Article Text

Download PDFPDF

Objective measurement of area differences in ‘private’ smoking behaviour: observing smoking in vehicles
  1. Vimal Patel,
  2. George Thomson,
  3. Nick Wilson
  1. Department Public Health, University of Otago, Wellington, New Zealand
  1. Correspondence to Dr George Thomson, Department Public Health, University of Otago, PO Box 7343, Wellington South, Wellington 6021, New Zealand; george.thomson{at}otago.ac.nz

Abstract

Objectives The objective is to (a) refine and use methods to measure the point prevalence of smoking and of secondhand smoke exposure in moving vehicles and (b) compare these prevalences (1) between two areas of contrasting socioeconomic status and (2) over time.

Methods The authors developed and tested a single-observer method and observed the point prevalence of smoking in vehicles in Wellington, New Zealand. The two observation sites represented high and low areas of socioeconomic deprivation (based on a small area deprivation index).

Results A total of 149 886 vehicles were observed. The mean point prevalence of smoking in vehicles at both sites combined was 3.2% (95% CI 3.1% to 3.3%). Of those vehicles with smoking, 4.1% had children present. Smoking point prevalence in vehicles was 3.9 times higher in the area of high deprivation than in the area of low deprivation (95% CI 3.6 to 4.2). The same pattern was seen for vehicles with only the driver at 3.6 times (95% CI 3.4 to 4.0), in vehicles with other adults at 4.0 times (95% CI 3.4 to 4.7) and in vehicles with children at 10.9 times (95% CI 6.8 to 21.3), with all results adjusted for vehicle occupancy.

Conclusions Observing smoking in vehicles using a single-observer method provides a feasible and objective indicator of the different smoking behaviours, especially around children, within an area. This study further supports the evidence from this country and internationally that adults and children from high-deprivation areas are much more likely to be exposed to secondhand smoke.

  • Data collection
  • poverty areas
  • smoking/epidemiology
  • socioeconomic factors
  • tobacco smoke pollution
  • public policy
  • tobacco control policy
  • taxation and price
  • economics
View Full Text

Statistics from Altmetric.com

Background

Secondhand smoke (SHS) is hazardous, especially for children.1 ,2 The hazards of SHS are likely to be exacerbated inside vehicles, even when the windows are down.3–8 A detailed review of these hazard levels made the following conclusions: ‘i) because of the confined cabin space, and ii) under the worst ventilation conditions, and iii) in terms of peak contamination, the evidence permits us to say that smoking in cars generates fine particulate concentrations that are, iv) very rarely experienced in the realm of air-quality studies, and that will thus constitute a significant health risk because, v) exposure to smoking in cars is still commonplace, and vi) children are particularly susceptible …’.9 Survey data from 2007 indicate that over 40% of smokers in the USA smoke in cars with non-smokers present.10 A number of jurisdictions require smoke-free vehicles when children are in the vehicle, including provinces and states in Australia, Canada and the USA.11

We have identified only two published studies that have observed smoking and SHS exposure in vehicles, by watching from roadsides (one conducted in New Zealand12 and the other in Italy).13 These respective studies reported on 16 055 and 5928 observed vehicles, and the point prevalence of smoking in the vehicles was 4.1% and 6.9%. The New Zealand study undertook an analysis by socioeconomic area and found that the point prevalence of smoking in vehicles in an area of high deprivation was 3.2 times greater than in an area of low deprivation. The Italian study found that children were present in 1.7% of those vehicles with observed smoking.

To further improve methods and the very limited existing data, we sought to (a) refine and use the methods to measure the point prevalence of smoking and SHS exposure in vehicles in Wellington, New Zealand and (b) to compare these prevalences (1) between two areas of contrasting socioeconomic status and (2) over time. The methods improvement sought was reliable single-observer data collection.

New Zealand context

There was a government-funded smoke-free vehicles media campaign during 2006–2008,14 and all workplace vehicles accessible by the public have been required to be smoke-free since 1990.15 In a 2006 survey of New Zealand high school pupils, 27% self-reported exposure to SHS in vehicles in the previous week.16 This proportion was higher among youth from areas of high socioeconomic deprivation (using a school-related deprivation measure).17 During 2006–2008, the overall exposure was the same, but the gap between high- and low-deprivation areas had widened, from 37% to 40% in areas in the four most deprived deciles compared with a change from 16% to 14% in areas of the two least deprived deciles.18 In a New Zealand survey of adults in 2008, 30% of those in high-deprivation area (deciles 8–10 measured by the NZDep 2006 index neighbourhood index of deprivation)19 reported that someone has smoked in a household vehicle in the last 7 days compared with 14% of those in low-deprivation areas (deciles 1–3).20

Smoking and socioeconomic differences

People in areas of higher socioeconomic deprivation are more likely to smoke and find it harder to quit,21 ,22 which may be related to several factors including deprivation-related stress, solo parenthood, education or ethnicity/minority status.23–25 Avoiding smoking around children appears to be more difficult with added stress, less access to information and lack of immediacy of SHS effects.26–28

Methods

Site selection

Solo observers on the roadside observed vehicles at two sites in the Wellington region over 15 km apart by road. The observation sites were chosen in the previous New Zealand study12 to be representative areas of high (Wainuiomata—NZDep deciles 7–9) and low socioeconomic deprivation (Karori—deciles 1–4). The 2006 census reported that the prevalence of daily smoking in adults was 30% in Wainuiomata29 and 11% in Karori.30 Māori (the indigenous people of New Zealand) comprise 6% of the Karori population and 27% of the Wainuiomata population.29 ,31 Both sites had topographical boundaries such that the observation sites were the major road in and out of each suburb. Site selection criteria included high traffic flows, low traffic speeds and good visibility of vehicle occupants (at both sites, observers were approximately only 1–2 m distance from the passing vehicles). The average flow was 935 vehicles per hour over the observation periods.

Sample times

Observations at both sites were made during high traffic periods (7:30–9:30 and 16:00–18:00) on 20 weekdays during February to April 2011 that were not in school holidays.

Data collection and analysis

A single observer at each site observed vehicles in the near lane (and adjacent lanes with traffic passing in the same direction). Two consecutive pairs of observers were used, one for 9 days and one for 11 days. Observers held a mechanical counter in one hand to count the total number of vehicles that fitted the sample frame (regardless of whether smoking was observed or not). For each vehicle with observed smoking, the observer recorded on a pre-formatted data sheet: the presence of smoking, the presence of other adults than the smoker and the presence of children. Observers swapped observation points every 2 days.

To assess and improve both the feasibility and accuracy of solo observer data collection, pilot tests of the methods were performed by one researcher (VP) at one site on a total sample of 733 vehicles. One of the first pair of observers (VP) trained all the other observers (during part of 1 day). This involved (1) providing a protocol for data collection (in advance) and (2) explaining, demonstrating and practicing data collection methods with observers at both sites until they could perform observations reliably and independently.

Inter-observer variation between observer pairs was assessed (for both observers at the same observation site in close proximity to each other). This took place prior to data collection, on three samples of 500 vehicles each for the first pair of observers and two samples of 500 vehicles each for the second pair (and additionally for the second pair on 1500 vehicles after data collection). These assessments were also used to improve observations by practice (and for the first pair only, to iteratively refine data collection methods).

We further refined the sampling frame used previously12 and did not observe (1) vehicles where it was difficult to see inside (for instance due to window tinting) and (2) all buses, taxis and trucks. Occupants appearing to be aged 12 years or younger were classified as children; otherwise they were recorded as adults. Smoking was defined as one or more people in a vehicle holding a cigarette, pipe or cigar in their hand or mouth.

Occupancy survey

To adjust for differences in vehicle occupancy between the two sites, we surveyed vehicle occupancy separately at both sites on two further single days during the same time periods (7:30–9:30 and 16:00–18:00). For every fifth vehicle in our sampling frame, the observer recorded on a pre-formatted data sheet: (1) the presence of the vehicle, (2) adult occupants other than the driver and (3) child occupants. Monte Carlo simulations were performed to determine the prevalence of smoking in vehicles that adjusted for differences in the proportions of vehicles carrying single occupants/other adults/children between Wainuiomata and Karori.

All data were entered into a Microsoft Excel database and analysed using Excel and OpenEpi (Emory University), with Monte Carlo simulations run in R 2.11.1 (R Foundation, Vienna, Austria). Ethical approval was obtained through the University of Otago (Category B ethics approval process).

Results

Evaluation of the method

The single-observer method appeared practical for the quantity and type of data collected. For the third (and final) assessment of inter-observer variation between the first pair of observers, there was perfect agreement between all the observational categories (both observers agreed that out of 500 vehicles, smoking was observed in 31 vehicles and that of vehicles with observed smoking, two carried other adults and one carried a child). For the final assessment of inter-observer variation between the second pair of observers (after data collection), there was at least substantial agreement between all the observational categories (κ values were (1) 0.99 for any smoking, (2) 0.87 for other adults in vehicles with smoking and (3) 0.80 for children in vehicles with smoking).

Results of the observational study

The occupancy surveys showed that the proportions of vehicles with (1) single occupants, (2) other adults and (3) child occupants in Karori were 0.9, 1.3 and 1.1 times than that in Wainuiomata, respectively (see table 1).

Table 1

Differences in vehicle occupancy by site from the vehicle occupancy survey (used to adjust smoking prevalence results)

A total of 149 886 vehicles were observed in 20 days (excluding trial periods). Large differences were found between the observation sites, particularly for smoking with children in the vehicle. An overall 3.2% point prevalence of any smoking in vehicles was observed (95% CI 3.1% to 3.3%; see table 2). The in-vehicle smoking prevalence in the area of high deprivation was 3.9 times greater than in the area of low deprivation (95% CI 3.6 to 4.2).

Table 2

Prevalence of smoking in vehicles by site

The point prevalence of smoking in vehicles was significantly higher in the morning (4.1%) than in the evening (2.4%; RR relative to the latter 1.7, 95% CI 1.6 to 1.8).

When adjusted for the variation in vehicle occupancy between the observation sites, the point prevalence of smoking in vehicles with only the driver was 5.5% in the area of high deprivation and 1.5% in the area of low deprivation (RR relative to the latter 3.6, 95% CI 3.4 to 4.0).

The unadjusted point prevalence of smoking in vehicles with other adults (other than the driver) was 0.6% (95% CI 0.58% to 0.66%). The adjusted prevalence in the area of high deprivation was 4.0 times greater than in the area of low deprivation (95% CI 3.4 to 4.7).

The unadjusted point prevalence of smoking in vehicles with children was 0.13% (95% CI 0.11% to 0.15%). Of all those vehicles with smoking, 4.1% had children present. The adjusted prevalence in the area of high deprivation was 10.9 times greater than in the area of low deprivation (95% CI 6.8 to 21.3).

Compared with data collected in the 2005 study at the same two observation sites, there was an absolute reduction in the point prevalence of smoking in vehicles of 1.1 percentage points (RR relative to the former 1.3, 95% CI 1.2 to 1.5) (see table 3). The relative reduction over time in the area of low deprivation was 1.2 times greater than in the area of high deprivation (95% CI 1.0 to 1.6).

Table 3

Trends over time in the point prevalence of all smoking in vehicles and smoking in the presence of others (in August 2005 and in February to April 2011)

There was an absolute reduction in the point prevalence of smoking in the presence of others in vehicles between 2005 and 2011 of 0.2 percentage points (RR relative to the former 1.3, 95% CI 1.1 to 1.6). The relative reduction over time in the low-deprivation area was 1.3 times greater than that for the high-deprivation area (95% CI 0.8 to 2.1).

Discussion

There are important advantages to using observed smoking in vehicles as an objective indicator of the exposure of children to SHS in an area. Vehicles are uniquely both confined and ‘private’32 but also observable from the outside (in contrast to the inside of homes and private workplaces). Also, a large number of vehicles can be observed in a short study period. Our results provide additional support to the idea that observed smoking in vehicles may be useful in a number of areas as a window on ‘private’ smoking and SHS exposure, indicating what may also be happening in homes.

The socioeconomic gradient

This study provides further evidence that people from high-deprivation areas are much more likely to be exposed to SHS.12 ,16 ,18 ,33–35 If such people are being exposed to SHS in the confined space of vehicles, then it seems plausible that home exposure is even more likely. In our study, the point prevalences of smoking, adult (other than the driver) and child exposure to SHS in vehicles were much greater in the area of high deprivation relative to the area of low deprivation (RRs relative to the latter 3.9, 4.0 and 10.9).

Besides the higher prevalence of smoking in the more deprived area, these differences may be partly explained by differences between the observation sites in (1) the normality of smoking around others, (2) cultural factors (eg, acceptability of non-smokers telling smokers not to smoke around them or near children), (3) mean journey length, (4) vehicle access for smokers, (5) greater consumption per smoker, leading to a greater chance of being observed in a journey and (6) home restrictions on smoking (eg, possibly impacting on levels of smoking in vehicles).

There is the further question of why the RR between the two areas was much greater for smoking with children compared with smoking alone. One possibility is that smokers in the more deprived area were not reached as effectively by social marketing about SHS risks as those in the less deprived area. For example, there may be variation in the reach of media campaigns to smokers in different socioeconomic groups—though results are mixed.36 ,37

The observed socioeconomic gradient of all smoking in vehicles was greater between the two areas (RR 3.9) than indicated by the 2006 census data, where the respective RR of regular smoking in the same two areas was 2.8.29 ,30 We observed an 11-fold differential in child exposure to SHS in vehicles between the two sites (of NZDep 2006 deciles 1–4 and 7–9). This would appear to be much greater than indicated by surveys of New Zealand high school pupils, where the RRs between the most deprived group (school deciles 1–4) and least deprived group (deciles 8–10) was 2.3 in 200616 and 2.8 in 2008.18 Even the lower limit of the 95% CI for the RR (6.8) was much greater than indicated by the surveys. This leads to questions about the relative accuracy of surveys and observations (see below) and about the extent that different things were being measured.

Strengths and weaknesses of this study

This study observed a sample of vehicles over nine times greater than the largest previous study (149 886 vs 16 055), allowing for much tighter CIs in the comparisons between different observation sites.

Quality of the method

Observation (even of point prevalence) may provide a more accurate picture of SHS exposure in vehicles in an area than survey data, particularly for the relative differences between areas and for changes over time. Surveys have problems from bias (eg, due to social desirability responses that increase the likelihood of smoking being denied or minimised), recall, response rates or other methods issues.38–41

Field observations of this nature are less likely to have significant bias (as in some other observation studies),42 as observers are separated from the observed. A bias (for instance against smokers or for children) might be relevant for instances where the age of children is in doubt, but generally bias would require observer cheating to have an effect. Observer cheating and inaccuracy (non-reliability) is of more concern than bias, with boredom and work avoidance needing to be countered by spot checks, immediate data transfer and rotation of observer placements.42 ,43 Well-designed field observations can be particularly useful in the unobtrusive investigation of disparities.44

The most subjective aspect of the observation, and the one with most inter-observer variation for the second pair of observers, was the judgement of whether children were aged 12 years or younger. One possible solution would be to train observers on photos of children of a known age that varied from 10 to 15 years.

The observational method used in this study overcomes some of the limitations of using self-reported data and refines established methods for the collection of observational data in vehicles. Further observation of smoking in vehicles would appear more efficient during the morning, as indicated by the higher prevalence of observed smoking, compared with in the evening. It appears unlikely that this unobtrusive observation would lead to behaviour modification by those observed, as can be a problem in some observation studies.42 ,45 Trained single observers appear to be a practical and reliable method, and with suitable and motivated recruits, the training can be effective in <1 day. Furthermore, this study observed both child and adult exposure to SHS in vehicles (which has only been collected in one other study worldwide, to our knowledge).13 Our study also appears to be the first to compare child SHS exposure between areas of different socioeconomic deprivation.

However, only two sites were sampled at specific times on weekdays over 2 weeks and in two seasons of the year (late summer and mid-autumn). Thus, our results may not be fully representative of smoking in vehicles in the Wellington region (or for elsewhere in New Zealand). Additionally, since the number of children observed in vehicles with smoking was relatively small, our RR of child exposure to SHS between the two sites is imprecise (as indicated by the wide 95% CI). There may be further uncertainty around this estimate due to the subjectivity in estimating the age of children observed.

Smoking prevalence

Our point prevalence for any smoking in vehicles of 3.2% is likely to be a marked underestimate of the true (population) prevalence of smoking in vehicle trips. Smoking could occur at any point during a vehicle trip, although further qualitative and survey data may help to establish if smoking is more likely at the beginning of a journey. Thus, the observation of smoking in cars at a single point in time will miss cases of ‘smoking in a trip’ (eg, especially when a person leaving work in central Wellington lit up early in the return trip home to either suburb). A survey of high school pupils16 appears to remain the best indicator of child exposure to SHS in a vehicle over a week. However, observing vehicles can help assess the relative differences between areas and changes over time.

Policy options

Increasing the prevalence of smoke-free vehicles has a number of benefits. Besides reducing the SHS hazard to both non-smokers and smokers themselves (including the mortality risk from vehicle crashes),46 smoke-free vehicles are likely to normalise smoke-free behaviour.47–50 In a number of countries, those in poorer and most socioeconomically deprived households are more likely to be exposed to SHS.27 ,51–53 The much higher exposure for those in more deprived areas in New Zealand, by observation, census and survey data, with apparently widening disparities, indicates that the current educational approach is not particularly effective for protecting the more socioeconomically deprived from SHS in vehicles. While social marketing is highly desirable as part of an overall approach, the stronger signal about SHS hazards that legislation gives (along with the accompanying unpaid media publicity) may be the most effective and cost-effective way to markedly improve protection for the most socioeconomically deprived. The regulatory options used around the world include requiring smoke-free vehicles while children or youth are in them (children/youth of a specified age or where they are in child safety devices) or requiring no smoking while driving.11

While the evaluation of smoke-free vehicle laws is sparse, there is some indication of effectiveness. An Australian study reported that the 2007 smoke-free vehicle law in South Australia (which only applies to vehicles carrying children) resulted in an increase in smoke-free vehicles with children from 69% in 2005 to 82% in 2008.54 There is majority and increasing support for smoke-free vehicle laws in a number of jurisdictions in the UK, North America and Australasia, suggesting that one element of the politics is in place,11 but the policy process to such laws may sometimes be complex.55 The relative success of seatbelt laws and child car seat laws56 are policy domains that may provide insights into how smoke-free vehicle laws might be best adopted. The experience of enforcing the use of particular types of child seats for cars, for those under particular ages, weights or heights (eg, in the UK, California),57 ,58 may help in preparing for the enforcement of smoke-free laws for cars carrying children under particular ages. Laws requiring vehicles to be smoke-free where there are more than one person, or requiring drivers to not smoke, would be much more simple to enforce. The limited effectiveness of bans on the use of handheld mobile phone in moving vehicles (because hands-free phones still distract)59 indicates the need for comprehensiveness in any law.

Further research

The priority for further research is to evaluate the impact of laws requiring vehicles to become smoke-free (especially when there are child passengers). These evaluations could include ongoing systematic observational work. Observational studies could particularly help to determine trends in disparities by area deprivation.

This type of study repeated internationally, using the same methodology, could provide objective comparisons of the point prevalence of smoking and SHS exposure in vehicles. One development could be a standardised smartphone application to enable ongoing data collection of observed smoking in vehicles by many observers at many observation sites internationally (a research topic we ourselves are exploring).

What this paper adds

  • Only two studies have observed smoking and secondhand smoke exposure in vehicles. But neither quantified child exposure to secondhand smoke in vehicles as well as conducting a socioeconomic analysis.

  • Our observations of almost 150 000 vehicles across two areas of contrasting socioeconomic status indicate large gradients in the prevalences of smoking and secondhand smoke exposure in vehicles. In particular, the child exposure to secondhand smoke was 11 times greater in the high-deprivation area (compared with low-deprivation area). When compared with data collected in 2005 (at the same observation sites), these findings indicate that while the overall prevalence of smoking in vehicles has fallen (by 34%), the reduction was markedly greater in the area of low deprivation.

  • Trained single observers appear to be a practical and reliable method. Observation of smoking in cars may provide a more accurate picture of local trends in smoking prevalence and secondhand smoke exposure in vehicles than in survey data.

Acknowledgments

We thank the other observers: Catherine Jones, Losa Moata'ane and Priyesh Patel, and thank Dr James Stanley who provided statistical advice and support. Our thanks to the Cancer Society of New Zealand who funded this study (but who had no role in the final content or decision to publish).

References

View Abstract

Footnotes

  • Funding Cancer Society of New Zealand.

  • Competing interests None.

  • Ethics approval University of Otago.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.