Measures of case mix are needed to control for patients' clinical status in studies assessing the process and outcomes of care. The Veterans Health Study (VHS) is a longitudinal study of determinants of health outcomes in ambulatory veterans. This study assessed the validity of a case-mix measure developed to quantify severity of illness in ambulatory type 2 diabetic patients. As part of the pilot phase of the VHS, 245 veterans using 4 primary care clinics in Boston were screened for diabetes and 5 other chronic illnesses when they presented for care. Subjects screening positive for diabetes returned to complete severity of illness and outcome measures. The variables for the diabetes case-mix module were chosen based upon the principles of parsimony, duration of follow-up, and clinical validity and credibility. Variables were selected to predict function, as measured by the Medical Outcomes Study Short-Form 36 (SF-36). The diabetic patients in this study had a heavy burden of chronic illness, with an average of 3.9 comorbid conditions and a mean general health perceptions score of 48 on the SF-36 (scored from 0 to 100, with 100 indicating best health). A summary variable called DMSEV was created for "diabetes severity". This included atherosclerotic complications(stroke, transient ischemic attack or myocardial infarction; chest pain frequency; and claudication), plus any history of eye, foot, or neuropathic symptoms. DMSEV correlated with all 8 outcome scales of the SF-36, and in particular was highly associated with physical function (r=0.49, P=.0001). Least squares linear regression analysis controlling for age and comorbidity confirmed the association of DMSEV with all 8 SF-36 scales. The correlation with physical function remained highly significant (P<.0001), with an R of 0.31. This patient-based self-assessment questionnaire and the summary variable DMSEV appear to be valid measures of severity of illness in ambulatory diabetic veterans with multiple comorbidities. After further testing, this case-mix measure may be suitable for controlling for severity of illness in ambulatory-based studies of diabetic patients with multiple chronic illnesses.