Propensity Scores: Method for Matching on Multiple Variables in Down Syndrome Research

Confounding variables can affect the results from studies of children with Down syndrome and their families. Traditional methods for addressing confounders are often limited, providing control for only a few confounding variables. This study introduces propensity score matching to control for multiple confounding variables. Using Tennessee birth data as an example, newborns with Down syndrome were compared with a group of typically developing infants on birthweight. Three approaches to matching on confounders--non-matched, covariate matched, and propensity matched--were compared using 8 potential confounders. Fewer than half of the newborns with Down syndrome were matched using covariate matching, and the matched group was differed from the unmatched newborns. Using propensity scores, 100% of newborns with Down syndrome could be matched to a group of comparison newborns, a decreased effect size was found on newborn birthweight, and group differences were not statistically significant.