Predictors of recall certainty of dates of analgesic medication use in pregnancy

Abstract

Recalled dates of medication use are difficult to validate, particularly for over-the-counter (OTC) medications. We evaluated mothers' recall certainty as an approximation of the accuracy of their recalled exposures. We used data from the Slone Epidemiology Center Birth Defects Study collected by retrospective interview of women about pregnancy medication use. For each medication reported, women were asked to report the dates of their use and how certain they were of their reported dates. For this analysis, 32,107 reports from 14,577 analgesic users were categorized as "high" or "low" recall certainty if they considered a reported date as exact or estimated, respectively. Patterns of analgesic use, maternal demographics, and health behaviors were explored as predictors of high recall certainty of dates of analgesic use. We used log-binomial regression with the generalized estimating equations extension to account for multiple reports per subject in estimating prevalence ratios (PR) and 95% confidence intervals. Relative to episodes of short-term routine use (≥4 times/week for ≤30 days), high recall certainty was more likely for episodes of a single dose (prescription PR = 1.92 (1.61, 2.27); OTC PR = 3.65 (3.03, 4.40)) and less likely for episodes of occasional use (prescription PR = 0.10 (0.05, 0.18); OTC PR = 0.65 (0.53, 0.79)). The association of chronic routine use with high recall certainty was inverse among prescription analgesics and positive among OTC analgesics (prescription PR = 0.56 (0.40, 0.80); OTC PR = 2.46 (1.92, 3.42)). Some characteristics that were associated with recall accuracy in previous studies were also associated with recall certainty in this study.