Precision screening for familial hypercholesterolaemia: a machine learning study applied to electronic health...

Area of Science:

• Genetics and Genomics
• Cardiovascular Medicine
• Artificial Intelligence in Healthcare

Background:

• Familial hypercholesterolaemia (FH) significantly increases the risk of early heart attacks and strokes.
• Despite its impact, an estimated 90% of individuals with FH remain undiagnosed in the USA.
• Early diagnosis and management are crucial for improving cardiovascular outcomes in FH patients.

Purpose of the Study:

• To develop and apply a machine learning model (FIND FH) to accelerate the early diagnosis of familial hypercholesterolaemia.
• To identify over 1.3 million undiagnosed individuals at high risk for cardiovascular events.
• To facilitate timely intervention for individuals with undiagnosed FH.

Main Methods:

• Trained the FIND FH machine learning model using deidentified health-care encounter data from four US institutions.
• Utilized diagnostic codes, prescriptions, and laboratory findings for model training and validation.
• Applied the validated model to large national and integrated health-care delivery system datasets.

Main Results:

• The FIND FH model identified 1,331,759 individuals likely to have familial hypercholesterolaemia in a national database.
• Expert review of flagged individuals confirmed high clinical suspicion warranting further evaluation and treatment.
• The model demonstrated a precision of 0.85 and an ROC AUC of 0.89.

Conclusions:

• The FIND FH machine learning model effectively identifies individuals with familial hypercholesterolaemia across large, diverse datasets.
• This approach can significantly improve the detection rate of undiagnosed FH.
• Accelerated diagnosis through machine learning can lead to earlier interventions and better cardiovascular outcomes.