Gary D Slade1. 1. Australian Research Center for Population Oral Health, Dental School, University of Adelaide, Adelaide, South Australia, Australia. gary.slade@adelaide.edu.au
Abstract
BACKGROUND: The purpose of this study was to provide interim results regarding the validity of six periodontal screening questions in predicting the prevalence of clinically assessed periodontitis among Australian adults. METHODS: The Australian National Survey of Adult Oral Health is a cross-sectional survey of Australians aged 15 years or older. Subjects were selected using a multistage probability sampling design. The interview included six questions to screen for periodontal disease and five demographic/health history questions that represented traditional risk indicators for the disease. Oral examinations were conducted by trained, calibrated dentists who measured periodontal recession and probing depth at mesio-buccal, mid-buccal, and disto-buccal sites on all erupted teeth. A computer algorithm determined three categories of periodontal case status: none/mild, moderate, or severe. Multivariable binary logistic regression models were constructed using six screening questions, five traditional risk indicators, and all 11 variables. RESULTS: This interim analysis used data from 2,999 subjects who resided in the first five states/territories that completed the survey. Based on clinical findings, 4.0% of subjects were classified with "severe" periodontitis and 25.2% were classified with "moderate" periodontitis. Five screening questions had non-response rates of <2%, whereas bivariate correlations among screening questions were weak, indicating little redundancy among items. The combined set of 11 questions achieved "useful" overall levels of prediction based on area under receiver operating characteristic curves. Values of combined sensitivity and specificity were modest when considered at a threshold needed for predicting prevalence. Larger values of combined sensitivity and specificity were achieved when higher prevalence was predicted. CONCLUSION: Six screening questions and five conventional risk indicators can be used readily in large population surveys, yielding useful levels of validity in predicting the prevalence of periodontitis.
BACKGROUND: The purpose of this study was to provide interim results regarding the validity of six periodontal screening questions in predicting the prevalence of clinically assessed periodontitis among Australian adults. METHODS: The Australian National Survey of Adult Oral Health is a cross-sectional survey of Australians aged 15 years or older. Subjects were selected using a multistage probability sampling design. The interview included six questions to screen for periodontal disease and five demographic/health history questions that represented traditional risk indicators for the disease. Oral examinations were conducted by trained, calibrated dentists who measured periodontal recession and probing depth at mesio-buccal, mid-buccal, and disto-buccal sites on all erupted teeth. A computer algorithm determined three categories of periodontal case status: none/mild, moderate, or severe. Multivariable binary logistic regression models were constructed using six screening questions, five traditional risk indicators, and all 11 variables. RESULTS: This interim analysis used data from 2,999 subjects who resided in the first five states/territories that completed the survey. Based on clinical findings, 4.0% of subjects were classified with "severe" periodontitis and 25.2% were classified with "moderate" periodontitis. Five screening questions had non-response rates of <2%, whereas bivariate correlations among screening questions were weak, indicating little redundancy among items. The combined set of 11 questions achieved "useful" overall levels of prediction based on area under receiver operating characteristic curves. Values of combined sensitivity and specificity were modest when considered at a threshold needed for predicting prevalence. Larger values of combined sensitivity and specificity were achieved when higher prevalence was predicted. CONCLUSION: Six screening questions and five conventional risk indicators can be used readily in large population surveys, yielding useful levels of validity in predicting the prevalence of periodontitis.