3 days) compared with controls (median 2 days; IQR 1-5 days) (p50.04) [5], suggesting a more rapid onset of symptoms in children at greater risk of hospital presentation for pneumonia. We now show that the same cases were less likely to have visited a GP for other illnesses in the previous year. The lack of association between these two variables (p50.8) suggests the two findings are unrelated.We therefore hypothesise that a child presenting to general practice promptly after illness onset, with a history of infrequent consulting, may be at increased risk for pneumonia or empyema. This is consistent with common knowledge regarding any illness that needs prompt and decisive action by the GP. However, our findings cannot inform GP decision making because they may be influenced by response bias. Larger, prospective studies are needed to test our hypothesis.