Evaluating Data Linkage: Creating longitudinal synthetic data to provide gold-standard linked data sets for robust methodological evaluation

When performing probabilistic data linkage on real-world data we, by the fact we need to link it, do not know the true linkage. Therefore, the success of a linkage approach is difficult to evaluate. Often small hand linked datasets are used as a ‘gold-standard’ for a linkage approach to be evaluated against. However, errors in the hand-linkage and the limited size, number, and availability of these datasets do not allow for robust evaluation.

This work looks to demonstrates the benefits of using longitudinal synthetic data to provide gold-standard linked data sets for linkage evaluation in the domain of Population Reconstruction. This research outlines:

– a model for generating population data sets

– an approach for statistical verification of the generated data sets

– an evaluation approach to verify the model across the range of possible inputs

– an approach for using the data to robustly evaluate linkage approaches

In application, the synthesised gold-standard data sets will be used to evaluate a range of preexisting linkage approaches. The evaluation will consider the performance of each linkage approach over variations in the population size, characteristics, and levels of data corruption.