Understanding data provenance when using electronic medical records for research: Lessons learned from the Deliver...

Area of Science:

• Health Informatics
• Primary Care Research
• Data Quality Management

Background:

• Choosing Wisely Canada recommendations aim to reduce unnecessary healthcare interventions.
• Primary care settings are key areas for implementing evidence-based practice changes.
• Evaluating the impact of guidelines requires robust data analysis.

Purpose of the Study:

• To assess the impact of Choosing Wisely Canada recommendations on reducing unnecessary health investigations and interventions in primary care.
• To investigate the quality and provenance of electronic medical record (EMR) data for longitudinal analysis.

Main Methods:

• Utilized the Deliver Primary Healthcare Information (DELPHI) database, containing deidentified EMR data from nearly 65,000 patients.
• Examined 10 years of data (2009-2019) for quality issues, focusing on data provenance and comparability after an EMR software change in 2012-2013.
• Employed probabilistic linkage to attempt record linking between different EMR software systems.

Main Results:

• A significant change in EMR software between 2012 and 2013 led to data quality issues.
• Observed challenges included limited record linkage, distorted procedure dates due to data migration, and unusual fluctuations in laboratory test and medication volumes.
• These issues impacted the comparability and reliability of longitudinal data analysis.

Conclusions:

• This study underscores the critical importance of assessing data provenance and quality before initiating research projects.
• Understanding data provenance is essential for anticipating and mitigating data quality issues, particularly in longitudinal EMR data analysis.
• Proactive data quality assessment is vital as longitudinal data analyses become more prevalent in healthcare research.