Adjusted weekly numbers of hospitalizations for acute COPD exacerbations and acute coronary syndromes (ACS). Footnote: Global P values for adjusted incidence rate ratio trend tests (reference = period 1, weeks 26 to 130), see Table 2. Curves displayed using smoothed values (Stata lowess function). doi:10.1371/journal.pone.0090417.g001 

Contexts in source publication

Context 1

... (Canton of Geneva, other or unknown), and diagnosis at admission. Because the definition of transient ischemic attack changed in 2009 [23], transient ischemic attack and ischemic stroke diagnoses were combined into a single category (i.e. ischemic cerebrovascular accident). As a quarter of patients included in the study had their residence outside of the Canton of Geneva, which could mitigate the impact of smoke-free legislations in Geneva, we also analysed data restricted to residents of the Canton of Geneva. The means and frequencies with standard deviation and 95% confidence intervals (95%CI) of study variables were computed. The mean number of weekly admissions for each diagnosis was calculated from the number of diagnosis-specific hospitalisations in each study period. We used Poisson regression models to compute the 95% CIs around these estimates and to test the hypothesis that there was a change in the mean number of weekly admissions for each diagnosis between the study periods corresponding to changes in smoking ban legislation. Dummy variables for the periods were created and the impact of smoking ban legislation was estimated by incidence rate ratios (IRR) of diagnosis-specific hospitalizations for the different periods, considering period 1 as the reference period. Some of the disease-specific hospital admissions, such as acute coronary syndrome, acute exacerbation of COPD, and pneumonia, are known to vary with seasons or to increase during the influenza epidemic [24]. Therefore, regression models were adjusted for season (defined by calendar months) and influenza epidemic periods using surveillance data of the Swiss Office of Public Health [25–28]. To account for the potential secular trends of diseases that would be independent of the legislation periods, we included a linear time trend in the models. Models were additionally adjusted for age and gender. The seasonal effect and secular trends were taken into account with methods previously used to allow comparison to previous large population-based studies that measured the impact of smoking bans on health outcomes [12,15,16,20]. We also examined whether the slope of the trend change across periods, by adding an interaction term between the study periods and the linear predictor for time to the models and using likelihood ratio tests [12]. For each outcome, we performed a global trend test of adjusted incidence rate ratio across four periods, using period 1 (weeks 26 to 130) as the reference period, with its result expressed as a global p-value. To estimate the events prevented, the admission rates obtained among Geneva residents were applied to the counts of population aged over 15, using census data of Geneva for each study period [29]. Hospital days and costs prevented by the smoking ban were estimated using the University Hospitals of Geneva administrative data. All statistical analyses were performed using Stata 11.0 (Stata Corp, College Station, USA). Overall, 5345 patients with a first hospitalization for acute coronary syndrome, ischemic stroke, acute COPD exacerbation, pneumonia and acute asthma were included (Table 1). Patients were predominantly males, had a mean age of 67 years, stayed on average for 11 days in hospital, and three-quarters of them lived in Geneva. Admissions for acute cardiovascular events were more numerous than for acute respiratory diseases. The socio-demo- graphic profile and admission diagnoses were similar across the 4 study periods and for the sub-sample of Geneva residents. The mean number of weekly hospital admissions for all studied conditions differed across periods for all population groups (Table 2). The number of hospitalisations for pneumonia and asthma was particularly low in each study period. For acute COPD exacerbation, the weekly number of hospitalisations significantly dropped over the 4 periods from 2.45 to 1.54 (p , 0.0001) (Table 2). The adjusted IRR decreased significantly in all periods after the initial smoking ban; it reached 0.54 [95%CI: 0.42–0.68] for all patients and 0.53 [95%CI: 0.41– 0.68] for Geneva residents during the final smoking ban (period 4). There was evidence of a change in the slope of the trend line across the study periods (P = 0.001 for periods x secular trends interaction). The reduction started before the legislative smoking ban and became stable in the second part of the period 3, when the ban was legislatively suspended but partially maintained in practice (Figure 1). Hospitalizations for acute coronary syndrome decreased only during the final smoking ban (period 4), where we measured a trend close to significance: adjusted IRR was 0.90 [95%CI: 0.80– 1.00] for all patients and 0.89 [95%CI: 0.78–1.02] for Geneva residents (Table 2). The slope of the trend line did not change across the periods (P . 0.05 for periods x secular trends interaction). The reduction started after the first legislative smoking ban was suspended but partially maintained (period 3) (Figure 1). Despite variations in the number of admissions, adjusted IRR of hospitalisations for pneumonia, acute asthma and ischemic stroke, did not significantly change throughout the 4 periods. This study shows that a legislative smoking ban is associated with a strong, highly significant reduction in hospitalisations for COPD exacerbation and a trend for decreased admissions for acute coronary syndrome. We observed no significant change in admissions for pneumonia, acute asthma and ischemic stroke. Among the strengths of this study is the large sample covering the large majority of hospital admissions in the community. It investigated the impact of a smoking ban on several types of acute respiratory and cardiovascular events, including some that were less explored. In addition, it used a robust Poisson regression analysis to better assess the specific effect of the smoking ban and minimize confounding effects, particularly secular trends and interaction with influenza epidemics. The use of data from a single hospital is a study limitation but the findings are probably generalizable for the Canton of Geneva. Indeed, hospitalisations for acute events in the Canton of Geneva are mainly centralized to the single acute care hospital of the University Hospitals of Geneva, which is the only public hospital of the Canton. For example, more than 90% of patients with an acute coronary syndrome are admitted in this institution. As we analysed retrospectively administrative hospital data, inaccurate coding of ICD-10 diagnoses at admission is possible and could potentially lead to misclassification bias. This might particularly affect patients admitted for combined acute COPD exacerbation and pneumonia, leading eventually to overestimate the decrease in admissions for COPD exacerbations. The hospital database only includes data on patients who were admitted to a ward and therefore patients who had ambulatory treatment were not considered. This latter limitation particularly affects data on patients admitted for acute asthma and pneumonia, who are mainly managed as outpatients. Moreover, as we do not have individual data on tobacco use and exposure to SHS, we were unable to determine if the observed decrease in COPD exacerbations was due to a decrease in cigarette consumption among active smokers or to a reduction in exposure to SHS, or both. Finally we did not conduct further stratified analyses (e.g. by age group) because of the limited number of events. This study is the first to show such a large decrease in admissions for acute COPD exacerbation after implementation of a smoking ban in public places. Although it was little explored, this effect was larger than expected, quickly detectable and progressive. A COPD exacerbation is a complex event involving bacterial or viral infections, constitutional factors, active smoking, exposure to SHS and other environmental pollutants, which synergistically induce a flare of inflammation in the lower airways [30]. The observed decrease in hospital admissions may have been due to reduced exposure to SHS and/or a decrease in active smoking induced by the ban [11] but the distinction was not possible in our study. In a Californian cohort, the risk of emergency department visit was 40% higher among COPD patients exposed to SHS than in non-exposed patients [31]. Exposure to SHS was an independent risk factor for readmission in a cohort of Spanish patients recruited during a hospitalization for COPD exacerbation [7]. The 46% reduction of admissions observed in our study was larger than the 27% decrease reported after a comprehensive stepwise smoking ban in Toronto [14] and the absence of change after implementation of the work place smoking ban in Ireland [16]. These differences might be partly explained by the different methods of adjustment between studies: the Canadian study adjusted only for secular trend while the Irish study also used a Poisson regression but adjusted for the seasonal temperature, levels of particulate matter PM 2.5 and PM 10 and the intensity of influenza epidemics. Finally, a secular trend is unlikely to explain this decrease as the number of hospitalisations for COPD in Switzerland declined by 27% from 2000 to 2006, but stabilized thereafter. In our study, the reduction in hospitalisations was larger (46%) and occurred later (2006–2010) [32]. Admissions for COPD exacerbations decreased progressively before the first smoking ban. This effect can be explained by the progressive introduction of smoke-free work and public places, from 2003, before our first 2-year study period preceding the first legislative smoking ban. Since 2003, the local tobacco control association has offered support to companies, public services and associations to implement smoke-free workplaces. Most of this activity took place before study period 1, between October 2003 and June 2006 and reached 93 groups and more than 65 9 000 persons. The activity decreased during the study ...

Context 2

... (ICD-10), 10th Revision [22]: acute coronary syndrome (ICD-10 codes: I 21); cerebrovascular diseases including ischemic stroke and transient ischemic attack (ICD-10 codes: I 63–67, G 45–46); COPD entities among chronic lower respiratory diseases (ICD-10 codes: J 40–44); pneumonia or influenza (ICD-10 codes: J 10–16), and asthma (ICD-10 codes: J 45-46). Like similar studies, we recorded only patients’ first hospital stay [12]. Extracted information included date of admission, gender, age, place of residence (Canton of Geneva, other or unknown), and diagnosis at admission. Because the definition of transient ischemic attack changed in 2009 [23], transient ischemic attack and ischemic stroke diagnoses were combined into a single category (i.e. ischemic cerebrovascular accident). As a quarter of patients included in the study had their residence outside of the Canton of Geneva, which could mitigate the impact of smoke-free legislations in Geneva, we also analysed data restricted to residents of the Canton of Geneva. The means and frequencies with standard deviation and 95% confidence intervals (95%CI) of study variables were computed. The mean number of weekly admissions for each diagnosis was calculated from the number of diagnosis-specific hospitalisations in each study period. We used Poisson regression models to compute the 95% CIs around these estimates and to test the hypothesis that there was a change in the mean number of weekly admissions for each diagnosis between the study periods corresponding to changes in smoking ban legislation. Dummy variables for the periods were created and the impact of smoking ban legislation was estimated by incidence rate ratios (IRR) of diagnosis-specific hospitalizations for the different periods, considering period 1 as the reference period. Some of the disease-specific hospital admissions, such as acute coronary syndrome, acute exacerbation of COPD, and pneumonia, are known to vary with seasons or to increase during the influenza epidemic [24]. Therefore, regression models were adjusted for season (defined by calendar months) and influenza epidemic periods using surveillance data of the Swiss Office of Public Health [25–28]. To account for the potential secular trends of diseases that would be independent of the legislation periods, we included a linear time trend in the models. Models were additionally adjusted for age and gender. The seasonal effect and secular trends were taken into account with methods previously used to allow comparison to previous large population-based studies that measured the impact of smoking bans on health outcomes [12,15,16,20]. We also examined whether the slope of the trend change across periods, by adding an interaction term between the study periods and the linear predictor for time to the models and using likelihood ratio tests [12]. For each outcome, we performed a global trend test of adjusted incidence rate ratio across four periods, using period 1 (weeks 26 to 130) as the reference period, with its result expressed as a global p-value. To estimate the events prevented, the admission rates obtained among Geneva residents were applied to the counts of population aged over 15, using census data of Geneva for each study period [29]. Hospital days and costs prevented by the smoking ban were estimated using the University Hospitals of Geneva administrative data. All statistical analyses were performed using Stata 11.0 (Stata Corp, College Station, USA). Overall, 5345 patients with a first hospitalization for acute coronary syndrome, ischemic stroke, acute COPD exacerbation, pneumonia and acute asthma were included (Table 1). Patients were predominantly males, had a mean age of 67 years, stayed on average for 11 days in hospital, and three-quarters of them lived in Geneva. Admissions for acute cardiovascular events were more numerous than for acute respiratory diseases. The socio-demo- graphic profile and admission diagnoses were similar across the 4 study periods and for the sub-sample of Geneva residents. The mean number of weekly hospital admissions for all studied conditions differed across periods for all population groups (Table 2). The number of hospitalisations for pneumonia and asthma was particularly low in each study period. For acute COPD exacerbation, the weekly number of hospitalisations significantly dropped over the 4 periods from 2.45 to 1.54 (p , 0.0001) (Table 2). The adjusted IRR decreased significantly in all periods after the initial smoking ban; it reached 0.54 [95%CI: 0.42–0.68] for all patients and 0.53 [95%CI: 0.41– 0.68] for Geneva residents during the final smoking ban (period 4). There was evidence of a change in the slope of the trend line across the study periods (P = 0.001 for periods x secular trends interaction). The reduction started before the legislative smoking ban and became stable in the second part of the period 3, when the ban was legislatively suspended but partially maintained in practice (Figure 1). Hospitalizations for acute coronary syndrome decreased only during the final smoking ban (period 4), where we measured a trend close to significance: adjusted IRR was 0.90 [95%CI: 0.80– 1.00] for all patients and 0.89 [95%CI: 0.78–1.02] for Geneva residents (Table 2). The slope of the trend line did not change across the periods (P . 0.05 for periods x secular trends interaction). The reduction started after the first legislative smoking ban was suspended but partially maintained (period 3) (Figure 1). Despite variations in the number of admissions, adjusted IRR of hospitalisations for pneumonia, acute asthma and ischemic stroke, did not significantly change throughout the 4 periods. This study shows that a legislative smoking ban is associated with a strong, highly significant reduction in hospitalisations for COPD exacerbation and a trend for decreased admissions for acute coronary syndrome. We observed no significant change in admissions for pneumonia, acute asthma and ischemic stroke. Among the strengths of this study is the large sample covering the large majority of hospital admissions in the community. It investigated the impact of a smoking ban on several types of acute respiratory and cardiovascular events, including some that were less explored. In addition, it used a robust Poisson regression analysis to better assess the specific effect of the smoking ban and minimize confounding effects, particularly secular trends and interaction with influenza epidemics. The use of data from a single hospital is a study limitation but the findings are probably generalizable for the Canton of Geneva. Indeed, hospitalisations for acute events in the Canton of Geneva are mainly centralized to the single acute care hospital of the University Hospitals of Geneva, which is the only public hospital of the Canton. For example, more than 90% of patients with an acute coronary syndrome are admitted in this institution. As we analysed retrospectively administrative hospital data, inaccurate coding of ICD-10 diagnoses at admission is possible and could potentially lead to misclassification bias. This might particularly affect patients admitted for combined acute COPD exacerbation and pneumonia, leading eventually to overestimate the decrease in admissions for COPD exacerbations. The hospital database only includes data on patients who were admitted to a ward and therefore patients who had ambulatory treatment were not considered. This latter limitation particularly affects data on patients admitted for acute asthma and pneumonia, who are mainly managed as outpatients. Moreover, as we do not have individual data on tobacco use and exposure to SHS, we were unable to determine if the observed decrease in COPD exacerbations was due to a decrease in cigarette consumption among active smokers or to a reduction in exposure to SHS, or both. Finally we did not conduct further stratified analyses (e.g. by age group) because of the limited number of events. This study is the first to show such a large decrease in admissions for acute COPD exacerbation after implementation of a smoking ban in public places. Although it was little explored, this effect was larger than expected, quickly detectable and progressive. A COPD exacerbation is a complex event involving bacterial or viral infections, constitutional factors, active smoking, exposure to SHS and other environmental pollutants, which synergistically induce a flare of inflammation in the lower airways [30]. The observed decrease in hospital admissions may have been due to reduced exposure to SHS and/or a decrease in active smoking induced by the ban [11] but the distinction was not possible in our study. In a Californian cohort, the risk of emergency department visit was 40% higher among COPD patients exposed to SHS than in non-exposed patients [31]. Exposure to SHS was an independent risk factor for readmission in a cohort of Spanish patients recruited during a hospitalization for COPD exacerbation [7]. The 46% reduction of admissions observed in our study was larger than the 27% decrease reported after a comprehensive stepwise smoking ban in Toronto [14] and the absence of change after implementation of the work place smoking ban in Ireland [16]. These differences might be partly explained by the different methods of adjustment between studies: the Canadian study adjusted only for secular trend while the Irish study also used a Poisson regression but adjusted for the seasonal temperature, levels of particulate matter PM 2.5 and PM 10 and the intensity of influenza epidemics. Finally, a secular trend is unlikely to explain this decrease as the number of hospitalisations for COPD in Switzerland declined by 27% from 2000 to 2006, but stabilized thereafter. In our study, the reduction in hospitalisations was larger (46%) and occurred later (2006–2010) [32]. Admissions for COPD exacerbations decreased progressively before the first smoking ban. This effect can be ...

Context 3

... is strong evidence that passive exposure to tobacco smoke is harmful to human health [1]. Tobacco smoke has pro- atherogenic and pro-thrombotic effects through various mecha- nisms [2]. A systematic review of 18 cohort and case-control studies demonstrated that exposure of non-smokers to secondhand smoke (SHS) is associated with a 25% increased risk of coronary artery disease and myocardial infarction [3]. Many studies and meta-analysis showed a moderate, consistent and dose- dependent association between exposure to SHS and the risk of stroke, which is increased by 25%, suggesting a causal relationship [4–6]. A strong and consistent association was reported between SHS exposure and several respiratory hazards: respiratory symptoms, worsening of lung function tests, prevalence of chronic obstructive pulmonary disease (COPD) and asthma, risk of COPD or asthma exacerbations and risk of hospital admission for COPD, asthma or pneumonia [7–9]. Exposure to SHS causes an important health and economic burden and is a major concern for health policy. Many countries have therefore introduced legislation for smoking bans in public and work places to reduce exposure to tobacco smoke. Numerous studies and 2 meta-analyses have shown that legislation for comprehensive smoking bans are associated with an average 15% reduction of hospital admissions for acute myocardial infarction and coronary events within a year [10–15]. Fewer studies and a recent meta-analysis showed that a comprehensive smoke-free legislation is followed by a 19% decrease in hospitalisations for ischemic stroke [10,14–16]. Workplace smoking bans are associated with an improvement of self-reported respiratory symptoms and lung function tests among hospitality employees heavily exposed to SHS [17–19]. Hospital admissions for acute respiratory diseases were globally reduced by 24% after implementation of comprehensive smoke-free laws [10,14–16]. The evidence is stronger for asthma among adults and children with a decrease in emergency visits and hospitalisations [8,10,14–16,20]. Two studies suggest a significant reduction of admissions for pneumonia and a controversial effect on COPD exacerbations, which does not emerge as significant in the meta- analysis [10,14,16]. A legislative smoking ban in public places was implemented after a popular vote on 1 July 2008 in the Canton of Geneva, Switzerland. Three months later, the smoke-free law was cancelled by the Supreme Court, which considered the implementing regulation as unlawful. However many public places remained smoke-free as managers maintained the ban voluntarily. A second vote led to a permanent smoking ban, which was applied from 30 October 2009 [21]. The Department of Health of the Canton of Geneva called for and sponsored this study to evaluate the effect of the public smoking ban on hospital admissions for acute respiratory and cardiovascular diseases at the University Hospitals of Geneva. The Ethics committee of the University Hospitals of Geneva approved the study (protocol 10-241R) on 18 February 2011 and waived the informed consent as only aggregate data from an existing database were used. We conducted this before and after intervention study in the University Hospitals of Geneva, Switzerland, a 1900-bed univer- sity and single public hospital in the Canton of Geneva, which is populated by about 450,000 inhabitants. We collected data in four periods defined according to the legislation on a public smoking ban in the Canton of Geneva (Figure 1): the 2-year period before the first smoking ban implemented on 1 July 2008 (Period 1); the 3-month period during the first smoking ban until its suspension by the Supreme Court on 30 September 2008 (Period 2); the 13- month period of suspended smoking ban from 1 October 2008 until the implementation of the final smoking ban on 31 October 2009 (Period 3); and the initial 14-month period of the final smoking ban from 30 October 2009 to 31 December 2010 (Period 4). From the hospital database of the University Hospitals of Geneva, we extracted data on patients aged 16 and over admitted between 1 July 2006 (104 weeks before the first legislation) and 31 December 2010 (61 weeks after the second legislation) with 5 primary diagnoses identified by the codes of the International Classification of Diseases (ICD-10), 10th Revision [22]: acute coronary syndrome (ICD-10 codes: I 21); cerebrovascular diseases including ischemic stroke and transient ischemic attack (ICD-10 codes: I 63–67, G 45–46); COPD entities among chronic lower respiratory diseases (ICD-10 codes: J 40–44); pneumonia or influenza (ICD-10 codes: J 10–16), and asthma (ICD-10 codes: J 45-46). Like similar studies, we recorded only patients’ first hospital stay [12]. Extracted information included date of admission, gender, age, place of residence (Canton of Geneva, other or unknown), and diagnosis at admission. Because the definition of transient ischemic attack changed in 2009 [23], transient ischemic attack and ischemic stroke diagnoses were combined into a single category (i.e. ischemic cerebrovascular accident). As a quarter of patients included in the study had their residence outside of the Canton of Geneva, which could mitigate the impact of smoke-free legislations in Geneva, we also analysed data restricted to residents of the Canton of Geneva. The means and frequencies with standard deviation and 95% confidence intervals (95%CI) of study variables were computed. The mean number of weekly admissions for each diagnosis was calculated from the number of diagnosis-specific hospitalisations in each study period. We used Poisson regression models to compute the 95% CIs around these estimates and to test the hypothesis that there was a change in the mean number of weekly admissions for each diagnosis between the study periods corresponding to changes in smoking ban legislation. Dummy variables for the periods were created and the impact of smoking ban legislation was estimated by incidence rate ratios (IRR) of diagnosis-specific hospitalizations for the different periods, considering period 1 as the reference period. Some of the disease-specific hospital admissions, such as acute coronary syndrome, acute exacerbation of COPD, and pneumonia, are known to vary with seasons or to increase during the influenza epidemic [24]. Therefore, regression models were adjusted for season (defined by calendar months) and influenza epidemic periods using surveillance data of the Swiss Office of Public Health [25–28]. To account for the potential secular trends of diseases that would be independent of the legislation periods, we included a linear time trend in the models. Models were additionally adjusted for age and gender. The seasonal effect and secular trends were taken into account with methods previously used to allow comparison to previous large population-based studies that measured the impact of smoking bans on health outcomes [12,15,16,20]. We also examined whether the slope of the trend change across periods, by adding an interaction term between the study periods and the linear predictor for time to the models and using likelihood ratio tests [12]. For each outcome, we performed a global trend test of adjusted incidence rate ratio across four periods, using period 1 (weeks 26 to 130) as the reference period, with its result expressed as a global p-value. To estimate the events prevented, the admission rates obtained among Geneva residents were applied to the counts of population aged over 15, using census data of Geneva for each study period [29]. Hospital days and costs prevented by the smoking ban were estimated using the University Hospitals of Geneva administrative data. All statistical analyses were performed using Stata 11.0 (Stata Corp, College Station, USA). Overall, 5345 patients with a first hospitalization for acute coronary syndrome, ischemic stroke, acute COPD exacerbation, pneumonia and acute asthma were included (Table 1). Patients were predominantly males, had a mean age of 67 years, stayed on average for 11 days in hospital, and three-quarters of them lived in Geneva. Admissions for acute cardiovascular events were more numerous than for acute respiratory diseases. The socio-demo- graphic profile and admission diagnoses were similar across the 4 study periods and for the sub-sample of Geneva residents. The mean number of weekly hospital admissions for all studied conditions differed across periods for all population groups ...