Statistical issues in assessing comorbidity.

I define three separate and distinct types of non-random comorbidity: epidemiologic, clinical and familial. These might exist singly or in any combination for a pair of disorders. The focus is on their definition and the measurement and interpretation of the types of comorbidity of most common concern: epidemiologic comorbidity. I discuss certain sources of epidemiologic comorbidity such as shared risk factors, or diagnostic 'fuzziness', and I indicate the directions of research design and analyses to disclose such sources.