Completeness of ascertainment of prenatal smoking using birth certificatesand confidential questionnaires - Variations by maternal attributes and infant birth weight

Birth certificate data frequently are used to monitor the prevalence of smo king during pregnancy. The authors used a two-sample capture-recapture meth od to estimate the completeness of ascertainment of prenatal smoking on bir th certificates and on confidential questionnaires in six US states. Comple teness of ascertainment was also examined according to maternal attributes and infant birth weight. The samples included white women who delivered a l ive infant between 1993 and 1995 in one of six states (Alabama, Alaska, Geo rgia, Maine, South Carolina; or West Virginia) and who responded to a quest ionnaire mailed to them 2-6 months postpartum as part of the Pregnancy Risk Assessment Monitoring System. State-specific sample sizes ranged from 2,64 7 to 4,795, The completeness of ascertainment ranged from 70.6% to 82.0% us ing birth certificates and from 86.2% to 90.3% using confidential questionn aires. In all six states, the birth certificates' completeness of ascertain ment varied by maternal education and infant birth weight, and the question naires' completeness varied by maternal age. Both birth certificates and qu estionnaires underestimated the true extent of smoking during pregnancy amo ng these white women. Differential reporting by birth weights recorded on b irth certificates would result in an overestimated association between tow birth weight and prenatal smoking.