Postal surveys of service use are likely to overestimate health service use and may also underestimate health needs in the population. A historical cohort study, using postal questionnaires and medical records, showed that non-respondents are registered at addresses in more-deprived wards, they are less likely to have attended a hospital diabetes clinic (38% vs 45%) and much less likely to have had a diabetes review in general practice (11% vs 26%). An analysis based on questionnaire respondents would only therefore both underestimate the level of material deprivation and overestimate the proportion receiving routine reviews in general practice in a population with a chronic condition.