More than 75% of American adults are estimated to have some form of periodontal (gum) disease. Those with periodontal disease are more likely to develop heart disease or have difficulty controlling blood sugar than those without periodontal disease. In addition, women with gum disease are more likely to deliver preterm, low weight babies than those without gum disease [22], and elderly adults who have experienced bone or tooth loss as a result of periodontal disease, have a history of stroke compared to those with healthy gums [20]. Prompt and correct detection of disease helps to identify patients who may have an early clinical benefit from treatment which may reduce complications of untreated periodontal disease such as these. Early diagnosis and treatment is also important in preventing tooth loss, improving the patients'quality of life, and reducing dental health costs. Diagnostic (measurement) error is prevalent in dental research including periodontal research. It has long been recognized that measurement error can bias estimates obtained from error affected independent variables. In periodontal research studies, the probed pocket depth (PPD) is one measure often used to assess periodontal disease status. Current methods of measuring PPD at specified sites by clinical means are subject to measurement error. The models in this proposal are designed to reduce the bias in estimating the incidence of peridontal disease through the correction of PPD measurement error. In a simulation study, we will evaluate the performance of estimates from four models designed to correct for measurement error: the adapted Carlos-Senning and Lu models, the MC-SIMEX corrected model, and newly proposed beta-regression model. We compare the performance of these estimates to estimates obtained from a naive model not accounting for measurement error.
The primary goals of this research is to reduce the bias of estimates associated with independent variables measured with error in large scale oral health studies where periodontal disease status is assessed. The study will quantify the effect of PPD measurement error on the estimation of periodontal disease incidence. Results from this research will provide a model for use in oral health studies in the estimation of disease incidence in the presence of measurement error.
