Abstract
BACKGROUND: An estimated 75% of the seven million Americans with moderate-to-severe chronic kidney disease are undiagnosed. Improved prediction models to identify high-risk subgroups for chronic kidney disease enhance the ability of health care providers to prevent or delay serious sequelae, including kidney failure, cardiovascular disease, and premature death. METHODS: We identified 11,955 adults > or =18 years of age in the Third National Health and Nutrition Examination Survey. Chronic kidney disease was defined as an estimated glomerular filtration rate of 15 to 59 ml/minute/1.73 m(2). High-risk subgroups for chronic kidney disease were identified by estimating the individual probability using beta coefficients from the model of traditional and non-traditional risk factors. To evaluate this model, we performed standard diagnostic analyses of sensitivity, specificity, positive predictive value, and negative predictive value using 5%, 10%, 15%, and 20% probability cutoff points. RESULTS: The estimated probability of chronic kidney disease ranged from virtually no probability (0%) for an individual with none of the 12 risk factors to very high probability (98%) for an older, non-Hispanic white edentulous former smoker, with diabetes > or =10 years, hypertension, macroalbuminuria, high cholesterol, low high-density lipoprotein, high C-reactive protein, lower income, and who was hospitalized in the past year. Evaluation of this model using an estimated 5% probability cutoff point resulted in 86% sensitivity, 85% specificity, 18% positive predictive value, and 99% negative predictive value. CONCLUSION: This United States population-based study suggested the importance of considering multiple risk factors, including periodontal status, because this improves the identification of individuals at high risk for chronic kidney disease and may ultimately reduce its burden.