Personalized treatment options for chronic diseases using precision cohort analytics.

Area of Science:

• Health Informatics
• Machine Learning in Medicine
• Clinical Decision Support

Background:

• Point-of-care decision-making is complex, often relying on generalized guidelines.
• Electronic Health Records (EHRs) contain vast amounts of patient data that can inform personalized treatment.

Purpose of the Study:

• To develop and evaluate a machine-learning workflow for precision cohort treatment options (PCTO).
• To support point-of-care clinical decisions by analyzing outcomes of past treatments for similar patient cohorts.

Main Methods:

• Developed a PCTO workflow involving data extraction, similarity model training, precision cohort identification, and treatment option analysis.
• Utilized EHR data from a large healthcare provider for hypertension (HTN), type 2 diabetes mellitus (T2DM), and hyperlipidemia (HL).
• A similarity model was trained to dynamically create cohorts of similar patients.

Main Results:

• The retrospective analysis showed that better treatment options were available for a majority of cases (75% for HTN, 74% for T2DM, 85% for HL).
• Models for HTN and T2DM were successfully deployed in a pilot study with primary care physicians.
• The workflow dynamically generates personalized treatment insights at the point-of-care.

Conclusions:

• A novel data-analytic workflow enables the creation of patient-similarity models for personalized treatment insights.
• Physicians can integrate real-world evidence from EHR data with clinical guidelines for improved medical decision-making.
• This approach enhances the ability to provide evidence-based, individualized care at the point-of-care.