Adult Tobacco Product Use Patterns across the Urban-Rural Continuum: The Longitudinal Population Assessment of Tobacco and Health (PATH) Study, 2016–2023

Abstract

Purpose:

To determine the association between adult tobacco/nicotine product use over time and residence location (urban, suburban, town, rural), controlling for demographics.

Methods:

Data from Waves 4–7 (2016–2023) of the Population Assessment of Tobacco and Health (PATH) Study (N=18,590 adults) were analyzed using survey-weighted logistic regression to evaluate location and time differences in likelihoods of current, daily, and established use of combustible tobacco, electronic nicotine delivery systems (ENDS), smokeless products, and poly use (≥2 products).

Findings:

Compared to those residing in urban locations, suburban residents were less likely to report current (adjusted odds ratio (aOR)=0.88), daily (aOR=0.73), and established combustible products use (aOR=0.77); town residents were more likely to use these products currently (aOR=1.20) and daily (aOR=1.42), and rural residents were more likely to use daily (aOR=1.25) and report established use (aOR=1.33). Rural residents had a lower likelihood of current ENDS use (aOR=0.90), compared with those in urban locations. Compared with urban residents, current smokeless products use was more likely among rural residents (aOR=1.63) and less likely among those in suburban locations (aOR=0.74). Participants living in suburban (vs. urban) locations were less likely to use ≥2 products currently (aOR=0.89) or daily (aOR=0.79), while rural residents were more likely to engage in daily poly use (aOR=1.39). Time was a significant factor in all models, with fluctuating patterns across waves.

Conclusions:

These findings highlight nuanced geographic differences in tobacco/nicotine product use patterns beyond simple urban-rural comparisons, informing efforts to eliminate tobacco/nicotine use disparities.

While tobacco use in rural populations has been documented in isolated United States (U.S.) geographic regions for several decades,^1–5 robust, large-scale investigation of rural tobacco use did not begin until the 2000s. This shift was in large part due to the CDC’s 2001 publication of the Urban and Rural Health Chartbook⁶ which used national data to show that both adolescent and adult smoking rates in the U.S. were higher among people living in rural places, compared with urban areas. Numerous other large-scale studies followed, corroborating rural disparities in tobacco use^7–10 and demonstrating their link to rural health disparities more generally.^9,11 Investigations have continued, as changes in healthcare, social climate, tobacco policy, and the tobacco industry necessitate continued surveillance and monitoring.^12–18

As the field of rural health disparities has progressed, understanding of rural tobacco use has become more nuanced. One emergent insight is that the definition of “rural” matters,¹⁹ and that the distinction between “rural” and “urban” is not a strict dichotomy.^20–22 Rather, there are varying levels of rurality and urbanicity. Moreover, the urban-rural continuum is not always linear in terms of risk. For example, suburban areas often house the most affluent and advantaged populations, putting them at lower risk for tobacco use compared with the most urban locations.^23,24 Extremely rural areas, while disadvantaged by some metrics (e.g., lack of access to healthcare) may also be relatively more protected by others (e.g., more restricted access to tobacco retail outlets). An evaluation of tobacco retailer density at the census tract level found that average density was greater in urban tracts with moderate to high disparities compared with the same subset of tracts in rural areas.²⁵

Further, more research is needed to examine variation in smoking patterns across the urban-rural continuum. Although daily tobacco use is common, many individuals also engage in non-daily or intermittent use. For example, adults who smoke non-daily now comprise approximately one-third of those who currently smoke cigarettes,²⁶ and this proportion is growing nationally.^26–29 Once thought to be a transitional pattern, stable non-daily tobacco use has become more common, particularly among young adults,^28,30–33 and is associated with deleterious health outcomes.^34–36 A similar delineation exists between those who use tobacco products experimentally (i.e., not yet ever having used the product ‘fairly regularly’ or not yet having reached a lifetime threshold for use, such as 100 cigarettes) versus those who have an established pattern of use. While the relationship between experimentation and eventual established use has typically been examined among adolescents,^37,38 experimental use of tobacco products among adults is also of interest, as they may initiate use of alternate tobacco products (such as e-cigarettes) as a possible smoking cessation strategy or as a new pattern of tobacco use.^39,40

The purpose of this study was to characterize prevalence and trends in tobacco use across varying levels of rurality and urbanicity, examining daily vs. non-daily and established vs. experimental use patterns among participants who reported current use (‘every day’ or ‘some days’). Data are from the Population Assessment of Tobacco and Health (PATH) Study, a large prospective cohort study examining tobacco use among a representative sample of U.S. adults.⁴¹ In addition to the indicators of geographic classification and time, the modeling included key sociodemographic characteristics as covariates. Detailed information on use group and tobacco/nicotine product categories (combustible, ENDS, smokeless, poly use) is displayed in Table 1.

Table 1.

Definitions of tobacco use variables.

Current use	Currently use the product ‘every day’ or ‘some days’
(a) Daily use	Currently use the product ‘every day’
(a) Non-daily use	Currently use the product ‘some days’
(b) Current established use	Has smoked ≥100 cigarettes in lifetime and currently smoke ‘every day’ or ‘some days’ (cigarettes only)
	Has ever used ‘fairly regularly’ and currently use ‘every day’ or ‘some days’ (for all other tobacco/nicotine products)
(b) Experimental use	Currently use ‘every day’ or ‘some days,’ but has not smoked 100 cigarettes in lifetime (for experimental use of cigarettes) or has not ever used the product ‘fairly regularly’ (for all other tobacco/nicotine products)
Current poly use	Currently use ≥2 tobacco/nicotine products, whether from the same product category (e.g., cigarettes and cigars) or different ones (e.g., cigarettes and ENDS)
(c) Daily poly use	Currently use ≥2 tobacco/nicotine products ‘every day,’ or use at least one of the products ‘every day’
(c) Non-daily poly use	Currently use ≥2 products only ‘some days’
(d) Current established poly use	Current established use of ≥2 products, or established use with at least one product
(d) Experimental poly use	Currently use ≥2 products; all products used at experimental level

Note: The observed frequencies for the two rows with “(a)” and the two rows with “(b)” each sum to the total number of current users; the observed frequencies for the two rows with “(c)” and the two rows with “(d)” each sum to the total number of current poly users.

Product categories: Combustible (cigarettes, traditional cigars, cigarillos, filtered cigars, pipes and hookah), ENDS (electronic nicotine delivery systems, including e-cigarettes and other electronic nicotine products), and smokeless (smokeless tobacco, snus pouches and other forms of smokeless tobacco) products. Poly use refers to the use of two or more products whether from the same or different product categories.

Methods

Design/Sample

This study comprised a longitudinal analysis of data from restricted-use files (RUF) of Waves 4–7 of the PATH Study. These data were collected between 2016 and 2023, using audio computer-assisted self-interviews that each participant completed in their home; for Wave 6, a telephone interview option was also provided due to concerns related to the spread of Covid-19. English and Spanish versions were available. Detailed information about the design and sample is available online.^42,43

The analytic sample for this study (N=18,590) comprised adults who completed Waves 4–7 of the survey and were not missing survey weight variables in the study, including specific tobacco product use indicators at all four waves, location of residence, and demographic variables. Participants were omitted from analyses if they had not completed the follow-up surveys (Waves 5–7). We utilized PATH-imputed demographic variables, ensuring minimal missing data for demographic (i.e., only age had any missing data, for 0.016% of participants) and residential location variables. The current combustible product use had no missing data. Missing data for other product types was minimal (0.002%-0.10).

Measures

The demographic factors used from PATH-imputed variables included age (<35 and ≥35 years), sex (male, female), race/ethnicity (Non-Hispanic White, Non-Hispanic Black, Non-Hispanic Asian, Non-Hispanic Other or Multi-racial, and Hispanic), and education (Less than high school or GED (General Educational Development), High school graduate, Some college or Associate’s degree, Bachelor's or Advanced degree); these were included as covariates in the multivariable models. The time variable was coded based on survey waves.

Location of participant residence.

Participant residence was classified using PATH Restricted Use Files (RUF) 4-level geographic classification scheme, including ‘Urban,’ ‘Suburban,’ ‘Town,’ and ‘Rural,’ which we used for the present analyses. This location variable was linked to each participant based on a modified National Center for Education Statistics (NCES) locale framework code⁴⁴ of their home address at the Census block level, based on the 2020 Census. While the original NCES location classification includes 12 subtypes, PATH collapsed these into four primary categories, with three subtypes within each of the four main types, which represent meaningful differences in population density, land use, and access to services while preserving analytic feasibility.

In this framework, 'Urban’ area is within a ‘Principal city’ and contained in an ‘Urbanized area’; ‘Suburban’ is outside of a ‘Principal city’ but inside an ‘Urbanized area’; ‘Town’ is a territory inside an ‘Urban cluster’; and ‘Rural’ includes any geographical area outside urban clusters that is Census-defined as a rural territory, regardless of its distance from an ‘Urbanized area’ or ‘Urban cluster.’ Conceptually, this classification reflects the addition of residential context relevant to tobacco-related exposures, including infrastructure, social environments, commercial density, and access to health resources.

This four-level classification differs from the ‘the binary ‘Urban’/‘Not urban’ measure previously used in PATH, which described location based on the 2010 Census. Compared with the binary location classification, the NCES measure provides greater geographic granularity by distinguishing ‘Suburban,’ and ‘Town’, which had largely overlapped with the prior ‘Urban’ in the binary categorization, while ‘Rural’ had greater overlap with the binary ‘Not urban.’²²

Tobacco/nicotine use.^42,46

Tobacco and nicotine products were classified into combustible (cigarettes, traditional cigars, cigarillos, filtered cigars, pipes, and hookah), ENDS (electronic nicotine delivery systems, including e-cigarettes and other electronic nicotine products), and smokeless (smokeless tobacco, snus, and other forms of smokeless tobacco) products. Poly use was defined as the use of two or more tobacco/nicotine products, whether from the same product category (e.g., cigarettes and cigarillos, both combustible) or from different ones (e.g., cigarettes and e-cigarettes, combustible and ENDS). Other variables included current use (use ‘every day’ or ‘some days’), daily and non-daily use (among those who currently use the product), and established and experimental use (among those who currently use the product).⁴⁷ These definitions are detailed in Table 1. The daily and established use variables are included here as indicators of intensity of use.

Data analysis

We summarized the full sample and divided this into the analytic sample (i.e., complete data for the outcomes included here) and those who were not able to be included in the multivariable analysis (i.e., those not retained because they were missing for one or more of the follow-up waves of data collection, Waves 5–7); the summary was done using unweighted (raw) frequencies and weighted percentages obtained using the SURVEYFREQ procedure in SAS (version 9.4) with replicate weights appropriate to each sample. The analytic and the incomplete sample were compared using the Rao-Scott Chi-square test, which evaluates differences in population-representative weighted distributions while accounting for the PATH Study's complex survey design. We used the SAS procedure SURVEYLOGISTIC for longitudinal logistic regression, a type of generalized estimating equations analysis strategy,⁴⁸ with repeated measures from the same individual accounted for in the model by including them in the same cluster. The purpose of this modeling was to determine whether location type and time were associated with the likelihood of tobacco use, with separate models for each tobacco use outcome (i.e., current, daily, and established use); we chose this analysis strategy so that the replicate weights could be included, reflecting the complex survey design.

We examined trends over time, first comparing individuals who currently use tobacco and nicotine (‘current use’) to all others (including both those who do not use tobacco (‘non-use’) and those who used tobacco formerly but not currently (‘former use’); then, within the current use groups, we evaluated differences between those who use daily vs. non-daily;⁴⁹ as well as differences between those with patterns of established use vs. experimental use⁵⁰ for specific product types (i.e., single and poly use of combustible tobacco, smokeless products, and ENDS). For each model, we included class variables of location type (with four categories; urban was used as the reference) and time (four waves of surveys; wave 4 was used as the reference); we also included the demographic covariates of age, sex, race/ethnicity, and education. We summarized the modeling results using adjusted odds ratios and 95% confidence intervals, and a statistical significance level of alpha=0.05.

We evaluated the significance of the location by time interaction within each of the models with a sufficient sample size for estimating this effect (including all current use models, and the daily and established use models for combustible products). For all models, the parameter estimates for the adjusted odds ratios and 95% confidence intervals were virtually identical whether the interaction was included or not. There was only one model (current use of combustible products) with a significant interaction effect, and the effect size of the interaction was small; none of the pre-planned pairwise comparisons were significant. Thus, the results described below include the main effects models only.

Findings

The weighted percentages in each category of the demographic variables are relatively consistent between the full sample of Wave 4 completions (N = 33,644) and the analytic sample (n = 18,590; see Table 2). It should be noted that only slightly more than half of participants completed all three follow-up waves of data collection (Waves 5–7) and were able to be retained in the analytic sample. In the weighted analytic sample, most participants were age 35 or older (68.5%), female (52.3%), non-Hispanic White (64.2%), and had at least some college education (61.6%). Most of the analytic sample lived in either an urban (31.2%) or suburban (38.5%) location, with 9.0% living in towns and 21.3% in a rural location. The comparisons between the analytic sample and the non-retained sample demonstrate that those who were 35 or older were more likely to be retained in the analysis, as were females and those with a higher level of education. The differences between the analytic sample and the non-retained sample were not significant for race/ethnicity and location.

Table 2.

Demographic and location characteristics for the full and analytic samples

Variable	Raw frequencies and weighted percentages (N=33,644)	Raw frequencies and weighted percentages (n=18,590)	Raw frequencies and weighted percentages (n=15,054)
	Full samplea	Analytic sampleb	Non-retained samplec	p (χ2)
Age				<.001
Age <35	18087 (30.2%)	9405 (28.2%)	8682 (33.1%)
Age ≥ 35	15551 (69.7%)	9183 (71.8%)	6368 (66.9%)
Sex				<.001
Male	16517 (48.1%)	8601 (45.7%)	7916 (51.2%)
Female	17125 (51.9%)	9989 (54.3%)	7136 (48.5%)
Race/Ethnicity
Non-Hispanic White	19137 (64.3%)	10651 (65.1%)	8486 (63.1%)	.0864
Non-Hispanic Black	5152 (11.7%)	2895 (11.6%)	2257 (11.8%)
Non-Hispanic Asian	894 (5.7%)	507 (5.7%)	387 (5.6%)
Non-Hispanic Other / Multi-racial	1787 (2.6%)	934 (2.5%)	853 (2.9%)
Hispanic	6672 (15.7%)	3603 (15.1%)	3069 16.5%)
Education
Less than High School/GED	6383 (15.9%)	3122 (13.5%)	3261 (19.6%)	<.001
High school graduate	8161 (23.8%)	4180 (21.4%)	3981 (27.2%)
Some college / Associate’s degree	11929 (31.0%)	6617 (31.2%)	5312 (30.7%)
Bachelor’s / Advanced degree	7169 (29.3%)	4871 (33.9%)	2498 (22.5%)
Location
Urban	11536 (31.1%)	6387 (31.1%)	5149 (31.1%)	.1286
Suburban	11947 (38.4%)	6730 (39.2%)	5217 (37.4%)
Town	3213 (9.1%)	1731 (8.8%)	1482 (9.4%)
Rural	6945 (21.4%)	3742 (20.8%)	3203 (21.1%)

^aThe full sample consists of participants in the PATH Wave 4 assessment without missing values for variables in the study, including sampling weights

^bThe analytic sample consists of participants in PATH who completed Waves 4–7 without missing values for variables in the study, including sampling weights

^cThe non-retained sample consists of participants in PATH who did not participate in at least one follow-up wave of the study (i.e., Waves 5 to 7)

Table 3 presents the weighted prevalence estimates for each use indicator (current, daily vs. non-daily, and established vs. experimental), by product category (combustible, ENDS, smokeless and poly use), for each location type (Urban, Suburban, Town, and Rural) and for each wave of the PATH survey (waves 4–7). These estimates provide an overview of use prevalence based on a longitudinal national survey using a location parameter that provides a finer distinction beyond a simple urban vs. rural dichotomy. Figure 1 displays the weighted prevalence estimates to summarize how use is changing over time by product type and by location, as well as for three distinct definitions of use.

Table 3.

Weighted Prevalence Estimates of Tobacco Product Use by Location and Time: Population Assessment of Tobacco and Health (PATH) Study, United States, 2016–2023

	Urban, weighted %(SE)				Suburban, weighted %(SE)				Town, weighted %(SE)				Rural, weighted %(SE)
	Wave4	Wave5	Wave6	Wave7	Wave4	Wave5	Wave6	Wave7	Wave4	Wave5	Wave6	Wave7	Wave4	Wave5	Wave6	Wave7
Combustible
Current Use	22.4(0.64)	21.3(0.62)	17.7(0.57)	18.1(0.53)	19.5(0.61)	18.5(0.56)	16.1(0.55)	15.2(0.56)	28.0(1.31)	26.0(1.35)	24.7(1.26)	24.4(1.38)	25.0(0.94)	23.9(1.07)	22.0(0.89)	20.9(0.86)
Daily Usea	12.5(0.53)	11.9(0.50)	11.2(0.45)	10.8(0.46)	11.0(0.47)	10.23(0.47)	9.8(0.45)	9.0(0.42)	21.0(1.01)	20.7(1.07)	19.2(1.03)	19.0(0.96)	19.0(0.84)	18.3(0.86)	17.2(0.75)	16.3(0.72)
Non-daily Usea	9.9(0.40)	9.4(0.40)	6.5(0.30)	7.4(0.36)	8.5(0.38)	8.2(0.38)	6.4(0.30)	6.2(0.29)	7.1(0.72)	5.3(0.60)	5.6(0.66)	5.4(0.79)	6.1(0.39)	5.6(0.41)	4.8(0.40)	4.6(0.42)
Established Useb	18.9(0.56)	18.4(0.57)	16.1(0.53)	16.2(0.50)	16.6(0.55)	15.5(0.50)	14.2(0.52)	13.3(0.54)	25.0(1.11)	23.8(1.24)	22.8(1.04)	22.5(1.07)	22.8(0.89)	22.3(1.05)	20.9(0.84)	19.9(0.81)
Experimental Useb	3.5(0.28)	2.8(0.26)	1.6(0.18)	1.9(0.16)	3.0(0.22)	3.0(0.27)	2.0(0.18)	1.9(0.15)	3.0(0.62)	2.2(0.44)	1.9(0.50)	1.9(0.46)	2.2(0.21)	1.6(0.20)	1.1(0.19)	1.0(0.17)
ENDS
Current Use	4.9(0.29)	6.5(0.33)	5.1(0.28)	6.1(0.30)	4.1(0.21)	6.0(0.28)	4.7(0.23)	5.3(0.27)	5.4(0.62)	6.4(0.54)	6.4(0.58)	6.8(0.50)	4.9(0.28)	5.8(0.34)	5.1(0.48)	6.2(0.43)
Daily Usea	1.8(0.20)	2.4(0.22)	2.4(0.21)	2.7(0.21)	1.6(0.13)	2.4(0.18)	2.3(0.17)	2.8(0.18)	2.0(0.38)	2.6(0.30)	3.7(0.39)	4.0(0.43)	1.7(0.18)	2.4(0.20)	2.7(0.32)	3.6(0.22)
Non-daily Usea	3.0(0.20)	4.1(0.23)	2.7(0.20)	3.3(0.22)	2.5(0.17)	3.6(0.21)	2.4(0.17)	2.5(0.19)	3.4(0.54)	3.8(0.49)	2.7(0.37)	2.7(0.41)	3.2(0.24)	3.4(0.27)	2.4(0.24)	2.6(0.34)
Established Useb	3.5(0.23)	4.5(0.27)	4.2(0.25)	4.8(0.25)	2.8(0.17)	4.4(0.24)	4.1(0.20)	4.6(0.22)	4.0(0.51)	4.8(0.45)	5.6(0.53)	5.8(0.50)	3.6(0.25)	4.2(0.29)	4.3(0.39)	5.4(0.4)
Experimental Useb	1.4(0.15)	2.0(0.16)	0.9(0.13)	1.3(0.16)	1.3(0.14)	1.6(0.16)	0.7(0.10)	0.8(0.12)	1.4(0.27)	1.7(0.32)	0.8(0.16)	1.0(0.23)	1.3(0.16)	1.5(0.17)	0.7(0.14)	0.8(0.11)
Smokeless
Current Use	2.0(0.24)	1.5(0.23)	1.2(0.21)	1.3(0.22)	1.9(0.15)	1.9(0.18)	1.6(0.17)	1.3(0.17)	3.7(0.49)	3.3(0.55)	3.0(0.56)	3.0(0.57)	6.2(0.48)	5.3(0.48)	5.2(0.41)	4.7(0.43)
Daily Usea	0.7(0.14)	0.6(0.16)	0.6(0.12)	0.6(0.15)	1.0(0.13)	1.0(0.14)	0.9(0.14)	0.8(0.10)	1.8(0.43)	1.9(0.48)	1.4(0.46)	1.8(0.51)	4.3(0.44)	3.8(0.45)	3.9(0.36)	3.6(0.40)
Non-daily Usea	1.3(0.21)	0.9(0.14)	0.7(0.14)	0.8(0.16)	0.9(0.12)	0.9(0.13)	0.7(0.13)	0.6(0.12)	1.9(0.33)	1.4(0.29)	1.6(0.25)	1.2(0.36)	1.9(0.27)	1.5(0.22)	1.3(0.18)	1.1(0.16)
Established Useb	1.6(0.22)	1.2(0.20)	1.1(0.20)	1.3(0.22)	1.8(0.14)	1.6(0.17)	1.5(0.17)	1.2(0.14)	3.2(0.45)	2.9(0.49)	2.7(0.55)	2.7(0.56)	5.5(0.51)	4.9(0.42)	4.9(0.39)	4.5(0.45)
Experimental Useb	0.4(0.09)	0.3(0.07)	0.1(0.04)	0.1(0.02)	0.1(0.03)	0.2(0.06)	0.1(0.03)	0.2(0.06)	0.4(0.14)	0.3(0.15)	0.4(0.09)	0.3(0.09)	0.7(0.24)	0.4(0.22)	0.3(0.07)	0.2(0.08)
Poly Use
Current Use	7.7(0.35)	7.5(0.35)	5.1(0.27)	5.6(0.29)	6.3(0.27)	6.5(0.31)	4.7(0.26)	4.0(0.27)	9.6(0.78)	7.8(0.59)	6.2(0.57)	5.9(0.63)	8.9(0.40)	8.6(0.53)	6.0(0.36)	6.5(0.50)
Daily Usea	0.8(0.11)	0.9(0.15)	0.9(0.11)	0.9(0.13)	0.8(0.12)	0.9(0.12)	0.5(0.09)	0.5(0.08)	1.6(0.31)	1.7(0.34)	0.6(0.16)	1.23(0.3)	1.9(0.26)	2.0(0.29)	1.4(0.18)	1.8(0.21)
Non-daily Usea	6.8(0.33)	6.5(0.34)	4.2(0.25)	4.8(0.23)	5.5(0.26)	5.6(0.29)	4.2(0.23)	3.5(0.23)	8.0(0.67)	6.1(0.46)	5.6(0.59)	4.7(0.59)	7.0(0.32)	6.6(0.49)	4.6(0.29)	4.6(0.47)
Established Useb	4.4(0.26)	4.5(0.26)	3.6(0.24)	3.8(0.22)	3.6(0.21)	4.0(0.24)	3.1(0.21)	2.8(0.20)	6.0(0.48)	5.2(0.46)	4.8(0.45)	4.1(0.45)	6.2(0.37)	6.1(0.44)	4.6(0.36)	5.1(0.43)
Experimental Useb	3.2(0.23)	2.9(0.22)	1.5(0.15)	1.8(0.16)	2.7(0.21)	2.5(0.21)	1.6(0.16)	1.2(0.15)	3.6(0.45)	2.6(0.38)	1.5(0.29)	1.9(0.34)	2.8(0.27)	2.5(0.27)	1.4(0.17)	1.4(0.16)

Note. Dates of Data Collection: 12/16–1/18 (Wave 4), 12/18–11/19 (Wave 5), 3/21–11/21 (Wave 6), and 1/22–4/23 (Wave 7).

^aAmong those currently using the product; daily and non-daily frequencies add to number of participants using that product currently

^bAmong those currently using the product; established and experimental frequencies add to number of participants using that product currently

Figure 1:

Weighted Prevalence Estimates of Tobacco and Nicotine Product Use by Location and Time: Population Assessment of Tobacco and Health (PATH) Study, United States, 2016–2023.

With emphasis on the independent class variables of location and time, Table 4 organizes these findings by product type to summarize the main effects logistic models for each definition of use (current use vs. former/never use; daily vs. non-daily use; and established vs. experimental use).

Table 4.

Weighted longitudinal logistic regression models^A to assess whether location and time associated with current, daily, and established use of combustibles, ENDS, smokeless and poly use.

Key variablesA	Tobacco/nicotine use categories
	Current use a (aOR; 95%CI)	Daily use b (aOR; 95%CI)	Established use c (aOR; 95%CI)
Outcome: Combustibles
Class variables:
Location (ref: Urban)
Suburban	0.88 (0.81–0.95) **	0.73 (0.67–0.80) ***	0.77 (0.67–0.90) **
Town	1.20 (1.09–1.31) **	1.42 (1.24–1.63) ***	1.01 (0.78–1.30)
Rural	0.97 (0.90–1.06)	1.25 (1.11–1.40) ***	1.33 (1.12–1.59) **
Time (ref: Wave 4)
Wave 5	1.08 (1.06–1.10) ***	0.93 (0.89–0.98) **	0.87 (0.80–0.95) **
Wave 6	0.91 (0.89–0.92) ***	1.13 (1.08–1.19) ***	1.25 (1.14–1.37) ***
Wave 7	0.88 (0.86–0.89) ***	1.03 (0.97–1.09)	1.15 (1.06–1.24) **
Outcome: ENDS
Class variables:
Location (ref: Urban)
Suburban	0.99 (0.91–1.08)	1.05 (0.91–1.21)	1.05 (0.90–1.22)
Town	1.01 (0.89–1.14)	1.06 (0.84–1.35)	1.04 (0.82–1.31)
Rural	0.90 (0.81–1.00) *	0.93 (0.78–1.10)	0.96 (0.79–1.17)
Time (ref: Wave 4)
Wave 5	1.15 (1.11–1.20) ***	0.80 (0.74–0.85) ***	0.70 (0.64–0.78) ***
Wave 6	0.93 (0.89–0.97) **	1.26 (1.14–1.40) ***	1.51 (1.30–1.77) ***
Wave 7	1.10 (1.06–1.15) ***	1.42 (1.29–1.56) ***	1.48 (1.32–1.66) ***
Outcome: Smokeless
Class variables:
Location (ref: Urban)
Suburban	0.74 (0.62–0.89) *	…d	…d
Town	1.12 (0.90–1.40)	…d	…d
Rural	1.63 (1.39–1.91) ***	…d	…d
Time (ref: Wave 4)
Wave 5	1.04 (0.98–1.10)	…d	…d
Wave 6	0.93 (0.87–0.99) *	…d	…d
Wave 7	0.86 (0.81–0.91) ***	…d	…d
Outcome: Poly use
Class variables:
Location (ref: Urban)
Suburban	0.89 (0.82–0.96) **	0.79 (0.66–0.94) *	0.88 (0.76–1.03)
Town	1.05 (0.94–1.18)	1.03 (0.80–1.34)	0.98 (0.83–1.15)
Rural	1.05 (0.96–1.16)	1.39 (1.14–1.71) **	1.20 (1.02–1.41)
Time (ref: Wave 4)
Wave 5	1.20 (1.16–1.24) ***	0.99 (0.87–1.14)	0.88 (0.81–0.95) **
Wave 6	0.82 (0.79–0.85) ***	0.96 (0.84–1.09)	1.22 (1.10–1.35) **
Wave 7	0.82 (0.78–0.86) ***	1.19 (1.05–1.36) *	1.27 (1.16–1.39) ***

^AAll models included the covariates of age, sex, race/ethnicity, and education (as shown in Table 1); these variables are suppressed in this table since they are not the focus of the analysis.

Abbreviations: ENDS = Electronic Nicotine Delivery System; Poly use = use of two or more tobacco/nicotine products; aOR = adjusted odds ratio; CI = confidence interval.

^aCurrent use vs. former/never use; sample size in all models is 18,590.

^bDaily use vs. non-daily among all current users (see Table 2); average sample sizes across the four waves: 5776, 1718, 583, and 1875 for combustibles, ENDS, smokeless and poly use, respectively.

^cEstablished use vs. experimental among all current users (see Table 2); average sample sizes across the four waves: 5776, 1718, 583, and 1875 for combustibles, ENDS, smokeless and poly use, respectively.

^dEstimates are not reliable for reporting due to small sample size within subgroup.

^*p ≤ .05;

^**p≤.005;

^***p≤.0001.

Combustible use.

With urban location as the reference and averaged over time, we found that suburban residents were less likely to report current use of combustible products (aOR=0.88, 95% CI: 0.81–0.95), while residents of towns were more likely to do so (aOR=1.20, 95% CI: 1.09–1.31; see Table 4). There was no difference in likelihood of current combustible use between urban and rural residents. With daily use as the outcome, suburban residents were less likely than those in urban areas to use combustible products daily (aOR=0.73, 95% CI: 0.67–0.80), while those living in towns and rural locations were more likely to use these products daily (aOR=1.42, 95% CI: 1.24–1.63; and aOR=1.25, 95% CI: 1.11–1.40, respectively), when compared with urban residents. With established use as the outcome and urban residence the reference, suburban residents were less likely to have developed a pattern of established use of combustibles (aOR=0.77, 95% CI: 0.67–0.90), while rural residents were more likely to have done so (aOR=1.33, 95% CI: 1.12–1.59); the difference in established use of combustibles was not significant between urban and town.

With Wave 4 (W4) as the time reference and averaged across locations, there was an increased likelihood of current use of combustible products in W5 (aOR=1.08, 95% CI: 1.06–1.10), followed by decreased likelihood in W6 (aOR=0.91, 95% CI: 0.89–0.92) and W7 (aOR=0.88, 95% CI: 0.86–0.89). The time comparisons for daily use and established use of combustible products were similar. With W4 as the reference for each other wave, there were decreases in likelihood of daily use (aOR=0.93, 95% CI: 0.89–0.98) and established use in W5 (aOR=0.87, 95% CI: 0.80–0.95), increases in likelihood in W6 (daily: aOR=1.13, 95% CI: 1.08–1.19 and established: aOR=1.25, 95% CI: 1.14–1.37, respectively), and an increase in W7 for established (aOR=1.15, 95% CI: 1.06–1.24), but not daily use.

ENDS use.

With urban as the reference and averaged over time, rural residents had a decreased likelihood of current use (aOR=0.90, 95% CI: 0.81–1.00; see Table 4), except for those residing in a suburban or town location. There were no location differences for the likelihood of ENDS daily use or established use.

Averaged over location categories, there were significant differences in current use likelihood of ENDS over time. Compared with W4, there was an increase in ENDS current use in W5 (aOR=1.15, 95% CI: 1.11–1.20), followed by a decrease in likelihood in W6 (aOR=0.93, 95% CI: 0.89–0.97), followed by an increase in W7 (aOR=1.10, 95% CI: 1.06–1.15). The patterns for likelihood of daily use and established use of ENDS were relatively consistent. Both outcomes demonstrated a significant decrease in use likelihood between W4 (reference) and W5 (aOR=0.80, 95% CI: 0.74–0.85 and aOR=0.70, 95% CI: 0.64–0.78 for daily use and established use, respectively), which was followed by increased likelihoods for both at W6 (daily: aOR=1.26, 95% CI: 1.14–1.40 and established: aOR=1.51, 95% CI: 1.30–1.77) and W7 (daily: aOR=1.42, 95% CI: 1.29–1.56 and established: aOR=1.48, 95% CI: 1.32–1.66).

Smokeless use.

As shown in Table 4, only the current use outcome for smokeless had sufficiently large subgroup sizes for both outcome categories for analysis; daily use and established use models for smokeless tobacco are not included due to PATH suppression rules. Averaged over the four waves, suburban residents had a decreased likelihood of current smokeless use, compared with urban residents (aOR=0.74, 95% CI: 0.62–0.89), while rural residents had an increased likelihood of current use for this product category (aOR=1.63, 95% CI: 1.39–1.91). With W4 as the reference and averaged over locations, the current use likelihood for smokeless declined in both W6 (aOR=0.93, 95% CI: 0.87–0.99) and W7 (aOR=0.86, 95% CI: 0.81–0.91).

Poly use.

Averaged over the four waves and relative to urban residents, those residing in suburban locations had a decreased likelihood of current use of two or more tobacco/nicotine products (aOR=0.89, 95% CI: 0.82–0.96; Table 4) and poly product daily use (aOR=0.79, 95% CI: 0.66–0.94). Compared with urban residents, rural residents had an increased likelihood of poly tobacco daily use (aOR=1.39, 95% CI: 1.14–1.71). There were no location differences in the likelihood of poly tobacco established use.

Examining the time effect averaged over the four locations and relative to W4, the likelihood of poly tobacco current use was higher in W5 (aOR=1.20, 95% CI: 1.16–1.24), but lower in W6 (aOR=0.82, 95% CI: 0.79–0.85), and W7 (aOR=0.82, 95% CI: 0.78–0.86). Poly daily use only differed in W7 vs. W4 (aOR=1.19, 95% CI: 1.05–1.36). However, likelihood of poly established use at W5 declined (aOR=0.88, 95% CI: 0.81–0.95), but increased in both W6 (aOR=1.22, 95% CI: 1.10–1.35) and W7 (aOR=1.27, 95% CI: 1.16–1.39) compared to W4.

Discussion

Among PATH Study adult participants followed in Waves 4–7 (collected 2016–2023), tobacco product use prevalence varied when considering differences in rural areas, towns, and suburban locations compared to urban places of residence. There were location and time differences in current use, daily use, and established use for most tobacco/nicotine products, though daily and established use patterns for ENDS products did not differ by location.

First, we found that residents of suburban locations were less likely than those in urban areas to report current, daily, and established use patterns for combustible tobacco products and the use of more than one tobacco/nicotine product, and to indicate current use of smokeless tobacco. While authors of an analysis of 2006–08 Behavioral Risk Factor Surveillance System data determined there was elevated likelihood of the use of cigarettes and smokeless tobacco in rural areas relative to both urban and suburban locations, their analysis did not examine urban vs. suburban prevalence differences.⁸ As suburban locales are often categorized together with urban locations, results from our study support a more nuanced approach to understanding tobacco use across the urban-rural continuum^10,20,21 that moves beyond a binary view of rurality and urbanicity. Variations in tobacco use patterns by location may reflect underlying socioeconomic differences among the location types examined, but literature in this area remains limited. More research is needed to understand the potential protective factors related to lower likelihood of tobacco use in suburban areas.

Second, we found that, compared to those in urban communities, residents of towns were significantly more likely to report current and daily use of combustible tobacco products. To our knowledge, this is the first time this association has been reported. These findings are important, however, since most of those who were identified with town as their location in this 4-level locale variable would have been classified as urban in a simple binary division.²² Compared with urban locations, the contrast of the lower use likelihood for both combustible products and poly use (regardless of use definition) and smokeless (for current use) in suburban areas coupled with higher combustible product use likelihood (for current and daily use) in towns underscores that combining these three locale categories together (i.e., urban, suburban, town) is an oversimplification of underlying use patterns. This finding provides further support for studying finer geographic distinctions to evaluate variations in tobacco/nicotine use within heterogenous locations, including those that are traditionally defined simply as rural.^10,20

Third, we found that relative to those living in urban areas, rural residents were not more likely to currently use combustible products, though they were more likely to use them daily and in an established pattern. The lack of difference in current use likelihood for combustibles between rural and urban residents is a relatively novel finding, though a recent paper based on an earlier wave of the PATH study described this result specifically for current cigarette use.²² From the same paper, rural residents, compared with urban residents, were less likely to currently use ENDS products. On the other hand, and consistent with many studies, those living in rural areas were more likely to currently use smokeless tobacco. They were also more likely to engage in daily poly use (two or more products), though their general prevalence of current use of multiple products did not vary from those living in urban locales.

An analysis of data collected in Wave 1 of the PATH study found a greater prevalence among rural adults relative to urban adults for cigarettes and smokeless tobacco, with no difference in ENDS,¹² but this was based on the binary distinction of urban-rural, which is less nuanced than the 4-level variable we have used. Another study with a simple urban-rural comparison found higher use likelihood for cigarettes, chewing tobacco, and snuff among rural populations nationally.⁵¹ The findings reported here further elucidate the importance of examining variations in rurality and urbanicity. And, they suggest that while rural residents were no more likely to report current use combustible products and to engage in poly use, and were less likely to currently use ENDS than those in urban places, rural residents may engage in more intense tobacco use patterns for the products they use; specifically, we found rural residents had a greater likelihood of daily and established combustible product use, and a greater likelihood of daily poly use, compared with urban residents. These differences in use behavior have important health implications because prior studies have found that cardiovascular disease risk increases with increasing cigarette use intensity⁵² and, compared with more frequent smoking, those who reduce their cigarette consumption have a lower risk of poor health outcomes.⁵³

Finally, use patterns over time (averaged over location) varied by product category and by use definition. Interestingly, for current use outcomes, there was an increase from W4 to W5 in use likelihood for combustible products, ENDS, and current poly use. For these three product categories plus smokeless use, likelihood of use was lower in W4 (referent) compared to W6. This latter finding may have been at least partly associated with the much larger proportion of interviews completed via telephone for W6, with social desirability perhaps influencing some responses relative to the usual mode of data collection.⁵⁴ The comparison of W4 to W7 suggested a decrease in likelihood of current combustible, smokeless and poly use over this time span, while ENDS use increased over the same period.

Strengths and limitations

One strength of this study is its analysis of a large, nationally representative sample of adults who use and do not use tobacco/nicotine products. Another is the inclusion of a more granular measure of location type, which has provided the ability to discern differences that would have been lost in the urban-rural dichotomy. An additional strength is the longitudinal design, demonstrating relative stability in use likelihood for most products over a 4-assessment timeline that was completed over the course of 7 years; the time-related changes in use pattern are consistent with what we know from other sources: in the U.S., cigarette use is declining over time while ENDS use is increasing.^55–57

The primary study limitation that only about half (55.2%) of those who completed Wave 4 of the PATH study also completed Waves 5–7, reducing the size of the analytic sample and potentially introducing bias. This concern is somewhat mitigated by the fact that the analysis used replicate weights, thus making the findings more representative. Lastly, the survey mode varied during Wave 6 due to the COVID-19 pandemic, when 68% of participants completed the survey by telephone vs. audio computer-assisted self-interviews completed in the participant’s home. Social desirability bias from telephone interviews may have impacted the findings as prior research has found that youth participants were less likely to report ENDS use during this time period,⁵⁴ though this phenomenon has not been evaluated among adults.

Conclusions

While most location-specific comparisons of tobacco use likelihood focus on differences between urban and rural areas, our study underscores that use patterns may vary within these binary categories. Compared to those living in urban areas, suburban residents were less likely to currently use combustible products, smokeless tobacco, and two or more products, while those living in towns were more likely to currently use combustible tobacco. Given that these three location categories would largely be classified as urban in PATH, this demonstrates variations within that category that may not be fully explored using the traditional dichotomy. Similarly, we found that those living in rural areas were not more likely than urban residents to currently use combustible products and they were less likely than those in urban areas to currently use ENDS. Our study has demonstrated that intensity of use varies by location, with an indication that relative to those living in urban areas, suburban residents tended to be less likely to use daily or in an established pattern, while those in rural areas and towns tended to be more likely to do so. A better understanding of the link between geographic location and likelihood of tobacco product use and use intensity provides the information needed to tailor public health approaches to reduce health inequities for those who are most at risk for tobacco use and its sequelae.