The Duke Severity of Illness Checklist (DUSOI) was evaluated on 414 primary care adult patients using data collected both by medical providers at the time of the patient visit and later by a chart auditor. Severity scores for individual diagnoses were determined by summing the ratings for four non-disease-specific parameters: symptom level, complications, prognosis without treatment, and expected response to treatment. Mean diagnosis severity scores (scale 0-100) among the 21 most prevalent diagnoses varied from a low of 13.9 for menopausal syndrome to a high of 43.0 for sprains and strains. An overall severity score was calculated by combining diagnosis severity scores and giving highest weights to the most severe diagnoses. Provider-generated overall severity scores (mean = 43.3) and auditor-generated overall severity scores (mean = 38.9) were significantly correlated (coefficient of agreement = 0.59, p < 0.0001). Diagnoses varied in their individual contribution to the overall severity score, from 8.9% for lipid disorder to 90.0% for sprains and strains. Separate comorbidity severity scores were calculated to measure the severity of all of each patient's health problems except the diagnosis under study. For example, patients with menopausal syndrome had co-existing health problems which generated a high mean comorbidity severity score of 43.2, while patients with sprains and strains had a low mean comorbidity score of 4.7. The DUSOI Checklist can be used in the clinical setting by both providers and auditors to produce quantitative severity scores (by diagnosis, overall, and for comorbidity) which are based entirely upon clinical judgment. This method should be useful in controlling for severity of illness in clinical studies and indicating the outcome of medical care in terms of reduction in severity of illness following medical interventions.