In a Norwegian cohort of middle-aged men and women, we found no adequate explanation for women’s higher likelihood of disability pension, except for a moderate impact of self-perceived health. Further adjustment for family situation, occupational class and working hours per week did not influence women’s higher disability risk in the total cohort. However, in analyses stratified by educational level, these factors fully explained women’s excess risk of disability pension among the highly educated, but not among the less educated.
Some have argued that health plays only a minor role in exit from working life in modern societies . However, several studies have shown that self-perceived health is a strong predictor of subsequent sickness absence and disability pension across gender and social class [6, 17, 24–26]. In the current study self-perceived physical health was strongly associated with disability pension in both genders, while self-perceived mental health was associated only among men. However, in the multivariate analyses both dimensions of self-perceived health contributed to explain women’s excess disability risk. This impact was found in the total cohort and in strata of educational level. A Norwegian study found that among individuals on sick leave for more than eight weeks due to musculoskeletal disorders, women less than 50 years of age had a higher risk of disability pension compared with men in the same age group . Self-perceived physical health may reflect musculoskeletal disorders in both genders, but the higher impact on women’s disability risk may be due to an earlier chronicity among women. Further, Laaksonen et al. found that self-perceived health and self-reported diagnoses explained the gender difference in long-term sickness absence among middle-aged . In a Norwegian study, self-perceived global health had little impact on women’s excess risk, while adding mental distress significantly reduced the gender difference . However, the addition of several self-reported health measures in the current study provided no further explanation of women’s excess risk of disability pension.
The relation between self-perceived mental health and risk of disability pension did not display a linear trend, a finding that may be related to the construction of the physical and mental summary SF-12 scales providing uncorrelated (orthogonal) factors, a construct that is theoretically unlikely . Using the multidimensional SF-36 questionnaire, Laaksonen et al. found a strong association between self-perceived physical health and sickness absence while mental health was only weakly associated . Studies using the Hospital Anxiety and Depression Scale (HADS), on the other hand, have found linear associations between higher scores on self-reported anxiety/depression and subsequent disability pension [28, 29].
In line with our findings, a Norwegian study found that living with children younger than 7 years of age diminished the likelihood of disability pension among women . However, the current study supports the general impression that marital status, children in the household or a retired spouse all have little impact on the gender difference in disability pension [4, 5, 8]. Women’s increased participation in working life has brought about some changes in traditional family roles and expanded the use of part-time jobs. With respect to the interface between work and family, findings indicate that both genders are at risk of sickness absence when experiencing negative interference from work on the family situation [30, 31].
Occupational social class
As confirmed in the literature, the risk of disability pension increases substantially with lower occupational social class [14, 22]. Some studies have suggested that women’s excess risk of sickness absence and disability pension is more likely explained by vertical gender segregation than by horizontal segregation [2, 32]. Vertical segregation refers to the unequal gender distribution in occupational social class, while horizontal segregation denotes the tendency of men and women to work in different occupations . In the current study, controlling for occupational class (vertical segregation) in the total cohort, did not contribute towards explaining women’s excess risk of disability pension, thus confirming the findings from a previous Norwegian study .
Nevertheless, in the analyses stratified by educational level we found that the risk of disability pension among highly educated women was considerably influenced by occupational class. One explanation is that higher education among women, more frequently than among men, is translated into occupational classes with less power, less autonomy and less status and thereby implies an excess risk of disability pension . In the current study, the different distribution of men and women by occupational class supports this interpretation. Among the lower educated, occupational class did not explain women’s higher disability risk, thus supporting studies that find male workplaces within the manual occupations particularly hazardous to health [32, 35].
In the current study, part-time work less than 27 hours per week was associated with increased risk of disability pension in both genders, but did not contribute to explain women’s higher risk of disability pension in the total cohort [5, 13]. One might question whether the risks associated with part-time work are linked with the reduced number of hours at the work place, and conversely the higher number of hours outside the work place, or are merely a reflection of poor health and reduced work ability among part-time workers. Considering the unequal distribution of men and women in part-time positions, reasons for part-time work most likely differ by gender. In the current study, the relatively high risk of disability pension among men working part-time indicates reduced work ability. Among women, however, part-time work is often preferred to ease the double burden of work and family obligations [7, 36]. When working hours per week did not influence the gender difference in disability pension in the total cohort, this may be related to the unequal gender distribution among part-time workers and the selection of men with reduced work ability into part-time positions.
However, in the analyses stratified by education, working hours per week contributed to explain women’s excess risk of disability pension among the highest and second highest educated, but not among the lowest educated. Again, the impact of gendered distribution, selection effects and adjustment for health make it difficult to interpret the findings. In conclusion, working hours per week is an ambiguous variable that needs to be combined with additional information when interpreting impact on gender difference in disability pension.
The unexplained risk
Population-based studies of risk factors for disability pension give valuable information about risks that are common for both genders and for men and women separately [11, 12, 14]. However, there is a need for studies that aim at examining the gender difference as such . In the current study, the substantial unexplained disability risk among women versus men, after adjusting for health, family and work factors, mirrors a complexity that may need a gender difference approach in order to be understood.
Firstly, occupational health research has accumulated far more knowledge of health hazards related to male working life than female . Secondly, widely used tools to assess the psychosocial working environment were developed within a male working force paradigm, and may lack essential perspectives related to women’s multiple roles in society [38, 39]. Emotional demands, rewards at work, management quality and role conflicts may be especially important for explaining the gender gap in long-term sickness absence and disability pension . Thirdly, education in modern society is not a onetime occurrence, but rather a continuous process throughout the working career. The ability to cope with changing work tasks, technology and decision processes is probably vital for conceiving work as a meaningful and significant part of one’s life. Lower education combined with part-time work gives women fewer options regarding jobs, less opportunity to attend training, less coping experience and a higher risk of exclusion as a result of reorganization and downsizing. These factors need to be addressed in further studies of gender difference.
Firstly, the response rate was higher for women than for men, 70% versus 57% respectively, which opens for possible selection bias related to genders. A study of nonparticipants in HUSK found that 5.5% of the male nonparticipants and 8.8% of the females were awarded disability pension after the HUSK survey was performed. Among participants the percentages were 2.7% for men and 5.2% for women. However, the study found no significant gender differences in the association between nonparticipation and awards of disability pension . This finding indicates that the gender distribution in cumulative incidence of disability recipients in the current study was probably not biased by the gender difference in nonparticipation.
Secondly, typical for non-participants in population-based studies are lower educational level and lower income . Since this under-representation related to level of education and income, is most likely similar for men and women, we do not think that our risk estimates are flawed by selective participation. Further, the relatively low incidence of male disability pensioners may impose some restrictions on the interpretation of the analyses. Finally, the information collected at baseline was cross-sectional, thus preventing identification of causal paths between independent variables.
Our study is based on a linkage between HUSK and the National Insurance Administration’s records of disability pensions awarded to Norwegian inhabitants from 1992 onwards. This register is complete, accurate and independent of exposure data obtained in the Hordaland Health survey. The study thus avoids the problem of attrition. The self-reported information at baseline was collected without participants or administrators being aware of future research hypotheses.
Further, the follow-up period from 1997 till 2004 was a period without major changes in Norwegian disability policy with the potential to alter the cumulative incidence by gender. Only minor fluctuations in incidence of disability pension awards were present during this period. The study design excluded individuals awarded disability pension up to 12 months after baseline, thus decreasing the risk of biased information from participants in the process of being awarded disability pension . Also, the current birth cohorts represent an age group with few incidences of pregnancies and maternity leaves that may influence baseline reporting of health and working hours per week.