To determine the prevalence of thyroid dysfunction in institutionalised elderly people in Cape Town and to assess the usefulness of an abnormal thyroid-stimulating hormone (TSH) concentration as a screening test in this group.Cross-sectional survey.Four old-age homes in Cape Town.Old-age home residents aged 60 years and over.Serum concentrations of TSH, free thyroxine and free tri-iodothyronine.Serum TSH estimations were performed on 658 participants, and were abnormal in 103 (15.6%)-41 (6.2%) being elevated (> 5.0 microU/ml) and 62 (9.4%) being low (> 0.4 microU/ml). There were 3 newly diagnosed cases of hyperthyroidism and 7 of hypothyroidism. Subclinical disease was diagnosed in 40 subjects. The overall prevalence of thyroid dysfunction in this population was 11.2%. In 22 (3.4%) this had previously been recognised, while in 50 (7.8%) the dysfunction was newly diagnosed by the current survey. The positive predictive value of a TSH concentration > 20 microU/ml in predicting hypothyroidism is 67%, while it will predict 100% of cases of subclinical hypothyroidism. A TSH concentration < 0.1 microU/ml will predict 23% of cases of hyperthyroidism, but 81% of cases of subclinical disease.The prevalence of thyroid dysfunction in institutionalised elderly people in Cape Town is similar to that reported for elderly people in other centres. Thyroid dysfunction had not previously been recognised in approximately two-thirds of the subjects in this study. The serum TSH concentration is a reliable screening test for thyroid dysfunction in the elderly, but is less useful if used to identify biochemical thyroid disease. An elevated TSH concentration is a better predictor of thyroid dysfunction in the elderly than a depressed TSH concentration.