AI Insight
The DePICtion study tested a decision support tool in 11 UK primary care practices to identify patients with potentially misclassified diabetes subtypes. The tool flagged 7% of diabetes patients as potentially misclassified, and among those who received definitive biomarker testing, 18% had their diagnosis changed, including 2 patients diagnosed with MODY and 11 reclassified between type 1 and type 2 diabetes. However, only about half of flagged patients received appropriate testing, indicating barriers to implementation despite the tool's diagnostic utility.
Why it matters
Correct diabetes classification is critical for appropriate treatment, as different subtypes require different management approaches. This tool could help identify the estimated 15% of patients receiving incorrect treatment due to misclassification, though improved practice engagement and data recording systems are needed to maximize its effectiveness in primary care settings.
Understand the Science
⚠️ Preprint – Noch nicht peer-reviewed
Dieser Artikel wurde noch nicht von unabhängigen Experten begutachtet. Die Ergebnisse sind vorläufig und sollten mit Vorsicht interpretiert werden.
Introduction: There are several subtypes of diabetes, including type1, type 2, and a genetic form called MODY. Classification can be challenging, particularly in young people, with up to 15% of people misclassified with the wrong subtype, resulting in inappropriate treatment. We assessed the feasibility of embedding a decision support tool (the DePICtion tool) in GP records to flag patients at risk of misclassification for further investigation and biomarker testing. Research Design & Methods: The DePICtion tool was run in 11 UK GP practices. The tool generated a list of patients at high probability of being misclassified for review by practice staff. After review, practices returned a follow-up data form, summarising the actions taken for these patients. Results: Overall, 7% (237/3501) of patients were flagged as potentially misclassified. Of 203 with follow-up data, 28 (14%) records had coding errors and 2 patients had died, leaving 173 (85%) suitable for biomarker testing to confirm the diagnosis. Around half (88/173) of these patients received new biomarker testing or had previous testing on the system (73 complete, 15 partial or unknown results). 17% (30/173) of patients declined/did not respond. For the remaining 55, the practice reason for not testing did not rule out an alternative diagnosis in the majority of cases (n=44). Among the 73 patients with definitive results, nearly 1 in 5 (18%; 13) changed diagnosis (2 diagnosed with MODY, 6 from type 1 to type 2, 5 from type 2 to type 1). Conclusions: The DePICtion tool identified people with diabetes that had been misclassified. However, many people flagged as potentially misclassified did not have classification biomarker testing. Better data recording and greater engagement with secondary care could improve uptake of testing to identify more misclassified patients and optimise their diabetes care.