4.3. Extra cost of comorbidity

The extra cost of comorbidity is estimated as the difference in the predicted costs for patients with both the “main” disease and comorbidity (as defined above) versus the predicted cost for patients with just a “main” disease:

(4.8)\[\begin{split}{\hat Cost}_{comorb} = P(C>0)E(C|>0, comorb=1, disease=1) - \\ P(C>0)E(C|C>0, comorb=0, disease=1)\end{split}\]

The parameters for this equation include not only the main disease parameters, but also disease interactions. However, as already mentioned, we could not estimate this model for every single disease, nor could we take into account the length of time since diagnosis because of the sample size limitations. Therefore at this step, we had to combine several diseases into groups (e.g. this applied to both types of strokes and all cancers). In addition, to reduce potential for residual confounding, a dummy was included to control for long-term chronic diseases (not specifically modelled) for Estonia and France analyses. An intercept in this model thus represents an average costs for people without any chronic diseases for people aged 18-39.