To our knowledge, this is the first study to provide comprehensive evidence of prison workers’ exposure to SHS throughout a country’s entire prison system.

Across a suite of measurement methods that include air sampling, biological markers of exposure, and subjective self-report, we have provided evidence of SHS exposure within cells, prison landings, halls, and other communal areas that is regular and systematic in all prisons, but varied by time of day, and between and within different prisons.

The 6-day PM2.5 concentrations measured in a residential hall of each prison are comparable with studies from other countries. The median value reported here was 31.7 µg m−3 (range 11–136 µg m−3) which is similar to the median value of 35.6 µg m−3 (range 27–70 µg m−3) reported from five prisons in England and Wales assessed in a near identical manner using the Dylos DC1700 device (Semple et al., 2015a). Other data from four prisons in England (Jayes et al., 2016) used a TSI Sidepak AM510 to measure PM2.5 concentrations over shorter periods (mean 6.5 hours) on residential landings and reported average concentrations of 43.9 µg m−3 on wings where smoking within cells was permitted. Given that Jayes et al.’s data were gathered during ‘daytime hours’, it is worth noting that the 6-day residential hall results from the present study were 36.5 µg m−3when restricted to daytime hours.

Studies from prisons in other parts of the world provide more divergent results. A study in a single New Zealand prison (Thornley et al., 2013) used the TSI Sidepak AM510 to measure PM2.5concentrations before the introduction of a tobacco ban and reported a GM value of 6.6 µg m−3 over a 14-day period. The device was positioned in the staff base adjacent to the four prison wings. Previous work examining PM2.5, again with the TSI Sidepak AM510, in six prisons in the USA (Proescholdbell et al., 2008) provided mean values of 93.1 µg m−3 from measurements in prison dormitory areas and lobbies. These 14 measurements were taken over short periods with between 43 and 91 minutes spent in each of the six prisons. A study in a Swiss prison (Ritter et al., 2012) reported PM10concentrations made in three prison areas with mean values of 30, 120, and 180 µg m−3, however, duration of measurement was not reported.

The 31.7 µg m−3 PM2.5 median in the current study can also be compared to other smoking and smoke-free environments. For context, the average values reported for smoke-free homes in Scotland is 3.1 µg m−3 (Semple et al., 2015c). Smokers’ homes in Scotland have a median value of 31 µg m−3 (Semple et al., 2015c)—very similar to the 6-day area value measured across the 15 Scottish prisons in this study. Data on PM2.5 concentrations measured in Scottish pubs and bars prior to smoke-free legislation in 2006 indicated a mean value of 246 µg m−3 (Semple et al., 2007b), nearly eight times greater than that measured in Scottish prisons.

The GM for the 86 mobile PM2.5 measurements was 24.1 µg m−3(GSD 4.2), very similar to that for 70 ‘spot’ measurements using a near identical protocol in six prisons in England and Wales (GM 24 µg m−3; GSD 3.5) in 2015 (Semple et al., 2015a). Time-course graphs of both the area and mobile monitoring results show the wide range of PM2.5 concentrations measured, by prison, time of day and specific locations and activities. The mobile measurement results suggest that some areas of most prisons, including health care, sports/gym facilities, teaching, and reception areas, are essentially smoke-free. Many workshop area measurements also indicate little, if any, SHS exposure. However, staff exposure is considerable in many other areas, particularly those close to cells. Staff offices, corridors, and landings show evidence of SHS drifting from prisoners’ cells to these communal areas.

Concentrations during recreation activities were particularly high. Activities involving cell unlocking, cell searches, cell fabric inspections, and cell maintenance generally suggest considerable exposure.

These activities may result in staff being exposed to concentrations that are several times higher than the WHO guideline for PM2.5 with some of these activity-based measurements indicating values comparable with those measured in Scottish bars when smoking was permitted (Semple et al., 2007b).

The airborne nicotine measurements reported in this study had a median of 0.32 µg m−3. These values are considerably lower than we would have anticipated given the PM2.5 results from the co-located Dylos DC1700 devices together with the data on likely nicotine concentrations from saliva samples.

We note that the ‘Rosetta stone’ equations developed by Repace and colleagues (2006) suggest that PM2.5 concentrations are roughly 10 times those of airborne nicotine in settings where SHS is present. Given the Dylos median of 31.7 µg m−3, we would anticipate an air nicotine median of about 3.2 µg m−3. In comparison, Hammond and Emmons (2005) measured weekly airborne nicotine concentrations in three US prisons before smoke-free rules were put in place. Their analysis of 84 locations indicated average values ranging between 3 and 11 µg m−3 in most living and sleeping areas within these prisons. Ritter et al. (2011) reported mean values of 7.0 µg m−3 in a Swiss prison, while work in smoking homes by Phillips and co-workers (1996) and by Butz et al. (2011) indicated airborne nicotine concentrations of 1.1 and 1.4 µg m−3, respectively. Both these studies (Butz et al., 2011; Ritter et al., 2011) also reported PM concentrations very similar to those measured by our Dylos DC1700 devices in prisons (39 and 35 versus 32 µg m−3). Our results using pre- and post-shift cotinine also suggest that prison workers’ nicotine intake matches with the 20–30 µg m−3 estimate of PM2.5 when using the Repace (2006) Rosetta Stone equations.

There are two possible explanations for the low concentrations of airborne nicotine we measured: firstly it is possible that the nicotine results we report are correct given that they were collected using a validated method; alternatively, it is possible that some systematic loss of nicotine occurred during the storage, transportation, or analysis of the filters.

While we acknowledge the possibility of the former, we consider that the latter is more plausible given the evidence of SHS exposure that we report here and the lack of alternative sources of the PM2.5 measured. We also note a strong and consistent relationship (R-squared = 0.91) between the airborne nicotine values and the PM2.5 concentrations suggesting that the measured PM2.5 was reflecting particle emissions that were linked to SHS. After extensive discussions with the laboratory to explore potential reasons for the low nicotine results, we identified that, for a week prior to shipping to the USA, the nicotine monitors were stored in a laboratory where temperatures regularly exceeded 27°C. We are also unaware of the environmental conditions in terms of temperature and pressure that the filters may have experienced during airfreight transport. We postulate that there may have been some systematic nicotine loss from the filters during either storage and/or air transportation to the USA after collection that resulted in a systematic error. Future work should aim to collect and analyse spiked samples to examine if such losses occur and calculate recovery efficiencies for this methodology.

The high level of agreement between the Dylos measured PM2.5results and the nicotine concentrations suggest that real-time measurement of PM2.5 with these low-cost devices presents considerable advantages over nicotine monitoring. The information on temporal changes of SHS concentrations, coupled with the simplicity of data collection with no laboratory analysis costs, provide significant practical benefits for future work in this area.

The salivary cotinine data taken at the end of the work-shift indicates a GM (GSD) value of 0.15 (2.48) ng ml−1. This compares to a GM (GSD) of 0.12 (3.39) ng ml−1 in 54 prison workers in England and Wales in 2015 (Semple et al., 2015a). A salivary cotinine GM of 0.09 ng ml−1 was reported in the most recent (2014/5) population-level survey of non-smoking adults in Scotland (Scottish Health Survey, 2015), while historically, the GM (GSD) value measured in bar workers in Scotland prior to smoke-free legislation in 2006 was 2.94 (2.28) ng ml−1 (Semple et al., 2007a). These data indicate that prison staff have exposure that is markedly higher than the general adult non-smoking population in Scotland, but also suggest that prison workers experience much lower exposures than those of bar workers prior to smoke-free legislation in 2006.

Using the difference between the pre- and post-shift saliva samples, we utilized Repace and colleagues’ (2006) ‘Rosetta Stone’ equations to estimate a PM equivalent exposure during the work-shift. The median increase in salivary cotinine for the 149 non-smoking workers to whom we could apply this method was 0.138 ng ml−1; this equates to a work-shift average of SHS-PM of 24.8 µg m−3. We acknowledge that this method excludes over 60% of those non-smoking prison staff who arrived at work with salivary cotinine levels <LOD and so may not be representative of the exposure of all prison workers. However, we note that the results generated by this approach are broadly in agreement with the personal PM2.5measurements made on 22 prison staff in England monitored for an average of 4.2 hours (Jayes et al., 2016) where a mean value of 23.5 µg m−3 was reported, and a study of six English prisons (Semple et al., 2015a) where the GM (GSD) personal exposure of 30 prison staff to PM2.5 was 19 (2.2) µg m−3.