Pattern of Agreement Among Medications Used During Pregnancy as Recorded in Self-Report and Administrative Claims...

Medication usage prior to conception and over the duration of pregnancy is common. Two frequent sources that record such medications are self-report and administrative databases. Our goal was to examine the pattern of agreement among medications reported in each source.

We linked all interviewed subjects recruited for the Birth Defects Study To Evaluate Pregnancy exposureS (BD-STEPS) in California to their Medi-Cal insurance claims. BD-STEPS is a population-based case-control study including livebirths, stillbirths, and terminations with a major birth defect and livebirth controls without defects. Among 195 California BD-STEPS participants that completed a structured interview and linked to their Medi-Cal claims, we compared the agreement of self-reported medication use with that documented in insurance claims using the simple kappa statistic and 95% confidence intervals.

Among the 195 subjects, accounting for both sources there were a total 128 medications recorded. Agreement between self-report and Medi-Cal claims based on the kappa statistic varied considerably; for example, we found that acetaminophen [k = 0.06 (-0.04, 0.15)], amoxicillin [k = 0.05 (-0.12, 0.21)], ibuprofen [k = 0.02 (-0.10, 0.14)], and ondansetron [k = 0.11 (-0.07, 0.28)] all had poor agreement. Five medications showed perfect agreement [k = 1.00 (1.00, 1.00)] including amlodipine, glyburide, methimazole, metoprolol, and mirtazapine. For some medications agreement varied by case and control status or time to recall, but estimates were often imprecise.

Medication reporting during pregnancy that comes from administrative claims and self-report each have their distinct advantages and limitations. Understanding how and why certain medications will populate one but not the other is an important consideration. Data drawn from multiple sources have the ability to complement one another, perhaps more so than confirm.