Applying Publicly Available Contextual Factors to Predict Smoking Relapse in a National Sample

Authors

  • Ross Kauffman Center for Interdisciplinary Studies, Ohio Northern University
  • Jeffrey S. Wilson Department of Geography, Indiana University
  • Timothy E. Stump School of Medicine, Indiana University
  • Patrick O. Monahan
  • Anna M. McDaniel College of Nursing, University of Florida

DOI:

https://doi.org/10.18061/ojph.v2i1.9040

Keywords:

Tobacco cessation, multilevel modeling, contextual factors

Abstract

Background: The ecological fallacy is broadly understood, though its complimentary problem, the individualistic or atomistic fallacy, is
less often considered. Multilevel models offer the statistical tools needed to avoid both errors by allowing simultaneous consideration
of individual, contextual, and policy factors. This study applies such methods to smoking cessation data. Tobacco control is of particular
concern in Ohio where the adult smoking prevalence remains around 22%.
Methods: Data from the 1,785 participants in the Technology Enhanced Quitline Study were used to test the theory that contextual
factors impact relapse rates and program effectiveness, employing a mixed-effects model to account for the nested nature of the data
while testing for the relationship between contextual factors and relapse, controlling for individual characteristics.
Results: No contextual factors or policy variables were significant predictors of smoking relapse in the sample, nor were any associated with the success of the intervention.
Conclusions: While this work could not identify specific influences of contextual and policy factors on smoking outcomes in our sample, it demonstrates the feasibility of adding such predictors to future clinical trials. This project clearly does not rule out the possibility that contextual and policy factors may influence smoking even after controlling for individual characteristics, but does not provide strong evidence of such a link. It is possible that these negative findings may be due to geocoded mailing addresses being a poor proxy for relevant contextual factors, use of the wrong geographic unit of analysis (modifiable areal unit problem), or a lack of temporal resolution in contextual variables.

Downloads

Additional Files

Published

2019-06-01

Issue

Section

Research Briefs