TY - JOUR
T1 - Who underreports smoking on birth records
T2 - A monte carlo predictive model with validation
AU - Land, Thomas G.
AU - Landau, Anna S.
AU - Manning, Susan E.
AU - Purtill, Jane K.
AU - Pickett, Kate
AU - Wakschlag, Lauren
AU - Dukic, Vanja M.
PY - 2012/4/24
Y1 - 2012/4/24
N2 - Background: Research has shown that self-reports of smoking during pregnancy may underestimate true prevalence. However, little is known about which populations have higher rates of underreporting. Availability of more accurate measures of smoking during pregnancy could greatly enhance the usefulness of existing studies on the effects of maternal smoking offspring, especially in those populations where underreporting may lead to underestimation of the impact of smoking during pregnancy. Methods and Findings: In this paper, we develop a statistical Monte Carlo model to estimate patterns of underreporting of smoking during pregnancy, and apply it to analyze the smoking self-report data from birth certificates in the state of Massachusetts. Our results illustrate non-uniform patterns of underreporting of smoking during pregnancy among different populations. Estimates of likely underreporting of smoking during pregnancy were highest among mothers who were college-educated, married, aged 30 years or older, employed full-time, and planning to breastfeed. The model's findings are validated and compared to an existing underreporting adjustment approach in the Maternal and Infant Smoking Study of East Boston (MISSEB). Conclusions: The validation results show that when biological assays are not available, the Monte Carlo method proposed can provide a more accurate estimate of the smoking status during pregnancy than self-reports alone. Such methods hold promise for providing a better assessment of the impact of smoking during pregnancy.
AB - Background: Research has shown that self-reports of smoking during pregnancy may underestimate true prevalence. However, little is known about which populations have higher rates of underreporting. Availability of more accurate measures of smoking during pregnancy could greatly enhance the usefulness of existing studies on the effects of maternal smoking offspring, especially in those populations where underreporting may lead to underestimation of the impact of smoking during pregnancy. Methods and Findings: In this paper, we develop a statistical Monte Carlo model to estimate patterns of underreporting of smoking during pregnancy, and apply it to analyze the smoking self-report data from birth certificates in the state of Massachusetts. Our results illustrate non-uniform patterns of underreporting of smoking during pregnancy among different populations. Estimates of likely underreporting of smoking during pregnancy were highest among mothers who were college-educated, married, aged 30 years or older, employed full-time, and planning to breastfeed. The model's findings are validated and compared to an existing underreporting adjustment approach in the Maternal and Infant Smoking Study of East Boston (MISSEB). Conclusions: The validation results show that when biological assays are not available, the Monte Carlo method proposed can provide a more accurate estimate of the smoking status during pregnancy than self-reports alone. Such methods hold promise for providing a better assessment of the impact of smoking during pregnancy.
UR - http://www.scopus.com/inward/record.url?scp=84860127095&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84860127095&partnerID=8YFLogxK
U2 - 10.1371/journal.pone.0034853
DO - 10.1371/journal.pone.0034853
M3 - Article
C2 - 22545091
AN - SCOPUS:84860127095
SN - 1932-6203
VL - 7
JO - PloS one
JF - PloS one
IS - 4
M1 - e34853
ER -