Relative Reduction in Prevalence (RRP): An Alternative to Cohen's Effect Size Statistics for Judging Alcohol, Cigarette, and Marijuana Use Prevention Outcomes.

Abstract

Jacob Cohen developed two statistical measures for judging the magnitude of effects produced by an intervention, known as Cohen's d, appropriate for assessing scaled data, and Cohen's h, appropriate for assessing proportions. These have been widely employed in evaluating the effectiveness of alcohol, cigarette, marijuana, and other drug prevention efforts. I present two tests to consider the adequacy of using these statistics when applied to drug use prevention programs. I used student survey data from grades 6 through 12 (N = 1,963,964) collected by the Georgia Department of Education between 2015 and 2017 and aggregated at the school level (N = 1036). I calculated effect sizes for an imaginary drug prevention program that (1) reduced 30-day alcohol, cigarette, and marijuana prevalence by 50%; and (2) maintained 30-day prevalence at a pretest level for multiple years. While both approaches to estimating intervention effects represent ideal outcomes for prevention that surpass what is normally observed, Cohen's statistics failed to reflect the effectiveness of these approaches. I recommend including an alternative method for calculating effect size for judging program outcomes. This alternative method, Relative Reduction in Prevalence (RRP), calculates ratio differences between treatment and control group drug use prevalence at posttest and follow-up, adjusting for differences observed at pretest. RRP allows researchers to state the degree to which an intervention could be viewed as efficacious or effective that can be readily understood by practitioners.