• bias;
  • longitudinal surveys;
  • older adults;
  • oral health

Abstract Longitudinal studies of the oral health of older adults are necessary for the measurement of disease incidence and identification of risk factors for oral disease. Loss of respondents to follow-up may, however, seriously bias longitudinal results. Using data from the 1989 and 1992 waves of the Ontario Study of the Oral Health of Older Adults, this paper examines loss to follow-up by comparing the characteristics of lost and retained respondents in terms of clinically-defined and self-rated oral health, sociodemographic characteristics, general health status, and health behaviours. Of the original 907 respondents who completed an interview and clinical examination, 611 participated in the 3-yr follow-up study. Study attrition rates were higher in the edentulous group. Although some statistically significant differences were found between those retained and lost, the magnitude of the differences was small and unlikely to seriously bias estimates of incidence or risk. These results illustrate the need to consider study attrition, response rates and comparisons of retained and lost to follow-up respondents in reporting the results of longitudinal studies. Without such information, confidence in the population estimates derived from those studies is undermined.