Growth Models of Maternal Smoking Behavior: Individual and Contextual Factors

Abstract

Background: Persistent maternal smoking during pregnancy, reduction or cessation during pregnancy, and smoking initiation or resumption postpartum impel further research to understand these behavioral patterns and opportunities for intervention. Objectives: We investigated heterogenous longitudinal patterns of smoking quantity to determine if these patterns vary across three maternal age groups, and whether the influence of individual and contextual risk factors varies by maternal age. Methods: Separate general growth mixture models were estimated for mothers ages 15–25, 26–35, and 36+, allowing different empirical patterns of an ordinal measure of smoking behavior at six time points, from preconception through child entry to kindergarten. Results: We identify five classes for mothers ages 15–25, four classes for ages 26–35, and three classes for ages 36+. Each age group presents classes of nonsmokers and persistent heavy smokers. Intermediate to these ends of the spectrum, each age group exhibited its own smoking classes characterized by the extent of pregnancy smoking reductions and postpartum behavior. In all three age groups, class membership can be distinguished by individual sociodemographic and behavioral characteristics. Co-resident smokers predicted nearly all smoking classifications across age groups, and selected neighborhood characteristics predicted classification of younger (15–25) and older (36+) mothers. Conclusions: The design, timing, and delivery of smoking prevention and cessation services, for women seeking to become pregnant and for women presenting for prenatal or pediatric care, are best guided by individual characteristics, particularly maternal age, preconception alcohol consumption, and postpartum depression, but neighborhood characteristics merit further attention for mothers at different ages.

Keywords:

• smoking
• pregnancy
• longitudinal trajectories
• neighborhood context
• general growth
• mixture model
• ordinal outcomes

THE AUTHORS

Dr. Mumford, Ph.D., M.H.S., B.A., is a social epidemiologist at NORC at the University of Chicago, a social science research organization that collects and analyzes data on key social issues. Dr. Mumford's research includes studies of tobacco epidemiology, tobacco control policies, and the relevant social ecology. Dr. Mumford also maintains an active research portfolio on the epidemiology and prevention of aggressive behavior in dating relationships.

Dr. Liu, Ph.D., M.A. B.A., is a Criminologist and Prevention Scientist interested in understanding the developmental vulnerability of risky behaviors in early stages of life, including maternal substance use and its impact on maternal and child physical, mental, and behavioral health. Dr. Liu is specialized in study design and advanced statistical methods with extensive experience in analyzing large longitudinal datasets. Prior to joining NORC, Liu completed an NIH funded postdoctoral fellowship in Prevention Science at Johns Hopkins University.

GLOSSARY

• Census tract: Geographic subdivisions of a county, usually coinciding with city or town limits, relatively homogenous with respect to population characteristics, economic status, and living conditions.

• General growth mixture models: A latent variable modeling technique that uses a categorical latent class variable in combination with continuous growth factors to explore population heterogeneity in the change process of the outcome of interest.