Adult-onset Type 1 Diabetes is frequently misdiagnosed as Type 2, delaying proper treatment and harming outcomes. To address this, IQVIA and Breakthrough T1D developed an AI model using de-identified EHR data to distinguish T1D from T2D. The algorithm analyzed markers such as autoantibody presence, C-peptide levels, insulin dependency, and overlooked predictors like lack of metformin history or higher specialist visits. In validation testing, the model achieved 35% precision at 5% recall, identifying more than 1 in 3 misdiagnosed patients correctly. This approach improves diagnosis, reduces healthcare burden, and advances T1D research.

Join for free to read