difference in nonsmoking-patient admissions between the counties during period 2 (5 admissions in Monroe County and 16 in Delaware County).
The study’s focus on nonsmokers strengthened its relevance for answering the question of the effect of a decrease in secondhand-smoke exposure, but exclusion of cases with comorbidities could exclude cases in which secondhand smoke triggered an event in a person predisposed to an acute MI, and it greatly reduced the sample size. Smoking status was determined on the basis of admission records, so there might have been misclassification. Most studies, however, including a review and meta-analysis of 26 published studies (Patrick et al., 1994) and more recent studies (Martinez et al., 2004; Studts et al., 2006), have demonstrated minimal or low underreporting of current smoking status, although others report that underreporting of smoking is significant in England and Poland but not in the United States (Lewis et al., 2003; West et al., 2007) or is rare but possibly increasing (Fendrich et al., 2005). A longer period of followup after implementation of the smoking ban would permit a fuller assessment of its impact on acute MI–related hospital admissions. In addition, Teo and Sorabi (2007) showed unusually small numbers of acute MI events in nonsmokers (for example, no admissions for acute MI in nonsmokers in Monroe County since January 1, 2005). With respect to the analysis, the authors compare the difference in acute MIs before and after the ban in Monroe County, and compare the number of acute MIs after the ban in Monroe County to Delaware County (a county with a similar population for which there were no significant differences in acute MIs prior to the ban in Monroe County, that did not implement a smoking ban). Both of those analyses, however, can have problems. Trends over time (for example, if the rate of acute MIs was decreasing prior to the implementation of the smoking ban) could confound the first analysis; differences between the two counties could confound the second analysis. A “differences-in-differences” analysis, which tests whether the differences between the decreases in the two counties are significant, would be a preferable analysis that would control for those potential confounders. Such an analysis is often conducted on observational data in social sciences to examine the effects of a program or policy change (Buckley and Shang, 2003).
The city of Bowling Green, Ohio, implemented a clean-indoor-air ordinance in March 2002 that banned smoking in all public places in the city except bars, restaurants with bars in isolated areas, and bowling alleys. Bars and bowling alleys allowed smoking at the owners’ discretion.