Case Study
Case Study
Patient Identification in Rare Diseases
Client Challenges
A global biotechnology company has developed a powerful therapy for rare metabolic disease patients, but many potential beneficiaries don’t know they have the disease until it’s progressed significantly. Our partner asked us to use our data assets and AI capabilities to identify more patients.
OM1 Solution
We built our proprietary OM1 Patient Finder to identify key patterns in medical histories that signal undiagnosed disease. Using OM1 Patient Finder for our client, we achieved:
Very strong overall predictive performance (AUC of 0.82)
Identification of patient subgroups at much higher risk than average, with more than 20x positive patients in the riskiest 1% of the population
Clarification of diffuse symptoms - like GI distress - that clinicians recognize but often aren’t specific enough to us in diagnosis
New hypotheses for better understanding and predicting disease progression and complications
Obstacle 1
Lack of quality data for understanding and identifying rare disease patients
The OM1 Patient Finder is integrated with the OM1 Real-World Data Cloud of over 250 million patients, which allows us to achieve strong predictive capability with high quality research-grade data.
Obstacle 2
Limited explainability of AI technologies
At OM1, we aim to achieve strong predictive performance while also maintaining clinical explainability. We can describe what's going on in the technology, so you're not left in the dark.
Obstacle 3
Lack of real-world application in clinical practice
At OM1, we conceptualize and calibrate Patient Finder from the start to be useful in clinical practice to find actual patients.
Results
The OM1 Patient Finder delivered tangible results, accurately identified patients, and improved clinical decision making:
The combination of strong data assets and a purpose-built AI platform delivered transparent and implementable results
We independently identified all the major clinical characteristics of this disease - and some that patients frequently bring up as under-appreciated, like mental health and addiction issues
Because Patient Finder is built to accept a range of data inputs, it can be implemented in a variety of environments
Platform Calibration
Key Milestones
Rapid initial calibration (3-4 weeks)
Go / no-go decision for both teams
Analytic outcomes and scenario simulation
Deep clinical insight into explanability
Clinical implementation into workflows to find patients
Cutting-edge AI technology integrated with high quality RWD can deliver real value and rare insights. The OM1 Patient Finder has demonstrated the ability to improve detection and diagnosis of rare disease patients. Diseases with long paths to diagnosis and diffuse, heterogeneously symptomology are especially good candidates.
