Information Bias & Data Quality

Summary

Develops the framework of differential versus non-differential misclassification, using Blair's pesticide biomarker validation to show how self-report attenuates associations toward the null and INTERPHONE's biologically nonsensical phone-side glioma results to illustrate recall bias. Covers social desirability through Midanik's alcohol-sales comparisons and Sjoding's pulse-oximetry calibration failure that left Black patients with occult hypoxemia three times more often than White patients during COVID. Argues that information bias is not just a methodological annoyance but an ethical and political problem in cause-of-death coding, Indigenous identity misclassification, gender-minority erasure, and the underrepresentation of non-European ancestries in genome-wide association studies, then closes with observer, detection, and surveillance biases through PSA screening and hormone replacement therapy.