Birth certificate data, collected for every birth in the United States, are frequently used for public health research. Such data have been used to track progress toward the Healthy People 2020 maternal and child health goals and to analyze socioeconomic disparities in health outcomes. A new Standard Certificate of Live Birth was issued in 2003, and all states had adopted the new form by 2015. 1 This brief presents data on the reliability of some data collected using the new birth certificate form, particularly variables critical for studying disparities in maternal and child health outcomes for women with Medicaid coverage at delivery. We evaluate six variables; three variables were new additions on the 2003 birth certificate (Medicaid enrollment at birth, prepregnancy body mass index, and breastfeeding initiation), and the other three variables have shown to be unreliable but are of significant policy research interest (prepregnancy smoking, diabetes, and hypertension). Until now, it has been impossible to evaluate the reliability of these variables for women2 with Medicaid across states. Data from three federal evaluations conducted from 2012 to 2016 provide a unique opportunity to evaluate the agreement between individual-level program evaluation data and data from birth certificates for infants born to Medicaid-enrolled women. The evaluations are here called MIHOPE, MIHOPE-Strong Start, and Strong Start. More information on these evaluations is available from three evaluation reports (Duggan et al. 2018; Lee et al. 2019; Hill et al. 2018). Each study targeted women with Medicaid, and each collected data for all six variables of interest. Because methods for the MIHOPE and MIHOPE-Strong Start evaluations were similar, we group data from those two evaluations for this analysis, here called MIHOPE/MIHOPE-Strong Start. All three studies validated information on Medicaid enrollment by checking states’ Medicaid enrollment records and collected data on the other five variables from either medical records or mothers’ self-reports. The three evaluations linked mothers’ evaluation data to birth certificates, allowing us to assess agreement between birth certificate and program evaluation data. Agreement between the two data sources increases confidence in the accuracy of the data, and disagreements reduce such confidence. However, neither data source constitutes a “gold standard,” so low agreement levels between birth certificates and evaluation data do not necessarily mean birth certificate data are inaccurate. Human errors, including those related to recall, omission, and misreporting, apply to the data sources examined. Data collection for the three evaluations, however, may have been more accurate than data from birth certificates for several reasons: evaluation data collection processes were more standardized, some risk factors (e.g., smoking) were less subject to recall bias in the evaluation data because they were collected during pregnancy, and evaluation data relied on directly accessed prenatal medical records for some medical and pregnancy risk factor variables.
Copyright:
Reproduced with permission of the copyright holder. Further use of the material is subject to CC BY-NC-DC license. (More information)