Assessing the Effect of Electronic Health Record Data Quality on Identifying Patients With Type 2 Diabetes:...

Area of Science:

• Health Informatics
• Clinical Data Science
• Chronic Disease Management

Background:

• Electronic health records (EHRs) are increasingly vital for healthcare data, necessitating robust data quality assessments.
• Accurate identification of patient populations, such as those with type 2 diabetes (T2D), is crucial for clinical research and operational decisions.
• Existing computable phenotype definitions for chronic conditions may be sensitive to EHR data quality variations.

Purpose of the Study:

• To evaluate the impact of EHR data quality issues on the performance of computable phenotypes for identifying type 2 diabetes (T2D) denominator populations.
• To assess how variations and robustness of phenotypes influence patient cohort identification.

Main Methods:

• Retrospective analysis of EHR data from approximately 208,000 patients with T2D at Johns Hopkins Medical Institution (2017-2019).
• Inclusion of 4 published phenotypes and 1 expert-defined phenotype for T2D.
• Simulation of data quality issues (incompleteness, inaccuracy, timeliness) across diagnosis, medication, and laboratory data types.

Main Results:

• Low population overlap (<23%) observed across different phenotypes.
• Simulated data incompleteness, inaccuracy, and timeliness issues significantly reduced the number of identified patients for each phenotype.
• Phenotype performance varied considerably under induced data quality degradation.

Conclusions:

• EHR data quality significantly affects the ability to reliably identify T2D patient cohorts using computable phenotypes.
• Findings highlight the need for common, robust T2D phenotype definitions to ensure consistency in clinical informatics and healthcare management.
• This study informs future efforts to standardize computable phenotypes for improved clinical research and operational decision-making.