1 INTRODUCTION

There has been a stark shift in the workspace from office to home in the U.S. to prevent the spread of the coronavirus disease 2019 (COVID-19) pandemic.1, 2 While this shift may have temporarily affected a select group of employed people, telecommuting may become the norm after the public health crisis is mitigated, leaving a more permanent impact on work-related behavior of individuals and communities. This is in view of many employers considering work-from-home arrangements for their employees permanently, with a majority of teleworkers preferring to work from home.3-5

The scope of occupational sectors offering work from home arrangements typically include management, education, computer, finance, and law, and exclude farm, construction and production. Recent studies examining the changes in smoking behavior following the onset of the COVID-19 pandemic in Belgium, Italy, Japan and the United Kingdom observed that people who work from home experienced increased smoking prevalence during the pandemic.6-9 While a few studies in the U.S. have investigated the changes in tobacco use during the pandemic compared to the pre-pandemic period, none of them examined the changes in smoking behavior specifically among telecommuters.10-15 Only one previous study, limited to a non-representative sample of U.S. adults from 2010 to 2011, examined change in smoking behavior, but found greater risk of tobacco use among non-telecommuters in the U.S.16

It can be hypothesized that concerns about secondhand or thirdhand smoke exposure and/or residential smoke-free policies may promote cessation or reduced consumption among smokers who work from home. Conversely, the absence of workplace smoke-free policies at home may potentially lead to smoking initiation, relapse, or increased consumption. Increased consumption or reduced cessation associated with telecommuting would add to potential increases in smoking triggered by COVID-19 when individuals may have used smoking as a coping mechanism, or stockpiled cigarettes in fear of supply bottlenecks.17 The net effect of working from home on smoking behavior is dependent on the relative strength of the smoke-free environments at both workplace and home experienced by individual workers.

In this paper, we examined whether telecommuting is associated with smoking behavior among U.S. wage and salary workers based on a national level survey from the pre-COVID-19 pandemic period. Findings may inform evidence-based public health interventions aimed at reducing smoking disparities and organizational health policies focused on employee well-being in the post-pandemic era.

2 METHODS In a cross-sectional study design, self-reported responses from U.S. wage and salary workers (aged 16–64 years) were used from the Tobacco Use Supplement of the Current Population Survey (TUS-CPS) July 2018 wave linked to the Leave and Job Flexibilities Module of the 2018 American Time Use Survey (ATUS), matched on household and individual identifiers. The following two-step model of smoking prevalence and intensity was used for estimation: (1) (2)where the outcome variables are smoking probability, SPi =1 if worker i smoked at least 100 cigarettes in their lifetime and currently smokes every day or somedays, and 0 otherwise; and smoking intensity, SIi = average number of cigarettes smoked per day by a current smoker i.

The exposure variable, telecommuting frequency, Xi, was coded ‘0’, indicating unable to work from home and 1–4 representing working from home ‘less than once a month’, ‘once a month’, ‘once every two weeks’, and ‘at least once a week’, respectively. We included an occupation variable, Wi, to differentiate the ‘structural zeros’ in Xi due to the nature of occupations that are typically not amenable to telecommuting (e.g., services, sales, office and administrative support, farming, construction) from the ‘random zeros’ in occupations that are amenable to telecommuting but for which employees did not report working from home during the survey period (e.g., management, business, finance, professional occupations).18, 19 Wi was coded ‘0’ if an employee's occupation belonged to the categories typically not amenable to telecommuting (more than 80% respondents in these occupations reported in the survey not being able to work from home) and 1 if an employee's occupation belonged to the categories amenable to telecommuting. The sample statistics by employees’ major occupation groups and telecommuting frequency are provided in Table A1 in the Supplementary File.

The coefficients β1 and β2 represent trait effect of telecommuting on the responses, SP and SI, all other things being equal, while the coefficients α1 and α2 measure the changes in responses per unit increase in Xi (telecommuting frequency) within the group whose jobs are amenable to telecommuting.20 The association between telecommuting status and smoking outcomes was tested using a composite linear hypothesis, H01:α1 = 0, β1 = 0 for smoking prevalence and H02:α2 = 0, β2 = 0 for smoking intensity. Statistical tests were 2-sided and considered significant at α = 0.10 due to the small sample size of the study, and the fact that, in small samples, meaningful results may fail to appear statistically significant at the conventional level of significance of 0.01 or 0.05.21

Covariates Zi included participants’ work status (part-time, full time, hours vary), socio-demographic characteristics (e.g., age, race/ethnicity, marital status, presence of children ages 0–5 in the household, annual family income, educational status), and regions from the CPS; smoke-free air policies from the State Tobacco Activities Tracking and Evaluation (STATE) System; and average cigarette price per pack (dollars) from the Tax Burden on Tobacco database corresponding to the states of residence.22, 23 The random error terms in the two equations are represented by e1i and e2i. Equations (1) and (2) were estimated using multivariable logit regression and generalized linear regression respectively. The regressions were weighted to be generalizable to the study population and to account for complex sampling design using replicate weights from the CPS (in STATA Version 15). Analyses were stratified by sex due to differences in preference for workplace flexibility between men and women employees.24 The analytical sample with non-missing values for all variables used in the analysis comprised 1,390 employees (690 men and 700 women). The study population and the stages of sample selection are shown in Figure 1.

1 Study population and sample selection

3 RESULTS

Nearly 12% of respondents were current smokers and smoked 14.6 cigarettes per day on average. The prevalence of smoking was higher among men (12.6%) than among women (10.9%). The intensity of smoking among those who were smokers was at the same level of 14.6 cigarettes per day among men and women. Although half of the employees (51.7% overall; 48.4% among men; 54.9% among women) were employed in occupations that are amenable to telecommuting, only 19.2% (18.0% among men; 20.1% among women) reported working from home in varied frequencies ranging from less than once a month to at least once a week. Two-thirds of the employees were full-time workers, about a tenth were part-time workers, and a quarter reported varying work hours. The summary statistics of all the covariates including demographic characteristics, socio-economic status, state-level tobacco control policy status and region of residence of respondents are provided in Table 1.

TABLE 1. Sample characteristics of U.S. wage and salary workers ages 16–64, 2018 Characteristics All Men Women Mean [95%–CI] Mean [95%–CI] Mean [95%–CI] Smoking outcomes Smoking prevalence (%) 11.7 10.0–13.4 12.6 10.1–15.1 10.9 8.5–13.2 Smoking intensity (number of cigarettes smoked daily) 14.6 13.3–15.9 14.6 12.8–16.4 14.6 12.6–16.5 Telecommuting frequency (%) 0 - None 80.9 78.8–82.9 82.0 79.2–84.9 79.7 76.7–82.7 1 - Less than once a month 3.1 2.2–4.0 3.8 2.3–5.2 2.4 1.3–3.6 2 - Once a month 2.9 2.0–3.8 3.6 2.2–5.0 2.1 1.1–3.2 3 - Once every 2 weeks 3.1 2.2–4.0 2.9 1.6–4.2 3.3 2.0–4.6 4 - At least once a week 10.1 8.5–11.7 7.7 5.7–9.7 12.4 10.0–14.9 Total 100.0 100.0 100.0 Occupations amenable to telecommuting (%) 51.7 49.0–54.3 48.4 44.7–52.1 54.9 51.2–58.6 Work status (%) Full-time 66.1 63.6–68.6 66.4 62.8–69.9 65.8 62.3–69.4 Part-time 9.1 7.5–10.6 8.1 6.1–10.1 10.0 7.8–12.2 Hours vary 24.8 22.5–27.1 25.5 22.2–28.8 24.1 21.0–27.3 Total 100.0 100.0 100.0 Women (%) 50.3 47.7–53.0 - -–- - -–- Age (years) 43.3 42.7–44.0 42.6 41.7–43.5 44.0 43.1–44.9 Presence of children (ages 0–5) in household (%) 14.9 13.0–16.8 16.1 13.3–18.8 13.7 11.2–16.3 Race/ethnicity (%) White, non-Hispanic 68.8 66.4–71.3 69.0 65.5–72.4 68.7 65.3–72.1 Black, non-Hispanic 11.6 10.0–13.3 10.0 7.7–12.2 13.3 10.8–15.8 Hispanic 12.9 11.1–14.6 12.9 10.4–15.4 12.8 10.4–15.3 Other 6.6 5.3–7.9 8.1 6.1–10.1 5.1 3.5–6.8 Total 100.0 100.0 100.0 Marital status (%) Married 46.7 44.1–49.3 51.0 47.3–54.8 42.4 38.8–46.1 Widowed 2.3 1.5–3.1 1.4 0.6–2.3 3.1 1.8–4.4 Divorced/separated 18.9 16.9–21.0 14.3 11.7–17.0 23.4 20.3–26.6 Never married 32.1 29.6–34.5 33.2 29.7–36.7 31.0 27.6–34.4 Total 100.0 100.0 100.0 Educational status (%) Less than high school 6.8 5.4–8.1 7.2 5.3–9.2 6.3 4.5–8.1 High school or equivalent 23.5 21.2–25.7 23.3 20.1–26.5 23.6 20.4–26.7 Some college education 28.3 26.0–30.7 29.7 26.3–33.1 27.0 23.7–30.3 College graduate 25.0 22.8–27.3 23.3 20.2–26.5 26.7 23.4–30.0 More than college 16.4 14.4–18.4 16.4 13.6–19.1 16.4 13.7–19.2 Total 100.0 100.0 100.0 Annual family income in relation to federal income poverty line (FIPL) (%) Below 100% of FIPL 5.5 4.3–6.7 5.5 3.8–7.2 5.4 3.7–7.1 100% to 200% of FIPL 11.5 9.8–13.2 11.2 8.8–13.5 11.9 9.5–14.3 200% to 400% of FIPL 28.3 25.9–30.6 31.2 27.7–34.6 25.4 22.2–28.7 Above 400% FIPL 54.7 52.1–57.4 52.2 48.4–55.9 57.3 53.6–61.0 Total 100.0 100.0 100.0 Tobacco control policy variables Average cigarette price per pack ($) 7.21 7.12–7.28 7.22 7.11–7.34 7.19 7.07–7.30 Smoking ban in Government worksite (%) 78.8 76.7–81.0 78.6 75.4–81.6 79.1 76.1–82.2 Private worksite (%) 73.1 70.8–75.4 73.2 69.9–76.5 73.0 69.7–76.3 Restaurant (%) 68.6 66.2–71.1 69.0 65.5–72.4 68.3 64.8–71.7 Bars (%) 58.0 55.4–60.6 60.0 56.2–63.5 56.1 52.4–59.8 Region (%) Northwest 16.3 14.4–18.3 16.8 14.0–19.6 15.9 13.1–18.6 Midwest 25.8 23.5–28.1 25.8 22.5–29.1 25.7 22.5–29.0 South 35.0 32.4–37.5 34.3 30.8–37.9 35.6 32.0–39.1 West 22.9 20.7–25.2 23.0 19.9–26.2 22.9 19.7–26.0 Total 100.0 100.0 100.0 Number of observations 1390 690 700 Note The mean and 95% CI of smoking intensity are based on responses from 105 current smokers. The sum of percentages of groups for each variable may not be exactly 100% due to rounding-off of individual group percentages. The percentages for the smoking ban variables for government worksite, private worksite, restaurant, and bar would not add up to 100% because these characteristics of tobacco control policy status of states are not mutually exclusive. One state can have smoking ban in one or more sites. Abbreviation: CI, confidence interval.

Smoking probability was negatively associated with employment in telecommuting amenable occupations, but in stratified analyses was significant among women but not men. The overall negative association was stronger with a higher frequency of telecommuting (Table 2). On average, the probability of smoking by an employee in a telecommuting amenable occupation was 0.52 percentage points lower than those in occupations not amenable to telecommuting and the probability decreased by 0.21 percentage points for each unit increase in telecommuting frequency. The combined effect of occupation and telecommuting frequency on smoking prevalence was negative and grew stronger in a dose-response fashion with a higher frequency of telecommuting (from −0.73 to −1.36 percentage points). Similar negative associations of smoking status with telecommuting frequency, employment in telecommuting amenable occupation, and the combined effects were found in women employees. Among men, the dose response negative association was observed between smoking status and telecommuting frequency, while the association with employment in telecommuting amenable occupations was not statistically significant.

TABLE 2. Estimates of the association of smoking prevalence and smoking intensity with telecommuting status of U.S. wage and salary workers ages 16–64 Smoking prevalence All Smoking prevalence, % (N) Men Smoking prevalence, % (N) Women Smoking prevalence, % (N) Marginal effect, % points (95% CI) p Marginal effect, % points (95% CI) p Marginal effect, % points (95% CI) p Telecommuting frequency −0.21 (−0.23, −0.18) <.001 −0.76 (−0.80, −0.72) <.001 −0.03 (−0.06, −0.00) .045 Employment in telecommuting amenable occupation −0.52 (−0.58, −0.45) <.001 0.01 (−0.09, 0.11) 0.862 −2.40 (−2.48, −2.31) <.001 Combined effect

0 – None

12.0% (1124) 12.9% (566) 11.1% (558) 1 – Less than once a month (%) 23.2% (43) −0.73 23.1% (26) −0.76 23.5% (17) −2.43 2 – Once a month (%) 7.5% (40) −0.94 8.0% (25) −1.52 6.7% (15) −2.46 3 – Once every 2 weeks (%) 11.6% (43) −1.15 10.0% (20) −2.28 13.0% (23) −2.49 4 – At least once a week (%) 7.1% (140) −1.36 7.5% (53) −3.04 6.9% (87) −2.52 Joint test of marginal effects of telecommuting frequency and employment in telecommuting amenable occupations = 0 Chi-square(2) = 734.05 <.001 Chi-square(2) = 1493.52 <.001 Chi-square(2) = 3995.75 <.001 Smoking intensity Telecommuting frequency 0.48 (−0.84, 1.80) 0.48 2.02 (−0.36, 4.42) 0.10 1.88 (−0.82, 4.59) 0.17 Employment in telecommuting amenable occupation −3.39 (−6.38, −0.41) 0.03 −0.36 (−5.79, 5.07) 0.90 −4.30 (−9.63, 1.03) 0.11 Combined effect −3.39 -