Outcome Measures (Dependent Variable)

The outcome of interest was a dichotomous smoking cessation medications use variable. Any use of first-line FDA approved smoking cessation medications including bupropion, varenicline, and nicotine replacement therapies were defined as the use of smoking cessation medication, this was identified from the prescribed medicine file.

Independent Variables

Predisposing characteristics included race/ethnicity, age, gender, marital status, and education. Age was categorized into three groups: 18-39 years, 40-59 years, and 60-85 years. Marital status was categorized as married or unmarried (unmarried included widowed, divorced, separated, never married). Education was categorized into three categories: < 12 years, 12-15 years, and >15 years. Race/ethnicity was categorized into four mutually exclusive categories: Non-Hispanic Whites, non-Hispanic Blacks, Hispanic, and others (including American Indian/Alaska Native, Asians, and Native Hawaiian/Pacific Islander).

Need characteristics included obesity, activities of daily living (ADLs), instrumental ADLs (IADLs), and general health status and Charlson comorbidity index. Obesity was categorized as normal weight, overweight, obese, and extreme obese based on body mass index value: normal weight: body mass index (BMI) <25; overweight: 25≤BMI<30; obese: 30≤BMI≤40, extreme obese: BMI>40. Presence of comorbidities and patients’ disease severity may affect the extent of health services and medication use. We calculated Charlson comorbidity index score and categorized it to reflect a low (score of 0-1), medium (2-3) or high (≥4) burden of comorbidity as previous studies. ADLs/IADLs were measured as dichotomous variables based on whether or not the individual needed help to perform any ADL/IADL. Self- perceived general health status was classified into two categories: Fail/poor and excellent/very good/good health status.

Statistical Analyses (Regression Model)

All statistical analyses were performed adjusting for MEPS survey design to generalize results at national level. Univariate logistic regression models were used to estimate the unadjusted association of each independent variable with smoking cessation medication use. Colinearity of all independent variables was tested using multicolinearity test and variables were removed if correlation coefficient analysis of the variables had a value of >0.7 and if the variance inflation factor was >10. Interaction assessment was carried out using the chuck test to assess the interaction between predictors, no interaction was identified. Therefore, no interaction term was included in the multivariate models. Multivariate logistic regression was performed to examine racial/ethnic disparity in smoking cessation medication utilization controlling for predisposing, enabling and need factors. All the independent variables with P < 0.20 in the univariate analysis were kept and included in the final multivariate regression model. Maldonado and Greenland (1993) suggest that potential confounders be eliminated only if P > 0.20, in order to protect against residual confounding.

A priori significance level at P < 0.05 was used in the final multivariate regression model. OR, and 95% confidence interval were reported. SAS 9.3 was used to carry out all the statistical analyses.