We found a significant decrease in overall mean admissions for AMI and in trend of admissions for angina in men. Although the trends for all cardiovascular diseases decreased non-significantly after the smoking ban, there were no other significant changes found in hospital admission rates for cardiac or respiratory diseases following the smoking ban. Compared with other studies, the difficulty identifying the effects of the smoking ban on hospital admission rates may be due to several reasons. This analysis examined admission rates 8 years prior to and 7 years after the smoking ban, longer than any other study reviewed. Meyers et al. reported that the size of the effect of the smoking ban decreased with the length of time post smoking ban captured by the study 
. This may account for the many non-significant changes in hospital admission rates in our long-term study. The relatively small size of the population resulted in substantial random variation in monthly admission rates and this may have obscured some of the trend. ARIMA models are excellent tools to deal with the complex correlation structures associated with time series data 
. The use of ARIMA models allowed changes resulting from the smoking ban to be separated from underlying trends in hospital admissions 
All of the cardiovascular and respiratory conditions examined have multiple risk factors such as environmental conditions, physical inactivity, inadequate nutrition, co-morbid conditions, active smoking and second-hand smoke exposure. Due to the nature of the data, no adjustment for these confounding variables was possible and this may have obscured some of the changes in hospital admissions due to the smoking ban. Changes in the distribution of confounders within the population may have changed the risk for admission with cardiovascular and respiratory conditions for the entire population, as well as at the individual level. Further studies examining different levels of these confounders may show additional population and individual level benefits of a comprehensive smoking ban. For example, the increasing levels of obesity and stagnant levels of physical inactivity in PEI from 2001 to 2007–08, well-known risk factors for cardiovascular conditions, may have caused admission rates to be unchanged despite any positive effects of the smoking ban 
. The predictions for monthly AMI admission rates in PEI with and without a smoking ban () demonstrate that, even in the setting of increasing hospitalization rates, public health interventions can lower the expected rate of hospitalizations.
The nature of the dataset did not allow for modeling of non-smokers and active smokers separately, despite active smoking being a risk factor for hospital admissions for both respiratory and cardiovascular diseases. It is possible that the observed changes in mean AMI admission rates and trend of angina admissions are the result of more active smokers quitting. Studies have shown that smoking cessation rates increase with a smoking ban and the PEI active smoking rate continues to drop 
. This drop in active smoking rate is a further benefit of the smoking ban.
Previous studies of cardiovascular and respiratory diseases and smoking bans have recognized and adjusted for the seasonal nature of hospital admissions 
. No seasonal pattern was apparent in the monthly time series of cardiovascular conditions. The respiratory conditions appeared to have a seasonal pattern on initial examination of the time series graphs, but on further exploration the pattern was irregular with cycles occurring every 7 to11 months. The cause of this irregular variation in admission rates warrants further exploration and could be related to the multiple triggers for exacerbations present in PEI, including agricultural pollens and pesticides, wood burning heat sources, changing air pollution levels, influenza activity and variations in temperature and humidity between seasons.
Ecological studies provide strong evidence for identifying causal associations at the population level and are important tools in evaluating the effect of public health policy 
. As well, the use of time series models did allow for evaluation of the smoking ban immediately and over time while controlling for time trends present in hospital admission rates before the smoking ban. The use of control conditions and a control province allowed for further comparison of time trends in hospital admissions. There were no decreases in the rates for the controls over the time period except for a small but significant reduction in the trend of admissions for bowel obstruction in all adults. Because of the design of the study, the analysis could not control for other factors such as additional tobacco laws at the federal and provincial levels and public health campaigns that may have affected smoking rates and subsequently hospital admission rates. In the control province (NB), it was not possible to exclude residents of Fredericton, NB who were exposed to a smoke-free law only 1 month after PEI and who represented 11.7% of the control province population. This may have biased the estimate of the effect of the PEI smoking ban towards the null as the changes in hospital admission rates for AMI resulting from the Fredericton smoking ban would have occurred at nearly the same time as in PEI.
This study examined the impact of comprehensive smoking ban within the unique socio-economic, cultural and climatic conditions of Atlantic Canada. This study is one of the first to examine the effect of smoking bans on respiratory disease, stroke, and angina hospital admission rates. The overall AMI results agree with previous research where decreases in overall AMI hospitalizations rates occurred after smoke-free laws were implemented 
. Our age- and sex- trends in AMI admissions, although not significant, were similar to those found in two Italian studies where men of all ages were more likely to benefit from smoking bans 15
. Recent studies have suggested that previous findings of significant changes in AMI hospitalizations may be the result of publication bias, model misspecification, including assumptions about linear trends, sampling bias and using too short of a time period after the implementation of a smoking ban 
. Previous work examining stroke hospitalizations found no significant associations with smoking bans while angina hospitalizations significantly decreased following smoking bans 
. Men appeared to benefit more from the smoking bans, as they did in two Italian studies showing a greater decrease in AMI admission in men 
. Similarly, we found a decrease in trend of admissions for angina in men, an associated cardiovascular condition. The current study did not match the findings of previous studies of smoking bans and respiratory conditions that showed large significant decreases in hospitalizations for pediatric and adult asthma and COPD 
. Shetty et al. have suggested that smoking bans may significantly reduce AMI hospitalizations in areas with limited voluntary private bans and high smoking prevalence but that the effect may be greatly reduced where these conditions do not exist 
. In addition, we suggest that the effect of the smoking ban in PEI may be reduced by the rural nature of the population leading to a potentially reduced exposure to SHS in public places compared to a more urban setting and by the decreasing active smoking prevalence 
. The current study was conducted using data from 14 years of hospital admissions, a validated population-wide database, and a robust model identification process, and likely presents a comprehensive picture of the effect of the smoking ban on overall cardiovascular and respiratory hospital admissions in PEI.