5.3. Validation of SHARE results with EHIS2 analysis¶
SHARE model results were validated for two main reasons: 1 – to see whether the employment likelihood results obtained for the 50-64 year age category are indeed similar to EHIS2 (European Health Interview Survey 2 – conducted in years 2013/14) results for the same age categories; and 2 – to see how employment likelihood for the same disease ranges by age group (within EHIS2) [European Commission. Eurostat., 2013 [68]].
A similar ZIP regression was applied in EHIS2; of note is the fact that EHIS2 is a cross-sectional survey, hence exposure was not lagged.
As shown in Table 5.7, employment likelihoods for people with diabetes are within five percentage points for the same age and sex groups, and SHARE estimates are more conservative and between the values of the younger and older age groups for EHIS2. The same is true for hypertension/ CVD-hypertension and COPD results. In EHIS2, the negaitve impact of CVD/stroke on employment is markedly higher than in SHARE, which is likely due to the fact that stroke is an acute disease. Since there is greater temporal confounding in the EHIS2 analysis, these results provide validation for the decision to utilise SHARE for the employment module, allowing for more conservative and better-adjusted results.
Since the employment impacts of the modelled diseases in SHARE lie between the values of the two age groups in EHIS2, this provides further reason to apply SHARE results obtained from ages 50-64 across all employed individuals in the microsimulation. From a theoretical perspective, the impact of a chronic disease on employment should be similar across ages – it is conceivable that a person undergoing cancer treatment at age 40 will miss similar days of work as a person at age 50.
Notably, EHIS2 is a cross-sectoral survey, therefore, the results are to be regarded as temporally confounded and only guideposts (not accurate validation points) for the SHARE analysis. Further, EHIS2 does not include information on cancer status, and the hypertension/stroke/CVD categories only partially match across SHARE-EHIS (Table 5.7).
EHIS2 |
SHARE |
|||||||||
|---|---|---|---|---|---|---|---|---|---|---|
Gender |
Age-group |
RF or Disease |
Employment RR |
LO95 |
HI95 |
pvalue |
Employment RR |
LO95 |
HI95 |
pvalue |
Women |
25-49 |
Hypertension (EHIS)/ CVD-Hypertension (SHARE) |
0.99 |
0.97 |
1.01 |
0.16 |
||||
50-64 |
0.88 |
0.87 |
0.89 |
<0.001 |
0.95 |
0.93 |
0.96 |
<0.001 |
||
Men |
25-49 |
0.99 |
0.98 |
1 |
0.11 |
|||||
50-64 |
0.91 |
0.9 |
0.92 |
<0.001 |
0.96 |
0.94 |
0.98 |
<0.001 |
||
Women |
25-49 |
Diabetes |
0.95 |
0.92 |
0.99 |
0.02 |
||||
50-64 |
0.84 |
0.82 |
0.86 |
<0.001 |
0.89 |
0.86 |
0.92 |
<0.001 |
||
Men |
25-49 |
0.96 |
0.92 |
0.99 |
0.01 |
|||||
50-64 |
0.89 |
0.87 |
0.91 |
<0.001 |
0.92 |
0.89 |
0.95 |
<0.001 |
||
Women |
25-49 |
COPD |
0.99 |
0.95 |
1.02 |
0.37 |
||||
50-64 |
0.87 |
0.85 |
0.9 |
<0.001 |
0.92 |
0.89 |
0.96 |
<0.001 |
||
Men |
25-49 |
0.9 |
0.87 |
0.94 |
<0.001 |
|||||
50-64 |
0.81 |
0.78 |
0.83 |
<0.001 |
0.83 |
0.79 |
0.87 |
<0.001 |
||
Women |
25-49 |
CVD (EHIS) Stroke (SHARE) |
insufficient sample |
|||||||
50-64 |
0.67 |
0.55 |
0.82 |
<0.001 |
0.83 |
0.78 |
0.89 |
<0.001 |
||
Men |
25-49 |
insufficient sample |
||||||||
50-64 |
0.66 |
0.55 |
0.8 |
<0.001 |
0.81 |
0.77 |
0.86 |
<0.001 |
||
Note: EHIS2 includes data from all EU countries, whereas SHARE includes fewer countries. The biggest difference is that the age of the survey population ranges from 19 – 65+ years in EHIS2, whereas it ranges from 50+ in SHARE. Further, EHIS2 does not include information on cancer, and the hypertension/ CVD categories do not fully overlap.