The distribution of episodes of illness--a research tool in general practice?