Predictors of opinions on prison smoking bans: Analyses of survey data from Scottish staff and prisoners

INTRODUCTION Policy-makers and practitioners need to understand characteristics associated with support for smoking restrictions to identify both potential allies and groups requiring particular support/targeted communication in the face of restrictions. Using data from prison staff and prisoners, we explored the structure and correlates of opinions relating to prison smoking bans. METHODS Questionnaires were completed by staff (online, N=1271; 27% return) and prisoners (paper-based, N=2512; 34%) in all 15 Scottish prisons in 2016–17. At that time, prisoners could smoke in their own cells and during outdoor recreation; staff smoking was prohibited anywhere on prison grounds. Staff and prisoner questionnaires included identical/very similar questions about opinions on smoking in prisons and prison smoking bans, own smoking behaviour, health and sociodemographic details. We also measured in every prison fine particulate matter (PM2.5) as a proxy for secondhand smoke (SHS) levels. RESULTS Principal components analysis identified two factors: ‘Positive about bans’ (higher scores among staff) and ‘Bans will be difficult’ (higher scores among prisoners). In multivariable analyses, ‘Positive about bans’ was associated with: not smoking (both staff and prisoners), better general health, more respiratory symptoms and working in an operational role among staff; and no asthma, more sensory symptoms, higher educational level and status/release date among prisoners. ‘Bans will be difficult’ was associated with: fewer sensory symptoms and lower prison SHS levels among staff and being a smoker among prisoners. In smoker-only analyses, heavier smokers were less positive about bans and more likely to believe bans will be difficult. CONCLUSIONS Results suggest it is possible to be positive about prison smoking bans whilst also recognising and/or concerned about potential operational difficulties, and that these opinions are associated with several characteristics additional to smoker status. Support for future prison bans may be stronger if staff have access to objective SHS exposure measures.


INTRODUCTION
Breathing in secondhand tobacco smoke (SHS) harms health, with SHS exposure estimated as causing around 1% of total worldwide mortality 1 . Restrictions to SHS exposure, which are most successfully achieved via smoke-free legislation, have clear health benefits 2 . Support for smoke-free policy is key to successful implementation 3,4 . It is therefore important to identify not only levels of support among those subject to smoking restrictions, but also characteristics associated with different levels of support, since these might enable those involved in implementation/ enforcement to identify potential allies 5,6 and better target measures to address expressed concerns and reduce potential problems.
Smoking restrictions have been increasingly introduced in public places since the 1970s, although with variation among jurisdictions in coverage. In many countries that have introduced workplace restrictions, exemptions occur for workplaces such as prisons, which are also regarded as 'homes'. In the absence of smoke-free policies, prisoner smoking rates are typically high, around 2-8 times those of the general population in studies internationally 7 . The health implications for prisoners who smoke and those (staff/prisoners) exposed to SHS have contributed to the introduction of increased restrictions or, in some jurisdictions, complete smoking bans in prisons 7 . We have previously documented Scottish prison staff and prisoner opinions on prison smoking bans, using survey and focus group data collected several months before the announcement that Scottish prisons would become smoke-free from November 2018 8 . This companion paper, based on survey data only, examines correlates of those opinions.
Surveys of both general populations and those in specific settings almost all find stronger support for hypothetical, proposed or currently implemented smoking restrictions among non-smokers. Examples within the general population include studies conducted in Australia 9 , Europe 10 , and Jordan 11 . Research within specific populations and/or settings also showing greater support for smoking restrictions among non-smokers is similarly international, including for example: Australian university staff and students 3 , Dutch psychiatric hospital staff 12 , and UK bar workers 13 .
Some studies have also identified other characteristics associated with greater support of smoking restrictions. These include older age 9,11,12 , being female 9 , and higher socioeconomic status 9 . Greater support of restrictions has also been found among those with more knowledge of, or stronger beliefs about, harms associated with smoking and SHS exposure 10,11 . Finally, although one study found little difference in levels of support according to self-rated health 10 , others have found greater support among those reporting that SHS irritates their eyes, that they dislike its smell and that increased restrictions would reduce SHS 'annoyance' 12 .
Some population-based or workplace-based studies have focused specifically on smokers. These have found more positive opinions about smoking restrictions and, variously, lower nicotine dependence, lighter smoking, fewer perceived smoking-related personal benefits, lower smoker identity, and greater intention to quit 5,6,9,14,15 .
Very few surveys have been conducted on opinions about smoking restrictions in prisons; also potentially relevant are surveys in secure hospitals. An Irish study surveyed 90 prison staff among those from other workplaces, reporting agreement with: 'Should there be a smoking ban in prisons?' (41% yes); prohibition of smoking in enclosed areas (79% agreed); and whether a complete ban would 'create more problems in the prison' (88% yes) 16 . As with population or other workplace-based studies, this study found greater support for restrictions among non-smokers. Greater support for smoking restrictions among non-smokers has also been found among Vermont prison staff (when there was a ban on indoor smoking among both staff and prisoners) 17 ; German prisoners (when smoking was allowed in cells) 18 ; Australian high security mental health inpatient facility staff 19 ; and UK forensic unit in-patients 20 . The Vermont study also found that support for smoking restrictions on prisoners was stronger among prison staff than among prisoners, and uniformed staff (likely to have more contact with prisoners) were more supportive of continuing to permit prisoners to smoke outdoors than non-uniformed staff 17 . Surveys in a US secure psychiatric unit before and after a total ban also found that at both time-points, staff were more likely than patients to support the ban 21 . policy changes; Phase 2 following the announcement, but before introduction of a smoke-free policy; and Phase 3 following policy implementation. These Phase 1 survey data were collected in late 2016/early 2017, several months before the (July 2017) announcement that all Scottish prisons would become completely smoke-free from 30 November 2018. The Scottish prison estate consists of 13 publicly and two privately managed prisons. Most hold a mix of offender types and include: 10 accommodating males only, one with females only and four with both males and females, one accommodating male young offenders (aged 16-21 years), and one open prison accommodating low-supervision adult male offenders. At the time of data collection, prisoners were allowed to smoke in their own cells and during outdoor recreation; staff and visitors were prohibited from smoking anywhere on prison grounds.
An online prison staff survey was conducted, open 1 November to 16 December 2016. The survey link and reminders were circulated to our staff contacts in all 15 Scottish prisons, with requests to forward these to other prison (but not NHS/visiting) staff. The prisoner survey was conducted via paper questionnaires between November 2016 and April 2017. In two prisons, TIPs staff were escorted around residential areas, distributed questionnaires to all prisoners who said they were willing to complete one, answered queries and helped with completion if necessary. In a third (the open prison), TIPs staff distributed questionnaires during an evening meal. In the remaining 12 prisons, questionnaires were supplied for prison staff to distribute to every prisoner and collect (in sealed envelopes protecting confidentiality), generally during an overnight lock-up. Staff and prisoner questionnaires included identical or very similar questions around opinions on smoking in prisons and prison smoking bans, own smoking, health and sociodemographic details.

Participants
Questionnaires were completed by 1271 staff (27%) and 2512 prisoners (34%). Response/return rates varied considerably among prisons: 10-38% for staff, and 10-60% for prisoners. The staff sample was identical with respect to proportions of males (71%) and females (29%), and very similar with respect to age, for Scottish Prison Service staff in post as at 31 March 2017 23 . The prisoner sample included a slightly higher proportion of females (7%) than within the Scottish prison population overall (5%) 23 , and prisoner smoking rates (74%) were slightly higher than those reported in the 2015 (72%) and 2017 (68%) Scottish prisoner surveys 24 . Table 1 shows question wordings and analytic categories (if appropriate) for all variables.

Staff years worked in prisons
Altogether, how many years have you worked in prisons in total?

Dependent variables
Prison smoking ban opinion items 9 were adapted from surveys of US prison staff opinions about restrictions to smoking in prisons 17 and Scottish bar workers' attitudes to smoke-free public places legislation 13 . They comprised seven items on whether prison smoking bans: 1) are a good idea, 2) cause a lot of trouble, 3) stop prisoners smoking long-term, 4) are hard to enforce, 5) are OK if stop smoking support available, 6) are OK if e-cigarettes available, and 7) are wanted by most staff; with five answer options and a single item on agreement with increased smoking restrictions in Scottish prisons.

Independent variables
Staff and prisoners were asked about their age, sex and education. The questionnaires also asked about: staff role, seniority band and number of years they had worked in prisons, and prisoner status (unconvicted, convicted and time to release). All participants were asked their current smoking status.
Smokers were asked about daily cigarettes  smoked, how much they had craved cigarettes today 25 , and whether they had tried a number of quit-smoking strategies in the past year.
All were asked health-related questions. These included health in general using a standard UK survey item 26 : 'has a doctor ever told you that you have asthma?'; and self-reported past month respiratory (wheezing/whistling, shortness of breath, morning cough, other cough, phlegm) and sensory symptoms (red/irritated eyes, runny nose/sneezing, sore/ scratchy throat). The symptom questions were based on International Union Against Tuberculosis and Lung Disease items previously included in studies of bar workers' health following smoke-free legislation 27 . The questionnaires also included the EQ-5D-5L, a standardised measure of health status including an item on current anxiety/depression 28 .
Between September 2016 and January 2017, TIPs also measured staff and prisoner SHS exposure in all Scottish prisons using multiple methods, including 6-day area measurement of fine Particulate Matter (PM2.5) 22 .

Analyses
Factor analyses (principal components analysis, varimax rotation) were conducted on the eight prison smoking ban opinion items. Exploratory analyses found one ('Prison smoking bans are OK if prisoners are allowed e-cigarettes or vapes') loaded on different factors for staff and prisoners (Supplementary Table 1). Analyses excluding this item resulted in two identical factors for staff and prisoners that explained 16% more total variance than an analysis constrained to a single factor (Supplementary Table  2). Factor analysis was therefore conducted of the remaining seven items for the prisoner and staff groups combined. The two resulting opinion factors (described in Results) were saved and formed the dependent variables in analyses of association.
Staff-prisoner differences on the prison smoking bans opinion factors (F-tests) and all other (independent) variables (chi-squared tests) were examined. Bivariable analyses (F-tests) examined differences in the two opinion factors according to each independent variable, separately for staff and for prisoners. ANOVAs entering the independent variable and one representing staff versus prisoner tested for significant interactions, indicating different associations within staff versus prisoners. Multivariable analyses (SPSS general linear models) entered those independent variables identified via bivariable analyses as significantly related (p<0.05) to each opinion factor, separately for staff and for prisoners. Similar analyses were conducted, restricted to smokers, entering cigarettes per day, craving and past-year quit attempt.
We tested for clustering by introducing a random intercept for prison in two key analyses (models entering staff versus prisoner, and smoker versus non-smoker; Supplementary Table 3). This was nonsignificant, therefore all analyses were conducted via standard linear models. As staff and prisoner questionnaire return rates varied among prisons, simple weights were derived to adjust for this. The results of final multivariable analyses based on unweighted and weighted data (Supplementary Table  4) were virtually identical, so results of unweighted analyses are presented. Table 2 shows how each opinion statement loaded on the two opinion factors, for staff, prisoners and both groups combined. The first factor ('Positive about bans', 45.8% variance explained) included five items loading 0.6 or greater (prison smoking bans: are a good idea, are OK if stop smoking support available, stop prisoners smoking long-term, are wanted by most staff, and in favour of increased smoking restrictions in Scottish prisons). The second factor ('Bans will be difficult', 22.4% variance explained) included two items loading 0.6 or greater (prison smoking bans: are hard to enforce, and cause a lot of trouble).

RESULTS
Supplementary Table 5 shows mean scores on the opinion factors and distributions on the independent variables. Staff scores were significantly (p<0.001) higher than those of prisoners on 'Positive about bans' (mean factor scores 0.647 and -0.351, respectively), but lower on 'Bans will be difficult' (mean scores -0.234 and 0.127, respectively). There were also significant staff-prisoner differences on all independent variables measured across both groups. Prisoners were more likely to be male and younger. Prisoners were also much more likely than staff to be smokers (74.2% vs 9.7%) and, among smokers, prisoners were heavier smokers, reported more craving and were less likely to have made a past-year quit attempt. Prisoners were more likely to report doctor-diagnosed asthma, more respiratory symptoms, fewer sensory symptoms and much higher rates of anxiety/depression. While similar proportions of staff and prisoners worked/lived in one of the prisons with lower measured SHS levels, staff were more likely than prisoners to be in one with higher levels. Most staff respondents had school-leaving or further educational levels and worked in operational and non-managerial roles. Most prisoners had left education by the age of 16 years and were convicted with three months or (much) longer until release. Table 3 shows non-smokers scored higher on the 'Positive about bans' factor among both staff and, particularly, prisoners. Among both, those reporting the best health, but also those reporting more sensory symptoms had higher 'Positive about bans' scores. Among staff, those reporting more respiratory symptoms were also more positive about prison smoking bans, while among prisoners, those reporting no asthma were more positive. Higher 'Positive about bans' scores were also seen for older prisoners and staff working in prisons with higher measured SHS; neither sex nor anxiety/depression were associated with 'Positive about bans' in either group. In staff-   Continued only analyses, those in operational roles had higher 'Positive about bans' scores. Among prisoners, those with more education and those with longer time until release were more positive, while unconvicted prisoners were least positive. Finally, among smokers only, those reporting lower cigarette consumption (both groups) and prisoners with fewer cravings and a past-year quit attempt were more positive about prison smoking bans. Table 4 shows that among both groups, younger people and smokers had higher 'Bans will be difficult' factor scores. Among staff, those reporting worse health, but also those reporting no respiratory or sensory symptoms were more likely to believe bans will be difficult. Prisoners reporting greater anxiety/ depression and staff in prisons with lower measured SHS had higher scores on this factor. Higher 'Bans will be difficult' scores occurred among staff with least experience of working in prisons and prisoners with least education and those who were unconvicted. In smoker-only analyses, prisoners with greater cravings, but also those with a past-year quit attempt, had   higher 'Bans will be difficult' scores.
In multivariable analyses, 'Positive about bans' factor scores remained significantly associated with not smoking for both staff and prisoners; with better general health, more respiratory symptoms and working in an operational role among staff; and with no asthma, more sensory symptoms, higher educational level and status (unconvicted versus convicted and release date) among prisoners (Table  5). Variance explained was much higher with respect to prisoners, attributable to the very strong association between own smoking and 'Positive about bans'.
Fewer variables were independently associated with 'Bans will be difficult', and variance explained was much lower. This was particularly the case for prisoners, where only current smoking was significantly positively associated in the multivariable model. Among staff, fewer sensory symptoms and lower prison SHS levels, but not own smoking, were independently associated with 'Bans will be difficult'.
Finally, multivariable analyses restricted to prisoners who smoked (no equivalent analyses for staff since multiple variables were not significantly associated with either opinion factor among staff smokers) showed that all variables significant in bivariable analyses of association remained so in the multivariable analyses. Thus, among prisoners who smoked, cigarettes per day remained inversely associated with 'Positive about bans', and craved cigarettes and past-year quit attempt remained  associated with both opinions factors. Given staff versus prisoners and smoker versus non-smoker differences in scores on the two smoking ban opinion factors, additional analyses (not shown in Tables) were conducted to see if the much higher prisoner smoking rate could account for differences in staff and prisoner opinions. Whilst the differences decreased substantially after adjustment for smoking status, they remained significant: 'Bans are positive' according to staff/prisoner F=1050.6 (p<0.001) before, and F=110.8 (p<0.001) after adjustment for smoking status; whilst 'Bans will be difficult' according to staff/prisoner F=108.8 (p<0.001) before, and F=10.1 (p=0.002) after adjustment.  prison smoking bans, collected before announcements about a smoke-free policy implementation date in all Scottish prisons, suggest opinions in both groups could be related to two underlying dimensions (factors). The first, accounting for most variance, represented 'positive' opinions towards prison smoking bans, and the second highlighted their potential difficulties. Importantly, this suggests it is possible to be both generally positive about prison smoking bans, whilst also recognising (and potentially concerned about) the operational difficulties they may bring. This was also evident in analyses of qualitative data from Scottish prison staff obtained around the same time that found 'Staff views were influenced by beliefs about: acceptability of the policy in principle and whether/how bans could be achieved' 8 .

Analyses of survey responses from Scottish prison staff and prisoners on items eliciting opinions about
Consistent with almost every other study within both general 9-11 , and specific populations or settings 3,12,13 , including prisons and forensic psychiatric units [16][17][18][19][20] , non-smokers were more likely to be positive about prison smoking bans. They were also less likely to suggest their introduction would bring difficulties. The effects of smoker status were particularly marked for prisoners, probably because staff were not allowed to smoke on prison premises at the time of the survey. This in turn would mean staff smokers, while potentially more sympathetic than non-smokers towards smoking prisoners, had little to lose from a smoking ban (and potentially gains in terms of their own health, reduced temptations to smoke and perceptions of unfairness from seeing prisoners smoking). Also consistent with studies of similar institutions 17,21 , staff were more positive about bans, and less likely to anticipate difficulties than prisoners. Many, but not all, of the staff and prisoner opinion differences were explained by prisoners' higher smoking rates.
Previous studies have identified a range of other characteristics associated with greater support for smoking bans (in a range of contexts), including: older age 9,11,12 ; being female 9 ; having higher socioeconomic status 9 ; dislike of, or symptoms associated with, SHS 12 ; and, among smokers, variables relating to heaviness/ dependence and intention to quit 5,6,9,14,15 . Although our analyses did not find differences between the opinions of males and females, there were differences according to age (older prisoners more positive; younger staff and prisoners more likely to identify potential difficulties) and, among prisoners, educational level (those with least education were least positive and most likely to identify potential difficulties). Although there were no differences according to staff band (managerial/ non-managerial), those in operational roles (so more likely to have regular exposure to prisoners' SHS) were more positive about prison smoking bans, and those who had worked for less time in prisons (i.e. less experienced) were more likely to suggest bans would bring operational difficulties. Unconvicted prisoners were least positive about prison smoking bans and most likely to believe they would cause difficulties, while those with the longest time to release were most positive. Given that, in the absence of smoking bans, smoking is a strong part of prison culture 29 , this finding is perhaps unexpected, but may be because those who are unconvicted have little incentive to 'buy into' smoking restrictions and/or are concerned about measures further restricting their freedoms. Among staff, but not prisoners, those in prisons where our objective SHS measurements showed higher SHS exposure levels 22 , were more positive about bans and less likely to believe they would cause difficulties. Although staff completed the survey before our SHS results were available, those working in prisons where levels were higher may have inferred this from their own experience and knowledge of their prison's architecture and ventilation systems.
Few studies have examined associations between health and opinions about smoking restrictions. Among both staff and prisoners, those reporting better general health were more positive about a ban, while staff reporting better health were least likely to say bans will be difficult, and prisoners with no asthma were more positive. Those already in poor health may disregard 'additional' health harms or, perhaps, have a more negative outlook generally, including in respect to a ban. However, some of these differences disappeared in the multivariable analyses, suggesting that they were partially accounted for by (non)smoker status. Importantly, experience of sensory symptoms (red/irritated eyes, runny nose/ sneezing, sore/scratchy throat) associated with SHS exposure, remained significant in multivariable analyses, suggesting that they may play an important role in reminding people of ongoing exposure to a potential health hazard.
The results of analyses restricted to smokers were also largely consistent with previous studies of attitudes towards smoking restrictions in other populations that have found more positive opinions about smoking restrictions among lighter, less dependent and/or committed smokers 5,6,9,14,15 . They are also in line with evidence that more nicotine dependent prisoners are a group that might be more likely to try to continue smoking after a prison smoking ban 30 . Heavier, more dependent smokers, without a recent quit attempt (potentially representing more committed smokers) were less positive about prison smoking bans. These associations were generally significant for prisoners, and in the same direction for staff, among whom numbers of smokers were relatively small. More dependent smokers and those reporting a recent quit attempt were also more likely to view bans as bringing operational difficulties.
Although counter-intuitive, perhaps the latter group were reflecting on their own difficulties in quitting. This study has both strengths and limitations. Strengths include its size and the fact that it encompasses both staff and prisoner views collected, for the first time, across all prisons within a country's criminal justice system. Another major strength is the fact that identification of two opinion dimensions in these quantitative analyses is consistent with the results of our analyses of independent qualitative data from prison staff on smoking bans in prisons. An important limitation is the relatively low response/ return rates and potentially unrepresentative samples. However, those included were very similar to all Scottish prison staff and prisoners with respect to characteristics on which comparable data are available. Another limitation is the cross-sectional nature of the self-report data, hence we cannot draw conclusions on causality, particularly on relationships between the health-related variables and opinions on prison smoking bans. Here, the relative lack of association between anxiety/depression and opinions provides some reassurance that the obtained associations were not simply the result of negative affectivity.
Previous studies have highlighted the importance of understanding characteristics associated with different levels of support for smoking restrictions as a way for policy-makers and practitioners to identify potential allies when introducing restrictions 5,6 , and to identify groups who may need particular support or more targeted communication in the face of restrictions.
Our analyses suggest that although smoker status and, among smokers, dependency and smoker identity are significant, the presence of sensory symptoms related to SHS exposure and working in environments where people are more heavily exposed to SHS were determinants of levels of support for smoking restrictions. The latter is a key finding, suggesting that future prison (or other workplace) bans internationally might be more strongly supported by staff if objective measurements of SHS exposure are available; such measurements could be a routine first step in the process towards smoke-free prisons. Identification of two opinion dimensions in our analysis, with the suggestion that it is possible to be both generally positive about prison smoking bans and concerned about potential associated operational difficulties, also has practical implications. It indicates that those who raise concerns over practical issues about implementation/enforcement of prison smoking bans are not necessarily unsupportive of their introduction and, equally, those who express strong support for prison smoking bans may, nevertheless, have some apprehension over unintended negative consequences. While this may be particularly important in respect to prison smoking bans where the operational difficulties are potentially very significant, it suggests policy-makers and implementers need to acknowledge the legitimacy of both dimensions of opinion in their measures to address concerns and mitigate potential problems.

CONCLUSIONS
Policy-makers and practitioners need to identify potential allies and adopt appropriate communication strategies when introducing smoking restrictions. Those operating within prison services would benefit from recognising the complexity of staff and prisoner opinions about changes to smoking rules in their communication strategy, since those who appear generally positive about prison smoking bans may also be concerned about associated potential operational difficulties. Associations between opinions and characteristics, including both symptoms related to SHS exposure and measured SHS levels, suggest support for future smoking restrictions in prisons (or other workplaces) may be stronger if links with symptoms are highlighted and objective SHS exposure measures included in the communication strategy.