1449-P: Identification of Type 1 Diabetes by an Electronic Health Records Algorithm from the CURE-CKD Registry

Introduction & Objective: Electronic health records (EHR) are a valuable source of real-world data for clinical research. Reliable methods are needed to identify type 1 diabetes (T1D) by EHR. The study aim was to develop an EHR algorithm for T1D using diagnosis codes and clinical characteristics validated by manual chart review.

Methods: Data from the Center for Kidney Disease Research, Education, and Hope (CURE-CKD) Registry were used to identify persons ≥12 years of age with T1D by diagnosis codes, ketoacidosis encounters, insulin pump and short-acting insulin use during 2013-2022. Manual chart review was performed on a subset of 100 participants to verify selection criteria; 40 from “very likely” category, 40 from “likely” category, and 20 from “unlikely”.

Results: T1D cohorts (N=31,695) were divided into likelihood categories based on confirming criteria (Figure). The cohort was 48.5% female and 75.7% White race identity. Their mean±SD for age was 42±19 years with HbA1c 8.2±2.1% and BMI 27.5±6.3 kg/m². Positive predictive values were 100% for very likely, 98% for likely, and 40% for unlikely categories.

Conclusions: T1D can be reliably identified by the CURE-CKD algorithm using clinical EHR data. Pending further validation, this algorithm can be used for T1D cohort selection in real-world studies of epidemiology, diagnosis, and treatment.

View largeDownload slide

View largeDownload slide

Close modal