6.5 Transitions into disability
6.5.4 Competing risks approach
The results until sofar have revealed a significant impact of initial physical and mental health on transitions into disability. We expect that initial health will have a less substantial influence on minor complaints than on more severe disabilities. Similarly, we expect initial mental health to be a very important predictor of disability due to mental problems. We therefore continue the analysis with independent competing risks models (e.g., Van den Berg, 2001). In our specific setting competing risks models boil down to MPH model applied to self-employed who enter disability due to a specific disorder; see also Markussen et al. (2011).
Because certain disorders are observed relatively infrequently, we cannot estimate separate MPH models for each specific ICD-10 disorder. We therefore restrict the com- peting risks analysis to three main categories. We first estimate a separate MPH model for self-employed who become disabled due to mental disorders (such as depression and burn-out). We also estimate a separate MPH model for policyholders who enter dis- ability due to muscular-skeletal disorders, injury and poisoning (henceforth referred to
184 Chapter 6. Health, Lifestyle and Disability Transitions of Self-Employed Workers as physical impairments, since poisoning occurs only once in our sample). This disease category encompasses disorders such as a fractured ankle and low back pain. Finally, we estimate a MPH model for policyholders who enter disability due to severe conditions (including cancer, diseases of blood and blood-forming organs, circulatory disorders, en- docrine and nervous diseases); see Smith (2004) who identifies severe conditions in a similar way.
Section 6.5.1 has shown that initial physical and mental ill-health, as well as being a smoker, significantly increases the risk of entering disability on the aggregate level. Panels 2 – 4 of Table 6.3 refine these results.17 There is considerable heterogeneity in terms of the disorder causing the disability. Physical ill-health, mental ill-health, smok- ing and a lack of physical exercise significantly increase the risk of entering disability due to a mental disorder. The positive effect of playing sports on disability outcomes (which is significant at the 10% level) is in line with results reported in the medical literature (e.g., Duijts et al., 2007; Hendriksen et al., 2010). It is also consistent with studies showing that physical exercise has a positive influence on avoiding depression (e.g., Mammen and Faulkner, 2013). The influence of physical exercise is not merely positive though. Doing sports increases the risk of becoming disabled due to physical impairments by 15%. This effect is significant at the 5% level. Since physical impair- ments include injuries such as a fractured ankle, this effect is likely to reflect sports injuries.
The economic relevance of the aforementioned health and lifestyle effects is substan- tial. An increase in the physical health score of 25 percentage points results in a 8.5% drop in the incidence rate of mental disorders; self-employed whose experienced mental disorders in the past have a 74% higher incidence rate than individuals who did not; smoking increases the risk of becoming disabled by 42%; and playing sports reduces the risk of entering disability due to a mental disorder by 17%. As expected, the effect of initial mental health on the risk of becoming disabled due to a mental disorder is significantly higher than the effect of initial mental health on the overall incidence rate. This confirms evidence from the medical literature that mental disorders tend to be persistent and recurring (e.g., Spijker et al., 2002).
Several other effects are worth mentioning. Initial physical ill-health significantly increases the disability risk due to physical impairments: an increase in the physical health score of 25 percentage points results in a 10% rise in the incidence rate. The risk of entering disability due to severe conditions is not significantly affected by previously experienced mental health problems, but is significantly increased by initial physical ill-health. As before, drinking alcohol does not have significant impact on the risk of becoming disabled.
6.5. Transitions into disability 185 Smoking significantly increases the risk of becoming disabled, regardless of the dis- order causing the incapacity. In particular, we find that smokers are exposed to an increased risk of entering disability due to a mental disorder. Although the literature has shown that the relation between smoking and mental health is complex, there is indeed evidence that smoking increases the risk of getting mental disorders (Cuijpers et al., 2007; West and Jarvis, 2005). The relation between smoking and disability due to severe conditions such as cancer and cardiovascular diseases is well-known from the medical literature.
There are no significant gender differences in the risk of entering disability due to physical impairments, but there are such differences with respect to the risk of becoming disabled due to mental problems and severe conditions though. Men face a significantly lower risk of entering disability than women in these two cases. The higher disability risk of women due to mental disorders is in line with evidence from the medical literature that women have a higher risk of getting mental disorders (WHO, 2000).18
Because of potential differences between men and women, we estimate the competing risks models for men only as a robustness check; see Table 6.7.19 The aforementioned results remain valid, with two main exceptions. For men, playing sports does not sig- nificantly decrease the risk of becoming disabled due to a mental disorder. This seems in line with the result in Table 6.5 that only for women playing sports significantly decreases the risk of becoming disabled in general. Second, for men previously experi- enced mental health problems significantly increase the risk of becoming disabled due to severe conditions. It is worth noting here that the severe conditions of the men and women in our sample differ a lot. For women we frequently observe breast and ovarium cancer as severe conditions, while for men cardiovascular diseases prevail.20 The rela- tion between mental disorders and cardiovascular diseases is confirmed by the medical literature, where it has been shown that depression is a risk factor for the onset of a wide range of cardiovascular diseases; (e.g., Van der Kooy et al., 2007).
We re-estimate all competing risks models and include interaction terms between the lifestyle indicators and the measures for physical and mental health. The interaction between the indicators for overweight and initial mental health problems is the only interaction term that turns out significant in some of the competing risks specifications. This interaction term is significantly positive in the MPH model for entry into disability 18We do not find significant unobserved heterogeneity for self-employed entering disability due to
a mental disorder; the estimated Heckman-Singer frailty distribution is degenerate. We establish unfavorable frailty (resulting in relatively slow recovery) for 98.7% of the policyholders that enter disability due to severe conditions.
19We do not have enough female self-employed in the sample to estimate the competing risks models
for women only.
20Because we do not have enough observations per disease category, it is not possible to use the more
186 Chapter 6. Health, Lifestyle and Disability Transitions of Self-Employed Workers due to physical impairments and in the MPH model for entry into disability due to severe conditions. In the latter two models overweight and mental ill-health exacerbate each other’s effect on the risk of becoming disabled. None of the interaction terms turns out significant in the competing risks models for men.