Diabetes Care, Vol 22, Issue 2 213-219, Copyright © 1999 by American Diabetes Association

ARTICLES

Performance of a predictive model to identify undiagnosed diabetes in a health care setting

CA Baan, JB Ruige, RP Stolk, JC Witteman, JM Dekker, RJ Heine and EJ Feskens
Department of Public Health, Erasmus University, Rotterdam, The Netherlands. baan@mgz.fgg.eur.nl

OBJECTIVE: To develop a predictive model to identify individuals with an increased risk for undiagnosed diabetes, allowing for the availability of information within the health care system. RESEARCH DESIGN AND METHODS: A sample of participants from the Rotterdam Study (n = 1,016), aged 55-75 years, not known to have diabetes completed a questionnaire on diabetes-related symptoms and risk factors and underwent a glucose tolerance test. Predictive models were developed using stepwise logistic regression analyses with the absence or presence of newly diagnosed diabetes as the dependent variable and various items with a plausible connection to diabetes as the independent variables. The models were evaluated in another Dutch population-based study, the Hoorn Study (n = 2,364), in which the participants were aged 50-74 years. Performances of the predictive models were compared by using receiver-operator characteristics (ROC) curves. RESULTS: We developed three predictive models (PMs), PM1 contained information routinely collected by the general practitioner, while PM2 also contained variables obtainable by additional questions. The third predictive model, PM3, included variables that had to be obtained from a physical examination. These latter variables did not have additive predictive value, resulting in a PM3 similar to PM2. The area under the ROC curve was higher for PM2 than for PM1, but the 95% Cls overlapped (0.74 [0.70-0.78] and 0.68 [0.64-0.72], respectively). CONCLUSIONS: Using only information normally present in the files of a general practitioner, a predictive model was developed that performed similarly to one supplemented by information obtained from additional questions. The simplicity of PM1 makes it easy to implement in the current health care setting.
CiteULike    Del.icio.us    Digg    Reddit    Technorati    What's this?