Table 2

ORs and diagnostic accuracy for type 2 diabetes over T years—simple model

 CARDIA (10 years) CARDIA-10 (10 years) ARIC (9 years) CHS (7 years) ORs Age group† 1.343* 1.242 1.078 0.848 Black (=1) 0.947 1.408* 1.323*** 1.265 Male (=1) 0.384*** 1.380* 1.575*** 1.512** BMI 1.087*** 1.143*** 1.138*** 1.144*** Parental history (=1) 1.661*** 2.357*** 1.871*** 1.324 Smoker (=1) 0.878 0.987 1.357*** 1.198 High SBP (>140 mm Hg=1) 3.846** 1.094 1.472*** 1.887*** High cholesterol (>240 mg/dL=1) 1.539 1.388* 1.002 0.948 Observations 4039 2813 8875 3094 Diabetes (N) 171 188 836 150 Diagnostic statistics AUROC (95% CI) 0.72 (0.69–0.76) 0.79 (0.76–0.83) 0.75 (0.73–0.77) 0.69 (0.65–0.73) PPV 50.00% 31.25% 34.92% 0.00% NPV 95.79% 93.46% 90.76% 95.15% Cumulative probability 4.23% 6.68% 9.42% 4.85% 1-year probability‡ 0.44% 0.73% 1.04% 0.53%
• *** p<0.01, ** p<0.05, * p<0.1

• †For CARDIA age (25–30)==1, CARDIA-10 age (35–40)=1, ARIC age (55–64)==1, CHS age (75+)==1.

• Supplementary appendix A-6 shows how to compute annual probabilities using the Logit model.

• T represents the maximum time in the sample. For CARDIA and CARDIA-10, T=10 years. For ARIC, T=9; and for CHD, T=7.

• For PPV and NPV the cut-off used is 5.

• Supplementary appendix A-5a shows non-exponentiated coefficients.

• BMI=(weight in kg)/(height in meters)2.

• PPV, p×sensitivity/(p×sensitivity+(1 p)(1−specificity)), where p is the prevalence.

• NPV, specificity×(1 p)/(p×(1−sensitivity)+(1 p)×specificity).

• ARIC, Atherosclerosis Risk in Communities; AUROC, area under the receiver-operating curve; CARDIA, Coronary Artery Risk Development in Young Adults; CHS, Cardiovascular Heart Study; BMI, body mass index; NPV, negative predictive value; PPV, positive predictive value; SBP, systolic blood pressure.