Comparison of two methods of estimating 48-month tooth loss incidence

This paper quantifies validity of self-reported tooth loss, compares incidence using two methods (semiannual self-report, biannual clinical examination), and compares conclusions about risk factors for tooth loss using these two methods.

The Florida Dental Care Study included persons who at baseline had at least one tooth. In-person interviews and clinical examinations were conducted at baseline, 24 months, and 48 months, with semiannual telephone interviews in between.

Agreement between self-reported and clinically derived tooth loss was high, although some statistically significant differences by certain baseline characteristics were evident. On a nominal scale (some tooth loss, none), kappa was 0.88 and percent concordance was 94 percent. On a ratio scale, Spearman's correlation was 0.90. Using self-report, the incidence estimate would have been 34 percent, as compared to 36 percent based on clinical examination. In a single bivariate (loss by self-report, loss by clinical examination) multiple logistic regression, conclusions about statistical significance and magnitude of seven risk factors for tooth loss did not differ.

Validity of self-reported incidence was excellent. The self-reported method allowed for semiannual estimates and was less resource intensive. Substantive conclusions about tooth loss using either method were similar, although validity did differ between persons with certain baseline characteristics.