OBJECTIVE: Studies of hypothyroidism are often based on patients referred to hospital. It is unknown, to what extent such studies are referral biased. Thus, the aim of the present study was to evaluate the magnitude of selection bias in a study of patients newly diagnosed with hypothyroidism. STUDY DESIGN AND SETTING: A computer-based system linked to laboratory databases identified patients with incident overt hypothyroidism (n=346) from 1997 to 2000 in Aalborg, Denmark. An electronic patient administrative system identified patients referred to Department of Endocrinology, Aalborg Hospital. Among patient characteristics recorded at the time of diagnosis-age, gender, nosological subgroup of hypothyroidism, thyroid stimulating hormone (TSH), total thyroxine (T4), total triiodothyronine (T3), thyroid peroxidase antibody (TPOAb), thyroglobulin antibody (TgAb), and calendar year--we searched for predictors of referral state. RESULTS: Of all hypothyroid patients, 86 (25%) were referred to our endocrine unit. The referred patients were younger (50.4 vs. 66.0 years, P<0.001), had higher serum TSH (53.6 vs. 32.6 mU/L, P=0.002) and lower serum total T4 (37.0 vs. 44.0 nmol/L, P=0.03) compared with nonreferred patients. In a multivariate model, only less age (P<0.001) and serum total T4 (P=0.03) were statistically associated with referral state. CONCLUSION: Hypothyroid patients referred to a specialized hospital unit were younger and marginally more hypothyroid than nonreferred patients. Thus, referral bias should always be considered in hospital-based studies of hypothyroid patients.