Applying Person-Centered Analyses to Identify Patterns in Smoking-Related Disparities: 2014 Kansas Behavioral Risk Factor Surveillance System

BACKGROUND:  Addressing tobacco related disparities remains a key goal for tobacco control programs and is crucial to achieving health equity. Traditionally the prevalence of tobacco use has been reported for specific demographic subpopulations. However, this approach treats characteristics that define the disparate subpopulations in isolation and ignores any confluence that may exist between them. Person-centered analyses, such as Configural Frequency Analysis (CFA), allow for a more holistic view by examining the co-occurrence of characteristics to identify typologies or groups of individuals that share the same attributes.

METHODS:  Data from the 2014 Kansas Behavioral Risk Factor Surveillance System (KS-BRFSS) was used to examine disparities in current smoking (i.e. having smoked at least 100 cigarettes and currently smoking every day or some days) by sex, race/ethnicity, education, annual household income, health insurance status, disability status, mental health status, and sexual orientation. Initial analyses used chi-squared tests to identify significant disparities in the weighted prevalence of current smoking for each sub-population. An index was then created that summed the number of disparity risk factors for each individual to examine the overall co-occurrence of disparate characteristics. The R package “confreq” was used to conduct CFA to identify “types” (patterns of disparate characteristics observed more frequently than expected by chance) and “anti-types” (patterns observed less frequently than expected by chance).

RESULTS:  Initial analyses identified significant disparities in the prevalence of current smoking within each of the socio-demographic characteristics examined. In addition, the disparity index indicated a substantial co-occurrence of disparity characteristics among current smokers: more than 40% scored four or higher on the index compared to only 18.2% of non-smokers. Results from the CFA identified a substantial number of types (n=29) and one anti-type however only five types met standard BRFSS suppression criteria for estimation of prevalence (i.e. numerator ≥5, denominator≥50 and RSE≤30). Among these types, women age 35 years and older who have poor mental health status, no health insurance, are living with a disability and have an annual household income of less than $25,000 had the highest prevalence of smoking (≥42.0%). The prevalence of current smoking was 11.4% or lower for the remaining three types.

CONCLUSIONS:  The characteristics that identify smoking related disparities do not occur in isolation and the use of person-oriented analyses to examine constellations of disparity-related characteristics may advance how programs can target their efforts to reach those in the population that are most impacted by tobacco use.