Screening for Impaired Fasting Glucose or Diabetes in Managed Care

Objective:To develop and validate equations to screen for impaired fasting glucose (IFG) or diabetes (FPG ≥ 110 mg/dl) using available health plan data.

Methods:We studied 28,925 members of a managed care organization without histories of diabetes:  1396 persons (5%) had FPG ≥ 110 mg/dl.  We developed predictive equations using multiple logistic regression for half of the population and validated the equations using the other half.  The equations used combinations of demographic data, claims and pharmacy data, lab values obtained from claims, and clinical data obtained from patient encounters.

Results:The independent predictors of IFG or diabetes were 1) age, sex, obesity, hypertension, dyslipidemia, and metformin use from claims and pharmacy data; 2) age, sex, hypertension, metformin use, and HDL level from claims, pharmacy, and lab data; and 3) age, sex, race, hypertension, metformin use, HDL level, triglyceride level, body mass index, and systolic blood pressure from claims, pharmacy, lab, and patient encounter data.  The optimal sensitivity and specificity for each of the three equations were 66% and 67%, 64% and 66%, and 60% and 76% respectively.  When applied to the validation sample, the sensitivity, specificity, and area under the curve (AUC) were 64%, 68%, and 0.72, 66%, 66%, and 0.73 and 61%, 76%, and 0.76 respectively.

Conclusion:These multivariate logistic regression equations can be applied in managed care using administrative claims and pharmacy data with or without laboratory and clinical data to identify persons as high risk for IFG or diabetes who might benefit from interventions to prevent or treat diabetes.