The multiple-record systems estimator when registrations refer to different but overlapping populations

Area of Science:

• Epidemiology
• Biostatistics
• Population Health

Background:

• Multiple-record systems estimation typically assumes all data sources pertain to the same population.
• This assumption is often violated in practice, with datasets covering different time periods or geographical regions.

Purpose of the Study:

• To develop and validate a method for multiple-record systems estimation when registration data relate to different populations.
• To assess the accuracy of population size estimates under these conditions.

Main Methods:

• Development of a novel statistical method for handling disparate population data in record linkage.
• Application of the Expectation-Maximization (EM) algorithm for generalized problem-solving.
• Utilizing the parametric bootstrap method for constructing confidence intervals.

Main Results:

• Demonstrated that under specific conditions, ignoring population differences yields correct population size estimates.
• The EM algorithm provides a flexible approach for more complex scenarios.
• The parametric bootstrap effectively constructs reliable confidence intervals.

Conclusions:

• The proposed method offers a robust solution for population size estimation with non-overlapping or partially overlapping datasets.
• This methodology enhances the accuracy and applicability of multiple-record systems in diverse epidemiological contexts.
• The study successfully applied the method to neural tube defect data across different time periods.