chevron_left
Building an ML model to identify undiagnosed patients with rare disease
Life sciences company partners with Clarivate Real-World Data team to identify patients with a disease prone to missed or delayed diagnosis
A North American biotech company collaborated with Clarivate to address the challenge of accurately identifying patients with a rare, inherited disease prone to missed or delayed diagnosis. By leveraging machine learning models on EHR and claims data from the prior five years, we achieved an identification accuracy range of 75-80%.
This success was underpinned by thorough groundwork and strategic feature selection, ensuring robust and reliable model performance. This case study explores the challenges and opportunities associated with the use of ML within rare disease space, probing such questions as:
- Can machine learning significantly improve rare disease patient identification?
- How did our ML approach boost patient identification accuracy up to 80%?
- How to avoid the pitfalls of a “garbage in, garbage out” scenario, biases, and inaccuracies?
- What steps are crucial to avoiding misclassification in rare disease detection?
- Why is feature selection essential for optimizing ML model performance?
