Is it possible to implement a rare disease case-finding tool in primary care? A UK-based pilot study

INTRODUCTION: This study implemented MendelScan, a primary care rare disease case-finding tool, into a UK National Health Service population. Rare disease diagnosis is challenging due to disease complexity and low physician awareness. The 2021 UK Rare Diseases Framework highlights as a key priority the need for faster diagnosis to improve clinical outcomes. METHODS AND RESULTS: A UK primary care locality with 68,705 patients was examined. MendelScan encodes diagnostic/screening criteria for multiple rare diseases, mapping clinical terms to appropriate SNOMED CT codes (UK primary care standardised clinical terminology) to create digital algorithms. These algorithms were applied to a pseudo-anonymised structured data extract of the electronic health records (EHR) in this locality to "flag" at-risk patients who may require further evaluation. All flagged patients then underwent internal clinical review (a doctor reviewing each EHR flagged by the algorithm, removing all cases with a clear diagnosis/diagnoses that explains the clinical features that led to the patient being flagged); for those that passed this review, a report was returned to their GP. 55 of 76 disease criteria flagged at least one patient. 227 (0.33%) of the total 68,705 of EHR were flagged; 18 EHR were already diagnosed with the disease (the highlighted EHR had a diagnostic code for the same RD it was screened for, e.g. Behcet’s disease algorithm identifying an EHR with a SNOMED CT code Behcet's disease). 75/227 (33%) EHR passed our internal review. Thirty-six reports were returned to the GP. Feedback was available for 28/36 of the reports sent. GP categorised nine reports as "Reasonable possible diagnosis" (advance for investigation), six reports as "diagnosis has already been excluded", ten reports as "patient has a clear alternative aetiology", and three reports as "Other" (patient left study locality, unable to re-identify accurately). All the 9 cases considered as "reasonable possible diagnosis" had further evaluation. CONCLUSIONS: This pilot demonstrates that implementing such a tool is feasible at a population level. The case-finding tool identified credible cases which were subsequently referred for further investigation. Future work includes performance-based validation studies of diagnostic algorithms and the scalability of the tool.