A new report out of the Ohio state Department of Health concludes that the statewide smoking ban implemented in May 2007 caused a significant decline in heart attacks during the period 2007-2009, based on an analysis of hospital discharge data for the diagnosis of myocardial infarction (heart attack).
The study compared hospital discharge data on patients with a primary diagnosis of acute myocardial infarction from before to after the state Smoke-Free Workplace Act went into effect. Data were obtained from all hospitals in Ohio for the years 2005 through 2009. Thus, the rate of change in heart attacks could be estimated for the year before the Act and for the first three years following implementation of the Act.
The report concludes: "The analysis determined a significant change in age‐adjusted rates of AMI discharges within one month after the enactment of the Smoke‐Free Workplace Act."
A press release accompanying the report also boasted that: "The analysis of discharge data from Ohio hospitals ... revealed a sharp decline in heart attack rates immediately following implementation of the law."
The conclusion that the smoking ban led to a significant decline in heart attacks was widely disseminated by the media (example 1; example 2; example 3).
The Rest of the Story
Unfortunately, the actual data presented by the report does not support the conclusion that the smoking ban led to an immediate and significant decline in heart attacks in Ohio.
The smoking ban went into effect in December 2006, although rules for implementation were not issued until May 2007. Thus, the ban was implemented starting in December 2006 but was in full effect by May 2007. Therefore, 2007 was the first year in which the ban was in effect. Any change in heart attacks from 2006 to 2007 could potentially be attributed to the smoking ban. The change in heart attacks from 2005 to 2006 represents the baseline trend in heart attacks in Ohio, prior to the smoking ban. The change in heart attacks from 2007 to 2008 and from 2008 to 2009 represents the 2nd and 3rd years of implementation of the ban.
So if the report's conclusion is correct, then the rate of decline in heart attack discharges in Ohio for the years 2006 to 2007, 2007 to 2008, and 2008-2009 should be substantially greater than the rate of decline in heart attacks from 2005 to 2006, prior to the ban. Let's look at the actual data:
The annual declines in the age-adjusted hospital discharge rates for acute myocardial infarction across hospitals in Ohio were as follows:
2006-2007 (baseline): -4.7%
2006-2007 (first year of implementation): -2.7%
2007-2008 (second year of implementation): -2.2%
2008-2009 (third year of implementation): -6.3%
Average annual decline post-implementation: -3.6%
Thus, what the data show is that the baseline annual rate of decline in heart attacks in Ohio was 4.7%, and the average post-implementation annual rate of decline in heart attacks was 3.6%.
In other words, the rate of decline in heart attack discharges in Ohio was greater prior to the smoking ban than it was in the first three years after the smoking ban.
This clearly does not support the conclusion that the smoking ban resulted in a large and immediate decline in heart attack discharges.
If one looks at the data graphically, one can see that there was little change in the heart attack trend in Ohio after the smoking ban. There was a gradual decline that clearly was present prior to the smoking ban went into effect. The rate of decline decreased slightly during the first two years of the ban and then accelerated slightly during the third year of the ban.
How, then, is the report able to come up with a conclusion that the smoking ban led to an immediate and significant decline in heart attacks?
The answer is simple: by using a cubic spline model rather than a simple linear model, which is what the report hypothesizes in its introduction section. The cubic model negates the large decline in heart attacks that was observed during 2005 (see Figure 1 of the report). It essentially "erases" the baseline trend of a substantial decline in heart attacks, zeroing it out so that any decline in heart attacks observed after the intervention will appear greater than prior to the intervention.
Had the report simply drawn two lines, one before and one after the intervention, it would have been forced to conclude that there was no significant change in the rate of decline in heart attacks (or if anything, the rate of decline decreased following the smoking ban).
The most accurate description of the trend in heart attacks in Ohio is that the rate was dropping rapidly prior to the smoking ban and that the rate of decline decelerated slightly during the first two years of smoking ban implementation and then accelerated during the third year. The presence or absence of the smoking ban appears to have nothing to do with the heart attack trend.
Another way in which the report makes it appear that the effect of the smoking ban is greater than actually observed is by plotting the data on a graph that fails to use 0 as the lower end of the y-axis (again, see Figure 1). This is a classic flaw in graph presentation that can deceive viewers into believing that an effect is present when it is not. Graphs should always have 0 as the lower end of the y-axis in order to avoid this problem.
Darrell Huff discusses this method of distorting data presentation in his classic book "How to Lie With Statistics." (see Chapter 5: The Gee-Whiz Graph)
Because the actual data show no significant increase in the rate of decline in heart attacks in Ohio after implementation of the smoking ban, I won't even get into the most important flaw of the analysis, which is the absence of any comparison or control group. Even if heart attacks did decline in Ohio during the past four years or so, one needs to know whether that is unique or whether heart attacks were also declining elsewhere during the past four years. Without any comparison group, it is impossible to make such a determination.
The rest of the story is that this report has manipulated the analysis in order to try to show an effect that is simply not present in the data. This doesn't do anti-smoking advocates any favors. While I agree with the ultimate conclusion of the health department director - that the smoking ban was a highly appropriate intervention that is having a positive health impact on Ohio residents - a large and immediate decline in heart attacks is not part of the reason. By placing such a large emphasis on the need to demonstrate an immediate decline in heart attacks (when it takes many years for secondhand smoke exposure to cause heart disease), and by distorting the science in order to try to show such effects, anti-smoking advocates are actually weakening, not strengthening the argument for workplace smoking bans.