A rare disease diagnosis currently takes years of specialist referrals and hospital tests, but an AI tool called MendelScan is being tested to spot these patients earlier by scanning GP records for hidden clues. This matters because rare disease patients often endure a long, costly diagnostic odyssey—seeing multiple specialists, undergoing repeated tests, and accumulating avoidable healthcare visits—before finally getting a diagnosis. MendelScan aims to cut that time dramatically by flagging patterns in routine primary care data that suggest an undiagnosed rare disease. If the trial succeeds, the tool could be rolled out across NHS primary care networks. GPs would receive automated alerts prompting them to refer patients for targeted genetic testing, catching conditions months or years earlier than current practice. This would reduce unnecessary hospital activity, improve patient outcomes, and free up specialist resources. The project is designed to generate the evidence on efficacy, economics, and acceptability needed for national commissioning—meaning it could quietly transform how the NHS finds rare disease patients, shifting from reactive specialist care to proactive detection in the GP surgery.
View original technical description
Background MendelScan has shown significant promise in: (1) detecting RD patients earlier than the current standard of care in retrospective studies on research datasets; and (2) detecting previously undiagnosed patients with very high probability of RD diagnosis on real world cohorts of c.200,000 NHS patients. Aims and Objectives This study aims to build on these initial findings to evaluate in real world NHS settings whether (vs current standard of care) MendelScan can: Reduce time-to-diagnosis for RD patients Increase overall RD diagnosis rates (find previously undiagnosed patients) Reduce avoidable healthcare activity associated with the pre-diagnosis phase for RD patients Be implemented practically, ethically and affordably within the NHS Work plan Four phases: Retrospective algorithm validation. Evaluating sensitivity, specificity and positive predictive value of RD case-finding algorithms by running them on very large research datasets (13m patients) of already-diagnosed RD patients. Implementation research. Studies on ethical and practical deployment issues including user acceptability, exploring the right user target and resource-intensity of deployment. Real world data collection. Deployment of MendelScan in real world primary care settings (initially 10 PCNs/practice groups, 750k patients, with potential to extend to c. 2m) to collect data on acceptance of recommendations by GPs, referrals, diagnoses, and resources / activity in the pre-diagnosis phase. Prospective research study, under protocol. Formal evaluation of the MendelScan tool using the NHS Clinical Research Network assessing the tool’s impact on diagnosis rates, time to diagnosis, and resource-intensity of the pre-diagnosis phase (vs current standard of care). Timelines for delivery 18 months. Phases 1-3, first 9 months, Phase 4 month 10-16. Evaluation and dissemination months 16-18. Anticipated Impact and Dissemination The study is designed to create the efficacy, economics and acceptability evidence base to justify broad national commissioning. Exit strategy after funding Deployment of MendelScan through ICSs (via PCNs) across the NHS.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know