Evaluation of Bias and Logistics in a Survey of Adults at Increased Risk for Oral Health Decrements

OBJECTIVES: Designing research to include sufficient respondents in groups at highest risk for oral health decrements can present unique challenges. Our purpose was to evaluate bias and logistics in this survey of adults at increased risk for oral health decrements. METHODS: We used a telephone survey methodology that employed both listed numbers and random digit dialing to identify dentate persons 45 years old or older and to oversample blacks, poor persons, and residents of nonmetropolitan counties. At a second stage, a subsample of the respondents to the initial telephone screening was selected for further study, which consisted of a baseline in-person interview and a clinical examination. We assessed bias due to: (1) limiting the sample to households with telephones, (2) using predominantly listed numbers instead of random digit dialing, and (3) nonresponse at two stages of data collection. RESULTS: While this approach apparently created some biases in the sample, they were small in magnitude. Specifically, limiting the sample to households with telephones biased the sample overall toward more females, larger households, and fewer functionally impaired persons. Using predominantly listed numbers led to a modest bias toward selection of persons more likely to be younger, healthier, female, have had a recent dental visit, and reside in smaller households. Blacks who were selected randomly at a second stage were more likely to participate in baseline data gathering than their white counterparts. Comparisons of the data obtained in this survey with those from recent national surveys suggest that this methodology for sampling high-risk groups did not substantively bias the sample with respect to two important dental parameters, prevalence of edentulousness and dental care use, nor were conclusions about multivariate associations with dental care recency substantively affected. CONCLUSIONS: This method of sampling persons at high risk for oral health decrements resulted in only modest bias with respect to the population of interest.