BACKGROUND: Although unreliability of periodontal measurements is considered mainly to concern attempts to detect true changes in longitudinal studies, it is also a source of variability that may substantially decrease the precision of estimates obtained in cross-sectional studies. This study assessed the effects of measurement errors on estimates of prevalence, extent, and severity of clinical attachment loss. METHODS: Four examiners performed repeat attachment level recordings in 128, 122, 134, and 133 adolescents, respectively, who participated in a study of clinical attachment loss among 9,162 high school students from Santiago, Chile. A total of 48,954 duplicate recordings were made. Two examiner teams carried out repeat examinations among 292 and 254 subjects, respectively, corresponding to a total number of 51,600 duplicate recordings for interexaminer reliability assessment. RESULTS: A sizeable imprecision was noted regarding prevalence estimates, particularly at the > or = 1 mm level, whereas examiners were reasonably consistent in the assessment of the number of sites with clinical attachment loss. Ninety-five percent of the differences regarding the number of sites with clinical attachment loss > or = 1 mm were in the range from -5 to 5 sites for intraexaminer comparisons, and from -9 to 8 sites for interexaminer comparisons. These differences should be seen in light of the number of sites affected, which ranged from 0 to 85 sites. CONCLUSIONS: Examiner variation may preclude comparisons of prevalence estimates between studies. Valid comparisons may be made between studies of extent estimates provided that uniform parameters are used to express the extent of periodontal breakdown.
BACKGROUND: Although unreliability of periodontal measurements is considered mainly to concern attempts to detect true changes in longitudinal studies, it is also a source of variability that may substantially decrease the precision of estimates obtained in cross-sectional studies. This study assessed the effects of measurement errors on estimates of prevalence, extent, and severity of clinical attachment loss. METHODS: Four examiners performed repeat attachment level recordings in 128, 122, 134, and 133 adolescents, respectively, who participated in a study of clinical attachment loss among 9,162 high school students from Santiago, Chile. A total of 48,954 duplicate recordings were made. Two examiner teams carried out repeat examinations among 292 and 254 subjects, respectively, corresponding to a total number of 51,600 duplicate recordings for interexaminer reliability assessment. RESULTS: A sizeable imprecision was noted regarding prevalence estimates, particularly at the > or = 1 mm level, whereas examiners were reasonably consistent in the assessment of the number of sites with clinical attachment loss. Ninety-five percent of the differences regarding the number of sites with clinical attachment loss > or = 1 mm were in the range from -5 to 5 sites for intraexaminer comparisons, and from -9 to 8 sites for interexaminer comparisons. These differences should be seen in light of the number of sites affected, which ranged from 0 to 85 sites. CONCLUSIONS: Examiner variation may preclude comparisons of prevalence estimates between studies. Valid comparisons may be made between studies of extent estimates provided that uniform parameters are used to express the extent of periodontal breakdown.