(B-040) Algorithmically Confirming Patients with Type 1 Diabetes Using Claims Data

Background: Diabetes and diabetic complications are a major cause of morbidity and mortality. Accuracy of diagnoses are thus critical for analyses using claims data. The 2021 ADA/EASD Consensus Report states misclassification of type 1 diabetes (T1D) as type 2 (T2D) in adults is common, and >40% of those diagnosed (Dx’d) with T1D after age 30 year are initially treated as T2D. A Workpartners Research Reference Database (RRDb) analysis found 22.5% of those with a T1D Dx also had a Dx for T2D or type 1.5 (T1.5D).

Objectives: To determine the rate of accuracy for a T1D Dx using claims data and to develop an algorithm to reclassify mis-Dx'd patients.

Methods: Retrospective analysis of Workpartners RRDb (2001-2024) of self-insured US employers. Patients (pts ) with T1D claims using ICD-9/-10 codes, index date=initial Dx date. Pts had no T2D Dxs pre-index, and >2 T1D Dxs (>1 month a part, < 1 yr post-Dx). Pts 1 yr post-index use of INS(ulin), oral anti-diabetic [OAD] prescriptions (Rxs), healthcare (HC) visits (1 yr), days with Rx fills, and Dxs for T1.5D and T2D were examined. Clinical review used to validate algorithm and Dx. Sensitivity analyses to examine differences based on ELIG(ibility) and pre-period washouts.

Results: We found 27,749 pts with a T1D Dx;48.2% [13,376] had no pre-index T2D Dx. Post-Index ELIG decreased pts to 9465 of which 75.6% (7154) had ≥2 T1D Dxs, 10.7% had no INS and mean (18.5 HC visits, 13.1 Rx fill dates) yet 84.3% had no OAD meds. The 6385 INS users had mean (17.7 HC visits, 21.5 Rx fill dates), 43.8% had ≥2 T2D Dxs and 14.8% of the confirmed (T1D & T2D) Dx’d pts used OAD meds; 1.9% had ≥2 T1.5D Dxs and 18.7% of the confirmed (T1D &T1.5D) Dx’d pts used OAD meds. Requiring 1 yr pre-index ELIG decreased the pts to 2432 and 51.9% (1263) had ≥2 T1D Dxs, 20.5% had no INS and mean (17.8 HC visits, 15.5 Rx fill dates) and 85.7% had no OAD meds. The 1004 INS users had mean (18.5 HC visits, 22.3 Rx fill dates), 35.9% had ≥2 T2D Dxs post-Index and 14.2% of the confirmed (T1D & T2D) Dx’d pts used OAD meds; 3.1% had ≥2 T1.5D Dxs and 16.1% of the confirmed (T1D & T1.5D) Dx’d pts used OAD meds. Our algorithms reduced the confirmed populations of pts with T1D to 67.5% using post-index data and 41.3% using pre-/post-index data. In addition, we identified pts with T1D and dual Dx for T1.5D or T2D.

Conclusions: Claims data are useful for analyzing the impact of conditions. However, clinicians and/or staff may enter the wrong Dx codes. Our algorithm's combination of confirmed Dx codes, Rx use, and pre- and post-index eligibility reduced mis-coded patients and provided confirmed cohorts of patients with different types of diabetes. These confirmed cohorts are better suited for research on outcomes, use of therapies, and the epidemiology associated with the different types of diabetes.