Abstract
Purpose:
This study compared two classifications of rurality and their associations with cigarette, e-cigarette, and smokeless tobacco (SLT) use among a nationally representative sample of 31,196 US adults.
Methods:
Data from Wave 1 of the Population Assessment of Tobacco and Health Study. Weighted descriptive statistics and multivariable logistic regressions assessed whether two classifications of rurality were differentially associated with past 30-day (P30D) cigarette, e-cigarette, or SLT use in separate models. Classifications were 1) the U.S. Census Bureau’s classification as urban/non-urban; and 2) the National Center for Education Statistic (NCES)’s classification as urban/suburban/town/rural. This study is reported in accordance with STROBE guidelines.
Findings:
With the Census Bureau classification, 79.3% were in urban areas. With the NCES classification, 34.3% were in urban, 35.1% in suburban, 9.4% in town, and 21.1% in rural areas. With the Census Bureau classification, non-urban (vs urban) residence was associated with reduced odds of e-cigarette use (AOR=0.79; 95% CI=0.70-0.88) and increased odds of SLT use (AOR=2.32; 95% CI=1.97-2.72). With the NCES classification with urban as reference, rural residence was associated with reduced odds of e-cigarette use (AOR=0.77; 95% CI=0.75-0.98); both town (AOR=2.16; 95% CI=1.69-2.78) and rural (AOR=2.75; 95% CI=2.16, 3.48) were associated with increased odds of SLT use. Location was not associated with cigarette use for either classification.
Conclusions:
Location was similarly associated with P30D e-cigarette and SLT use across both classifications in adjusted models. The use of classifications with more categories may be beneficial to understand nuanced location differences in tobacco use.
Keywords: Rural, measurement, urban, suburban, tobacco
INTRODUCTION
Rural areas in the United States (U.S.) are disadvantaged in terms of their risk for tobacco use1 and related disease (e.g., lung cancer),2 which is the result of a pro-tobacco environment (e.g. targeted advertising) coupled with poor access to resources that discourage tobacco use.3 In 2021, rates of current cigarette smoking and SLT use were 18.0% and 4.5%, respectively, in U.S. rural areas, as compared to 10.5% and 1.8%, respectively, in urban areas.1 Some work also shows higher rates of electronic cigarette (e-cigarette) use for rural areas,1 while others suggest the opposite pattern4 or that e-cigarette use rates do not differ by geographic region.5 Importantly, rural areas are heterogeneous, and cross-study differences may be due to how geographic areas are delineated.
Existing classifications of rurality are numerous and take into account different characteristics of the geographic area. For instance, some are based primarily on population density, with rural areas being defined as those with relatively sparse populations (e.g. <2,500 residents, U.S. Census Bureau, 2010). These classifications fail to consider areas’ access to health-promoting resources like jobs and healthcare. Indeed, other classifications incorporate proximity to more urbanized areas, such as through the use of commuting flows (i.e., home to workplace travel; RUCA) or their distance from those areas.6 This latter characteristic has been used within continuous measures to better differentiate rural areas that are more versus less isolated6 or rural.7 Tradeoffs between the various classifications has been described elsewhere, with the more useful classification being contingent upon the topic being studied.6–8
Only one study has directly compared geographic location classifications to determine their impact on tobacco use, and that study focused on cigarette smoking-related outcomes.9 While all classifications revealed a similar pattern of results (e.g., higher smoking in rural areas), those that more clearly delineated rural areas were most sensitive to rural-urban differences.9 More work is needed to investigate the varying classifications of geographic location to better understand nuanced differences in tobacco disparities. Using data from the national-level Population Assessment of Tobacco and Health (PATH) Study, we compared rates of cigarette, SLT, and e-cigarette use using two classifications of rurality.
2. METHOD
2.1. Study Design and Sample
We conducted a cross-sectional analysis using data from adults (aged 18+) in Wave 1 (W1) of the PATH Study (2013-2014), a national longitudinal cohort survey of U.S. youth and adults. W1 was used because it is the only wave of PATH Study data that includes two geographic location classifications for direct comparison. Details about the PATH Study design, methods, and instruments are publicly available.10 The current analysis used the W1 Adult Restricted Use Files10 and was limited to adults who were not missing data on any study measures (N=31,196). This study qualified as exempt per the Westat Institutional Review Board and was conducted following STROBE guidelines.
2.2. Measures
2.2.1. Geographic location
Geographic location classifications were provided by the PATH Study.10 One classification (Census Bureau) was dichotomous (urban, non-urban). For the Census Bureau classification, respondents’ geographic locations were categorized as “urban” if the majority of the sampling area’s total population resided in areas classified as urban according to the 2010 U.S. Census (i.e., ≥2,500 residents) and “non-urban” otherwise. The other classification, from the National Center for Education Statistics (NCES’s modified classification), included four categories, which were: urban, suburban, town, or rural. For this classification, categories were assigned to individual respondents at the U.S. Census block level using 2020 Census block definitions. Respondents’ geolocations were coded to match 12-level locale classifications from the NCES prior to being collapsed by the PATH Study into the four categories listed.10 Additional information can be found at: https://nces.ed.gov/programs/edge/Geographic/LocaleBoundaries.
2.2.2. Past 30-day (P30D) tobacco use
Primary outcomes were P30D cigarette, e-cigarette, or SLT use (coded as 0 for no and 1 for yes for each). For SLT use, use of traditional SLT (e.g., chewing tobacco, snuff, dip) and snus were combined.
2.2.3. Sociodemographic covariates
Sociodemographic covariates were categorized as in Supplemental Table 1. These included age, sex, race, ethnicity, educational attainment, total household income, and U.S. region. Missing data on age, sex, and race were imputed as described in the PATH Study Restricted Use Files User Guide.10
2.3. Statistical analysis
We first examined the weighted crosstab between the Census Bureau and modified NCES classifications to examine overlap between the two. Then, weighted descriptive statistics and logistic regressions were used to evaluate associations between geographic location and P30D cigarette, e-cigarette, or SLT use in separate models. Unadjusted models and models adjusted for sociodemographic covariates were examined to estimate odds ratios (ORs) and adjusted ORs (AORs), respectively. Three pairs of adjusted logistic models were considered, with each based on one of the three outcomes. One model in each pair was based on the Census Bureau classification for geographic location and the other with the modified NCES classification. To account for multiple comparisons, p-values were adjusted using the Benjamini-Hochberg procedure with a false discovery rate of q=0.05.11 First, raw p-values for all geographic location comparisons (a total of 24) were ranked in ascending order. Then, adjusted p values for each ranked comparison were calculated individually by multiplying each comparison’s raw p value by the total number of comparisons and then dividing by the comparison’s rank. For example, for the comparison ranked as #10 (p=0.006), the adjusted p value was calculated as: (0.006*24)/10. Comparisons with an adjusted p value of <0.05 were considered statistically significant.
All analyses were weighted using the W1 single-wave survey weights, which included full sample and 100 replicate weights, to produce nationally representative estimates. Variances were computed using the balanced repeated replication (BRR) method with Fay’s adjustment set to 0.3 to increase estimate stability. Details on how replicate weights were constructed by the PATH Study can be found in the PATH Study Restricted Files User Guide.10 All analyses were conducted using Stata/MP 17.0 (www.stata.com/statamp/).
3. RESULTS
3.1. Geographic location classification comparison
The two classifications of geographic location provided similar values for respondents from larger metropolitan areas, with 99.7% and 98.2% of urban and suburban respondents (from the modified NCES classification), respectively, being classified as urban with the Census Bureau classification (Table 1). The classifications did not match quite as well for respondents from smaller areas, with 89.3% and 10.9% of town and rural respondents (from the modified NCES classification), respectively being classified as urban with the Census Bureau classification.
Table 1.
Weighted crosstab showing overlap in respondent samples between geographic location classifications (Census Bureau and modified NCES) among adults at Wave 1 of the Population Assessment of Tobacco and Health (PATH) Study, N=31,196.
| Modified NCES Classification | ||||
|---|---|---|---|---|
| Urban (n=10,711) | Suburban (n=10,951) | Town (n=2,942) | Rural (n=6,592) | |
| Weighted column % | ||||
| Census Bureau Classification | ||||
| Urban (n=24,752) | 99.7 | 98.2 | 89.3 | 10.9 |
| Non-urban (n=6,444) | 0.3 | 1.8 | 10.7 | 89.2 |
NCES=National Center for Education Statistics.
3.2. Sample characteristics
Participants were 52.2% female, largely of White race (78.5%), and the mean age was 47.8 years. There was a relatively even distribution of respondents across household income and education levels. According to the Census Bureau classification of geographic location, 79.3% of adults were located in urban areas and 20.7% were in non-urban areas. According to the modified NCES classification, 34.3% were in urban, 35.1% in suburban, 9.4% in town, and 21.1% in rural areas (Supplemental Table 1).
Comparing weighted characteristics by the Census Bureau classification of geographic location, more respondents in urban (vs non-urban) areas had a bachelor’s or advanced degree (30.9 vs 18.3%) and household incomes of $100,000+ (17.6 vs 12.0%). Urban (vs non-urban) areas also had larger Hispanic (17.8 vs 4.7%), Black (13.3 vs 6.8%), and Other/multiple race (11.2 vs 3.6%) populations. Similar findings were observed when comparing sample characteristics by the modified NCES classification. Characteristics for respondents from town and rural areas were more similar to one another and distinct from urban and suburban respondent characteristics whereas characteristics for respondents from urban and suburban areas were more similar to one another (Supplemental Table 1).
3.3. Geographic location and P30D tobacco use
3.3.1. Census Bureau classification of geographic location
Adults in non-urban (vs urban) areas had higher prevalences of P30D cigarette (24.3 vs 21.8%) and SLT use (6.0 vs 2.0%), but not P30D e-cigarette use (6.2 vs 6.8%; Supplemental Table 1).
Non-urban (vs urban) residence was significantly associated with increased odds of P30D cigarette use in unadjusted (OR=1.16), but not adjusted analyses (AOR=0.92). Non-urban (vs urban) residence was not significantly associated with P30D e-cigarette use in unadjusted analysis (OR=0.90) but became significantly associated with reduced odds in adjusted analyses (AOR=0.79). Non-urban (vs urban) residence was significantly associated with increased odds of P30D SLT use in unadjusted (OR=3.11) and adjusted models (AOR=2.32) (Table 2).
Table 2.
Unadjusted and adjusted weighted logistic regressions examining associations between geographic location and past 30-day cigarette, e-cigarette, or SLT use among adults at Wave 1 (2013-2014) of the Population Assessment of Tobacco and Health (PATH) Study, N=31,196.
| Unadjusted | Adjusted with Census Bureau Classification1 | Adjusted with Modified NCES Classification1 | |
|---|---|---|---|
| OR (95% CI); adjusted p value | AOR (95% CI); adjusted p value | ||
| Outcome: Past 30-day Cigarette Use | |||
| Geographic location: Census Bureau Classification | |||
| Urban | Ref | Ref | \ |
| Non-urban | 1.16 (1.04, 1.28); 0.014 | 0.92 (0.85, 1.00); 0.068 | \ |
| Geographic location: Modified NCES Classification | |||
| Urban | Ref | \ | Ref |
| Suburban | 0.81 (0.74, 0.88); 0.022 | \ | 0.93 (0.86, 1.01); 0.107 |
| Town | 1.17 (1.02, 1.35); 0.046 | \ | 0.97 (0.85, 1.11); 0.731 |
| Rural | 1.07 (0.95, 1.20); 0.319 | \ | 0.91 (0.82, 1.00); 0.087 |
| Outcome: Past 30-day E-cigarette Use | |||
| Geographic location: Census Bureau Classification | |||
| Urban | Ref | Ref | \ |
| Non-urban | 0.91 (0.81, 1.02); 0.120 | 0.79 (0.70, 0.88); 0.011 | \ |
| Geographic location: Modified NCES Classification | |||
| Urban | Ref | \ | Ref |
| Suburban | 0.86 (0.76, 0.96); 0.024 | \ | 0.95 (0.85, 1.06); 0.416 |
| Town | 0.99 (0.81, 1.22); 0.946 | \ | 0.83 (0.69, 1.00); 0.089 |
| Rural | 0.85 (0.75, 0.97); 0.038 | \ | 0.76 (0.67, 0.86); 0.007 |
| Outcome: Past 30-day Smokeless Tobacco/Snus Use | |||
| Geographic location: Census Bureau Classification | |||
| Urban | Ref | Ref | \ |
| Non-urban | 3.12 (2.70, 3.60); 0.005 | 2.32 (1.98, 2.72); 0.004 | \ |
| Geographic location: Modified NCES Classification | |||
| Urban | Ref | \ | Ref |
| Suburban | 1.06 (0.82, 1.38); 0.702 | \ | 0.98 (0.77, 1.25); 0.885 |
| Town | 2.77 (2.12, 3.61); 0.004 | \ | 2.17 (1.69, 2.78); 0.003 |
| Rural | 3.85 (3.05, 4.86); 0.003 | \ | 2.75 (2.17, 3.48); 0.002 |
Adjusted for age, sex, race, ethnicity, educational attainment, household income, and U.S. region
\ = Not applicable.
OR=odds ratio. AOR=adjusted odds ratio. CI=confidence interval. NCES=National Center for Education Statistics. SLT=smokeless tobacco. Bolded values denote statistical significance after false discovery rate correction for multiple comparisons, p<0.05.
3.3.2. Modified NCES classification of geographic location
Adults in suburban areas had the lowest weighted prevalence of P30D cigarette use (19.6%), followed by urban (23.2%), rural (24.4%), and town (26.1%) areas. Adults in rural areas had the highest prevalence of P30D SLT use (6.0%), followed by town (4.4%), suburban (1.7%) and urban (1.6%) areas. Prevalence of P30D e-cigarette use was similar across geographic locations: 7.2% for urban, 6.3% for suburban, 7.2% for town, and 6.2% for rural (Supplemental Table 1).
Suburban (vs urban) was significantly associated with reduced odds (OR=0.80) and town (vs urban) was significantly associated with increased odds of P30D cigarette use (OR=1.17) in unadjusted analyses, but both associations became nonsignificant after adjusting for covariates (AOR=0.93 and AOR=0.97, respectively). Suburban (vs urban) was significantly associated with reduced odds of P30D e-cigarette use in unadjusted (OR=0.85), but not adjusted analysis (AOR=0.95); and rural (vs urban) was significantly associated with reduced odds of P30D e-cigarette use in unadjusted (OR=0.86) and adjusted models (AOR=0.77). Town (vs urban) and rural (vs urban) were consistently associated with increased odds of P30D SLT use in unadjusted (OR=2.70 and OR=3.85, respectively) and adjusted models (AOR=2.16 and AOR=2.75, respectively) (Table 2).
4. DISCUSSION
In this large nationally representative study of U.S. adults, we directly compared two classifications of geographic location. The two classifications were aligned with one another relatively well for respondents from larger metropolitan areas, but the Census Bureau classification classified respondents from smaller areas of the U.S. as “urban” frequently. Still, geographic location was similarly associated with P30D cigarette, e-cigarette, and SLT use across both classifications in adjusted models. Specifically, rurality was associated with higher rates of SLT use and lower rates of e-cigarette use; there was no association between rurality and cigarette use in adjusted models. To our knowledge, this is only the second study to directly compare how two different geographic location classifications may be differentially associated with tobacco use using a nationally representative sample, and the first to investigate e-cigarette and SLT use as outcomes.9
The use of varying classifications for geographic location can lead to bias in scientific measurement and research findings, and there are numerous classifications for geographic location with widespread use in tobacco research. Although the two classifications examined in this study were overall similar in their associations with P30D cigarette, e-cigarette, and SLT use, we observed nuanced differences that suggest the modified NCES classification is a more sensitive measure of respondents’ geographic location in this context. Specifically, we found that non-urban (vs urban) when dichotomized according to the Census Bureau and rural (vs urban) as well as town (vs urban) when classified according to NCES were significantly associated with increased odds of P30D SLT use. This finding suggests that “town” respondents may be more similar to “rural” respondents in terms of their SLT use, although the crosstab shows that ~89% of town respondents were classified as urban with the Census Bureau classification. Due to their demographic and tobacco use similarities, it may be appropriate to combine respondents from town and rural locations (as opposed to combining town with suburban and urban) when smaller cell sizes are an issue.
Our study had several strengths including a large representative sample and comparable data from different definitions, but also had limitations. First, there are place-associated factors that put people in rural areas of the U.S. at increased risk for tobacco use (e.g., lower tobacco excise taxes34; greater social norms around tobacco), many of which were not specifically measured or included here. It is possible that the associations found in our study could be more directly accounted for by such unmeasured factors. Second, the PATH Study only includes the two classifications of geographic location examined at W1 (2013-2014) and results may not generalize to other geographic classifications. Finally, our analysis included only U.S. adults; findings may not generalize to youth or adults in non-U.S. countries.
5. CONCLUSIONS
In this large nationally representative study among U.S. adults, two different classifications of geographic location (Census Bureau, modified NCES) agreed with one another relatively well for respondents from larger metropolitan areas, but the Census Bureau classification characterized respondents from smaller areas of the U.S. as “urban” frequently. This may suggest that some nuance is lost when dichotomizing a heterogenous concept like geographic location. Geographic location was similarly associated with P30D cigarette, e-cigarette, and SLT use across varying classifications in adjusted models, but classifications with more (vs fewer) categories may be beneficial to better understand nuanced differences in tobacco use, particularly SLT use, across geographic locations.
Supplementary Material
Statement of Funding Support:
Research reported in this publication was supported by grant numbers U54DA036151 from the National Institute of Drug Abuse [NIDA] and Food & Drug Administration [FDA] Center for Tobacco Products [CTP]. Additional funding for authors was provided by the Centers for Disease Control and Prevention [CDC] by grant number U48DP006391 to the West Virginia Prevention Research Center [author MDB], NIDA K01DA055696 and NCI R01CA289551 [author BKH], NIDA and FDA CTP U54DA058256 [authors SH and MKR], NCI R01CA273206 and NCI and FDA CTP U54CA287392 [author MER]. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH, FDA, or CDC. This work is a cross-institution collaborative project from the Marketing Influences Special Interest Group supported, in part, by U54-DA046060 from the Center for Coordination of Analytics, Science, Enhancement and Logistics (CASEL) in Tobacco Regulatory Science (National Institute of Drug Abuse [NIDA] and the Food and Drug Administration’s Center for Tobacco Products [FDA CTP]).
Footnotes
DECLARATION OF INTEREST
All authors have no conflicts of interest to disclose.
REFERENCES
- 1.Cornelius ME, Loretan CG, Wang TW, Jamal A, Homa DM. Tobacco Product Use Among Adults - United States, 2020. MMWR Morb Mortal Wkly Rep. Mar 18 2022;71(11):397–405. 10.15585/mmwr.mm7111a1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Harrington RA, Califf RM, Balamurugan A, et al. Call to Action: Rural Health: A Presidential Advisory From the American Heart Association and American Stroke Association. Circulation. Mar 10 2020;141(10):e615–e644. 10.1161/cir.0000000000000753 [DOI] [PubMed] [Google Scholar]
- 3.Ozga JE, Romm KF, Turiano NA, et al. Cumulative disadvantage as a framework for understanding rural tobacco use disparities. Exp Clin Psychopharmacol. Oct 2021;29(5):429–439. 10.1037/pha0000476 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Blank MD, Ozga JE, Romm KF, et al. Geographic isolation predicts tobacco product use among youth: A latent class analysis. J Rural Health. Mar 2022;38(2):373–381. 10.1111/jrh.12583 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Lewis-Thames MW, Langston ME, Fuzzell L, Khan S, Moore JX, Han Y. Rural-urban differences e-cigarette ever use, the perception of harm, and e-cigarette information seeking behaviors among U.S. adults in a nationally representative study. Prev Med. Jan 2020;130:105898. 10.1016/j.ypmed.2019.105898 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Doogan NJ, Roberts ME, Wewers ME, Tanenbaum ER, Mumford EA, Stillman FA. Validation of a new continuous geographic isolation scale: A tool for rural health disparities research. Soc Sci Med. Oct 2018;215:123–132. 10.1016/j.socscimed.2018.09.005 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Rayens MK, Kingsolver AE, Christian WJ, et al. Considering Relative Rurality in Tobacco Regulatory Science. Nicotine Tob Res. Dec 19 2024; 10.1093/ntr/ntae303 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Long JC, Delamater PL, Holmes GM. Which Definition of Rurality Should I Use?: The Relative Performance of 8 Federal Rural Definitions in Identifying Rural-Urban Disparities. Med Care. Oct 1 2021;59(Suppl 5):S413–s419. 10.1097/mlr.0000000000001612 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Roberts ME, Doogan NJ, Tanenbaum E, et al. How should we define “rural” when investigating rural tobacco use in the United States? Subst Abus. 2021;42(4):788–795. 10.1080/08897077.2020.1856292 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.United States Department of H, Human Services. National Institutes of Health. National Institute on Drug A, United States Department of H, Human Services F, Drug Administration. Center for Tobacco P. Data from: Population Assessment of Tobacco and Health (PATH) Study [United States] Restricted-Use Files. 2024. doi: 10.3886/ICPSR36231.v40 [DOI] [Google Scholar]
- 11.Benjamini Y, Hochberg Y. Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society: Series B (Methodological). 1995/January/01 1995;57(1):289–300. 10.1111/j.2517-6161.1995.tb02031.x [DOI] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.