Collecting data for dental caries studies is costly. In countries where uniform patient records are available for virtually the whole population, it is tempting to use them as a data source. Our aim was to compare data collected from patient records to those obtained by trained examiners. In 1992 and 1995, dentists who were specially trained and calibrated examined random samples of 12- and 15-year-olds living in two towns in Finland. The dental record of each child was obtained from public dental clinics, the dental status was entered into a computer file, and the DMFS value was calculated. Data were available for 824 children. In the two data sets, 1.3% of the tooth surfaces were recorded differently (DMF vs. sound) with the related kappa value being 0.70. In two thirds of the discrepancies, the reason was that a filling was marked in only one of them, which confirms the known difficulty in discerning a white filling. For 48% of the subjects, the DMFS values calculated from the two sets of data were equal. The difference was 1 and 2 surfaces for 28 and 11%, respectively. Public health dentists had almost equally often registered more and less DMF surfaces compared to trained examiners. The results suggest that data collected from public health records are not decisively inferior to those obtained from examinations by trained examiners. In large enough settings, data obtained from patient records could possibly be used as a replacement for separate surveys.