Exposure measurement error when assessing current glucocorticoid use using UK primary care electronic prescription...

Area of Science:

• Pharmacovigilance
• Real-world data analysis
• Rheumatology research

Background:

• Accurate assessment of medication exposure is crucial for pharmacovigilance and clinical research.
• Electronic health records, such as UK primary care prescription data, are valuable but may contain misclassification errors.
• Understanding the extent of misclassification is vital for reliable interpretation of study findings.

Purpose of the Study:

• To quantify misclassification in glucocorticoid (GC) exposure using UK primary care prescription data.
• To compare participant-reported GC use with electronic prescription records.
• To assess the impact of misclassification on observed effect sizes.

Main Methods:

• Cross-sectional study of rheumatoid arthritis patients prescribed oral GCs.
• Comparison of self-reported GC use (paper diary) against electronic prescription data (CPRD).
• Generation of a hypothetical population dataset to simulate misclassification bias.

Main Results:

• 86% of participants were correctly classified for current oral GC use.
• Imprecise estimation of current GC dose (correlation coefficient 0.46).
• Poor concordance for GC injections (kappa statistic 0.14); significant bias observed in simulated data.

Conclusions:

• Low misclassification of current oral GC use can lead to substantial bias in research.
• Researchers must carefully evaluate potential exposure misclassification in their analyses.
• Findings highlight the need for cautious interpretation of prescription data in epidemiological studies.