ABSTRACT
OBJECTIVE: To evaluate the errors and costs of a sampling and survey design for producing reliable estimates of oral health in a demographically diverse local population with a potentially high non-response rate. BASIC RESEARCH DESIGN: A two-stage cluster sample of 45-54 year old adults was drawn from a sampling frame of patients registered with local GPs and cross checked against the electoral register. The modes of data collection were interview/examination and questionnaire. Statistical properties of the estimates were evaluated and costs of the survey described. MAIN OUTCOME MEASURES: The frequency of categories of non-response were described. Sample characteristics were compared with (1) the local population's (1991) census distribution with respect to gender, ethnic composition and socio-economic group (SEG); (2) local utilisation data. The bias effects on oral health values of gender, SEG and non-response, were estimated. The design effect was described for mean number of teeth and proportion with 21 teeth or more. Optimal primary and secondary sample sizes, and non-response follow-up fractions, were estimated. CONCLUSION: This paper describes and costs a sampling and survey design which efficiently produced statistically acceptable estimates of oral health, using small samples from a selected age group in a district, posing demographic and non-response problems in ensuring reliable epidemiological results. The findings indicate that national surveys may give reasonable projected estimates for even richly diverse local populations.
