The Impact of Name Transformation on Match Rates Within a Large Consumer Database

Area of Science:

• Health Informatics
• Data Science
• Biostatistics

Background:

• Accurate patient record linkage is crucial for healthcare research.
• Privacy-preserving record linkage methods are sensitive to input data quality.
• Name features are vital for effective record linkage.

Purpose of the Study:

• To evaluate the impact of name transformation strategies on patient matching accuracy.
• To assess the effectiveness of various name transformation techniques.
• To determine optimal methods for enhancing patient record linkage.

Main Methods:

• Utilized a large commercial dataset of 68 million records (59 million unique individuals).
• Implemented and evaluated eight distinct name transformation strategies.
• Measured performance using precision, recall, and F1 scores.

Main Results:

• Transforming names to include common nicknames significantly increased recall.
• This nickname strategy maintained high precision while boosting the F1 score to 0.905 (vs. 0.807 without transformation).
• Other tailored feature transformation strategies also showed improvements.

Conclusions:

• Name transformation strategies enhance the precision and recall of patient matching.
• Optimized name transformations are key to creating high-quality, linked datasets for research.
• These methods improve the reliability of patient data linkage.