Algorithmic identification of atypical diabetes in electronic health record (EHR) systems

Abstract

Aims: Understanding atypical forms of diabetes (AD) may advance precision medicine, but methods to identify such patients are needed. We propose an electronic health record (EHR)-based algorithmic approach to identify patients who may have AD, specifically those with insulin-sufficient, non-metabolic diabetes, in order to improve feasibility of identifying these patients through detailed chart review.

Methods: Patients with likely T2D were selected using a validated machine-learning (ML) algorithm applied to EHR data. "Typical" T2D cases were removed by excluding individuals with obesity, evidence of dyslipidemia, antibody-positive diabetes, or cystic fibrosis. To filter out likely type 1 diabetes (T1D) cases, we applied six additional "branch algorithms," relying on various clinical characteristics, which resulted in six overlapping cohorts. Diabetes type was classified by manual chart review as atypical, not atypical, or indeterminate due to missing information.

Results: Of 114,975 biobank participants, the algorithms collectively identified 119 (0.1%) potential AD cases, of which 16 (0.014%) were confirmed after expert review. The branch algorithm that excluded T1D based on outpatient insulin use had the highest percentage yield of AD (13 of 27; 48.2% yield). Together, the 16 AD cases had significantly lower BMI and higher HDL than either unselected T1D or T2D cases identified by ML algorithms (P<0.05). Compared to the ML T1D group, the AD group had a significantly higher T2D polygenic score (P<0.01) and lower hemoglobin A1c (P<0.01).

Conclusion: Our EHR-based algorithms followed by manual chart review identified collectively 16 individuals with AD, representing 0.22% of biobank enrollees with T2D. With a maximum yield of 48% cases after manual chart review, our algorithms have the potential to drastically improve efficiency of AD identification. Recognizing patients with AD may inform on the heterogeneity of T2D and facilitate enrollment in studies like the Rare and Atypical Diabetes Network (RADIANT).

Fig 1. Flow diagram–base and branch algorithms for identifying patients with possible atypical diabetes (AD) in the Massachusetts General Brigham (MGB) Biobank.

MGBB: Mass-General Brigham Biobank; T2D: type 2 diabetes; BMI: body mass index; HDL: high-density lipoprotein; T1D: type 1 diabetes; NPV: negative predictive value; PPV: positive predictive value.

Fig 2. Method for classification of diabetes type.

Research assistants summarized and endocrinologists manually reviewed identified charts to classify patients into one of three categories: atypical diabetes (AD), not AD (including typical type 1, type 2, LADA, MODY, pancreatic, steroid-induced, and secondary diabetes), and need more information (NMI). Methodology for classification of diabetes based on clinical data including interpretation of genetic testing, lab values, medication use, and medical history. MGB: Mass-General Brigham; MGBB: Mass-General Brigham Biobank; T2D: type 2 diabetes; HDL: high-density lipoprotein; CF: cystic fibrosis; T1D: type 1 diabetes; EMR: electronic medical records; MODY: maturity-onset diabetes of the young; dx: diagnosis; yrs: years; NMI: need more information; BMI: body mass index.

Fig 3

Polygenic scores compared between individuals in the AD, ML T1D, and ML T2D cohorts: (a) T1D rsPS, and (b) T2D gePS. (a) Restricted to significant polygenic score (rsPS) for T1D did not differ significantly between the atypical diabetes (AD), machine learning (ML) type 1 diabetes (T1D), and ML type 2 diabetes (T2D) cohorts. (b) The global extended polygenic score (gePS) for T2D was significantly higher in the AD group in comparison to the ML T1D cohort.