We evaluated the effect of a dosage card signed by both patient and health worker on compliance in a tuberculosis ward. Compliance was assessed by testing a urine specimen from each patient for the presence of isoniazid and rifampicin pre-intervention and at two weeks, six weeks and 16 months post-intervention. Pre-intervention compliance was 62% and increased to 88%, 91% and 93% in the short, medium and long-term, respectively (P = 0.03, P = 0.02, P < 0.01). The assumption that directly observed therapy is successfully practised in hospital is not always correct. Simple inexpensive measures can improve compliance and have an impact on tuberculosis control.