relationship between IQ and risk of long-term sick leave as well as DP has been reported in previous studies.2 ,4 ,5 The underlying mechanisms behind the association between IQ and DP might be related to health (ie, IQ>health>DP). Individuals with lower IQ have been shown to have an increased risk of morbidity and mortality.6–10 Furthermore, educational level might modify the association between IQ and DP (ie, IQ> education>DP). Actually, having a higher education seems to be a protective factor irrespective of IQ level on the risk of attaining a DP.2 In addition, IQ might also directly affect DP (ie, IQ>DP).
The time trends between IQ and DP have not yet been addressed, that is, if low IQ has increased as a risk factor for DP over time. Henderson et al4 examined the association between IQ and long-term sick leave (DP) in three birth cohorts (born in 1946, 1958 and 1970). However, information on sickness absence was self-reported and sickness absence was measured at different ages in the cohorts, not exploring period effects. Thus, even though the association between IQ and long-term sick leave (DP) was shown to be relatively stable, time trends remain to be studied.4
Over time, the demands of working life have altered; manual work has changed towards more qualified and computerised work. It might therefore be hypothesised that higher cognitive demands are required in working life today as compared with 20 years ago. It is therefore possible that, over time, low IQ has become a stronger risk factor for DP. Even if IQ has increased slightly over time in absolute sense11 the proportion of individuals with low IQ has remained fairly stable and has not increased over time. Therefore, low IQ must have become a stronger risk factor over time if it is to explain the increase in DP over time. If not, factors other than IQ must be behind the increase in DP.
Material and methods
Population and record linkage
This study covers the entire Swedish male population born between 1951 and 1976, eligible for military conscription; in total 1 317 249 men. Conscription examinations were mandatory for all men with Swedish citizenship. Men with severe chronic diseases or mental retardation were exempted from conscription examinations. During the study period over 90% of the male population attended conscription examinations. Missing information on IQ could be the result of not attending conscription (at all or partly) or due to administrative loss of data.
Information about the study subjects was obtained by linkage between several of the national registers in Sweden using the unique personal identification number ascribed to all individuals with permanent residence in Sweden. The Register of the Total Population was linked with the Military Conscription Register, the Population and Housing Censuses (PHCs), and the longitudinal integration database for health insurance and labour market studies (LISA). Information on DP was collected from the Swedish Social Insurance Agency. Mortality data was obtained from the Cause of Death Register and migration data from RTP.
The study population was followed in the registers mentioned above from conscription until the end of 2006. Age was calculated based on birth year, month and day, which means that everyone was followed until exactly 30 years of age.
Ethical considerations
The study has been approved by the regional Ethics committee at the Karolinska Institutet, Stockholm, Sweden (reference number 2005/1501-31/3).
Outcome variables
Information on DP was available between the years 1971 and 2006. According to Swedish legislation, DP can be granted by the Swedish Social Insurance Agency if working ability is chronically reduced with at least 25% due to disease. To be eligible for DP a certificate issued by a physician must affirm the disability. Until 2003, individuals in the age span 16–64 years were eligible for DP. In 2003, the administration of DP was transferred from the Swedish pension system to the Swedish sickness insurance system. The lower age limit was set to 18 years of age. In 2003, the youngest birth cohort in the study population (born 1976) was 27 years of age; why these changes should not be of importance in this study.
Explanatory variables and confounders
IQ was measured at military conscription at a mean age of 18.3 years (SD 0.55 years). The IQ test, which has been described in detail elsewhere,12–14 consisted of four subtests: a logical (capacity to understand written instructions and apply them to solving a problem), verbal (knowledge of synonyms), spatial (respondents were shown a schematic diagram of an object in its preassembled, two-dimensional state and were then required to identify the correct three-dimensional object from a series of drawings) and a technical (mathematical/physics problems, measuring a component of general knowledge) test. All test scores—including a global IQ score (measuring general ability, g) derived from a summation of the results of the four subtests—were standardised against the entire population of conscripts from the previous years to follow a Gaussian distribution between 1 and 9, with a mean of 5 and SD of 2. The stanine scale corresponds to the traditional IQ scale (with mean 100 and SD of 15) so that 1 equals IQ under 74, 2 equals IQ 74–81, 3 equals IQ 82–89, 4 equals IQ 90–95, 5 equals IQ 96–104, 6 equals IQ 105–110, 7 equals IQ 111–118, 8 equals IQ 119–126 and 9 equals IQ 126 and over.15 ,16
Information on highest attained education was collected from PHCs for 1970 and 1990 and from LISA between 1990 and 2004. In Sweden, 9 years of schooling was compulsory during the study period. After 9 years of schooling, individuals could continue by choosing a practically oriented education or a theoretically oriented education. In the analyses, level of education was used as a categorical variable, rated as: low education, corresponding to 9 years of schooling, medium education corresponding to additional secondary schooling of maximum of 3 years and higher education corresponding to more than 3 years of secondary schooling (mostly university education).
Statistical methods
First we analysed the association between IQ and DP over time by calculating several descriptive measures (mean values, proportions, etc). In order to capture the period effect, five cohorts were constructed by clustering 5-year birth cohorts: 1951–1955, 1956–1960, 1961–1965, 1966–1970 and 1971–1976. Next, the descriptive measures were presented stratified by educational level—since educational level has changed over time and shown to be a risk factor for DP. Finally, Cox proportional hazards regressions17 were calculated in order to estimate the relative risk (RR) of low IQ vs high IQ on DP for each of the birth cohorts. In the regression models, the RR of low IQ (score 1 and 2 on the stanine scale) in comparison with that of medium (3–6 on the stanine scale) and high score (7–9 on the stanine scale) was calculated. Univariate as well as multivariate models (adjusted for education) were calculated. Analyses were additionally adjusted for region of living and parental education, but it had very limited impact on the estimates and was thus not included in the final models. The follow-up ended at the date of DP, date of emigration or on 31 December 2006, whichever came first. HRs were estimated together with their 95% CIs, using the PHREG procedure in SAS V.9.3.
Results
The cohort included all those born between 1951 and 1976 who were registered in the military conscription register, and totalled 1 317 249 Swedish men. After exclusion of men with missing information on IQ (87 903 men) the study population consisted of 1 229 346 men.
Table 1 presents descriptive information about the outcome (DP before age 30), the exposure (IQ) and educational level. The proportion that received DP before the age of 30 has increased over time, from 0.68% in the cohort born between 1951 and 1955 to 0.95% in the cohort born between 1971 and 1976. The mean IQ of those who received DP was higher in the younger cohorts than in the older. Additionally, the mean age of those who received DP was somewhat higher in the younger cohorts than in the older. The proportion of low-IQ-scoring individuals among those who had received DP was larger in the older cohorts than in the younger, whereas the proportion of higher scoring individuals remained fairly constant. The educational level increased dramatically during the study period. For example, the proportion of low educated decreased from 51% in the cohort born between 1951 and 1955 to 30% in the cohort born between 1971 and 1976. This change was not only evident among those who received DP but also among the total population/entire birth cohorts, where the corresponding decline was 23% to 7%. The proportion with high education also changed over time, however, not as much as the low-educated group decreased—and the change was of similar size in the group having received DP as it was in the general population, see table 1.
Descriptive table of disability pension (DP), IQ and education by birth cohort
Table 2 shows the proportion in each cohort that received DP in three strata of IQ, low, medium and high IQ, that is, the prevalence of DP over time. The prevalence increased in all three IQ strata across the cohorts, from 2.5% to 3% in the low-IQ group, and from 0.28% to 0.4% in the high-IQ group. Table 3 shows the proportion that received DP in three educational strata. The increase of DP over time was similar in all three educational strata, even if the prevalence was higher among low-educated individuals.
Proportion of disability pension before age 30 in stratum of IQ
Proportion of disability pension before age 30 in stratum of education
Table 4 presents the RR of low IQ and medium IQ versus high IQ for DP in five cohorts. Model 1 presents the unadjusted RR and shows that the RR of receiving DP among individuals with low IQ in comparison with individuals with high IQ has decreased from 11.09 (95% CI 9.43 to 13.04) in the oldest cohort to 6.64 (95%CI 5.81 to 7.59) in the youngest cohort. This means that, over time, the risk for receiving DP among those with a low IQ decreased from a factor of 11 to 6 compared with those with a high IQ. Adjusting the model for educational level reduced the effect from IQ but did not change the gradient. The RR was then 5.68 (95% CI 4.71 to 6.85) for the cohort born between 1951 and 1955 and 2.62 (95% CI 2.25 to 3.05) for the cohort born between 1971 and 1976. No gradient was seen for medium IQ versus high IQ over time.
Relative risk with 95% CIs for disability pension before the age of 30
Since the causal pathways between DP and education may go both ways, that is, having a low education might be a cause of DP or early DP might be the cause of low educational achievement, and this may affect the analyses of IQ and DP, sensitivity analyses were performed. In these analyses those who received a DP under the age of 25 were excluded in order to allow enough time for everyone to achieve a high education. The estimates changed somewhat but the gradient and the main conclusions remained the same (data not shown).
Table 5 presents the RR of low IQ versus high IQ for DP stratified by educational level. The relative effect of low IQ over time/cohort was seen in all educational strata, from 4.69 (95% CI 3.37 to 6.51) to 2.48 (95% CI 1.59 to 3.89) in the group with the lowest education, 5.35 (95% CI 4.12 to 6.93) to 2.69 (95% CI 2.18 to 3.32) in the medium educational group and from 8.81 (95% CI 3.96 to 19.60) to 4.60 (95% CI 2.86 to 7.41) in the group with the highest education.
Relative risk with 95% CIs for disability pension before the age of 30 of low IQ (1 and 2) versus high IQ (7–9) for different cohorts and in three educational strata
Discussion
The basis for this study was the idea that having a low IQ might have increased as a risk factor for DP over time, in parallel with increasing demands in working life. However, our results gave no support to this idea. Even though the effect of low IQ was substantial for the risk of DP before the age of 30, in the same magnitude as previously reported,2 ,5 ,18 the increased prevalence of DP has been fairly evenly distributed across all IQ and educational strata. The regression analyses of IQ and DP showed that low IQ has become less important as a risk factor over the time period between the early 1970s and 1990s.
However, since educational level is correlated with IQ and is also an important risk factor for DP, analyses need to take education into account in order to accurately mirror the importance of IQ. The educational level in the population has increased over the study period making the group with a low IQ and a low education proportionally smaller over time. Since education is protective against DP and since education has increased relatively more in the low-IQ group than in the high-IQ group, the RR of low IQ has decreased over time. Looking into absolute risks revealed that the absolute risk of early DP for low-IQ-scoring individuals with a low education has actually increased over time but since this group has become smaller over time it has caused the RR of IQ to decrease (results not shown but available on request). Educational level thus appeared to partly mediate the association between IQ and long-term DP (sickness absence), which has also been found in earlier studies.4 Since education is such an important factor, and attaining an early DP might hinder attainment of a higher education, sensitivity analyses were performed excluding all those who had received their DP before the age of 25. The results of the sensitivity analyses did not change the main results. Thus, even if reverse causation may still occur to a certain extent, it cannot explain the time trends of low IQ on early DP.
If low IQ cannot explain the increase in DP among young individuals over time, explanation must be sought elsewhere. One such explanation could be that younger cohorts might be more prone to apply for DP.19 If this is the case, and if willingness to apply for benefits is correlated with higher IQ, the time trends might be explained by a behavioural change among younger cohorts where application for benefits has become more accepted.
Another explanation could be that mental disorders have increased over time and may have become a more accepted reason to apply for DP. The Swedish Social Insurance Agency has analysed the prevalence of the diagnoses for which DPs are approved, which is presented in a Swedish report, (http://www.forsakringskassan.se/wps/wcm/connect/1b385241-a51d-421d-a369-46ffef1a0faa/redovisar_2007_03.pdf?MOD=AJPERES). This report states that, in principle, the entire increase in DP can be attributed to psychiatric disorders. Earlier research has found Swedish men with lower IQ to have higher admission rates for mental disorders.20 In addition, Upmark et al21 reported that having a lower IQ was a strong predictor of DP with psychiatric diagnoses. However, our results showed that the increase of DP has taken place in all IQ and educational strata. One explanation might be that the association between IQ and psychiatric disorders has diminished over time, like the association between IQ and DP.
Strength and limitations
The most important strength of this study is the access to register data on the exposure and the outcome. In addition, high-quality register data were available for the entire Swedish population diminishing the bias of non-response. This is important since both the exposure and the outcome are likely to be correlated to the probability of answering or participating in surveys. The use of register data also minimises loss to follow-up. If misclassification is present, it is likely to be randomly distributed; hence only driving the estimated risks towards null. The measure of IQ at conscription examinations has been shown to provide valid estimates with good prediction values for many later outcomes.6 ,22 ,23 It should also be stressed that the present study excludes all those with diagnosed mental retardation since they are exempted from conscription examinations. Inclusion of these individuals would probably strengthen the association between IQ and DP but would not add information to the analysis of time trends. A limitation of the study is that women are not included in the study since female conscription examinations were rare during the study period. The results cannot be generalised to women and it would be of importance to study how the association between IQ and risk of DP has changed over time among women as well.
Conclusion
Over time, having a low IQ has become less important as a risk factor for early DP as compared with having a high IQ. Thus, the explanations for the increased prevalence of young adults with DP are not likely to be found from an association of changing of demands of working life and IQ. The study emphasises the importance of educational level as a protective factor. The results are important for policy makers in the design of preventive actions.
What is already known on this subject
The prevalence of disability pension among individuals below 30 years of age has increased in several countries. Low IQ has been shown to be an important risk factor for attaining a disability pension in several cohorts.
What this study adds
Over time, low IQ has become less important as a risk factor for early disability pension as compared with a high IQ. The explanation seems to be that the increase of disability pension has been evenly distributed across all IQ strata and that educational level has increased over time acting as a protective factor. Thus, low IQ cannot explain the increased prevalence of early disability pension.
References
Footnotes
Competing interests None.
Ethics approval The regional Ethics committee at the Karolinska Institute, Stockholm, Sweden (reference number 2005/1501-31/3).
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement The data underlying this study are protected by an agreement between Karolinska Institutet (KI), (FR) and Statistics Sweden. Data cannot exit KI without ethical approval and assurance that all data are handled according to Swedish legislation. However, full documentation of all data and analyses are stored within KI.