Quantifying bias reduction with fixed-duration versus all-available covariate assessment periods

Area of Science:

• Epidemiology
• Health Services Research
• Biostatistics

Background:

• Cohort studies in healthcare databases require assessing confounders over a covariate assessment period (CAP) before cohort entry.
• The amount of available baseline time can differ between exposure groups, potentially impacting confounder measurement.

Purpose of the Study:

• To compare the bias introduced by fixed-duration versus all-available CAPs for confounder adjustment.
• To evaluate these CAP approaches under conditions of similar and discrepant available baseline time between exposure groups.

Main Methods:

• Simulated cohorts of 10,000 patients with binary confounder, exposure, and outcome variables.
• Baseline time simulated based on observed data from statin users versus nonusers.
• Compared bias from fixed-duration and all-available CAPs under varying baseline time conditions.

Main Results:

• All-available CAPs showed less bias than fixed-duration CAPs when baseline time was similar between groups.
• When baseline time differed, the preferred CAP depended on confounding direction and covariate sensitivity.
• In settings with differential baseline time, all-available CAPs could be more biased than fixed-duration CAPs.

Conclusions:

• The choice between fixed-duration and all-available CAPs is influenced by confounding and misclassification biases, which are often unknown.
• Comparing available baseline time between exposure groups is recommended.
• A fixed-duration CAP is advised when there's a large discrepancy in baseline time to mitigate potential bias from differential data availability.