3044 — Who is Missed When Screening for Diabetes and Prediabetes with a New VA/DOD Algorithm?
Olson DE, Atlanta VAMC and Emory University School of Medicine; Jackson S, Emory University; Tomolo A, Atlanta VAMC and Emory University School of Medicine; Mohan A, and Barb D, Emory University School of Medicine; Zhu M, and Long Q, Emory University Rollins School of Public Health; Phillips LS, Atlanta VAMC and Emory University School of Medicine;
A new VA/DoD screening algorithm for diabetes and prediabetes using fasting plasma glucose (FPG) and hemoglobin A1c (A1c) in patients with risk factors could improve clinical decision-making and care by translating new knowledge about A1c and easier testing into practice. We compared the VA/DoD algorithm to ADA diagnostic guidelines to see if subjects would be missed by the VA algorithm.
We applied the VA algorithm to datasets of Americans without known diabetes with FPG, A1c, and oral glucose tolerance tests (OGTT) from a prospective Screening for Impaired Glucose Tolerance Study (SIGT) and the NHANES 2005-2006 survey (NH0506), including subjects if risk categories of age, race/ethnicity, BMI, lipids, blood pressure, and family history were known. Diabetes (5.8% in SIGT, 7.7% in NH0506) or prediabetes (45.3% in SIGT, 48.1% in NH0506) were classified per ADA diagnostic guidelines. If FPG >=126 mg/dl or A1c >=7%, they received a VA/DoD diagnosis of diabetes, or a VA/DoD diagnosis of prediabetes if the FPG >=100 mg/dl or A1c >=5.7%.
95% of 1581 SIGT subjects and 93% of 1723 NH0506 subjects would be screened by the VA/DoD algorithm, including all subjects with diabetes and 98% with prediabetes. The VA/DoD algorithm only identified 54-60% of diabetic subjects and 75-79% of prediabetic subjects. FPG determined VA/DoD diabetes 91% of the time and A1c contributed rarely. However, A1c contributed to over half of the VA/DoD prediabetes diagnoses. Relying on A1c as the primary test was more likely to miss older, white, and male subjects.
The VA/DoD algorithm includes all subjects that would eventually be diagnosed with diabetes, but does not effectively reduce the number of screened subjects, while missing a large proportion of diabetics vs. standard ADA criteria. The algorithm identifies relatively more subjects with prediabetes that could undergo further testing.
As VA policy, the current algorithm appears not to reduce how often primary care providers would screen for diabetes. Screening without OGTT misses early diabetes; performs differently according to race, age, and gender; and does not identify prediabetic subjects with impaired glucose tolerance known to benefit from best practices of lifestyle and medication interventions.