Problem:

• Interoperability and record sharing will compound the patient matching problem by increasing the volume of data that must be matched to the correct patients.
•  New sources of data are emerging that will need to be integrated into the patient profile, including the patient-generated health information submitted through websites and wearables.
• Correlating official medical records with all this additional information will become almost impossibly complex.

Solution:

• The deterministic rules are defined to identify the exact and potential matches of records based on similar scores.
• A machine learning model is implemented to make decisions on potential matching pairs and predict whether they are a match or not to combine patient health records.
• Automatically selecting the most challenging pairs for users to manually resolve, then retraining ML models based on resolved data.

Tech Stack:

• Python
• Azure Machine learning algorithm
• Azure Statistics
• Azure SQL

Benefits:

• To obtain a comprehensive view of the patient's health record.
• Accurate patient matching is a top priority to get complete visibility into the care continuum. Correct patient matching powers our care coordination workflows; it can lower costs, enhance clinical decision-making, and improve patient care and coordination.
• It helps physicians track a patient’s treatment history, assists patients in making more informed decisions about their treatment options, and assists health care professionals in determining which form of treatment will best suit their patients’ overall needs.