One in 17 of us will develop a rare disease. Would you want to know if you’re one of them?

Hiba Junaid, Data Scientist at Barts Life Sciences and Suthesh Sivapalaratnam, Consultant Haematologist at Barts Health, spoke at the Barts Health flagship event celebrating International Clinical Trials Day and its contribution to detecting rare diseases through the use of AI. 

Rare diseases affect over 3.5 million people in the UK alone, yet it can take nearly 5 years on average for patients to receive a diagnosis. Some people wait decades, enduring an odyssey of diagnoses. 

One of these rare diseases is Gaucher’s disease, an often-misdiagnosed condition that mimics common blood disorders. Symptoms like low blood counts, fatigue, and enlarged spleens often lead doctors towards more familiar conditions, meaning Gaucher’s can go either undetected or misdiagnosed. 

At Barts Life Sciences, we developed an algorithm incorporating AI to identify people at higher risk of rare diseases, starting with Gaucher. 

Working with Sanofi using a specially developed algorithm, we scanned 2.5 million patient records from our hospital level datasets and combined these with North East London GP data, including blood test results, diagnoses, and clinical notes. We looked at a range of indicators such as diagnostic codes (ICD-10, SNOMED CT), key lab results (e.g. enzyme activity), genetic markers (eg the GBA gene), and uses the natural language processing tool MedCAT to scan free-text notes.  

If one clinician reviewed the free text notes for 160,000 patients, it would take them 3 years to do so – without taking a break. The natural language processing tool incorporated in this project, took a few hours to review the same amount of free text. 

We identified 8 new confirmed Gaucher diagnoses, and 16 high-risk individuals flagged for clinical review and further investigation. Each diagnosis brings someone closer to appropriate care and often a much better quality of life.  

We are now planning to expand this model framework to more than 10 rare diseases, including immunological conditions. 

What comes next?  

However, the use of AI in clinical settings raises some serious questions. If you were flagged as high-risk, would you want to know?  And who should contact you? Your GP? A specialist? An AI-generated message? We asked the audience and the answers varied, showing how important it is to involve the public and clinicians in shaping these pathways. 

And, of course, we face other big questions: Who owns healthcare data? How do we keep it secure? What role should tech companies play? This requires us to take a more holistic view of AI in clinical care. 

Using AI to surface what’s hidden could dramatically shorten the diagnostic journey for thousands of patients. But only if we build it with care, collaboration, and consent. 

And what did our audience say? Did they want to know? 82% of our audience said yes. 

Would you want to know?