A social-contextual investigation of smoking among rural women: multi-level factors associated with smoking status and considerations for cessation

Introduction: The social-contextual model of tobacco control and the potential mechanisms of the maintenance or cessation of smoking behavior among disadvantaged women, including rural residents, have yet to be comprehensively studied. The purpose of this study was to determine the association between selected individual, interpersonal, workplace, and neighborhood characteristics and smoking status among women in Appalachia, a US region whose residents experience a disproportionate prevalence of tobacco-related health disparities. These findings may assist in efforts to design and test scientifically valid tobacco control interventions for this and other disadvantaged populations.

Methods: Women, 18 years of age and older, residing in three rural Ohio Appalachian counties, were recruited using a two-phase address-based sampling methodology for a cross-sectional interview-administered survey between August 2012 and October 2013 (N=408). Multinomial logistic regression was employed to determine associations between select multilevel factors (independent variables) and smoking status (dependent variable). The sample included 82 (20.1%) current smokers, 92 (22.5%) former smokers, and 234 (57.4%) women reporting never smoking (mean age 51.7 years).

Results: In the final multivariable multinomial logistic regression model, controlling for all other significant associations, constructs at multiple social-contextual levels were associated with current versus either former or never smoking. At the individual level, for every additional year in age, the odds of being a former or never smoker increased by 7% and 6% (odds ratio (OR) (95% confidence interval(CI)): 1.07 (1.0-1.11) and 1.06 (1.02-1.09)), respectively, as compared to the odds of being a current smoker. With regard to depression, for each one unit increase in the Center for Epidemiologic Studies Depression Scale score, the odds of being a former or never smoker were 5% and 7% lower (OR(95%CI): 0.95(0.91-0.999) and 0.93(0.88-0.98)), respectively. Five interpersonal factors were associated with smoking status. As the social influence injunctive norm score increased by one unit, indicating perception of smoking to be more acceptable, the odds of being a former or never smoker decreased by 23% and 30%, respectively. For every one unit increase in the social participation score, indicating past-year engagement in one additional activity type, the odds of being a former or never smoker increased by 17% and 36%, respectively. For every 10% increase in the percentage of social ties in the participant's advice network who smoked, the odds of being a former or never smoker were 24% and 28% less, respectively. For every 0.1 unit increase in the E/I index, indicating increasing homophily on smoking in one's social network, the odds of being a former or never smoker were 20% and 24% less, respectively, in the time network, and 18% and 20% less, respectively, in the advice network. At the neighborhood level, for every one unit increase in neighborhood cohesion score, indicating increasing cohesion, the odds of being a former smoker or never smoker were 12% and 14% less, respectively.

Conclusions: These findings indicate that a social-contextual approach to tobacco control may be useful for narrowing a widening trajectory of smoking disparity for rural women. Interpersonal context, in particular, must be considered in the development of culturally targeted cessation interventions for Ohio Appalachian women.