Background
Asthma and chronic obstructive pulmonary disease (COPD) are prominent chronic respiratory conditions that incur significant health and financial costs in the United States [
1,
2]. Both diseases entail airflow obstruction and airway inflammation that can progressively worsen and require long-term clinical management [
3,
4]. Past surveillance and research efforts have shown the American Indian/Alaska Native (AI/AN) population to suffer a greater prevalence of these conditions [
5‐
7]. The health disparities experienced by AI/AN peoples in various chronic diseases including respiratory diseases was shown to be associated with indicators of low socioeconomic status such as poverty [
8,
9] and health risk behaviors such as tobacco use [
10,
11]. These disease correlates are typically included and hence tracked in population-based surveys such as the Behavioral Risk Factor Surveillance System (BRFSS).
Yet additional correlates of chronic respiratory disease exist which may contribute to disease disparity observed in the AI/AN population. In particular, AI/AN children sustain higher rates of respiratory syncytial virus (RSV) infection [
12] which is associated with development and/or exacerbation of asthma [
13]. Further, persistent exposure to poor indoor air quality is also increasingly recognized to be associated with adverse respiratory health [
14‐
16]. Lastly, AI/AN individuals were more likely to be exposed to occupational inhalants which adversely impact respiratory health [
17‐
19]. These variables relate to the etiologies of asthma and COPD and are not included in the BRFSS. Consequently their potential contribution to explain the disease disparity is not known. In this study, we address the contributing influence of race/ethnicity (being AI/AN or non-Hispanic white) on disease status using the most recent multi-year BRFSS data. We hypothesize that if differences in disease etiology are associated with disease prevalence, their effect could be observed as a significant association between the race/ethnicity variable and chronic respiratory disease. Alternatively, if socioeconomic factors are the major drivers of disparity, then AI/AN race/ethnicity will not be independently associated with chronic respiratory disease after adjusting for socioeconomic covariates.
Methods
We used data from the BRFSS survey conducted in the years 2011–2018. The BRFSS is an annual random-digit dialed telephone health survey of adult US residents aged 18 and older [
20]. It is conducted across all US states and territories by state health departments in collaboration with the Center for Disease Control and Prevention. In 2017, the AI/AN population was oversampled in 11 states with historically high AI/AN relative populations to increase understanding of their health status [
21]. These states were: AK, AZ, MN, MT, NE, NM, NC, ND, OK, SD, and WI. We therefore restricted our analysis to responses from residents of these 11 states. We generated a map showing the locations of these states and their overlap with federally recognized and statistical AI/AN entities by using the open source ‘tigris’ package [
22] in R version 3.6.1. The shapefile containing the geographic information of these federally recognized AI/AN entities used for map plotting was obtained from the US Census Bureau [
23]. The analysis of publicly available, de-identified data does not constitute human subjects research as defined in federal regulations, and thus this study did not require Institutional Review Board (IRB) review.
The primary outcome, chronic respiratory disease status, was dichotomized as Negative if respondents gave 0 affirmative answers for the following questions or as Positive if they gave at least 1 affirmative answer: ‘Have you ever been told you had asthma’ and ‘Have you ever been told you have chronic obstructive pulmonary disease, COPD, emphysema, or chronic bronchitis?’ The exposure variable, race, included AI/AN or non-Hispanic White (hereafter termed White) as indicated by the BRFSS computed race variable. Respondents of other races were not included in the analysis. The number of responses analyzed per year averaged 75,029 (62878–87,350) which included approximately 5% AI/AN respondents (4.5–6.3%).
Demographic, socioeconomic, and behavioral variables were included as covariates in the analysis. Demographic covariates with their respective levels included: sex (male, female); age (18–24, 25–34, 35–44, 45–54, 55–64, 65 and older). Socioeconomic covariates included marital status (currently married, never married, formerly married); education level (some high school and lower, high school graduate, some college, college graduate); annual household income (<$10,000, $10,000 - < $15,000, $15,000 - < $20,000, $20,000 - < $25,000, $25,000 - < $35,000, $35,000 - < $50,000, $50,000 - < $75,000, ≥$75,000, unreported income); and access-to-care (adequate, inadequate). Behavioral covariates included smoking status (smoker, non-smoker); and weight morbidity as defined by body mass index (BMI) (underweight, normal weight, overweight, obese). Access-to-care was a composite variable aimed to assess both healthcare access and utilization. It was dichotomized as Inadequate if respondents gave 0–1 favorable answers for the following questions or as Adequate if they gave ≥2 favorable answers: ‘Do you have any kind of health care coverage, including health insurance, prepaid plans such as HMOs, or government plans such as Medicare, or Indian Health Service?’, ‘Do you have one person you think of as your personal doctor or health care provider?’, and ‘Was there a time in the past 12 months when you needed to see a doctor but could not because of cost?’ Respondents were categorized as smokers if they gave a positive answer to either of the questions: “Have you smoked at least 100 cigarettes in your entire life?”, and “Do you now smoke cigarettes every day, some days, or not at all?”
Responses with missing values for any variables except annual household income were excluded from analysis. Due to a substantial proportion of responses with missing annual household income, this variable could not be assumed as missing at random. We therefore assigned these responses to an “unreported income” category and included it as a level during regression.
Bivariable and multivariable logistic regression analyses were performed on individual years of BRFSS data from 2011 to 2018. Rao-Scott χ2 test was used to test for statistical difference between categorical variables. 2-tailed p values < .05 are considered statistically significant. To select for significant covariates for the logistic regression models, a stepwise forward modeling process was used. Odds ratios indicate associations when confidence intervals (CI) exclude 1.
To account for the complex sampling design of the BRFSS, data analysis was performed with the survey package [
24] using R version 3.6.1. Specifically, the yearly adjusted prevalence of chronic respiratory disease with 95% CI for AI/AN or white populations was calculated using yearly survey weights available as part of the BRFSS data.
Discussion
In this cross-sectional analysis using annual BRFSS surveys spanning 2011–2018, we observed higher prevalence of chronic respiratory disease in AI/AN respondents compared to non-Hispanic white respondents, and sought to characterize factors contributing to this disparity.
We found that being AI/AN was not independently associated with chronic respiratory disease. Adjusting for sociodemographic and behavioral covariates equalized the odds ratios for chronic respiratory disease between AI/AN and white populations. By contrast, socioeconomic factors such as low annual household income and educational attainment were strong determinants of chronic respiratory disease status. These findings were consistent for almost all years included in the analysis despite 2017 being the only year when the AI/AN population was oversampled. The exception was 2015 when the AI/AN sample could have been too small to reach statistical significance. Regardless, these consistent observations underscored the importance of low socioeconomic status in determining chronic respiratory disease status [
27,
28], while further showing that the AI/AN characteristic retained no disparity after accounting for these covariates. These data also validate the effectiveness of such public health surveillance effort in capturing critical disease correlates.
Why might AI/AN people exhibit higher prevalence of chronic respiratory disease? It has been well documented that a greater proportion of AI/AN people live in poverty compared to other races. In the most recent 5-year estimate (2014–2018) from the American Community survey, the median household income for the AI/AN population was $41,879 compared to $60,293 for the nation as a whole [
29]. In the same period, 25.8% of AI/AN people lived below poverty level compared to 14.1% of the general population [
30]. AI/AN levels of education attainment similarly lag behind that of other races [
31] and AI/AN high school graduation rate was found to be lowest among all races/ethnicities [
32]. Smoking prevalence was also highest for the AI/AN group compared to other race/ethnicities [
33,
34]. The increased proportion of AI/AN people with low socioeconomic status is worrying. Low socioeconomic status has been consistently shown to be associated with worse disease outcome for both COPD [
28,
35,
36] and asthma [
37,
38],
In summary, our findings confirm that the AI/AN population still exhibits higher prevalence of chronic respiratory disease compared to the non-Hispanic white population. Positive disease covariates include established socioeconomic variables while the AI/AN racial characteristic is not independently associated with disease. Our study therefore recommends that efforts to further promote cooperative mobilization of public health and social service infrastructures may make progress to address this disease disparity.
Strengths and limitations
Our study leveraged the oversampled data of the AI/AN population in the 2017 BRFSS to gain insight into the disparity of chronic respiratory disease. Despite 2017 being the only year with oversampling, our finding was consistent for all 7 years of the BRFSS where the AI/AN sample was found to be sufficient. However, we acknowledge a number of limitations in this study. Due to the cross-sectional nature of the BRFSS survey, temporal relationships and causality relationships of determinants cannot be established, some of which may also be bidirectional. This study relied on self-reported data in regards to health-risk behavior (smoking), disease status, and race classification. These types of self-classification might include nonrandom misclassification, and can present a limitation to the study, such as when self-reported smoking underestimates true smoking behavior [
39]. However, some studies suggest that self-reported rates of smoking are generally reliable for large national US surveys [
40,
41], and the present study confirms the expected observation that smoking status is positively associated with chronic respiratory disease. For self-reported asthma and COPD status, previous research has demonstrated general concordance between self-reported status and clinical diagnosis using spirometry [
42,
43]. Lastly, our analysis was limited to covariates available from the BRFSS survey, which may fail to capture all confounders, some of which could be ethnoculturally unique to the AI/AN population.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.