Deterministic vs. Probabilistic: Best Practices for Patient Matching Based on a Comparison of Two Implementations

Area of Science:

• Health Informatics
• Data Management
• Medical Imaging

Background:

• Sharing patient data across healthcare systems requires reliable patient record linkage.
• Ontario has four centralized diagnostic imaging repositories (DIRs) facilitating regional data sharing.
• Two regional DIRs in south-central and southeast Ontario utilize different patient matching methodologies.

Purpose of the Study:

• To analyze and compare the effectiveness of a deterministic patient matching approach against a probabilistic one.
• To evaluate patient identity domain performance in consolidating regional diagnostic imaging repositories.
• To identify insights from comparing deterministic and probabilistic patient matching strategies.

Main Methods:

• Analysis of patient data matching techniques used by two distinct diagnostic imaging repositories (DIR1 and DIR2).
• DIR1 employs a deterministic matching method based on a single specific value.
• DIR2 utilizes a probabilistic scorecard, weighing various patient demographics for matching.

Main Results:

• The study measured the performance of a deterministic patient identity domain against an established standard.
• Comparative analysis aimed to reveal differences in accuracy and efficiency between the two matching methods.
• Insights were gathered on the strengths and weaknesses of both deterministic and probabilistic patient matching.

Conclusions:

• Understanding the nuances of different patient matching algorithms is essential for effective health data integration.
• The findings provide a basis for optimizing patient record linkage in multi-institutional health systems.
• This research contributes to improving the reliability of patient data sharing across diagnostic imaging repositories.