UC San Diego

Can an intervention program that is highly effective in reducing the prevalence of an unhealthy behavior in the general population also reduce the disparity among its subgroups? That depends on what measure of disparity is used. Using simple algebraic models, this study demonstrates that disparity measured in terms of relative difference between two groups tends to increase when the prevalence of the behavior is in decline. The study then shows an empirical example, by analyzing the effects of the California tobacco control program on smoking prevalence of two education groups, the lowest (less than 12 years) and the highest (16 years or more). It examines the data from four California Tobacco Surveys covering the years 1996, 1999, 2002, and 2005. The effects of three components of the tobacco control program known to be effective in decreasing prevalence (media, worksite policy, and price) on the two education groups are assessed. The smoking prevalence for the two groups is obtained from these four surveys and a regression line is computed for each education group from 1996 to 2005. Results show that the California program is effective with both low education and high education groups and that the rate of decline in smoking prevalence from 1996 to 2005 is no smaller for the low education group than for the high education group. The paper then discusses that an analysis of disparity based on relative difference, however, could result in misleading recommendations that an intervention like the California tobacco program needs to change from its current whole-population approach to one that focuses on targeting subgroups because it has not reduced disparity. It proposes that research should focus more on increasing the rate of change among less advantage groups and less on the relative disparity compared to some other group.