Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
June 7, 2026Diabetes

All eight EHR-based algorithms showed poor T1D identification with PPV ranging from 23-79% and modest discrimination (ROC AUC 0.65-0.78).

View Full Paper
Ask AI
Bookmark
Share

Why the study?

Do EHR-based classification algorithms accurately identify biomarker-defined adult-onset T1D in patients with adult-onset diabetes?

Population

7081 adult-onset diabetes cases from the MyCode cohort, mean age at diagnosis 59.8 years, 55% female, 95%…

Comparison

Eight widely used EHR-based classification… vs Biomarker-defined diabetes subtypes

Design

Cohort

Follow-up

median diabetes duration of 9.0 years at biomarker…

Authors

JLJIANG LIRSRICHARD STAHLAGALICIA GOLDEN

Discussion

Loading...

Member takes

Overview

EHR-based algorithms substantially misclassify adult-onset Type 1 Diabetes compared to biomarker definitions, highlighting the need for biomarker-informed classification in research and clinical care.

Key Points

  • To validate EHR-based diabetes classification algorithms against biomarker-defined adult-onset T1D.
  • Analyzed 7081 adult-onset diabetes cases from the MyCode cohort, including 3365 insulin-treated and 3152 non-insulin-treated individuals.
  • Measured random C-peptide and islet autoantibodies to define T1D and T2D.
  • Evaluated eight EHR-based algorithms for positive predictive value (PPV), negative predictive value (NPV), and ROC AUC.
  • Identified 3.1% adult-onset T1D and 71% T2D based on biomarkers.
  • Algorithms demonstrated poor identification of T1D, with PPV ranging from 23-79% and high NPV (96-97%).
  • The lipid model by Lynam et al. had the highest PPV at 79%, but ROC AUC was modest (0.65-0.78), especially lower for insulin-treated individuals.

Structured PICO

Do EHR-based classification algorithms accurately identify biomarker-defined adult-onset T1D in patients with adult-onset diabetes?

P
Population
7081 adult-onset diabetes cases (3365 insulin-treated and 3152 non-insulin-treated) from the MyCode cohort, mean age at diagnosis 59.8 years, 55% female, 95% Europeans.
I
Intervention
Eight widely used EHR-based classification algorithms for adult-onset T1D
C
Comparator
Biomarker-defined diabetes subtypes (C-peptide and islet autoantibodies)
O
Outcome
Positive predictive value (PPV), negative predictive value (NPV), and ROC AUC for identifying adult-onset T1D

EHR-based algorithms substantially misclassify adult-onset Type 1 Diabetes compared to biomarker definitions, highlighting the need for biomarker-informed classification in research and clinical care.

Cite This Study

LI et al. (2026) studied this question.

synapsesocial.com/papers/6a250ae37def13d035e1afd2https://doi.org/10.2337/db26-2214-p
View Full Paper
Ask AI
Bookmark
Share

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 12222-P: Prospective Validation of AI for Detecting Misclassified Adult Type 1 Diabetes: Insights on Precision, Clinical Workflows, and Adoption2026
  2. 22231-P: RECLASS-T1D: Real-World Evaluation of AABBCC Criteria for Diabetes Classification: Experience from 6.7 Million U.S. Patients2026
  3. 32223-P: Harnessing Machine Learning to Forecast Two-Year Progression to Stage 3 Type 1 Diabetes (T1D) in Individuals with Multiple Type 1 Diabetes Autoantibodies: A Noise-Resistant, Multidimensional Clustering Approach2026
  4. 42321-P: Machine-Learning Modeling for T2DM Prediction in over 3 Million Adults2026
  5. 52401-P: Genetic Regulation of Molecular Traits Enhances Precision Risk Stratification in Youth-Onset Type 2 Diabetes2026