Using propensity score matching with doses in observational studies: An example from a child physical abuse and sleep quality study

Both physical abuse and poor sleep quality are public health concerns among adolescents, particularly in mainland China, but examining any causal effect of physical abuse on adolescents' sleep quality using a randomized controlled trial is not possible for obvious ethical reasons. Researchers have proposed the use of propensity score matching with doses to minimize overt bias and estimate the effect of multidose treatments or varying degrees of risk exposure in observational studies. In this paper, we demonstrate the propensity score methods with a focus on matching with doses in an examination of the relationship between physical abuse levels (frequency and number of perpetrators) and self-reported sleep quality among adolescents. Secondary analyses were conducted using data from the China Jintan Child Cohort. The sample comprised 707 adolescents (13.16 +/- 0.90 years old) who had complete data on physical abuse, sleep, and covariates. Propensity scores were computed from eight covariates and used to carry out pair matching, matching with the frequency of abusive experience, and matching with the number of perpetrators. The standardized differences of covariates suggested an acceptable balance between groups after matching. The results derived from matching sets consistently indicated that adolescents being physically abused by parent(s) have worse sleep quality. Despite its inherent limitations, propensity score matching with doses provides a useful tool for nurse researchers analyzing observational data.