Data Visualization Tools of Tobacco Product Use Patterns, Transitions and Sex Differences in the PATH Youth Data.

Abstract

Evaluations of multiple tobacco product use and temporal changes in patterns of use are complicated by a large number of combinations and transitions. Visualization tools could easily identify most common patterns and transitions. Set intersection bar plots describe ever use of five tobacco products among 12-17 years old youth in wave 1 of the Population Assessment of Tobacco and Health (PATH) study (N = 11 497). Heat maps visualize unweighted frequencies of transitions from ever use at wave 1 (2013-2014) to past 12-month use at wave 2 (2014-2015). Weighted calibrated heat maps assess differences in relative frequencies of transitions by pattern at wave 1 and identify differences in transitions by sex. The most common tobacco product ever use patterns in wave 1 were of cigarettes only, e-cigarettes only or hookah only, followed by ever use of both cigarettes and e-cigarettes. Initiation of use between waves was uncommon. The most frequent transition among those who reported use at wave 2 but not at wave 1 (N = 971) was to e-cigarette use (N = 301). However, among e-cigarette-only ever users at wave 1 (N = 260), about half did not report any product use at wave 2. Use of three or more products remained stable. Adolescent girls compared to boys appeared more likely to report hookah use at both waves. Set intersection bar plots and heat maps are useful for visualizing tobacco product use patterns and transitions, especially for multiple products. Both techniques could identify common problematic tobacco use patterns across and within populations. Given the growing complexity of the youth tobacco use landscape, approaches to efficiently communicate patterns of multiple tobacco product use should be used more often. This study introduces set intersection bar plots and modified versions of heat maps to the tobacco product literature and illustrates their use in the PATH youth sample. These techniques are useful for visualizing absolute and relative frequencies of multiple possible patterns and transitions. They also suggest targets for subsequent statistical inference such as sex differences in hookah use. The methods can be applied more generally for data visualization wherever large number of combinations occurs.