Effect of smoke-free workplaces on smoking behaviour: systematic review

Abstract

Objective

To quantify the effects of smoke-free workplaces on smoking in employees and compare these effects to those achieved through tax increases.

Design

Systematic review with a random effects meta-analysis.

Study selection

26 studies on the effects of smoke-free workplaces.

Setting

Workplaces in the United States, Australia, Canada, and Germany.

Participants

Employees in unrestricted and totally smoke-free workplaces.

Main outcome measures

Daily cigarette consumption (per smoker and per employee) and smoking prevalence.

Results

Totally smoke-free workplaces are associated with reductions in prevalence of smoking of 3.8% (95% confidence interval 2.8% to 4.7%) and 3.1 (2.4 to 3.8) fewer cigarettes smoked per day per continuing smoker. Combination of the effects of reduced prevalence and lower consumption per continuing smoker yields a mean reduction of 1.3 cigarettes per day per employee, which corresponds to a relative reduction of 29%. To achieve similar reductions the tax on a pack of cigarettes would have to increase from $0.76 to $3.05 (€0.78 to €3.14) in the United States and from £3.44 to £6.59 (€5.32 to €10.20) in the United Kingdom. If all workplaces became smoke-free, consumption per capita in the entire population would drop by 4.5% in the United States and 7.6% in the United Kingdom, costing the tobacco industry $1.7 billion and £310 million annually in lost sales. To achieve similar reductions tax per pack would have to increase to $1.11 and £4.26.

Conclusions

Smoke-free workplaces not only protect non-smokers from the dangers of passive smoking, they also encourage smokers to quit or to reduce consumption.

What is already known on this topic

Smoke-free workplaces are associated with lower cigarette consumption per continuing smoker

What this study adds

Smoke-free workplaces reduce prevalence of smoking as well as consumption

The combined effects of people stopping smoking and reducing consumption reduces total cigarette consumption by 29%

To achieve similar results through taxation would require cigarette taxes per pack to increase from $0.76 to $3.05 in the United States and from £3.44 to £6.59 in the United Kingdom

Introduction

Passive smoking is linked with cancer, heart disease, respiratory illness¹,² and is the leading source of indoor air pollution.³ In the United States, passive smoking has been linked to the deaths of at least 53 000 non-smokers each year, about one non-smoker for each eight smokers that tobacco kills.²,⁴ By August 2001, 234 US communities had enacted local ordinances that required all workplaces to be completely smoke-free (185 communities included restaurants; American Nonsmokers' Rights Foundation, local ordinance database), and many businesses implemented voluntary policies creating smoke-free workplaces. By 1998-9, 69% of US workers employed indoors outside the home had smoke-free workplaces.⁵

Smoke-free workplaces not only protect non-smokers, they also create an environment that encourages smokers to cut back⁶ or quit. Since as early as the 1980s the tobacco industry has recognised that smoke-free workplaces have a major effect on cigarette consumption.⁷ In 1992 Phillip Morris Tobacco Company privately estimated that if all workplaces were smoke-free, total consumption would drop about 10%, through a combination of quitting and cutting down.⁸

Estimating the effect of creating smoke-free workplaces on total cigarette consumption is important because many places are implementing tobacco control programmes with money from dedicated taxes^9–18 or with funds from the settlement of lawsuits against the tobacco industry.¹⁹ There are many potential elements of such a programme, including increased taxes, legislation on smoke-free workplaces and public places, mass media education programmes, youth access laws, school based programmes, community programmes, and cessation assistance.^20–22 Outside the United States, restriction of tobacco advertising is also an option. A quantitative comparison of the effects of these interventions would enable public health policy makers to make maximum use of the (usually limited) funds available for tobacco control.

We investigated the effects of smoke-free workplaces on cigarette consumption and compared these effects with those obtained by raising taxes.

Methods

Study selection

We located studies on the effects of totally smoke-free workplaces on prevalence of smoking and daily cigarette consumption through Medline, Science Citation Index, Social Sciences Citation Index, Current Contents, and PsychInfo; from reviews⁶,²¹,²³,²⁴; and from references in the papers we located.

We included 26 studies reported in 24 papers. Worksite studies measured changes in smoking that accompanied regulations in individual workplaces assessed prospectively,^25–32 in sequential cross sections,²⁶,^33–38 or retrospectively.³⁷,^39–43 Population studies compared behaviour of smokers employed in workplaces with differing smoking policies and were cross sectional.^44–48

We excluded a further 16 studies because they evaluated policies that were not totally smoke-free^49–64 and 11 others because they did not report the desired outcomes.⁵⁵,^65–74 We excluded one study because the surveys carried out before and after the non-smoking policy were separated by eight years, so results could be contaminated by confounding factors.⁷⁵

Data analysis

We computed differences in consumption (per smoker and per employee) and prevalence before and after workplaces became smoke-free (in workplace studies) or between comparable samples with and without regulations (in population studies).

Using t tests we found that workplace and population studies did not yield different effects (P>0.2).⁷⁶ We compared the results from different study designs with analysis of variance. Sequential cross sectional studies yielded significantly smaller changes in number of cigarettes per smoker than the other study designs (P=0.003), but there were no significant differences among the study types for prevalence (P=0.081), consumption per employee (P=0.219), or relative change in consumption (P=0.143). We therefore pooled all studies in a random effects meta-analysis.⁷⁶,⁷⁷

If standard errors for consumption and prevalence change were not reported we estimated them (see table 1). We did not conduct a meta-analysis for the consumption per employee because we could not compute the standard errors necessary for the meta-analysis. There was no evidence of publication bias as assessed with funnel plots (figure).

Table 1.

 Summary of studies of effects of smoke-free workplace policies on cigarette consumption and prevalence of smoking

	Year of survey	Setting	Detail	Time to follow up (months)	Consumption per continuing smoker			Prevalence of current smokers			Consumption per employee (consumption per current smoker×prevalence)
					No of participants*	Absolute change (cigarettes/day) (SE) and method for calculating SE†		No of participants*	Absolute (%) change (SE) and method for calculating SE†		Absolute change (cigarettes/day)	Relative change (%)
Worksite studies
Prospective cohort:
Borland, 199025	1988	Government offices, Australia	—	6	170	−5.2 (0.6)/workday (method 1)		2113	−1.0 (0.9) (method 2)		−0.4	−28
Borland, 199126e	1988	Telecom company, Australia	Longitudinal data and cross sectional data at baseline	6	106	−2.6 (0.9)/workday (method 2)		620	−3.5 (1.8) (method 2)		−1.7	−27
Brigham, 199427	1989	Hospital, Baltimore, MD	—	1	34	−5.1 (0.9)/workday (method 2)		—	—		—	—
Broder, 199328	1989	University, Toronto, Canada	—	11	13	−2.2‡ (2.3) (method 3)		137	0.0 (2.8) (method 2)		−0.3	−11
Gomel, 199329	1989	Ambulance service, Sydney, Australia	—	1.5	20	−5.2 (1.3) (method 2)		—	—		—	—
Hudzinski, 199030	1986	Hospital, New Orleans, LA	—	12	—	—		684	−8.0 (1.5) (method 2)		—	—
Hudzinski, 199431	1986	Hospital, New Orleans, LA	—	18	18	−3.7‡ (1.9) (method 3)		—	—		—	—
Stillman, 199032	1988	Hospital, Baltimore, MD	—	6	355	−5.4‡§ (0.4) (method 3)		2877	−5.5 (0.6) (method 2)		−1.6	−44
Sequential cross section
Becker, 198933	1987	Hospital, Baltimore, MD	—	6	210	0.0‡§ (0.7) (method 2)		1466	−1.6 (1.8) (method 2)		−0.2	−11
Borland, 199126	1988	Telecom company, Australia	Longitudinal data and cross sectional data at baseline	18	—	—		2513	−3.1 (1.8) (method 2)		—	—
CDC, 199034	1989	Hospital, Pueblo, CO	—	12	73	−1.8¶ (1.0)/workday (method 3)		1777	−4.0 (2.1) (method 2)		−1.1	−23
Gottlieb, 199035	1988	Government offices, TX	—	6	—	—		448	−8.5 (4.0) (method 2)		—	—
Mullooly, 199036	1986	Health maintenance organisation, northwest US	—	12	2571	+0.01§ (0.95) (method 1)		—	—		—	—
Offord, 199237	1987	Hospital, Rochester, MN	Cross sectional and retrospective data	24	—	—		15817	−2.5 (0.6) (method 2)		—	—
Tsushima, 199138	1987	Hospital, Honolulu, HI	—	12	276	−2.1§‡‡ (0.8) (method 2)		1711	−1.9 (1.8) (method 2)		−0.6	−24
Retrospective:
Baile, 199139	NR	Hospital, Tampa, FL	—	4	—	—		349	−1.4 (2.3)** (method 2)		—	—
Daughton, 199240	NR	Hospital, Omaha, NE	—	5	187	−2.3 (0.8)/workday (method 2)		1076	−1.5 (1.2) (method 2)		−0.7	−21
Offord, 199237	1987	Hospital, Rochester, MN	Cross sectional and retrospective data	24	—	—		8778	−3.6 (0.4) (method 2)		—	—
Olive, 199641	1988	Hospital, Dayton, OH	—	12	117	−2.6§‡‡ (1.0)/workday (method 3)		—	—		—	—
Rosenstock, 198643	1984	Health maintenance organisation, Puget Sound, WA	—	9	—	—		434	−0.7 (1.7)** (method 2)		—	—
Stave, 199142	1989	Hospital, Durham, NC	—	9	67	−4.5‡§‡‡ (1.4) (method 2)		368	5.4 (2.1)** (method 2)		−1.8	−41
Population studies
Brenner, 199244	1987	Household survey, Germany	Reported separately for women and men, average of two given here	—	295	−1.8** (1.7) (method 3)		—	—		—	—
Farelly, 199945	1992-3	Current population surveys, US	—	—	13091††	−2.7** (0.4) (method 1)		66211††	−5.7 (.4) (method 1)		—	—
Glasgow, 199746	1993	Community intervention trial for smoking cessation, US	—	—	2884††	−2.8 (0.4) (method 1)		—	—		—	—
Kinne, 199347	1989-90	Household survey, WA	Reported separately for women and men, average of two given here	—	127	−2.5‡§ (1.6) (method 2)		689††	−12.5 (3.4) (method 2)		−2.5	−53
Woodruff, 199348	1990	California tobacco survey, CA	—	—	1508††	−3.4§ (0.6) (method 2)		7290††	−6.3 (1.0) (method 2)		−1.6	−42
Pooled estimate	—	—	—	—	—	−3.1 (0.4)		—	−3.8 (0.5)		−1.3	−29

NR= not reported. 

^*Cohort size for longitudinal studies (allowing for loss to follow up) or sum of two samples for cross sectional studies. 

^†Method: 1—SE or 95% CI for change reported in the paper; 2—SE calculated based on average of SD of unrestricted and smoke-free estimates; 3—SE calculated based on average of SD estimates generated by methods 1 and 2 from other studies of same type. For methods 2 and 3, SE estimates calculated as √(SD²/n) where n is size of cohort (allowing for loss to follow up) for longitudinal studies and √(SD²/(1/n₁+1/n₂)) where n₁ and n₂ are two sample sizes for cross sectional analyses. 

^‡Average of workday and leisure day consumption estimates ((5×workday+2×leisure)/7). 

^§Per current smoker (smoker at time of survey) rather than continuing smoker (smoker at both baseline and follow up). 

^¶Cohort of smokers. 

^** Estimated from initiation and cessation rates. 

^‡‡Reported consumption as categorical variable; converted here to continuous variable by using midpoints. 

^††Sample sizes computed by multiplying total sample size by proportions of individuals in smoke-free and unrestricted workplaces. 

Results

Effects of smoke-free workplaces

Implementation of totally smoke-free workplace policies was associated with a reduction in absolute prevalence of 3.8% (95% confidence interval 2.8% to 4.7%) and a decrease in consumption of 3.1 (2.4 to 3.8) cigarettes per day per continuing smoker (table 1). Combination of the effects of stopping smoking (lower prevalence) and the lower consumption per continuing smoker means that 1.3 (range 0.2-1.8) fewer cigarettes were smoked per day per employee (smokers and non-smokers), which corresponds to a 29% (11%-53%) relative reduction.

The time between implementation of the totally smoke-free workplace policies and the follow up survey ranged from 1 to 24 months (mean 10 months, median 9 months) in the 21 workplace studies. This sample size has an 80% power for detecting a correlation of plus or minus 0.064. The correlation between length of follow up and effect was not significant (prevalence r=0.08, P=0.75; consumption per smoker r=0.45, P=0.09; consumption per employee r=0.28, P=0.43). The effects of smoke-free workplaces after they were implemented remained stable over time.

Comparison with tax increases

Increasing the price of cigarettes by 10% leads to a 4% reduction in consumption per capita (which reflects changes in both consumption per smoker and prevalence).²³ To obtain the 29% drop in employee consumption resulting from smoke-free workplaces would require an increase in the price of cigarettes of 73% (29%/0.4). Such an increase would require the average tax per pack to be increased from $0.76 to $3.05 in the United States and from £3.44 to £6.59 in the United Kingdom (table 2).

Table 2.

 Price and tax changes necessary to obtain same effect as smoke-free workplaces

Location	Effect of smoke-free workplace on consumption	Equivalent price effect (per pack)		Price/tax increase	Equivalent tax effect (per pack)
		Relative change*	Absolute change		Relative change	Absolute change
Within workplace (consumption/employee)
US	−29%†	+73%	$3.12‡ to $5.41	$2.29	+300%	$0.76‡ to $3.05
UK	–29%†	+73%	£4.30§ to £7.45	£3.15	+92%	£3.44§ to £6.59
In population (consumption/capita)
US	–4.5%¶	+11%	$3.12‡ to $3.47	$0.35	+47%	$0.76‡ to $1.11
UK	–7.6%¶	+19%	£4.30§ to £5.12	£0.82	+24%	£3.44§ to £4.26

Note: Numbers may not add up because of rounding. As of June 2002: £1=$1.50=€1.55; $1=£0.67=€1.03. 

^*Smoke-free workplace effect/price elasticity of cigarette consumption (−0.4²³) (see text). 

^†From table 1. 

^‡Average price or tax in United States in 2000.⁸¹ 

^§Average price or tax for 20 Marlboros in United Kingdom in 2001 (Clive Bates, personal communication). 

Effect on consumption per employee of smoke-free workplace (−29%)×proportion employed indoors (50%^{78 79})× proportion not yet smoke-free (31%⁵) for US; for UK figures are −29%×50%^{78 79}×52%.⁸⁰ 

The effect of workplace policies on consumption in the general population will be equivalent to a smaller tax increase because making workplaces smoke-free affects only those employed indoors and outside the home (about half of the US adult population⁷⁸,⁷⁹) and whose workplaces are not already smoke-free (31% in the United States⁵ and 52% in the United Kingdom⁸⁰), while tax increases will affect all tobacco buyers. The marginal effect on the entire population of making all workplaces smoke-free would be a reduction in per capita cigarette consumption of 4.5% in the United States and 7.6% in the United Kingdom. To achieve these reductions would require an increase in tax from $0.76 to $1.11 per pack in the United States and from £3.44 to £4.26 in the United Kingdom (table 2).

Effect on tobacco company revenues

In the United States about 103 million people are employed indoors outside the home,⁷⁸,⁷⁹ 69% of whom are already in smoke-free workplaces.⁵ The marginal effect of the remaining workplaces becoming smoke-free would be 40 million fewer cigarettes smoked per day (103 million×31%×−1.3 cigarettes per day per employee). In the United States in 2000 the average pretax price for 20 cigarettes was $2.36⁸¹ so this reduced consumption would cost the tobacco industry $1.7bn per year in lost sales. Likewise in the United Kingdom 29.9 million people are employed indoors,⁸² with 48% already working in smoke-free workplaces. In 2001 the pretax price of 20 Marlboro cigarettes was £0.86 (Clive Bates, personal communication) so the marginal effect of all work sites becoming smoke-free would be 20 million fewer cigarettes smoked per day, worth £310m a year to the tobacco industry.

Discussion

In this review we found that smoke-free workplaces are associated with a decrease in prevalence of tobacco consumption of nearly 4%, a decrease not due to underlying secular trends in prevalence. From 1987 to 1995 prevalence in the United States fell at about 0.46% per year,⁸³ and the follow up time for the studies we analysed was generally less than one year. Our results for consumption per smoker (3.1 fewer cigarettes per day) are comparable with those of Chapman et al,⁶ who estimated that smoke-free policies were associated with a fall in consumption of 3.5 cigarettes per day per continuing smoker. They did not quantify effects on prevalence.

Totally smoke-free workplaces versus smoke-free areas

Three of the population studies allowed us to compare the effects of totally smoke-free policies with those of partially smoke-free policies (where smoking is allowed in some areas other than work areas) (table 3).⁴⁵,⁴⁶,⁴⁸ Totally smoke-free workplaces had about twice the effect on consumption and prevalence as policies that allowed smoking in some areas. Internal research at Phillip Morris reached similar conclusions in 1992: “Milder workplace restrictions, such as smoking only in designated areas, have much less impact on quitting rates [than totally smoke-free workplaces] and very little impact on consumption.”⁸

Table 3.

 Comparison of the effects of totally smoke-free versus partially smoke-free workplaces^*

Study	Setting	Consumption/smoker†			Prevalence (%)
		Smoke-free v unrestricted	Partially smoke-free v unrestricted		Smoke-free v unrestricted	Partially smoke-free v unrestricted
Woodruff, 199348	1990 statewide survey of California	−3.3	−2.8		−6.3%	−3.5%
Glasgow, 199746	1993 study of 22 North American communities	−2.8	−1.2		NR	NR
Farelly, 199945	1992-93 survey of US workers	−2.7	−1.5		−5.7%	−2.6%

NR: Not reported. 

^*Totally smoke-free workplaces allow no smoking inside buildings, partially smoke-free workplaces allow smoking in designated areas. 

^†Cigarettes per day. 

Effects of legislation

To protect the health of non-smokers US local and state governments have enforced legislation restricting smoking in public and in workplaces. Seven studies reported the effects of these laws on adult consumption or prevalence.^84–90 The five studies that characterised laws according to extensiveness of their coverage found decreases in consumption per capita or prevalence associated with more extensive laws compared with no laws ranging from 0.16 to 0.73 fewer cigarettes per day per capita and 3.7% to 4.5% reduction in absolute prevalence. ⁸⁴,^86–88,⁹⁰

Local clean air laws are stronger and more comprehensive than state legislation.⁹¹ Strong local ordinances in California in 1990-1 were associated with an absolute quit rate (over the previous six months) 7.6% higher than in areas with no workplace laws.⁸⁶ A Canadian study in 1990-1 found a 21% reduction in the odds of being a smoker in areas with high versus low coverage of smoking bylaws.⁸⁹ A 1995 Finnish law that prohibited smoking in public areas in workplaces (with the option of creating separately ventilated smoking rooms) was associated with a 4.5% drop in prevalence of smoking and three fewer cigarettes smoked per day among continuing smokers.⁵⁷

Effects of smoke-free workplaces and clean air legislation on teenagers

Teenagers respond to smoke-free environments by decreasing smoking. One study reported the effects of workplace policies on employed teenagers.⁹² Teenagers who worked in totally smoke-free work sites were 68% (95% confidence interval 51% to 90%) as likely to ever smoke than those who worked in less restricted work sites. Eight studies examined the effect of state and local clean air laws on smoking in young people.⁶,⁹⁰,^93–98 Studies that compared the effects of extensive versus no laws found absolute reductions in prevalence of 2.3% to 46.0%,⁹⁶,⁹⁹ a relative reduction in prevalence of 17.2%,⁹⁷ and a relative reduction in per capita cigarette consumption of 50.4%.⁹⁰ Clean air laws can have substantial effects on smoking in teenagers, even though few are employed in ways that directly subject them to the laws.

Voluntary action to make homes smoke-free leads to lower rates of smoking among US high school students (odds ratio 0.79, 95% confidence interval 0.67 to 0.91, for 30 day prevalence).⁹⁷ Teenagers living in a smoke-free home were 74% (62% to 88%) as likely to ever smoke compared with those who lived in households with no smoking restrictions, after adjustment for demographics and smoking status of other household members.⁹² Among young people aged 14 to 22 years old in the United States the only significant predictor of planning to stop smoking was the belief that passive smoking harms non-smokers; this belief more than doubled the chances of planning to stop or of having stopped already.¹⁰⁰ These effects, as well as the workplace effects, probably act by reinforcing the social unacceptability of smoking.

As interventions designed to restrict the purchase of cigarettes have no effect on the prevalence of smoking among teenagers,¹⁰¹ clean indoor air is an effective strategy for reducing tobacco consumption among teenagers.

Weaknesses of study

We included studies with different methods carried out in different settings at different points in time. There were, however, no systematic differences in the results obtained in the workplace and population based studies, and the random effects model in the meta-analysis allows for any heterogeneity in study design.

We used changes in consumption per continuing smoker to measure the effect of the policies on consumption. Eight of the studies, however, reported consumption per current smoker (that is, including in the baseline measurement people who quit after the policy was implemented and in the follow up those who started after the policy).³³,⁴¹,⁴²,^44–48 We used these data as consumption per continuing smoker, although there is evidence that those who stop smoked less¹⁰²,¹⁰³ and therefore that changes in consumption among current smokers are smaller on average than those among continuing smokers. We may therefore be underestimating the effect on continuing smokers.

Interpretation of results

Smoke-free workplaces not only protect non-smokers from passive smoking but also encourage smokers to quit or reduce their consumption, reducing total cigarette consumption per employee by 29%. If all workplaces that are currently not smoke-free in the United States and the United Kingdom were to become smoke-free, consumption per capita (for the entire adult population) would drop by 4.5% and 7.6%, respectively. Achieving the same result with a tax increase would require a 47% tax increase in the United States and a 24% increase in the United Kingdom. While producing benefits for non-smokers by eliminating passive smoking³²,⁵⁷,¹⁰⁴,¹⁰⁵ and making it easier for smokers to reduce or stop smoking, smoke-free workplaces substantially reduce tobacco industry sales. This loss in revenues explains why the industry fights so hard against legislation to ensure that workplaces become smoke-free.⁷,⁸,¹⁰,^106–110

Figure.

Funnel plot used to assess publication bias