Potential Misclassification and Bias Due to Low Birth Weight Category in the Population Level

Hyojun Park, University of Wisconsin-Madison

A considerable misclassification and bias may be introduced in causal inference when low birth weight (LBW) category is used since LBW is a mixture of effects from the duration of the child’s gestation and intrauterine growth. Small for gestational age (SGA) that captures only intrauterine growth is known as a better measure. Using the Early Childhood Longitudinal Study-Birth Cohort, this study identifies the degree of misclassification of LBW model and determines the potential bias in various models that LBW was used as a dependent, independent, or mediating variable. The results show that the magnitude of misclassification was considerable and differed by preterm and term birth. In addition, the magnitude and direction of bias greatly varied in the LBW models as a dependent, independent, or mediating variable. The results strongly suggest that LBW may introduce bias in the causal inference, so that SGA category should be utilized whenever possible.