In this study, we examined the association between diagnosis-specific LTSA and all-cause disability retirement, as well as with disability retirement due to the same or due to some other diagnostic group than that recorded in a pre-retirement LTSA spell. Clear differences were found between the LTSA diagnostic groups in predicting disability retirement. The association was the strongest when both the LTSA and disability retirement were due to the same diagnostic group, but significant associations were also found when disability retirement was caused by a different diagnostic group. In general, upper non-manual employees consistently had the highest HRs, whereas manual workers had the lowest HR for disability retirement.
In general, LTSA due to any of the four diagnostic group was strongly associated with all-cause disability retirement among men and women. LTSA due to musculoskeletal diseases and mental disorders had the strongest associations with all-cause disability retirement, in accordance with previous studies [4,5,6, 9, 10, 21, 22]. A novel finding was that those with LTSA due to mental disorders and musculoskeletal diseases transferred more often to disability retirement due to the same diagnostic group than due to some other diagnostic group. Additionally, LTSA due these diagnostic groups had slightly stronger association, in terms of HRs, with all-cause disability retirement than LTSA due to respiratory or circulatory diseases. While the differences between LTSA diagnostic groups were small, LTSA due to mental disorders increased the risk of disability retirement slightly more than musculoskeletal diseases in men, whereas in women the order was the opposite. This has also been found in a previous study .
A higher proportion of those who had an LTSA due to respiratory or circulatory diseases transferred to disability retirement due to some other diagnostic group. However, in comparison to those who had no LTSA due to these diagnostic causes, having an LTSA was more strongly associated with disability retirement due to the same diagnostic group than due to some other diagnostic group, resembling the results pertaining to mental and musculoskeletal diseases. The results regarding circulatory diseases are in accordance with previous findings [4, 21] while the results regarding respiratory diseases partly corroborate results of previous studies that have found evidence that demonstrated both a strong association [4, 6, 23] and no association  between LTSA due to respiratory diseases and disability retirement. The mixed results are probably due to differences in the study population, and/or variance in the definitions of the length of LTSA and that of diagnostic groups. Respiratory diseases vary between subgroups – ranging from a mild cold to chronic obstructive pulmonary disease – across which the risk of disability retirement is likely to vary. As this study only included long-term LTSAs, the respiratory diseases were probably severe. However, as only a relatively small number of individuals had an LTSA and/or disability retirement due to respiratory diseases in this study, especially when stratified by occupational class, caution must be practiced in interpreting the magnitude of this association.
Contrary to previous studies that have studied the association between LTSA and disability retirement, we also included information on the diagnosis of disability retirement, revealing that the diagnostic groups were also associated with disability retirement due to some other diagnostic group. The general mechanisms that govern the association between LTSA and disability retirement due to some other diagnosis than that during the LTSA spell can only be hypothesized in this study. First, a disability retirement can be admitted only due to one main diagnosis. Therefore, in the case of multimorbidity, only one condition can be chosen to be the main diagnosis. Due to a potentially long period between the LTSA measurement and the event of disability retirement in this data set, individuals could have also suffered from other diseases or disorders during the period between the measured LTSA spell and their eventual disability retirement. It is also possible that somatic diseases are recognised first while mental diagnoses remain unreported . Multimorbidity is strongly associated with poor health  and work disability [26, 27], rendering people vulnerable to additional illnesses. For example, respiratory diseases are associated with low work ability  and poor self-rated health . They also commonly co-occur with diabetes  and depressive symptoms , and result in an increased risk of disability retirement especially when combined with another chronic disease or depression . These results imply that disability retirement can be preceded by multimorbidity that clinical practitioners should take into account when planning rehabilitation measures.
Occupational class differences
In comparison to other occupational classes, a higher proportion of manual workers transferred to disability retirement. However, among upper non-manual employees having an LTSA increased the risk of disability retirement the most. On the other hand, this increase in risk as the least among manual workers. Previous studies have found similar results. For example, studies found that upper non-manual employees that have taken sickness allowance have a higher relative risk of disability pension than other occupational classes, especially when the LTSA was due to mental disorders , and due to musculoskeletal diseases . Additionally, a French study found that those in a higher occupational position had a higher risk of poor health after sick leave due to cancer and mental disorders compared to those in a lower occupational position . The higher disability retirement risk in upper non-manual employees is related to the relative differences within occupational classes. However, in absolute terms, the risk of disability retirement is the highest among manual workers.
There are several possible explanations for this. Manual workers have typically higher rates of LTSA and disability retirement [11, 12], but, they may more frequently have an LTSA that does not lead to disability retirement. Instead, they may be more likely to be forced out of the workforce (due to unemployment, for example) than non-manual employees , their applications for disability retirement are more often rejected , or they may need a longer LTSA to recover from illnesses or diseases. Manual workers’ work environment can force them to take sickness absence even if they have less severe work disability. Contrarily, upper non-manual employees’ work environment can provide more flexibility and enables them to continue working even with a disease or disorder. Since the reference groups in this study consisted of those with no LTSA due to the examined diagnostic groups in each occupational class, our results may reflect the fact that despite generally having less work ability problems, upper non-manual employees are more likely to transfer to disability retirement once they avail sickness absence. Their employers may prefer that they remain on sickness absence until fully recovered, as positions in such occupations can be difficult to replace . Further, upper non-manual employees often have psychologically demanding jobs [36, 37], which are especially difficult to return to after mental health problems. More research should be conducted on the possible explanations regarding why upper non-manual employees have a relatively higher risk of disability retirement than manual workers once they undergo LTSA.
Among women, LTSA due to musculoskeletal diseases increased the risk of all-cause disability retirement the most in lower non-manual employees and manual workers. Meanwhile, LTSA due to mental disorders was the strongest predictor of disability retirement among upper non-manual employees and one of the strongest predictors of the same among the self-employed. This is likely to reflect the fact that many women work sectors that are physically demanding such as the health care sector. Among upper non-manual employee men, LTSA due to circulatory diseases had a relatively strong association with disability retirement compared to other occupational classes. According to a previous study, cardiovascular diseases are more strongly associated with disability retirement among those in low occupational classes than in higher occupational classes . However, while the previous study  studied the additional and synergistic effects of both occupational class and cardiovascular diseases, we examined the effect of LTSA through stratification according to occupational class. Thus, while manual workers have the highest risk of disability retirement across all occupational classes, having an LTSA due to circulatory diseases increases the risk of disability retirement more among upper non-manual employees.
Strengths and limitations
The main strength of this study was its ability to utilize a large data set that comprised several combined high-quality national register sources, including a 70% random sample of the Finnish employed population that consisted of almost 1.5 million people. These data ensured an almost complete follow-up of the same individuals with negligible and random missing information and the ability to use date-specific information pertaining to both LTSA and disability retirement. However, the data lacked information on short-term sickness absence (less than 10 days), previous work trajectories, work environment, subjective health, health behaviour and the type of disability pension (i.e. information on part-time or full-time, and permanent or fixed-term disability retirement) which may have explained some of the observed associations. No causal effects could be established due to the observational research setting. Since the analysis was conducted only with Finnish data, whether these results hold in other countries is yet to be ascertained. The results may be generalized with caution to other countries with reasonably similar social security systems with respect to work disability. Future studies should examine the possible explanations for the occupational class differences in the transition from sickness absence to disability retirement due to different diagnoses. Introducing information on factors such as health behaviour and work environment can help to understand why some diagnoses are more important than others in predicting disability retirement, and where the occupational differences in this association emerge. The role of comorbidity in the transition to disability retirement and the occupational class differences in it should be studied more in detail, for example, by adding information on an individual’s medical history. Lastly, the research design used in the current study could be used to examine the role of more specific diagnoses in the transition to disability retirement.