Assessment of potential bias mechanisms in the analysis of large electronic medical record datasets
Year: 2026
Type: article
Abstract: Electronic medical records aid general practitioners in the follow-up of their patients. The vast amounts of patient data that are being collected through such electronic medical records offer opportunities to use them for other purposes such as epidemiological and (pre)clinical research. But the use of electronic medical record data for research purposes raises concerns about potential biases, since these data are typically not collected within a study design specifically developed to address a predefined research question, which may instead emerge during or after data collection. In addition, data collection practices and reporting behaviours can vary between general practitioners, while patients themselves differ in their likelihood of seeking care or reporting symptoms. Although a large patient population may mitigate some of these random variations through the law of large numbers, systematic biases related to the non-randomized sample of the participating practices and heterogeneity in patient characteristics may persist. If not appropriately accounted for, these sources of heterogeneity and potential bias can lead to different or potentially conflicting research findings. Using a large electronic medical records dataset from Flanders, Belgium, we assess how simple statistical approaches, such as adding practice-level random effects and changing the analytical entity, can help identify potential bias mechanisms and assess the sensitivity of estimated associations to them.
Source: Scientific Reports
Institutions: KU Leuven, Vrije Universiteit Brussel, Hasselt University
Language: English
Subfield: Statistics and Probability
Field: Mathematics
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,690
OpenAlex ID: W7233609854
