Methods

Questionnaire development and implementation

A questionnaire designed to assess the prevalence and characteristics of hospital smoking policies was developed based upon a review of the literature and the input of smoking cessation experts. The questionnaire was pilot tested at 20 volunteer hospital sites, 10 of which were recruited because of their expressed interest in smoke-free campus initiatives and 10 that were recruited from a cohort of Pennsylvania hospitals in various stages of implementing a smoke-free campus. Although pilot sites were not selected based on their smoke-free campus status, eight of these sites reported a smoke-free campus and the remaining 12 indicated that they did not currently have a smoke-free campus. Modifications were incorporated into the final version of the questionnaire.

The questionnaire was organised into discrete domains addressing: (1) current smoking policy and policy exceptions; (2) the smoke-free campus policy and its implementation; (3) compliance and enforcement issues; (4) smoking cessation services available to patients and employees; (5) other employee considerations; (6) champions and barriers; and (7) community and marketing considerations.

All respondents were instructed to select one of six statements that “best” described their hospital’s current smoking policy. Based upon their selection, respondents were then split into two groups: hospitals that had implemented a smoke-free campus policy and those that had not. A hospital was considered to have a “smoke-free campus” if the survey respondent selected either of these options:

(a) “Smoking is prohibited on all hospital owned or leased property/facilities (indoors and outside). We have no designated smoking areas on the campus and are a totally smoke-free campus”.

(b) “Smoking is prohibited on hospital owned or leased property/facilities (indoors and outside); We have no designated smoking areas on the hospital campus, but there are remote locations outside the smoke-free perimeter of the campus (eg, parking lots, storage warehouses, etc) that are not covered by the smoke-free policy”.

All other hospitals were designated as “not-smoke-free campus”. Hospitals that had not implemented a smoke-free campus policy were further differentiated based upon their answer to the question: “Does your hospital plan to adopt a completely smoke-free campus policy?” Those respondents that indicated “planning is under way” were directed to the full set of questions in the subsequent six domains that were similar to those presented to the “smoke-free” group. Respondent hospitals without plans to implement a smoke-free campus policy were not asked questions concerning policy implementation, compliance or enforcement; however, they were asked to identify factors that influenced maintaining the current policy as well as factors that might influence changes to current policy. No hospital was asked more than 40 items in total in the questionnaire.

Participants

A total of 4494 Joint Commission-accredited hospitals were invited to complete the web-based questionnaire which was administered in an electronic format through SurveyMonkey (www.SurveyMonkey.com) and distributed as a URL (uniform resource locator) link embedded in the email invitation. The email introduced the research project, solicited voluntary participation and reassured the recipients that the questionnaire and their participation (or lack thereof) was not associated with Joint Commission-accreditation activities. Links to the information sheet/informed consent form and the web-tool navigation instruction sheet were also embedded in the message.

Joint Commission-accredited hospitals represent approximately 80% of all US hospitals and 90% of hospital beds and include general medical-surgical, paediatric, behavioural health, critical access, long-term care and rehabilitation facilities in the US, Puerto Rico, the US territories and military facilities. The invitation to participate in the study was emailed to The Joint Commission-accreditation contact or the chief executive officer (if the contact person’s email was not available) with the request that the survey be completed by a staff member most familiar with the smoking policies. Seventy-five of these emails could not be delivered despite multiple attempts; no other contact with these hospitals was pursued.

To encourage a greater response rate, a total of seven reminder emails were sent at 2-week intervals throughout the data collection period to all non-responding hospitals. Each reminder note encouraged participation, repeated the salient points from the initial email and included the embedded links to the questionnaire, information sheet and web-tool navigation instructions. Reminders were also e-mailed to those hospitals that had started but not completed the questionnaire. At the end of the data collection period (6 November 2007–22 February 2008), responses were reviewed for incomplete, duplicate, or inconsistent records from the same organisation. Hospital personnel were contacted to clarify and correct discrepancies in the data.

Definition of measures

Smoking cessation counselling performance

The rate of smoking cessation counselling for participating hospitals was determined using nationally standardised performance measures developed by The Joint Commission.24 Since 1 July 2002, in order to comply with The Joint Commission hospital-accreditation requirements, accredited hospitals have submitted monthly performance measure data on a quarterly basis. Among the measure data regularly submitted by hospitals are rates for three identically defined measures that address the provision of smoking cessation counselling to adult patients being treated for acute myocardial infarction, heart failure and pneumonia. These measures target all inpatients, 18 years of age or older, with a history of smoking cigarettes anytime during the year before hospital admission. Rates are calculated based upon the number of patients receiving some form of smoking cessation counselling (that is, documentation of any direct discussion with the patient, brochures, handouts, videos, referral to smoking cessation resources and/or prescription for a smoking cessation aid) before discharge. Patients who are transferred to another hospital or to hospice care, leave against medical advice or die during the stay are excluded from the population, since the hospital may not have had an opportunity to provide these patients with appropriate counselling. It is an expectation, therefore, that all the patients who are included in the measure population should receive smoking cessation advice/counselling. These measures have been reviewed and endorsed by the National Quality Forum,25 have been adopted for use by the Hospital Quality Alliance and have been used in other published research.23 26

Demographic variables

Demographic characteristics for bed size, ownership (profit, not-for-profit, federal government and non-federal government) and hospital service type (general-full service, critical access, behavioural health/psychiatric or specialty hospital) were obtained from data supplied in each hospital’s accreditation application to The Joint Commission. Additional demographic characteristics (consolidated metropolitan statistical area (CMSA) and AMA-resident status were obtained from the American Hospital Association Annual Survey Database for Fiscal Year 2003 Data (American Hospital Association, 2005).

Statistical analysis

SAS was used for all data analysis. All differences between the groups were compared using a χ² test; p values were deemed significant if p<0.05. Performance measure rates were calculated at the patient-level, aggregating all numerator and denominator cases within groups, based upon a 12-month period from January–December 2007. Secondary analyses of performance measure rates incorporated the same approach, but relied upon performance measure data collected for patients discharged during 2005. The relation among smoke-free campus status and various demographic characteristics was analysed using a logistic regression model, using smoke-free campus status as the dependent variable. The relation between smoke-free campus status and each of the smoking cessation counselling measure rates (acute myocardial infarction, heart failure and pneumonia) were analysed using analysis of variance (ANOVA), with smoke-free campus status and the demographic variables as the covariates. In order to evaluate the relation between the measure rates and hospitals considered to be “early adopters” of a smoke-free campus policy, a third analysis was conducted. Hospitals were ranked based upon the year in which they adopted their policy, and hospitals in the lowest decile were identified as “early adopters” (policy adopted before 2005). The relation between early adopter status and each of the smoking cessation counselling measure rates (acute myocardial infarction, heart failure and pneumonia) were analysed using analysis of variance (ANOVA), with early adopter status, smoke-free status and the demographic variables as the covariates.

Results

A total of 1916 hospitals (a response rate of 42.6%) completed the smoking policy questionnaire. Immediately following the survey administration period, a matched, random sample of 179 non-responder hospitals was contacted by telephone in order to assess the effects of non-respondent bias; 105 of these hospitals (58.7%) were successfully contacted and each agreed to answer the first three questions from the survey which addressed the current smoking policy and policy exemptions.

Among hospitals that responded to the questionnaire, 865 (45.2%) reported a smoke-free campus policy with no designated smoking areas and 1051 (54.9%) reported a policy that permitted smoking either outdoors or within designated indoor areas or both. The called non-responders reported a similar distribution of smoking policies. Among these hospitals, 48 (45.7%) reported a smoke-free campus policy and 57 (54.3%) reported that they did not have a smoke-free campus policy (table 1). There was no statistically significant difference between the responders and called non-responders with respect to their smoke-free status.

Responders did not differ from all survey recipients with respect to their smoking cessation counselling performance measure rates. There were no statistically significant differences between these two groups on the demographic characteristics of hospital service type, size, teaching status and urban or rural setting. Statistically significant differences (p = 0.044) were noted between responders and all survey recipients with respect to ownership type. The proportion of responders versus all survey recipients was higher for not-for-profit hospitals (0.536 of survey responders vs 0.525 of survey recipients) and non-federal government (that is, state, county or other local government) hospitals (0.165 vs 0.145) compared to federal government owned (that is, Veterans Administration and military) hospitals (0.048 vs 0.056) and for-profit hospitals (0.251 vs 0.274).

To facilitate data analysis, all 1916 respondent hospitals were divided into two groups—those with a smoke-free campus policy and those without. These groups were then further subdivided into one of four subgroups: (1) smoke-free campus, (2) smoke-free campus with exceptions, (3) planning to implement a smoke-free campus policy, and (4) no smoke-free campus and no plans to change (table 2). The hospitals that reported a smoke-free campus policy (n = 865; 45.2%) were assigned to one of the first two subgroups, based upon their answers to the policy exception questions. Those reporting both a smoke-free campus policy and no exceptions to the policy were assigned to the smoke-free campus subgroup (n = 657; 35.7% of all respondent hospitals). Hospitals reporting a smoke-free campus with exceptions (either for specific patient populations or granting physicians or administrators authority to over-ride the smoke-free campus policy) were assigned to the smoke-free campus with exceptions subgroup (n = 208; 11.3% of all respondent hospitals). Of the 865 smoke-free hospitals, 804 (92.9%) answered the survey question that assessed whether or not their policy extended to all tobacco products. A total of 691 (85.9%) hospitals reported that their policy covered all tobacco products—that is, cigars and smokeless tobacco.

Respondent hospitals initially identified as not having a smoke-free campus policy (n = 1051; 54.9%) were also assigned to one of two subgroups based upon whether they were actively planning to adopt a smoke-free campus policy. Hospitals that reported that they were actively planning to implement a smoke-free campus policy were assigned to the planning to implement a smoke-free campus policy subgroup (n = 283; 15.4% of all respondent hospitals). Those reporting no active plans (or considering the option but without specific plans) to implement a smoke-free campus policy were assigned to the final subgroup, no smoke-free campus and no plans to change (n = 692; 36.6% of all respondent hospitals) (table 2). Seventy-six hospitals could not be assigned to a subgroup, as they failed to complete the question related to planning for a smoke-free campus. While these hospitals were included in the analyses comparing smoke-free campus to non-smoke-free campus hospitals, they were excluded from the subgroup analyses.

Demographic comparisons

Smoke-free campus and non-smoke-free campus hospitals were compared across several demographic categories. Table 3 displays the absolute rates of smoke-free versus non-smoke-free campus hospitals across all demographic variables studied. Logistic regression revealed that ownership type was the only demographic variable that was statistically significantly associated with a hospital’s smoke-free campus status. Private, non-profit hospitals were three times as likely as for-profit hospitals to have adopted a smoke-free campus policy (odds ratio (OR) 3.14, confidence interval (CI) 2.24 to 4.40). Hospitals owned by the federal government were the least likely to have adopted a smoke-free campus policy. In comparison, for-profit hospitals were almost 10 times more likely (OR 9.86, CI 2.29 to 42.33), public (non-federal government) hospitals were almost 16 times more likely (OR 15.60, CI 3.66 to 66.48) and private non-profit hospitals were almost 31 times more likely (OR 30.91, CI 7.38 to 129.40) than federal government-owned hospitals to adopt a smoke-free hospital campus. Differences observed among the other demographic variables (that is, hospital service types, teaching status, bed size and rural or urban location) did not provide a statistically significant contribution to the model.

Geographical distribution of hospitals with smoke-free campus policies

The percentage of hospitals reporting a smoke-free campus policy (with or without exceptions) was calculated for each state and is displayed in figure 1. Variation in the adoption of smoke-free hospital campus policies was observed across states, ranging from 0%–88.9% (mean  = 47.9%). States were grouped based upon the percentage of smoke-free campus policies reported within each state: those reporting ⩾75% smoke-free campus policies (nine states); those reporting 50%–74% smoke-free campuses (14 states); those reporting 25%–49% smoke-free campuses (17 states); and those with <25% smoke-free campuses (10 states). Rates are not reported for states with both fewer than 10 survey responders and a questionnaire response rate lower than the survey’s national response rate of 43% (five states), as percentages derived from such small numbers were deemed too small to make any inferences regarding the overall smoke-free campus policy prevalence in the state.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Percentage of hospitals reporting a smoke-free campus, by state (based on all survey respondents, as of February 2008). The percentage of hospitals reporting a smoke-free campus policy was calculated for each state based on the total number of survey respondents from each state. A total of 1916 Joint Commission-accredited hospitals responded to the survey. Percentages are not displayed if a state had fewer than 10 survey respondents and a state response rate that was less than the national survey response rate of 43%. The smoke-free hospital campus designation was based upon the hospital’s self-report of current smoking policies (that is, smoking is prohibited on all hospital owned or leased property/facilities (indoors and outside), and no designated smoking areas are available within the perimeter of property controlled by the hospital).

Performance on standardised inpatient smoking cessation counselling measures

Using data from 2007, initial comparisons of smoking cessation counselling performance measure rates, between hospitals that had adopted a smoke-free campus and hospitals that had not adopted a smoke-free campus showed a statistically significant difference (p<0.001). Smoke-free campus hospitals (n = 865) had mean rates for the AMI, heart failure and pneumonia measures of 98.4%, 96.5% and 94.3%, respectively. Hospitals that did not have a smoke-free campus (n = 975) had measure rates of 98.0%, 95.2% and 93.1%, respectively. Comparisons of smoking cessation counselling performance measure rates also revealed statistically significant differences across subgroups (p = 0.001) for all three measures (table 3). A second analysis using ANOVA, however, to compare performance measure rates with smoke-free campus status and the demographic variables as the covariates, revealed that these differences were better explained by hospital demographic factors than by the hospital’s smoke-free campus status.

Given the high national measure rates, and the likely influence of ceiling effects, an additional analysis was performed to investigate differences in performance measure rates among early adopters of smoke-free campus policies. In order to identify hospitals that had adopted a smoke-free campus policy earlier than their peers, all hospitals with a smoke-free campus policy were ranked based upon the year in which they adopted their policy (table 4). Hospitals in the lowest decile were identified as “early adopters” (which included 77 hospitals that adopted their policy before 2005). Using performance measure data from 2005 (the following year), the smoking cessation counselling measure rates for the 77 early-adopter hospitals were compared to the smoking cessation counselling rates of all hospitals that currently did not have a smoke-free campus or plans to adopt one (n = 692). Once again, data were analysed using analysis of variance (ANOVA), with early adopter status, smoke-free status and the demographic variables as the covariates. Although statistically significant differences were observed between smoke-free campus and non-smoke-free campus hospitals for all three measures (p<0.05), the smoking policy did not account for these differences. Differences were associated with teaching status, ownership type and/or bed size, rather than the smoking policy.

Discussion

To the best of our knowledge, this study provides the first systematic estimate of the prevalence and characteristics of hospitals that have, or have not adopted smoke-free campus policies. When The Joint Commission’s indoor smoking ban was put into effect in 1992, only 2.7% of hospitals reported having implemented a smoke-free campus.1 Today that number is 45.2%, and another 15.4% of hospitals report that they are actively planning to implement such a policy. This trend may be of interest to hospital executives who may be considering the implementation of a smoke-free campus policy. Coinciding with this trend, there is also a growing number of resources available to assist hospitals with their efforts.15 16 27

Demographic differences that were observed between hospitals that have adopted or have not adopted a smoke-free campus policy were both expected and surprising. Previous research has indicated that more restrictive smoke-free policies reduce employee smoking and increase their cessation rates.1 23 Given the frequently voiced concerns related to the impact of more restrictive smoking policies on employee retention and patient and family member satisfaction, it was not surprising that for-profit hospitals appeared somewhat less inclined to adopt a smoke-free campus policy. Similar differences between for-profit and non-profit hospitals have been observed in several recent studies.28 29 30 31 Lower rates of adoption were also observed among federal hospitals. This was probably because of the influence of legislation governing all Veterans Administration (VA) hospitals. The Stagger’s Amendment to the Veterans Health Bill, (HR5192) which was passed on 1 October 1992, stipulates that “…each Department medical facility maintain a suitable patient indoor smoking area and provide access to such areas for patients and residents desiring to use tobacco products”.32 33 Such legislation makes it virtually impossible for VA hospitals to adopt a completely smoke-free campus.

Given the often cited belief that it can be counterproductive to encourage patients to quit smoking during psychiatric treatment, it was not surprising that psychiatric hospitals had lower rates of smoke-free campus adoption than other types of hospitals34 (although these differences were also accounted for by the hospital’s ownership type rather than the hospital type). What was surprising about this finding, however, was that as many as 31.5% of psychiatric hospitals have adopted a smoke-free campus policy and another 14.3% reported that they have active plans to adopt such a policy. Such adoption rates may encourage the efforts of smoking cessation advocates who are beginning to address the significant health risks that accompany the disproportionately high rates of smoking among psychiatric patients.34 35 The geographical distribution of hospitals that have adopted smoke-free campuses was also interesting. Expectations that lower smoke-free campus adoption rates might be observed in tobacco-producing states were not supported.

In this study, we also hypothesised that more restrictive hospital smoking policies would be reflected in measure rates of hospital compliance with clinical treatment guidelines regarding inpatient smoking cessation counselling. While statistically significant differences were initially observed between smoke-free campus hospitals and non-smoke-free campus hospitals, those differences were very small: 0.4%, 1.3% and 1.2% for AMI, heart failure and pneumonia, respectively, and they were better explained by differences in hospital demographic factors, rather than the hospital’s smoking policy. Similar findings for comparisons among early-adopter hospitals and non-adopters of smoke-free campus policies, suggests that there may be little relation between the adoption of smoke-free campus policies and the provision of smoking cessation counselling to patients being treated by the hospital. While we hypothesised that patients would be more likely to receive smoking cessation counselling in hospitals that adopted stricter smoking policies, the implementation and enforcement of a hospital’s non-smoking campus policies appears to have very little influence on the actions of clinical staff during the treatment process.

Limitations

The findings from our study need to be considered in light of several methodological limitations. First, the study population was limited to Joint Commission-accredited hospitals. While The Joint Commission accredits the vast majority of hospitals in the United States, survey results cannot be considered representative of non-accredited hospitals. The analysis of performance measure data was similarly restricted to only the acute care medical-surgical hospitals that submit smoking cessation measure data for AMI, heart failure and pneumonia to The Joint Commission. Psychiatric, specialty and critical access hospitals do not collect or submit these data, so these types of hospitals were not represented in the performance measure rate analysis. We do not know, therefore, what influence smoke-free campus policies may have on smoking cessation efforts within these types of facilities. Second, despite the similar distribution of survey responders and the non-responders on demographic characteristics and performance on quality measures, it is always a concern that survey non-responders will differ from those that voluntarily completed the questionnaire. This concern was addressed, in part, through the sample of non-responders that were contacted by telephone. Similar results between survey responders and non-responders offer some support for the generalisability of the results, but they do not eliminate the possibility of response bias; likewise, the 75 hospitals that failed to receive an email could have contributed to selection bias as they may have been demographically different from the group.

Third, although 865 (45.2%) respondent hospitals reported a smoke-free campus policy, 116 of them also identified exceptions to their policy. This raises the question: to what extent are these campuses really smoke-free? The study was not designed to evaluate the rigour and consistency with which smoke-free campus policies might be enforced nor to verify the self-reported smoke-free status. Subgroup analysis found no differences between the smoke-free and smoke-free-with-exceptions groups with respect to demographic characteristics or smoking cessation counselling measure rates, but it is not possible to rule out other differences between the groups.

It should also be noted that while the survey was sent to The Joint Commission-accreditation contact at each hospital, there was no mechanism to ensure that this individual would be the actual individual completing the survey. The survey instructions encouraged the recipient to forward the survey to the person best suited to answer it. This, coupled with the use of a web-based survey instead of employing another survey method, such as a paper or telephone survey, could have reduced or biased survey responses.36 Finally, it is important to note that the study represents a single snapshot in time.

Administration of the survey was completed in February 2008. Hospitals that implemented smoke-free campus policies since that date are not captured in the rates reported above. To estimate the number of hospitals that may have adopted a smoke-free campus policy by the end of 2008, it may be possible to add hospitals that indicated specific plans to adopt such a policy. Ongoing efforts to promote tobacco-free hospital campuses in North Carolina, for example, indicate that, as of October 2008, 87.3% of North Carolina hospitals had adopted tobacco-free campus policies.15 This number is entirely consistent with the combined percentage of survey responders from North Carolina who reported in February 2008 that they had already adopted a smoke-free hospital campus (66.1%) or had plans to adopt one (17.9%). Applying this estimation approach on a national scale suggests that by 2009, as many as 60% of US hospitals may have adopted a smoke-free hospital campus policy.

Conclusions

A growing number of US hospitals are choosing to adopt smoke-free hospital campus policies. As of February 2008, over 45% of US hospitals report adopting a smoke-free campus policy and over 15% report that they are actively pursuing the adoption of such a policy. Contrary to study hypotheses, the adoption of smoke-free campus policies appears to have very little influence on the rate at which hospitalised patients receive cessation counselling. Hospitals that have adopted these policies do not differ appreciably from those that have not adopted a smoke-free campus with respect to bed size, teaching status or urban and rural setting, although private, non-profit hospitals were more likely to adopt a smoke-free campus than other types of hospitals. Given the increasing prevalence of smoke-free hospital campuses, and the complexity that often accompanies the implementation and enforcement of these policies, research that can identify successful implementation and enforcement strategies may be of significant benefit to the healthcare community.