Analyses were stratified by race, as pooled analyses, with or without interaction terms, are statistically compromised given the substantial imbalance in socioeconomic status between AAs and EAs in our sample which could impact the cogency of risk scores (see Imbens and Rubin, 2015).
