OBJECTIVE To describe the prevalence of known diabetes in an area of New Zealand with a large Pacific Islands and Maori population. METHODS A cross sectional door to door census with identification of those with known diabetes was conducted between April and October 1992. The data was validated by comparison with available local general practice diabetes registers and data from a repeat visit to a randomly selected 5% of houses. RESULTS Interviews were completed at 92.7% of the 5081 households, containing 22,651 residents (1417 European, 5606 Maori, 14,802 Pacific Islands). The Pacific Islands population was 40% larger than that predicted from the 1991 census. The age adjusted prevalence of known diabetes in adults (aged > or = 20 years) was 2.8% (95% CI 1.9-3.9) in Europeans, 6.9% (95% CI 6.0-7.9) in Maori and 4.6% (95% CI 4.1-5.1) in Pacific Islands people. The greatest differences in prevalence were found in those aged 40-59 years. Interviews at 185/280 houses revisited showed that 13% of households had moved completely in the 2-8 months between visits. Comparison between participating general practice registers and the door to door survey database showed that 11.2% (26/232) of diabetic individuals were missed by the door to door survey, and 23.7% (55/232) were not on the register of the named general practitioner. CONCLUSIONS The prevalence of known diabetes in the community is much higher than that in the workforce. Differences between Maori and Pacific Islands people may be due to more undiagnosed diabetes or a lower risk of diabetes in the latter.