We omitted interaction testing in our analyses because predictors and key-risk behaviors were correlated, thereby limiting our understanding of any effect modification that may have occurred among key variables. Additionally, we assumed that missing data were random, and performed longitudinal analyses with a population-average model. We acknowledge that these missing data may have led to underestimating the effect for some key predictors.