CHETS Wales was a repeated cross-sectional study of year 6 (10-11 years old) school children in Wales. Data were collected immediately pre-ban (from 31st January 2007 to 30th March 2007) and one year after the initial data sweep (from 31st January 2008 to 28th April 2008).
The study aimed to include students from a nationally representative sample of 80 Welsh primary schools. Based on the hypothesis that among children who live in households with other smokers, there will be no change in home-based SHS exposure post-legislation, the sample size was identified to have 80% power to demonstrate equivalence within 0.15 sd, assuming an intra cluster correlation (ICC) of 0.05, or within 0.2 sd with an ICC of 0.14. The same schools were approached pre-and post-legislation to minimise differences in the sample between survey years. The sampling frame, which included all state maintained schools in Wales, was stratified according to high or low free school meal entitlement (above or below population mean of 17.12%) and Local Education Authority. Within each stratum, schools were selected on a probability proportional to size (total number of students in the school). Where schools declined to participate, replacement schools were randomly identified from within the same stratum.
In each school, one year 6 (age 10-11) class was randomly selected to participate. Due to the small size of many Welsh primary schools, or due to requests from schools to include all of year 6, the whole of year 6 was often included in the survey. Some classes also included year 5 (i.e age 9-10), students who were taught alongside year 6 children. Pre- and post-legislation however, 97.6% and 97.4% of participating children were aged 10 or 11 years at the time of data collection.
Consent was sought at three levels. First, a letter was sent to the headteacher of each selected school inviting them to participate. Schools were followed up by telephone if they did not respond. When schools consented, further details of the study were provided by telephone and arrangements for data collection made. Consent letters were then sent to parents/carers of all potential participants (including those in year 5) via the school. An 'opt out' system was adopted in most schools, whereby parents/carers were asked to inform the school if they did not wish their child to participate. Seven schools requested that an 'opt-in' consent procedure was used. A researcher contacted schools to ascertain numbers of eligible participants. At each data collection, students were given the opportunity to withdraw from the study.
Data were collected in the classroom by trained CHETS Wales staff. Class teachers were asked to be present, but not to intervene in the data collection unless asked to do so by the researcher. All students were asked to complete an anonymous questionnaire regarding their own smoking behaviour and that of friends and household members (including parental figures), smoking norms and attitudes, recent SHS exposure in a variety of public and private locations, and asthma. In order to enable comparability of results, most questionnaire items were identical to those used in CHETS . Students also provided an anonymous saliva sample for cotinine assay using a cotton wool swab of a salivette®. Attempts were made to ensure that data collections in each school were conducted at a similar time of the week and day at each datasweep although this was not always possible due to other school commitments.
Anonymous saliva samples were linked to questionnaires by unique identification numbers. Absentees were followed up where absenteeism at data collection was greater than 40%. The study protocol and consent procedures employed were approved by the School of Social Sciences Ethics Committee, Cardiff University.
The year of data collection (2007 or 2008) was used as a proxy for the primary independent variable, introduction of smoke-free legislation
Salivary cotinine concentrations
The study's primary outcome measure, salivary cotinine (a metabolite of nicotine), has been identified as the most reliable and suitable biomarker of exposure to tobacco smoke in the previous 72 hours [27–29]. Saliva samples were assayed using capillary gas chromatography with a specific nitrogen/phosphorus detector from a 100 μl sample  and which had a detection limit of 0.1 ng/ml. Assays were conducted by the laboratory used by CHETS, ensuring comparable results.
Respondent smoking behaviour was measured using the HBSC scale . Children who gave a response other than 'I do not smoke', or with salivary cotinine concentration above 15 ng/ml  were classified as smokers and excluded from analyses relating to cotinine.
Perceived frequency of SHS exposure
Children were asked to indicate how often they were in a place where people are smoking.
Locations of perceived SHS exposure
Children were asked if they had been in a car, a café, someone else's home, a bus or a train, an indoor leisure facility and the home on the previous day, and whether anyone was smoking there. They were also asked to indicate the amount of time spent there.
Parental smoking in the home
Parents were classified as smoking if the child reported that they 'smoke every day' or 'smokes sometimes'. Children were asked to identify whether parental figures (mother, father, stepfather or mother's partner, and stepmother or father's partner) smoked in the home. Children were subsequently categorised according to the number of parent figures who smoked within the child's home (neither parent figure, father figure only, mother figure only, or both).
Children were asked to indicate year and month of birth and age in years on the day of data collection was calculated (date of birth set to the 14th of the month). Responses to items on the Family Affluence Scale (FAS: ), which includes measures of bedroom occupancy, car ownership, holidays and computer ownership, were summed and taken as a marker of socioeconomic status (SES). Measures of material affluence are typically completed more accurately by children than alternative markers of SES such as parental occupation or education . The FAS has been validated favourably against other measures of SES such as parental occupation .
Time of day data collected
The time of data collection was included as a control variable in adjusted regression analyses. Times were divided into three categories (9-11 am, 11 am-1 pm and 1 pm to 3 pm).
Analyses replicated those of Akhtar et al . However, compared to the Scottish dataset, the Welsh dataset had a higher degree of positive skew, with 47% of samples containing cotinine concentrations below the limit of detection. Hence, a number of supplementary analyses were included. All statistical analyses were conducted using the svy settings of Stata version 9, to adjust for the clustered nature of the data sample.
Population change in secondhand smoke exposure
Saliva samples containing cotinine concentrations below the limit of detection (0.10 ng/ml) were assigned a random value from the left tail of a truncated log-normal distribution. Percentages of each sample whose cotinine concentrations were below 0.10 ng/ml or above noteworthy cutpoints [7, 35] were calculated.
Median cotinine concentrations pre- and post-legislation were presented alongside modelled estimates of geometric mean concentrations (both unadjusted and adjusted for age, family affluence and time of day data collected) for each survey year. To adjust for the time of day saliva samples were collected, dummy variables were used for 11 am-1 pm collections and 1 pm-3 pm collections, with 9-11 am collections acting as the reference category. Modelled estimates, and significance of change, were assessed using linear regression models, with log-transformed cotinine values as the dependent variable. Since linear analyses were compromised by reliance on imputation for 47% of cases, the distribution was split into approximately equal tertiles ('low' <0.10 ng/ml, 'medium' = 0.10-0.50 ng/ml, 'high' >0.50 ng/ml) and analysed using multinomial logistic regression.
Perceived exposure to secondhand smoke
Frequencies and percentages of children reporting each category of perceived frequency of SHS exposure pre- and post-legislation were calculated. Significance of change over time was assessed using multinomial logistic regression analysis.
Frequencies and percentage responses to items relating to locations of self reported exposure to SHS were calculated and significance of change over time assessed using design adjusted chi-squared analyses, comparing the proportion of children reporting that someone was smoking in the location in question against all other responses (including missing values).
Displacement of parental smoking
Sub-group analyses examined cotinine concentrations and perceived frequency of SHS exposure, pre- and post-legislation, by number of parent figures who smoked within the home (no parent figures, father figure only, mother figure only, or both). One hundred and twelve children pre-legislation and 122 post-legislation responded only to items relating to a mother figure or a father figure, and most reported that this parent smoked in the home. Thus it was assumed that this question had been interpreted as requiring a response only if the parent smoked in the home and these children were classified as having one parent who smoked in the home.
Frequencies and percentages of children in pre- and post-legislation samples within each subgroup were calculated and significance of change in percentages of children with parents who smoke in the home were assessed using design adjusted chi-squared analyses. Finally, the linear and multinomial regression analyses described above were repeated for each subgroup.