TReND: Measurement and Smoking Patterns in National Data

Visit TReND’s web portal to learn more about research, programs, policies, and resources relevant to tobacco and health disparities.

Return to TReND Projects

About the Project

Rationale: The latent class model is a statistical model for the identification of subgroups in a population. In the context of cigarette smoking, these latent classes could correspond to different patterns of smoking or types of smokers. These different smoking patterns may not only provide an important description of smoking within priority populations, but they may also be important in understanding other smoking related behavior, for example identity and perceived need for treatment or cessation.

Purpose: The purpose of this study is to investigate and identify smoking patterns across several priority populations by conducting latent class analysis on data from the National Survey of Drug Use and Health (NSDUH), 2003—2005. Latent class analysis will be run separately for groups defined by race/ethnicity, gender and age. Results from this study will be a significant improvement beyond simply saying that members of a particular group smoke less than some other group. Rather, smoking history and current smoking information will be used to characterize common patterns of use (e.g., heavy smoking versus light and intermittent smoking) within priority populations. Small sample sizes will be addressed by pooling respondents in the same demographic groups across adjacent years (e.g., all non-Hispanic Black women between 18 and 25 from 2003 through 2005 surveys) and determining the appropriateness of this pooling strategy by assessing trends in smoking patterns for the parallel White group and identifying any broader, societal-level changes in smoking patterns across the years. A further approach to addressing small sample sizes in these models is to reduce the size of the model by using appropriate parameter restrictions. Once the classes are identified for each demographic group, we will also examine relations between class membership and other variables such as education, age of first cigarette, type of cigarettes smoked, and items designed to measure nicotine dependence.

Impact: Results from this study will increase our understanding of the various smoking patterns among priority populations, and thus, will help inform the development and implementation of more appropriately tailored prevention and cessation programs rather than applying a "one size fits all" approach. TReND’s investigators will disseminate the study findings through peer-reviewed journal publications and presentations at major professional conferences.

Research Team

Brian Flaherty, Ph.D. (Principal Investigator)
University of Washington

Cara J. Kiff, MA
University of Washington

Dennis R. Trinidad, Ph.D., M.P.H.
Claremont Graduate University

Donna Vallone, Ph.D., M.P.H.
American Legacy Foundation

K. Vish Viswanath, Ph.D.
Harvard School of Public Health
Dana-Farber Cancer Institute

Contact Us

Principal Investigator
Brian Flaherty, Ph.D.Exit Disclaimer
Assistant Professor
Department of Psychology
University of Washington

Project Activities and Findings


Not yet available.


Flaherty BP (September 2009). Cigarette smoking patterns and covariates in U.S. National data within three racially classified social groups. Legacy, Washington DC.

Flaherty BP (July 2010). Identifiability and model selection in a mixed indicator latent class model. The 2010 International Meeting of the Psychometric Society, Athens, GA.

Flaherty BP (July 2009). Model selection in latent class and mixture models: What’s to be preferred? The16th International Meeting of the Psychometric Society, Cambridge, UK.

Flaherty BP (April 2010). Model selection in latent class and mixture models: What is to be preferred? The 31st Annual Meeting of the Society of Behavioral Medicine, Seattle, WA.

Flaherty B P (February 2008). Smoking patterns among a sample of U.S. adult recent smokers. The 14th Annual Meeting of the Society for Research on Nicotine and Tobacco, Portland, OR.

Flaherty B P (2008). Smoking patterns in U.S. national data. Invited presentation to the Center for Statistics and the Social Sciences seminar series, University of Washington, Seattle, WA.

Flaherty BP, Kiff C (February 2010). African American and Hispanic adult smoking patterns, U.S. 2004. The 16th Annual Meeting of the Society for Research on Nicotine and Tobacco, Baltimore, MD.

Kiff C, Flaherty BP (August 2010). Inference challenges in latent class models. The118th Annual Meeting of the American Psychological Association, San Diego, CA.

Kiff CJ, Flaherty BP (June 2009). Variation in Smoking Patterns among African American and Hispanic U.S. Adults. Poster presentation at the Center for Statistics and the Social Sciences' Tenth Anniversary Conference, University of Washington, Seattle, WA.

Other Resources and Publications

Data and Instruments

National Survey of Drug Use and Health (NSDUH), 2003—2005: Methodology Reports and Questionnaires Exit Disclaimer

Select Scientific Publications and Articles

Flaherty BP (2002). Assessing the reliability of categorical substance use items with latent class analysis. Drug Alcohol Depend, 68S, 7–20.

The National Cancer Institute (NCI) and American Legacy Foundation are proud to fund the Tobacco Research Network on Disparities (TReND). Previous support has also been provided by the Department of Health and Human Services (DHHS) Office on Women’s Health, NCI Office of Women’s Health, and the NCI Center to Reduce Cancer Health Disparities.