Oral abstracts

Aims: <br/>Misclassification of diabetes types (type 1, type 2 and MODY) is common, affecting 7%–15% of type 1 and 2 cases and 77% of MODY cases. We developed a decision support tool using validated prediction models to automatically search primary care systems and identify patients with diabetes at high probability of misclassification, and who may benefit from further investigations or referral to secondary care. This study evaluates the feasibility and acceptability of using this tool to improve diabetes classification in adults diagnosed ≤50 years.<br/><br/>Methods: <br/>A mixed-methods, non-randomised intervention study across 12 primary care practices in the Southwest and East Midlands, UK. Feasibility is assesed through practice uptake, data quality, misclassification rates and practitioner workload. Acceptability is explored via one-to-one qualitative interviews with up to 40 practice staff and patients and thematically analysed.<br/><br/>Results: <br/>Eleven practices have run the tool (mean practice size: 15,623), identifying a mean of 27 patients with diabetes per practice with potential misclassification: 3 with a high probability of MODY, 12 coded as type 1 but more likely type 2 and 12 coded as type 2 but more likely type 1. Preliminary qualitative data suggest the tool can help to audit miscoding and improve patient care. Practice staff reported that the tool was easy to use, though embedding in primary care systems was preferred.<br/><br/>Conclusions: <br/>The DePICtion tool shows promise for improving diabetes classification and is potentially acceptable to both practitioners and patients. Further refinement could ensure better diagnosis and management for more patients.