Reduce Diagnostic Bias in Rare Diseases with AI-Powered Patient Identification
Download our scientific poster presented at ATS 2024 and discover how machine learning models help identify likely undiagnosed patients with Alpha-1 Antitrypsin Deficiency (AATD), especially among underrepresented racial and ethnic groups.
What’s Inside the Poster PDF
Clear visualizations of disparities in AATD diagnosis by race
Model architecture and validation results
Real-world implications for trial design and patient targeting
Actionable strategies for integrating AI into clinical workflows
Get the poster and explore how AI improves patient identification across clinical cohorts
Diagnostic Inequity in Rare Diseases
Despite advances in data and diagnostics, patients from Asian, Black, Hispanic, and other minority populations remain underdiagnosed in many rare diseases, including AATD.
Clinical trials and development pipelines risk being built on incomplete patient profiles — delaying time-to-diagnosis and treatment access.
Machine Learning for Early & Equitable Patient Identification
Our team developed a prediction model that analyzes over 330 million de-identified patient records and flags likely undiagnosed cases of AATD based on:
Medical claims data
Treatment and diagnosis history
Race and ethnicity context
Comorbidity patterns
This model was designed to improve diagnostic reach and close gaps in identification across racial groups.
What the Model Uncovered
16,431 likely undiagnosed AATD patients identified
48% of flagged patients were non-White
Median age at flagging: 10 years younger than diagnosed patients
Significant underrepresentation of Black, Hispanic, and Asian populations in current diagnosed cohorts
Higher asthma prevalence in undiagnosed flagged patients
+42% earlier identification compared to traditional approaches
Better diagnosis starts with better data.
Learn how AI helps uncover underdiagnosed patients across racial and ethnic groups — and what it means for your trials.
What’s Inside the Poster PDF
Clear visualizations of disparities in AATD diagnosis by race
Model architecture and validation results
Real-world implications for trial design and patient targeting
Actionable strategies for integrating AI into clinical workflows