A Longitudinal Analysis of Respiratory Illness and Tobacco Use Transitions

Abstract

Introduction:

Among individuals with chronic respiratory conditions, transitions between patterns of tobacco product use are not well understood. This study examines how transitions, including quitting altogether, differ over time between those who do or do not have chronic respiratory conditions.

Methods:

Data from youth and adult participants of the longitudinal Population Assessment of Tobacco and Health (PATH) Study (2013–2018) were analyzed. Youth ages 12–17 were included if they had aged into the adult sample by Wave 4. Stratified polytomous regression models built under a first-order Markov assumption modeled the probability of transitioning between different states/patterns of tobacco product use (exclusive current e-cigarette use, exclusive current combustible tobacco product use, current dual use of combustible products and e-cigarettes, and no current tobacco product use) at each wave. Marginal transition probabilities were computed as a function of ever or past-year diagnosis of a respiratory condition (separately for asthma and a composite variable representing chronic bronchitis, emphysema, and/or COPD). Analyses were conducted in 2020–2021.

Results:

Most individuals, regardless of respiratory condition, maintained the same pattern of tobacco use between waves. Exclusive combustible tobacco product users, including those with or without a respiratory condition, were not likely to become exclusive e-cigarette users or to quit using tobacco entirely.

Conclusions:

Although combustible tobacco use negatively impacts management and prognosis of respiratory illnesses, combustible tobacco users who were recently diagnosed with a chronic respiratory condition were not likely to quit using tobacco. Efforts to encourage and support cessation in this medically vulnerable population should be increased.

INTRODUCTION

Individuals who use tobacco products, including combustible cigarettes and e-cigarettes, are more likely to develop chronic respiratory conditions such as asthma, chronic obstructive pulmonary disease (COPD), emphysema, and chronic bronchitis.^1–8 Further, continued use of combustible tobacco products negatively impacts the management and prognosis of respiratory illnesses,^9,10 making it critical for individuals with these conditions to abstain from using these products.

E-cigarette use may also negatively impact the management and prognosis of these conditions. While e-cigarettes may expose users to lower levels of some harmful tobacco-related constituents than conventional cigarettes, a recent review concluded that there is insufficient evidence to determine whether e-cigarettes pose less harm to respiratory health than combustible tobacco products, citing studies that found measurable adverse biologic effects on the cellular, organ, and population levels.¹¹ Nonetheless, many U.S. adults believe that e-cigarettes are less harmful than conventional cigarettes,¹² and many people who smoke who use e-cigarettes report doing so in an attempt to quit or reduce smoking or to mitigate the health risks of cigarettes.^13–16 However, despite the importance of quitting tobacco use completely for users with chronic respiratory conditions, it is unknown whether patterns of transition between combustible cigarette use, e-cigarette use, dual use, and no tobacco use differ among individuals with or without these conditions.

This study uses Markov transition models to analyze data from the Population Assessment of Tobacco and Health (PATH) Study, a nationally-representative, longitudinal cohort study of tobacco use among U.S. youth and adults, to prospectively characterize transitions between different tobacco use patterns among individuals with and without chronic respiratory conditions.

METHODS

Study Sample

Using a stratified, address-based, area-probability sampling design, the PATH study collects data on the civilian, non-institutionalized U.S. population aged 12 years and older.¹⁷ This study analyzed public-use data from 26,072 individuals who participated in all of the first 4 waves of the PATH Study: Wave 1 (W1, collected 2013–2014), Wave 2 (W2, 2014–2015), Wave 3 (W3, 2015–2016), and Wave 4 (W4, 2017–2018). Participants from the youth cohort were included in the analytic population if they had aged into the adult cohort (i.e., turned age 18 years) by W4. Further restrictions applied to define the analytic population are detailed below. Analyses were conducted in 2020–2021.

The NIH IRB determined this study to be exempt from review because data are de-identified. Adult respondents and parents/legal guardians of youth respondents provided informed consent; youth respondents provided assent. Results are presented according to STROBE reporting guidelines.

Measures

Participants were asked whether they currently use any of the following: cigarettes, e-cigarettes, cigars (including traditional cigars, filtered cigars, and cigarillos), pipe tobacco, hookah, smokeless tobacco and snus. For adults, current cigarette use was defined as having smoked at least 100 cigarettes in their lifetime and now smoking every day or some days. Current use of other tobacco products was defined as using the product every day or some days. For youth, current use of all tobacco products was defined as use in the past 30 days without the lifetime use threshold. Based on these items, a variable representing current tobacco use for each wave was created with the following categories: (1) exclusive current e-cigarette use, (2) exclusive current combustible tobacco product use (i.e., cigarette, cigar, hookah, pipe tobacco), (3) current dual use of combustible products and e-cigarettes, (4) no current use of any tobacco product (including never users and former users; called “non-users”). Because of the relatively small number of smokeless tobacco and snus users, individuals who reported current smokeless tobacco use at any wave were excluded (n=1,775).

For each respiratory condition, “ever diagnosis” was ascertained for all adult participants at W1 and for new adult participants at W2–W4 using the item, “Has a doctor, nurse, or other health professional ever told you that you had (asthma/chronic bronchitis/emphysema/COPD)?” “Diagnosis in the past year” for each respiratory condition was ascertained at W2–W4 for continuing adult respondents using the item “In the past 12 months, has a doctor, nurse, or other health professional told you that you had (asthma/chronic bronchitis/emphysema/COPD)?” Parents/guardians provided responses to identical questions about their child/children’s asthma and chronic bronchitis diagnoses. Data on emphysema and COPD were not collected in the PATH Youth/Parent survey. Youth who aged into the adult group were assumed not to have been diagnosed with emphysema or COPD prior to joining the adult cohort.

Based on a participant’s responses to the illness-specific questions, a composite variable was created to represent having ever received a diagnosis for any of the following respiratory conditions at each wave: chronic bronchitis, emphysema, or COPD. At W1, this variable represented an ever diagnosis of any of the 3 conditions. At W2–W4, this variable represented an ever diagnosis (for aged-up youth taking their first adult survey) or a diagnosis in the previous 12-months (for continuing respondents). Asthma, a respiratory condition that is not necessarily caused by smoking and that is prevalent in all age groups,⁸ was analyzed separately from other respiratory conditions but was similarly defined according to an ever or in the past 12-month diagnosis. In contrast to asthma, smoking is a major cause of chronic bronchitis, emphysema, and COPD,⁸ conditions that are more prevalent in older age groups.¹⁸ Participants with missing data on any of the 3 conditions were assigned a missing value for the composite variable and subsequently excluded from the models (diagnosis of a respiratory condition had n=262, 81, and 60 missing observations for W1, W2, and W3, respectively; diagnosis of asthma had n=80, 82, and 61 missing observations for W1, W2, and W3, respectively).

Demographic variables included age at W1 (12–17, 18–24, 25–54, ≥55 years), sex (male, female), and race and ethnicity (Black, Hispanic, Other [including American Indian or Alaska Native, Asian Indian, Chinese, Filipino, Japanese, Korean, Vietnamese, Other Asian, Native Hawaiian, Guamanian or Chamorro, Samoan, Other Pacific Islander, and multiple races and ethnicities], White). The imputed versions of these variables, provided by the PATH Study, were used. In addition, time-varying variables included education level (youth participant/not yet finished with education, adult with some high school to a high school diploma, adult with some college or more), likelihood of having substance use disorder, likelihood of having an internalizing disorder, and likelihood of having an externalizing disorder. Behavioral health factors were measured using the Global Appraisal of Individual Needs-Short Screener (GAIN-SS). Following GAIN-SS guidelines,¹⁹ scores were calculated for substance use, internalizing behaviors (e.g., depression, anxiety, trauma), and externalizing behaviors (e.g., attention deficits, hyperactivity, impulsivity). For each, individuals were classified into 2 groups: low/moderate vs high likelihood of having a diagnosis.

Statistical Analysis

The probability of transitioning between different states (S_t) of tobacco product use (exclusive current e-cigarette use, exclusive current combustible tobacco product use, current dual use of combustible products and e-cigarettes, no current tobacco product use) across waves (t=1, 2, 3, 4) was modeled using stratified polytomous regression. Models were built under a first-order Markov assumption (i.e., assuming that the state S_t at wave t depends only on S_t−1 but not on the states S_t−j, where j>1). Specifically, let p_rs(t) = P(S_t = s|S_t−1 = r) denote the transition probability of changing tobacco product use behavior from S_t−1 = r at wave t − 1 to S_t = s at wave t, where:

$$p_{rs}(t)=\frac{\text{exp}\left\{{\mathit{X}}^T{\mathit{\beta }}_{rs}(t)\right\}}{1+{\sum }_{s\ne r}\text{exp}\left\{{\mathit{X}}^T{\mathit{\beta }}_{rs}(t)\right\}}.$$ (1)

In equation (1), X and β_rs(t) denote a vector of W1 covariates and corresponding association parameters, respectively. Equation (1) was also fit allowing X to vary between waves (i.e., using X(t)). However, as the results were very similar, this study focuses on the results with X from W1, and results with X(t) are in the Appendix. At each transition, all cases with respiratory health and tobacco use information for both consecutive waves were included, even if they were missing this information at other timepoints. Models were estimated at each wave t stratified by tobacco product use at wave t−1. Stratification allows transition probabilities to differ across states of tobacco product use and time. A visual representation of the Markov models can be found in Appendix Figure 1. A separate set of models was built for the composite respiratory condition variable and for asthma. For both, models were estimated with 2 different adjustments: a full set of covariates (sex, age group, race and ethnicity, educational attainment, and likelihood of having internalizing, externalizing and substance use disorders) and a minimal set of covariates (sex, age group, and race and ethnicity). All models were restricted to respondents with complete data on the model variables.

Marginal transition probabilities were computed from the transition probabilities as a function of ever (t=2) or past-year (t=3, 4) diagnosis of the respiratory condition by averaging over other model covariates. All analyses incorporated PATH W1–W4 All-Wave Weights, which account for loss-to-follow-up and non-response. SEs and 95% CIs of estimated model coefficients were computed using the balanced repeated replication method²⁰ with a Fay’s value of 0.3; models were implemented using the functions ‘svrepdesign()’ and ‘svymultinom()’ in the R packages ‘survey’ and ‘svrepmisc,’ respectively. These were subsequently transformed to the probability scale to obtain CIs for the estimates of the transition probabilities p_rs and the corresponding marginal probabilities.

RESULTS

Among 24,111 participants included in analyses, 9,931 were youth (12–17 years) and 14,178 were adults (aged ≥18 years) at W1 (Table 1). A majority of the population was female (54.1%) and White (64.2%), and 56.9% were adults with at least some college education. The numbers of respiratory condition diagnoses and of individuals reporting use of each tobacco product type at each wave are shown in Appendix Table 1.

Table 1.

Description of the Sample^a (N=24,111) at Wave 1

Characteristic	Unweighted N	Weighted % (95% CI)
Age category, years
12 to 17	9,931	5.0 (4.9, 5.0)
18 to 24	4,481	12.1 (12.0, 12.2)
25 to 54	5,235	49.6 (49.1, 50.2)
≥55	4,462	33.3 (32.7, 33.8)
Missing	2	0.0 (0.0, 0.0)
Sex
Male	10,883	45.9 (45.7, 46.1)
Female	13,228	54.1 (53.9, 54.3)
Missing	0	0.0 (0.0, 0.0)
Race and ethnicity
Black	3,807	12.0 (11.9, 12.1)
Hispanic	4,915	15.9 (15.8, 16.0)
Otherb	1,868	7.9 (7.7, 8.0)
White	13,521	64.2 (64.0, 64.5)
Missing	0	0.0 (0.0, 0.0)
Level of education
Youth; not yet finished with education	4,481	5.0 (4.9, 5.0)
Adult; less than high school to high school degree	8,107	37.7 (37.5, 38.0)
Adult; some college or more	11,443	56.9 (56.6, 57.1)
Missing	80	0.4 (0.3, 0.6)
Substance use disorder likelihoodc
Low-to-moderate likelihood of having a diagnosis	22,435	95.2 (94.9, 95.5)
High likelihood of having a diagnosis	1,381	3.6 (3.3, 3.9)
Missing	295	1.2 (1.0, 1.4)
Internalizing disorder likelihoodc
Low-to-moderate likelihood of having a diagnosis	16,419	77.0 (76.2, 77.7)
High likelihood of having a diagnosis	7,424	21.8 (21.1, 22.6)
Missing	268	1.2 (1.0, 1.5)
Externalizing disorder likelihoodc
Low-to-moderate likelihood of having a diagnosis	20,060	89.3 (88.8, 89.9)
High likelihood of having a diagnosis	3,687	9.0 (8.5, 9.5)
Missing	364	1.7 (1.4, 2.0)

^aAnalytic sample includes individuals who completed Waves 1–4 of PATH, had tobacco use data available for at least 2 consecutive waves, and were in the adult sample by Wave 4.

^bIncludes American Indian or Alaska Native, Asian Indian, Chinese, Filipino, Japanese, Korean, Vietnamese, Other Asian, Native Hawaiian, Guamanian or Chamorro, Samoan, Other Pacific Islander, and multiple races and ethnicities.

^cDetermined using the GAIN-SS.

Appendix Figures 2 and 3 show results from the fully-adjusted models for respiratory illness and asthma, respectively. Models adjusted for the full set of covariates, including time-varying covariates, yielded similar results to those adjusted for the minimal set. Thus, results reported here focus on the minimally-adjusted models.

Figure 1 shows results from minimally-adjusted models using the composite respiratory illness endpoint. Among W1 exclusive current e-cigarette users, those with a respiratory illness were most likely to either quit all tobacco products (35.8%) or to remain exclusive e-cigarette users by W2 (31.2%). W2 exclusive e-cigarette users with a respiratory illness were most likely to remain exclusive e-cigarette users (43.3%) or quit all tobacco products by W3 (32.8%), while W3 exclusive e-cigarette users with a respiratory illness were most likely to remain exclusive e-cigarette users (51.4%) or transition to dual combustible tobacco and e-cigarette use by W4 (34.2%). Exclusive e-cigarette users without a respiratory illness diagnosis at W1 were most likely to remain exclusive e-cigarette users at W2 (46.0%). A similar pattern was observed at subsequent transitions: e-cigarette users without a respiratory illness were most likely to remain in the same tobacco user group (42.1% for the W2-to-W3 transition and 44.0% for the W3-to-W4 transition).

Figure 1.

Marginal transition probabilities between tobacco user groups between consecutive waves by combined respiratory diagnosis endpoint, including chronic bronchitis, emphysema, and COPD.

Notes: Transition probabilities presented here were averaged over model covariates, which included sex, age group, and race and ethnicity. Sample sizes exclude those with missing information on respiratory diagnoses. Sample sizes for transitions are as follows: Wave 1 to Wave 2 N=22,945; Wave 2 to Wave 3 N=23,527; Wave 3 to Wave 4 N=23,801. The composite respiratory illness endpoint includes chronic bronchitis, emphysema, and COPD. At Wave 1, “respiratory diagnosis” refers to an ever diagnosis. For waves 2–4, this refers to a past-12-month diagnosis for returning participants or an ever diagnosis for new adult participants.

Among exclusive current combustible tobacco users, transition patterns were consistent across all waves, irrespective of a respiratory illness diagnosis. Exclusive combustible tobacco users were most likely to continue combustible tobacco use at subsequent waves, although marginal probabilities were slightly higher among those with a respiratory illness diagnosis compared to those without an illness. Exclusive combustible tobacco users with and without respiratory conditions were unlikely to become exclusive e-cigarette users: transition probabilities were ≤2.1% across waves, irrespective of a respiratory illness diagnosis. While generally low, transition probabilities from combustible tobacco use to no use (i.e., quitting) were lower among those with a respiratory illness diagnosis compared to those without (W1-to-W2: 9.2% vs 17.6%; W2-to-W3: 9.1%–15.0%; W3-to-W4: 9.9% vs 14.8%). Transition probabilities to dual use were also relatively low across all waves, though they were marginally higher among those with a respiratory illness diagnosis compared to those without (W1-to-W2: 15.7% vs 11.9%; W2-to-W3: 8.7% vs 5.9%; W3-to-W4: 6.2% vs 5.4%, respectively).

Regardless of a respiratory illness diagnosis at W1, current dual users were most likely to continue using both products or to transition to exclusive combustible tobacco use at all subsequent waves, with the highest probability of becoming an exclusive combustible tobacco user. Across all waves and irrespective of a respiratory illness diagnosis, transition probabilities from dual use to exclusive e-cigarette use were quite low (≤6.6%) and tended to be lower among those with a respiratory illness diagnosis (≤4.7%).

Across all waves and regardless of a respiratory illness, non-users of tobacco products were most likely (≥91.5%) to remain non-users at subsequent waves. A greater proportion of non-users at W2 and W3 with a respiratory illness diagnosis reported initiating combustible tobacco use at subsequent waves than non-users without a diagnosis.

Figure 2 shows results from minimally-adjusted models using the asthma endpoint. Among exclusive current e-cigarette users with asthma, those at W1 and W2 were most likely to remain exclusive e-cigarette users at subsequent waves (54.6% and 45.3%, respectively). Similar patterns were observed among those without asthma, with somewhat lower transition probabilities (42.4% and 41.9% at W1 and W2, respectively). At W3, exclusive e-cigarette users with asthma were most likely to remain exclusive e-cigarette users (30.0%) or to quit using tobacco products entirely (30.1%). Exclusive e-cigarette users without asthma were most likely to remain exclusive e-cigarette users (45.6%).

Figure 2.

Marginal transition probabilities between tobacco user groups between consecutive waves by asthma diagnosis.

Notes: Transition probabilities presented here were averaged over model covariates, which included sex, age group, and race and ethnicity. Sample sizes exclude those with missing information on asthma diagnoses. Sample sizes for transitions are as follows: Wave 1 to Wave 2 N=23,127; Wave 2 to Wave 3 N=23,526; Wave 3 to Wave 4 N=23,800. At Wave 1, “asthma diagnosis” refers to an ever diagnosis. For waves 2–4, this refers to a past-12-month diagnosis for returning participants or an ever diagnosis for new adult participants.

Among exclusive current combustible tobacco users, similar patterns were observed for asthma as for the composite respiratory illness measure. Most users continued combustible tobacco product use at subsequent waves. Irrespective of an asthma diagnosis, the probability of transitioning to exclusive e-cigarette use was very low (≤1.8%). Compared to the composite respiratory illness models, differences in transition probabilities for those with and without an asthma diagnosis were less marked.

As with the composite respiratory illness measure, similar transition patterns were observed across all waves and among current dual product users with and without an asthma diagnosis, with about 40% continuing use of both a combustible product and an e-cigarette and about 40%–50% transitioning to exclusive combustible tobacco use. Again, transition probabilities were quite low for exclusive e-cigarette use (≤6.8%), and there were no observable differences by asthma diagnosis.

As with the composite respiratory illness measure, non-users of tobacco products were most likely (>95%) to remain non-users at subsequent waves. Results were consistent across waves and among users with and without an asthma diagnosis.

DISCUSSION

Given that combustible tobacco product use negatively impacts the management and prognosis of respiratory illnesses,^9,10 it is important that individuals with these conditions abstain from combustible tobacco use and that clinicians promote and assist with cessation attempts among individuals who smoke. However, among those with a self-reported diagnosis of chronic bronchitis, emphysema, COPD, or asthma, this study did not observe large probabilities of transitioning from exclusive combustible tobacco use to no use of any tobacco product. In addition, between 10.4% and 35.8% of exclusive e-cigarette users with a chronic respiratory condition became dual-product users at the following wave. These findings have important clinical and population health implications. While health concerns may increase motivation to quit and thus increase quit attempts,²¹ motivation to quit does not necessarily result in abstinence after a quit attempt.²² Indeed, combustible tobacco product users with emphysema or COPD, which take years to develop, may be longer-term or heavier users and therefore may find it more difficult to quit.^23–25 However, quitting smoking is the only established intervention for reducing the loss of lung function among patients with COPD, and the 2020 Surgeon General’s Report on Smoking Cessation recommends that clinicians should counsel asthma patients who smoke on cessation.²⁶ The importance of cessation is further highlighted by current and former smokers being at increased risk of severe COVID-19 and death.²⁷ Therefore, clinicians should intensify their efforts to provide evidence-based cessation treatment to patients who report any form of tobacco use (e.g., brief clinical interventions such as the 5As and FDA-approved pharmacologic interventions)—particularly patients with chronic respiratory conditions.²⁶

While some may hypothesize that dual-product use could be a transitory stage between exclusive combustible tobacco product use and either exclusive e-cigarette use or complete abstinence, a majority of dual-product users in this study, irrespective of a respiratory condition, maintained their use of both product types or transitioned to other harmful patterns of tobacco use (e.g., exclusive combustible tobacco use). This is concerning because the morbidity and mortality associated with smoking is well established, and also because emerging evidence demonstrates an increased net toxicant exposure among dual users of cigarettes and e-cigarettes,²⁸ as well as immunological, respiratory, and cardiovascular harms from e-cigarette aerosol exposure.^11,28

Surprisingly, the proportion of non-users who had initiated combustible tobacco product use by subsequent waves was higher among those with chronic bronchitis, emphysema, or COPD than among those without. To reinforce prevention efforts, future research should seek to understand why patients with chronic respiratory conditions might be differentially initiating combustible tobacco product use.

This study has several strengths. It is the first to explore tobacco use trajectories over time among individuals with chronic respiratory conditions, a timely and important topic in tobacco control research. Using sophisticated statistical models, this study analyzed data from the nationally-representative and longitudinal PATH Study, which allowed it to draw conclusions generalizable to the U.S. population. Further, these conclusions are strengthened by the consistent results of sensitivity analyses controlling for a more extensive list of covariates, including time-varying covariates.

Limitations

There are some limitations to this study. First is the Markov assumption, which dictates that only the current state is considered when modeling transitions. Therefore, transition probabilities presented here do not account for individuals’ behavior in the more distant past. This study was also limited by the small number of chronic bronchitis, emphysema, and COPD diagnoses reported at each wave. These conditions were combined for analysis, but future research should explore the relationship between each individual respiratory condition and tobacco use transitions. Finally, the models presented here do not account for some factors that may influence both the development of a respiratory condition and tobacco use behaviors following diagnosis, such as pack-years of smoking or intention to quit among current and former tobacco users. Future studies should explore the role of these constructs.

CONCLUSIONS

This study finds that combustible tobacco product users who have recently been diagnosed with a chronic respiratory condition are no more likely to quit than those without a recent diagnosis. More must be done to encourage and support cessation of, and abstinence from, all tobacco products, including e-cigarettes, in this medically vulnerable group of individuals.