The development of matching criteria for epidemiological studies using record linkage techniques

Area of Science:

• Data Science
• Record Linkage
• Biometrics

Background:

• Accurate identification of individuals across datasets is essential for public health and research.
• Existing methods for record linkage often rely on multiple data points for confirmation.

Purpose of the Study:

• To test the hypothesis that identical surnames, given names, and birth dates reliably indicate the same person.
• To quantify the probability of false matches in record linkage based on shared name and birth date information.

Main Methods:

• Compiled frequency distributions of common birth date digits between matched records from two registers.
• Compared empirical distributions with computer simulations modeling different individuals.

Main Results:

• A significant divergence between observed and simulated distributions confirmed the hypothesis.
• When surname, two given names, and at least four birth date digits matched, only 0.012% of records represented different people.
• With surname, two given initials, and six birth date digits matching, 0.1% of records were false matches.

Conclusions:

• High confidence in same-person identification is achieved with matching surnames, multiple given names/initials, and a substantial number of birth date digits.
• The study provides a quantitative basis for setting thresholds in probabilistic record linkage systems.