(247) Validating Type II Diabetes Case Definitions in Electronic Health Records: A Comparison of Sensitivity and Specificity Using HbA1c as the Gold Standard

Publication Number: 907

Background: Accurate identification of Type II diabetes (T2D) cases in electronic health records is critical for implementing effective public health programs.

Objectives: This study aims to evaluate the validity of various case definitions using HbA1c level as the gold standard.

Methods: We identified patients who visited family medicine clinics between 2019 and 2022 and randomly selected 200 patients with T2D ICD-10 codes and 200 patients without, matched on the index-date of ICD-10 diagnosis. We collected data on demographics, HbA1c measurements, oral hypoglycemic agents, insulin use, diabetes complications, and reasons for visiting family medicine. We applied three T2D case definitions: 1) T2D ICD-10 code on or before index-date, 2) any oral hypoglycemic agent on or before index-date, and 3) T2D ICD-10 code AND any oral hypoglycemic agent on or before index-date. The gold standard definition of T2D was HbA1c level >= 6.5 according to the American Diabetes Association guideline. We estimated sensitivity, specificity, and positive predictive value (PPV) for each definition.

Results: Patients with T2D ICD-10 codes were significantly older than patients without ICD-10 (59 vs. 46, P-value < 0.05). The gender distribution was similar between the groups (60% female). Patients with T2D ICD-10 codes had significantly higher proportions of retinopathy (15% vs. 11%) and hypoglycemia (30% vs. 15%) cases (P-value < 0.05). Patients without ICD-10 had a significant proportion of heart attack (8% vs. 2.5%) and stroke (9.5% vs. 7%) cases (P-value < 0.05). The T2D ICD-10 code alone identifies 77% of true cases (sensitivity) while capturing 10% false T2D cases (90% specificity). Oral hypoglycemic prescription alone captures 95% of true cases (sensitivity) while the specificity is 80%. Combining both definitions yields 75% sensitivity, 95% specificity, and 95% PPV.

Conclusions: T2D case definition in electronic health records should be multifaceted, as relying solely on ICD-10 or oral hypoglycemic prescriptions increases the proportion of false positives. Combining both definitions should preserve sensitivity and significantly increase specificity.