The contents of 200 letters of referrals from physicians to a clinical obesity unit were compared with patient-reported data and analysed for concordant and discordant information. For major comorbidities (diabetes, hypertension, joint problems) concordant information was found in about 80%, whereas discordant information was found in 46% for smoking, and 66% for medication. Forty-®ve percent of psychological problems, described by patients, were not identi®ed by the referring physician. The role of the referral letter as a tool in medical communication has received little attention and comparative data are generally lacking, but our data suggests that there is room for considerable improvement of the quality of the referral letter.