What is the clinical evidence from this study?

Study design: Observational. Population: Misclassified Type 1 Diabetes. Intervention: AI algorithm for detecting misclassified T1D. Primary outcome: Rates of suspected and confirmed T1D cases.

What does this research mean for the field?

An AI algorithm integrated into clinical workflows can identify adults with misclassified Type 1 diabetes with reasonable precision, though real-world adoption is hindered by operational barriers such as chart review time constraints and data gaps. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

What question did this study set out to answer?

This study aims to validate an AI algorithm for identifying adults misclassified as having type 2 diabetes instead of type 1 diabetes.

June 7, 2026

2222-P: Prospective Validation of AI for Detecting Misclassified Adult Type 1 Diabetes: Insights on Precision, Clinical Workflows, and Adoption

Resultado clave

An AI algorithm integrated into clinical workflows helped clinicians identify suspected misclassified T1D patients with reasonable precision, though specific quantitative results were not reported.

Puntos clave

This study aims to validate an AI algorithm for identifying adults misclassified as having type 2 diabetes instead of type 1 diabetes.
The AI algorithm was applied to electronic medical records data from two US health systems.
Endocrinologists and PCPs reviewed flagged cases over a 6-month period and provided feedback on the integration process.
Quantitative measures assessed rates of suspected and confirmed T1D cases.
Providers reported the algorithm effectively flagged suspected T1D cases, facilitating diagnosis.
Barriers included time constraints for chart reviews and incomplete data in the health information exchange.
Operational challenges regarding ownership and prioritization affected clinician engagement.

Diseño del estudio

Tipo

Observational

Multicéntrico

Sí

PICO estructurado

Does an AI algorithm integrated into clinical workflows improve the detection of misclassified adult Type 1 Diabetes?

Población

Adults at high risk for misclassified T1D evaluated over 6 months at two US health systems.

Exposición

AI algorithm trained to identify adults with misclassified T1D from EMR data, integrated into clinical workflows via the HSX health information exchange

Resultado

Real-world precision, utility, and barriers to adoption (rates of suspected and confirmed T1D cases)

An AI algorithm can help clinicians identify misclassified adult Type 1 Diabetes with reasonable precision, though operational and workflow barriers to adoption remain.

Limitaciones

Chart review time constraints
Gaps in HIE data
Operational challenges in ownership, prioritization, and care gap intervention by physician peers

Resumen

Introduction and Objective: Misclassification of T1D as T2D is common in adults, potentially leading to ineffective treatment and adverse outcomes. This study prospectively validates an AI algorithm trained to identify adults with misclassified T1D from EMR data, integrating it into clinical workflows at two US health systems via the HSX health information exchange (HIE). The aim is to assess its real-world precision, utility, and barriers to adoption. Methods: The algorithm was run on HSX data to flag patients at high risk for misclassified T1D at the two sites. Endocrinologists and PCPs reviewed these patients' charts, intervened on suspected T1D cases over 6 months, and shared feedback on data quality and workflow integration. Quantitative measures included rates of suspected and confirmed T1D cases. Results: Figure 1 shows the quantitative results obtained by the end of the follow-up period. Providers found the algorithm useful for flagging suspected T1D cases but also identified barriers: chart review time constraints, gaps in HIE data, and operational challenges in ownership, prioritization, and care gap intervention by physician peers. Conclusion: Our study shows that AI can help clinicians identify suspected T1D patients with reasonable precision. In the future, inclusion criteria and EMR data can be refined to improve precision and reduce clinician burden. More research and clinical engagement are also needed to address operational challenges. Disclosure C. Anastasopoulou: Speaker's Bureau; Current; Ascendis Pharma A/S. Speaker's Bureau; Ended; Sanofi. Consultant; Ended; Radius Pharma. I. Brusini: Employee; Current; IQVIA Inc. A.D. Rao: Advisory Panel; Ended; Corcept Therapeutics. M. Hackenberg: Employee; Current; IQVIA Inc. A. Sees: Employee; Current; IQVIA Inc. C.J. McFadden: Employee; Current; IQVIA Inc. B. Meyer: Employee; Current; IQVIA Inc. S. Lee: Employee; Ended; IQVIA Inc. E. Latres: Stock/Shareholder; Current; Regeneron Pharmaceuticals Inc. G. Liptak: None. N. Leavitt: Employee; Current; IQVIA Inc. Funding Breakthrough T1D (2-PAR-2023-1293-I-X)

Me gusta

Guardar