Linking longitudinal health data to track care following emergency department presentation: challenges, solution and an exemplar in early pregnancy complications.

Abstract

Background: Linking routinely collected patient-level health data supports service planning and research while protecting privacy, though it poses several challenges. This paper demonstrates the process used to produce high-quality linked data for early pregnancy complication care for patients presenting to the Emergency Department (ED).

Methods: Ten years of regional health data were merged using deterministic linkage in five steps: 1) applying strict inclusion/ exclusion criteria via extraction code, 2) collecting data, 3) refining and pre-processing, 4) preparing datasets, and 5) linking to create the final dataset. Many challenges arose throughout this process, and pragmatic solutions were co-developed with the research team and data custodian.

Results: Challenges included handling multi-supply of encrypted, complex datasets; inconsistent health data systems; limited formal support for data interpretation; incomplete and conflicting records; and misalignment between data and research questions. Solutions involved automated data management, clinically guided extraction and exclusions, and a time-based grouping method to improve linkage yield, address missing data and absent linking terms.

Conclusion: Using an exemplar, methods of data collection, pre-processing and linking have been described. A transferable multi-step process and key lessons support efficient use of health data and data driven local policy decisions for the care of patients presenting with early pregnancy complications to the ED.