The methodology of this study followed the processes of the international "Global Air Monitoring Study" being coordinated by the Roswell Park Cancer Institute in New York [4]. However, it included a number of additional methodological features including the more structured sampling of pubs, bars and restaurants.
Selection of urban pubs and restaurants
Our initial sampling method involved the random selection of bars and restaurants from the Central Business District (CBD) of Wellington City. This is an area based, with simplifications for some boundaries, on the official "Central Area Boundary" of the Wellington City Council [5]. This sample frame was also used for other settings detailed below (with the exceptions being for the hospital, the airport and for some bus and taxi routes that extended beyond the CBD). We aimed to sample equal numbers of bars and restaurants with and without linked "outdoor" smoking areas. Our sampling ceased once we had four venues in each of the four different categories. The selection of map coordinates of the CBD was based on random number tables. These tables also determined the direction to take on foot from the selected map coordinate. The first bar or restaurant encountered was then assessed and sampled. Bars were defined as venues that had any "bar area" that served alcohol and with selling alcohol being considered to be the main purpose of the establishment. Restaurants were defined as venues having sit-down meals with menus and with selling food being considered to be the main purpose of the establishment. Data collection occurred on Friday and Saturday nights in the 5 pm to 10 pm time slot during May/June 2006 (late autumn and early winter in New Zealand). To facilitate outdoor sampling of adjacent "outdoor" smoking areas we selected days when it was not raining.
Additional selection of urban "bars"
After field experience with the sampling method detailed above, we also undertook another method that was more orientated towards sampling more traditional "bars" where the focus was on serving drinks rather than food. This was because we suspected that these venues may be more likely to have clientele with higher smoking prevalence rates, and hence infringements of the law may be more likely in these venues. The selection process involved searching the electronic yellow pages (Telecom, 2006) under the categories of "bars & brasseries" and "hotels & taverns" and within the category of "Wellington CBD". From these entries (n = 48+88 respectively) we selected those bars that met the following criteria: (i) the word "bar" was in the name; (ii) there were none of the following additional words in the listed name of the bar: "café", "restaurant", "grill" or "pool saloon"; (iii) the bar was not also listed under the "adult entertainment" category of the yellow pages. These steps generated a list of 14 bars from which we randomly selected 11 using random number tables. Data collection occurred on one day, Saturday 24 June 2006 (from 2 pm to midnight). By chance, none of the sampled bars had adjacent "outdoor" smoking areas (ie, no "sit down" outside tables).
Selection of small town and rural pubs
The South Wairarapa area was selected as a convenience sample of a rural area (to minimise travel time and travel costs from Wellington). A search of the "Yellow Pages" Directory found 10 entries for "Hotels & Taverns" in "Featherston & Districts" and five entries in "Carterton". From these entries a purposeful sample of the more remote and more traditional pubs was made ie, those that focused on serving drinks rather than food. This selection was based on website information and included those venues with the words "tavern" or "hotel" or "inn" in their name (since these terms tend to reflect more traditional establishments rather than more tourist-orientated or "up-market" establishments). The nine selected pubs were in small towns (n = 6) and in more remote rural settings (n = 3). All these pubs were visited on a single day (Friday, 16 June 2006) starting at midday.
Sampling of other indoor and outdoor settings
We also used convenience sampling in Wellington city (mainly in the CBD) to investigate a range of ten other smokefree venues, including: cafés, shops, offices, a hospital, a smokefree pedestrian arcade; and nine transport settings (buses, a taxi, a train, bus and train stations, and an airport). Finally we identified six outdoor settings in the Wellington CBD (ie, walkways, roadside areas, and a city park) that were not designated smokefree except for one of the walkways.
Sampling of relatively enclosed "outdoor" smoking areas
To allow for comparisons with indoor areas, we also included outdoor smoking areas where these were available at the included venues. This was provided there was at least one occupant present other than the investigator and at least two cigarettes were smoked during the data collection period. Four of the these outdoor areas were from the 34 included pubs, bars and restaurants, while two others were purposively included as they were known to be relatively enclosed (ie, with walls on at least three sides).
Air quality data collection
As per the processes of the international "Global Air Monitoring Study", indoor venues were visited unannounced by either one or two of the investigators. Also to avoid affecting occupants' behaviour, we behaved as normal customers (ie, bought drinks or food in the pubs and restaurants). At each indoor sampling site, the busiest room in the pub or restaurant was identified and an attempt was made to obtain central seating (though in some busy settings this was not always possible). We ensured that the sampling was not within two metres of any open doors or windows leading to the outside or kitchen areas. For "outdoor" smoking areas we aimed to sample from the middle of the area and not in the immediate vicinity (< 1 metre) of anyone smoking.
Data were collected on fine particulates which are defined as being 2.5 μm in diameter or less (ie, "PM2.5"). The data were collected using a TSI SidePak AM510 (TSI, Inc, St Paul, USA) portable real-time air quality monitor (a photo of which is at: [6]). This device recorded average levels of these particulates (PM2.5) over one minute periods. The use of the monitor followed a protocol modified from one developed for a US study [7] and as used in a previous UK study by one of the authors (RE) [8]. This monitor was zero-calibrated prior to use and was fitted with a 2.5 μm impactor with an air flow rate of 1.7 l/min. The air flow rate has been validated in the New Zealand setting using a pneumotachograph (Hans Rudolph 4813 pneumotachograph, Vacuumed differential pressure transducer 4500, Vacumetrics, California, USA), and was within 10% of the stated flow rate.
A length of Tygon™ tubing was attached to the inlet of the SidePak, with the other end left protruding (slightly) outside the bag. The bag with the sampling equipment was carried or placed on a seat or table wherever possible to sample the ambient air. Recording occurred for 35 minutes (to ensure a 30 minute sample). Where indoor sampling occurred in hospitality settings, we also sampled for at least 5 minutes from directly outside the venue (but 5 metres away from any people smoking on the street).
Observation data collection
At each indoor setting, additional information was systematically collected on a preformatted data collection sheet. This included recording number of people in room/area (at 0, 15 and 30 minute intervals) and the number of lit cigarettes (at 0, 15 and 30 minute intervals). A sketch was drawn of the layout of the room and the location of the sampling site in relation to doors and to any "outdoor" smoking areas.
Data analysis
A calibration factor of 0.32 was applied to the measured data based on calibration work with a ThermoMIE personalDataRAM model pDR-1200 real-time aerosol monitor (ThermoAndersen, Inc, Smyrna, GA, USA) and as used in other studies using the SidePak monitor [7–10]. This type of monitor has also been calibrated against standard pump-and-filter gravimetric methods [10]. The recorded measurements were downloaded to a personal computer for analysis using TRAKPRO version 3.4 and EpiInfo (CDC Atlanta). Mean time-weighted averages and peak levels for each setting were calculated.
Ethics approval
Ethics approval was obtained through the University of Otago's ethical review system (Category B approval). At the international level the "Global Air Monitoring Study" has obtained ethical approval from the Roswell Park Cancer Institute Ethics Review Board.