The National Health and Nutrition Examination Survey (NHANES) is a nationally representative, cross-sectional survey designed to monitor the health and nutrition of the civilian, non-institutionalized, resident U.S. population.¹⁷ NHANES is conducted through a household interview and in-person examinations within a mobile examination center (MEC). The survey design has included oversampling of various groups over time, such as Mexican-Americans prior to 2007–2008, Hispanic persons starting in 2007–2008, and non-Hispanic, non-Black Asian persons starting in 2011–2012.^18,19 Signed informed consent for participation in NHANES was obtained from participants who had reached the age of maturity in their state (usually age 18) and from parents or guardians of participants who had not yet reached the age of maturity. Written consent or assent was additionally obtained from children ages 7–17 years. For most questionnaire items, participants ages ≥16 years were interviewed directly while adult proxies provided information for participants ages <16 years.²⁰ For smoking and recent tobacco use questions, participants ages ≥12 years answered questions themselves without use of proxy.²⁰ Serum was collected from those who participated in the NHANES examination.²¹ The cumulative examination response rate for all NHANES participants for 2007–2018 was 65.6%, declining from 75.4% during 2007–2008 to 48.4% during 2017–2018.

Data from linkage-eligible NHANES participants during the years 2007 through 2018 were linked with HUD administrative records on public and assisted housing programs for years 1996 through 2019. Participants were considered eligible for linkage if they consented to linkage and provided sufficient data for linkage. Prior to 2009, a refusal to provide a 9-digit social security number was considered an implicit refusal for data linkage.²² HUD data were available through the “NCHS Data Linked to HUD Housing Assistance Program Files”.²³ Data from NHANES participants were linked with HUD administrative data from Housing Choice Vouchers, Multifamily Housing, and Public Housing, HUD’s three largest housing assistance program categories. The percentage of all MEC participants eligible for linkage by survey cycle ranged from 78.5% to 93.8% for survey years used within the analysis.²⁴ Linkage was performed with both deterministic and probabilistic approaches using identifiers, including social security number, name, birth date, sex, and zip code and state of residence.²² Timing of HUD program participation was assessed alongside survey participation to determine concurrency. Additional details about data linkage are available.²² Linkage data and rural-urban coding were available through NHANES Limited Access Data.²⁵ All other data were available through NHANES online public use data. The National Center for Health Statistics (NCHS) Ethics Review Board approved both NHANES and the data linkage between NHANES and HUD.

NHANES participants who self-reported active tobacco or nicotine use or who met defined criteria for active tobacco or nicotine use based on serum cotinine level were excluded from analyses. Self-reported active tobacco or nicotine use was defined based on available questionnaire data for those ages 12 years and above as those who smoked at least one cigarette in the last 30 days or used any tobacco or nicotine product in the previous 5 days. Serum cotinine levels were available for NHANES participants ages 3 years and older.²⁶ The lower limit of detection was 0.015 ng/mL. Consistent with prior published U.S. Centers for Disease Control estimates, SHS exposure was defined as a serum cotinine of ≥0.05 ng/mL for those ages 3–11 years and 0.05–10.0 ng/mL for those ages ≥12 years.^3,5,6 Those with serum cotinine <0.05 ng/mL were considered to have no SHS exposure. Individuals ages ≥12 years with a serum cotinine >10 ng/mL were assumed to have active tobacco or nicotine use. Children ages 3–11 years with serum cotinine >10 ng/mL were rare (n < 20), and most of these children were ages <8 years. These children were assumed to have SHS exposure rather than active or nicotine tobacco use. As the definition for SHS exposure is based on serum cotinine, other secondhand sources of exposure to nicotine beyond tobacco smoke, such as secondhand aerosol from nicotine-containing e-cigarettes will contribute.^27,28

Those concurrently receiving assistance from a HUD program at the time of their NHANES examination were defined as HUD-assisted while others were defined as HUD-unassisted.

Various sociodemographic, housing, and survey characteristics were used in analysis: sex (male, female), age (children: 3–11, 12–17 years; adults: 18–44, 45–61, ≥62 years), race and Hispanic origin (Black non-Hispanic, White non-Hispanic, other non-Hispanic, and Hispanic), income to poverty ratio (IPR), education of household reference person for children (categorization based on publicly available data categories: less than high school graduation, high school graduate or some college, college graduate), education for adults (less than high school graduation, high school graduate, some college, college graduate), health insurance (government program, private, none), marital status of household reference person for children or marital status for adults (married or living with partner, widowed or separated or divorced, never married), household size (1 person, 2–4 people, ≥5 people), examination season (November–April, May–October), urban-rural classification (urban, rural), 4-year survey cycle (2007–2010, 2011–2014, 2015–2018), and presence of household members who smoke within the home (0, ≥1).

Race and Hispanic origin were self-reported and categorized as released by the NHANES survey in four categories: Black non-Hispanic, White non-Hispanic, “other, non-Hispanic,” and Hispanic. Those included in the “other, non-Hispanic” category were persons not of Hispanic origin reporting other races or more than one race. The study design for the included years does not allow for reliable disaggregation of race and Hispanic origin into other categories. Therefore, analyses were limited to the race and Hispanic origin groups included in the NHANES design for all survey cycles used in this analysis. IPR was calculated using the Department of Health and Human Services poverty guidelines, which varies by family size and geographic location.²⁹ Eligibility for HUD assistance is based, in part, on income limits defined by percentage of median family income for the area.³⁰

To improve comparability in income between HUD-assisted and HUD-unassisted individuals, a comparison group of HUD-unassisted individuals included those with an IPR <2.0 (hereafter referred to as “low-income”).³¹ Differences in income eligibility by location, differences in defining income by IPR vs. percentage of median household income, and additional eligibility criteria for HUD assistance are not accounted for in this categorization scheme. Urban-rural classification was defined according to the 2013 NCHS Urban-Rural Classification Scheme for Counties.³² Urban was defined as residing within one of the four metropolitan county categories (large central metro, large fringe metro, medium metro, or small metro) and rural was defined as residing within one of two nonmetropolitan county categories (micropolitan or non-core).

All analyses were performed using R version 4.0.3 (R Foundation for Statistical Computing), including the R survey package and accounted for the survey’s complex, multistage probability design.³³ Analyses used eligibility-adjusted MEC examination weights which were adjusted for nonresponse to linkage within age, sex, and race and ethnicity domains.²² All linked participants with a probability above a threshold of 0.9225 were included in this analysis.

Prevalence estimates of SHS exposure were calculated during 2007–2010, 2011–2014, and 2015–2018 among 2 groups: 1) HUD-assisted persons and 2) low-income HUD-unassisted persons. Confidence intervals for prevalence estimates were obtained using the Korn and Graubard method.³⁴ Linear and quadratic terms for time trends were assessed using logistic regression. When quadratic time terms were significant, changes between 2007–2010 and 2011–2014 and between 2011–2014 and 2015–2018 were assessed separately. The relationship of receipt of HUD assistance with exposure to SHS among individuals with low-income was assessed using odds ratios (OR) calculated with bivariate logistic regression as well as multivariable logistic regression to account for potential confounding and mediating variables. Confounding variables were selected a priori including sex, age, race and Hispanic origin, IPR (continuous), education and marital status (of the household reference person for children and adolescents), health insurance status, household size, exam season, urban-rural category, and survey cycle. A sensitivity analysis within the supplementary materials additionally included presence of members of the household who smoke inside the home within the multivariable models as a mediating variable associated with both HUD assistance and exposure to SHS.

NHANES participants ages 3 years and older, who were HUD-linkage eligible, were examined in the MEC, and had an available serum cotinine level were included in analysis (Figure 1). Participants with active tobacco or nicotine use were excluded from analyses. Analyses were stratified into two groups: 1) children and adolescents ages 3–17 years and 2) adults ages 18 and above. Because exposure to SHS varies by age,^5,6 prevalence estimates and time trends were further stratified by age, based on categories often used in HUD reports: 1) 3–11, 2) 12–17, 3) 18–44, 4) 45–61, and 5) ≥62 years.^22,35,36 HUD-assisted individuals were compared with low-income HUD-unassisted individuals, using bivariate and multivariable models and complete case analysis. To further improve comparability between HUD-assisted and unassisted groups, the small number of HUD-assisted participants with an IPR ≥2.0 were excluded from models.

Of the 16,653 participants ages 3–17 years examined in the NHANES MEC during 2007–2018, 14,947 (89.8%) were linkage-eligible and had adjusted examination weights available. Among these, 2,969 (19.9%) did not have cotinine data and 429 (2.9%) were excluded for active tobacco or nicotine use. Of the remaining 11,549 children and adolescents ages 3–17 years, 6,743 (955 HUD-assisted and 5,788 low-income HUD-unassisted) were included in analysis while 4,806 (4,055 with HUD-unassisted with IPR ≥2 and 751 HUD-unassisted with no IPR data) were excluded (Figure 1). The distribution of social and demographic characteristics among the HUD-assisted and low-income HUD-unassisted groups are shown in Table 1 and Table S1.

The prevalence of exposure to SHS among HUD-assisted children and adolescents did not significantly change from 2007–2010 through 2015–2018 (80.2% [95% CI: 73.2%–86.1%] to 73.2% [95% CI: 63.2%–81.7%], P = 0.17; Figure 2A). There was a significant decrease in SHS exposure from 2007–2010 through 2015–2018 among those ages 12–17 years receiving HUD assistance (78.6% [95% CI: 68.0%–87.0%] to 64.3% [95% CI: 52.5%–74.9%], P = 0.04) but not among those ages 3–11 years. There was a significant quadratic trend for SHS exposure from 2007–2010 through 2015–2018 among low-income HUD-unassisted children and adolescents ages 3–17 years and those ages 12–17 years with decreases in SHS exposure from 2007–2010 through 2011–2014 (all ages: 57.4% [95% CI: 51.4%–63.3%] to 46.2% [95% CI: 42.2%–50.1%], P = 0.01; ages 12–17 years: 52.2% [95% CI: 44.3%–60.0%] to 39.9% [95% CI: 33.9%–46.1%], P = 0.01) followed by no change from 2011–2014 through 2015–2018. For low-income HUD-unassisted children and adolescents ages 3–11 years there was a significant decrease in SHS exposure from 2007–2010 through 2015–2018 (60.6% [95% CI: 54.5%–66.4%] to 50.8% [95% CI: 44.5%–57.1%], P = 0.02)

In bivariate analyses of low-income children and adolescents, those receiving HUD assistance were more likely to have exposure to SHS than those not receiving HUD assistance for the entire study period (2007–2018: OR = 3.39 [95% CI: 2.58–4.47]) and for each 4-year period (2007–2010: OR = 3.08 [95% CI: 2.17–4.38]; 2011–2014: OR = 4.02 [95% CI: 2.45–6.60]; 2015–2018: OR = 3.01 [95% CI: 1.77–5.12]; Figure 3A). After adjustment for sociodemographic and other characteristics, this relationship remained significant overall (OR = 1.60 [95% CI: 1.11–2.30]) and for 2007–2010 (OR = 1.82 [95% CI: 1.13–2.95]). Findings remained similar in sensitivity analyses adjusting weights for differential missingness in serum cotinine data by age, sex, and race and Hispanic origin (Figure S1A) and adjusting for presence of people within the household who smoke inside the home (Figure S2A).

Of the 35,162 participants ages ≥18 years examined within the MEC during the study period, 31,845 were linkage-eligible and had adjusted examination weights available. Among these, 1,768 (5.6%) did not have cotinine data available and 8,269 (26.0%) were excluded for active tobacco or nicotine use. Of the remaining 21,808 adults ages 18 years and over, 8,950 (878 HUD-assisted and 8,072 low-income HUD-unassisted) were included in this analysis while 12,858 (11,013 with HUD-unassisted with IPR ≥2 and 1,845 HUD-unassisted with no IPR data) were excluded (Figure 1). The distribution of sociodemographic characteristics among HUD-assisted and low-income HUD-unassisted adults are shown in Table 1 and Table S1.

The observed decrease in prevalence of exposure to SHS among HUD-assisted adults was not significant, from 2007–2010 through 2015–2018 among all ages (58.9% [95% CI: 50.7%–66.7%] to 50.1% [95% CI: 40.1%–60.1%], P = 0.18; Figure 2B) nor among specific age groups; however, some estimates by time period did not meet NCHS standards of statistical reliability.³⁷ Among low-income HUD-unassisted adults, trends in SHS exposure had a quadratic relationship for all ages combined, those ages 18–44 years, and those ages 45–61 years. Among low-income HUD-unassisted adults in these age groups, decreases in SHS exposure were seen from 2007–2010 through 2011–2014 (all ages: 43.3% [95% CI: 39.9%–46.8%] to 31.6% [95% CI: 28.2%–35.1%], P < 0.001; ages 18–44 years: 45.9% [95% CI: 42.5%–49.4%] to 34.5% [95% CI: 29.7%–39.4%], P < 0.001; ages 45–61 years: 46.5% [95% CI: 39.3%–53.8%] to 32.2% [95% CI: 27.4%–37.3%]; P = 0.001) followed by no change from 2011–2014 through 2015–2018. For ages ≥62 years SHS exposure decreased from 2007–2010 to 2015–2018 (35.8% [95% CI: 31.0%–40.9%] to 24.0% [95% CI: 19.4%–29.0%], P < 0.001).

Bivariate analysis showed that adults receiving HUD assistance were more likely to have SHS exposure than low-income adults not receiving HUD assistance during 2007–2018 (OR: 2.11 [95% CI 1.66–2.68]) and for each 4-year period examined (2007–2010: OR = 2.18 [95% CI: 1.49–3.21]; 2011–2014: OR = 2.13 [95% CI: 1.40–3.23]; 2015–2018: OR = 2.23 [95% CI: 1.50–3.31]; Figure 3B). After adjusting for various social, demographic, and survey characteristics, odds ratios were attenuated but remained significant overall (OR = 1.54 [95% CI: 1.20–1.98]), for 2011–2014 (OR = 1.52 [95% CI 1.00–2.32]), and for 2015–2018 (OR = 1.59 [95% CI: 1.04–2.43]). Results did not differ meaningfully when adjusting weights for missingness of cotinine data or for presence of household members who smoke within the home (Figures S1B and S2B).