Notwithstanding that this difference [remained] even after adjusting for factors such as social connections and health behaviors, including smoking and exercising, that is, the usual negative correlation between religiosity and high-risk behavior was accounted for, 21% is nowhere near enough to be meaningful.