The Performance of a Risk Score in Predicting Undiagnosed Hyperglycemia

Abstract

OBJECTIVE—Type 2 diabetes is a serious disease that is commonly undetected and for which screening is sometimes advocated. A number of risk factors are associated with prevalent undiagnosed diabetes. The use of routinely available information on these factors has been proposed as a simple and effective way of identifying individuals at high risk for having the disease. The objective of this study was to assess the effectiveness of the Cambridge risk score in a large and representative population. RESEARCH DESIGN AND METHODS—A risk score derived from data in a previous study was tested for its ability to detect prevalent undiagnosed hyperglycemia as measured by a GHb ≥6.0, 6.5, or 7% in 6,567 subjects aged 39–78 years in the European Prospective Investigation of Cancer−Norfolk cohort. RESULTS—For a specificity of 78%, the risk score predicted a GHb of ≥7.0% in subjects aged 39–78 years, with a sensitivity of 51% (95% CI 40–62). The areas under the receiver-operating characteristic (ROC) curve for GHb ≥6.0, 6.5, and 7% were 65.7% (63.8–67.6), 71.2% (68.4–75.2), and 74.2% (69.5–79.0), respectively. The area under the ROC curve was not significantly reduced if data on family and smoking history were unavailable for any of the cut-offs for GHb. CONCLUSIONS—The risk score performed as well as other previously reported models in all age groups. We concluded that a simple risk score using data routinely available in primary care can identify people with an elevated GHb with reasonable sensitivity and specificity, and it could therefore form part of a strategy for early detection of type 2 diabetes.