Background. Clinicians and patients are often uncertain about the likely clinical course of community-acquired lower respiratory tract infection (LRTI) in individual patients. We therefore set out to develop a prediction rule to identify patients at risk of prolonged illness and those with a benign course. Methods. We determined which signs and symptoms predicted prolonged illness (moderately bad symptoms lasting > 3 weeks after consultation) in 2690 adults presenting in primary care with LRTI in 13 European countries by using multilevel modelling. Results. 212 (8.1%) patients experienced prolonged illness. Illness that had lasted > 5 days at the time of presentation, > 1 episode of cough in the preceding year, chronic use of inhaled pulmonary medication and diarrhoea independently predicted prolonged illness. Applying a rule based on these four variables, 3% of the patients with 90%) recover within 3 weeks. A prediction rule containing four clinical items had predictive value for the risk of prolonged illness, but given its imprecision, appeared to have little clinical utility. Patients should be reassured that they are most likely to recover within three weeks and advised to re-consult if their symptoms persist beyond that period.