According to data from the Healthcare Cost and Utilization Project (HCUP) database, there was a dramatic decline in heart attack hospital admissions in Nebraska in 2004 and a sharp decline in heart attack admissions in South Carolina during the same year. These declines in heart attacks coincide precisely with sharp cuts in funding for anti-smoking programs in each of these states.
In Nebraska, heart attack admissions fell by 28.5% from 2003 to 2004. This is in marked contrast to the existing trend in heart attacks in the state. Heart attack admissions were increasing by an average of 2.3% per year during the period 2001-2003 in Nebraska.
In South Carolina, heart attack admissions fell by 12.5% from 2003 to 2004. This is also in marked contrast to the existing trend in heart attacks in that state. Heart attack admissions were increasing by an average of 3.0% per year during the period 2001-2003 in South Carolina.
These declines in heart attacks are much greater than what would have been expected based on trends in other states. In all other states for which data are available, heart attack admissions fell by 5.1% from 2003 to 2004.
The dramatic decline in heart attack admissions in Nebraska coincides with a dramatic cut in funding for the state's tobacco control programs. According to the state: "The Nebraska legislature did away with the funding for this [state tobacco control] program in the spring of 2003. Tobacco Free Nebraska's $21 million ($7 million in each of three years) was cut. Third year funding for Tobacco Free Nebraska was eliminated and the TFN staff resources must rely on state general funds and CDC funds to sustain a skeletal program. Local coalition funding was virtually eliminated as well. ... Tobacco Free Nebraska has launched a $700,000 media campaign to air radio and television commercials with anti-smoking themes. ... This effort has been eliminated by the historic legislative cut of the spring 2003."
It appears, then, that the historic legislative cut in anti-smoking programs in 2003 resulted in the greatest decline in heart attack admissions observed in Nebraska and in the successful reversal of the increasing trend in heart attacks in the state.
The huge decline in heart attack admissions in South Carolina, which reversed that states trend of increasing heart attack admissions during the previous two years, coincided with the complete elimination of what had been a successful anti-smoking program conducted during the previous four years.
These declines in heart attacks, which buck the national trend, occurred in states which have enacted no statewide smoking bans, few if any local smoking bans, and which both received grades of F in 2004 from the American Lung Association for the category of Smokefree Air in its State of Tobacco Control 2004 Report Card.
The Rest of the Story
Of course the observed declines in heart attacks in Nebraska and South Carolina in 2004 were not due to the cuts in anti-smoking programs. But this demonstrates the danger of using data from one year after an event to draw conclusions regarding the effect of that event on the phenomenon of interest, especially when the sample is small and there is great baseline variation.
This is why the claims being made from Helena, Pueblo, and Saskatoon are so suspect. In these studies, data from one year (6-18 months) after a smoking ban was implemented are being used to suggest that the smoking ban resulted in that single year's observed decline in heart attack admissions. Essentially, you have a single data point that is being used not only to assume that a definite trend is present but to ascertain the cause of the change in the statistic for that single year.
The reasoning used here is exactly the same as that used in the Helena, Pueblo, and Saskatoon studies. Baseline trends in heart attack admissions for a period of roughly four years prior to the event of interest were studied to establish the baseline trend in heart attack admissions. Then the heart attack data for the first year following implementation of the event of interest were used to assess the effect of the event on heart attack admissions. A dramatic decline in heart attacks was observed that coincided perfectly with the event in question. No similar decline (to the same extent) was observed in a number of control states that did not have the event of interest. Using the same reasoning as in Helena, Pueblo, or Saskatoon, one would conclude that the cut in anti-smoking programs resulted in the dramatic 28.5% decline in heart attack admissions in Nebraska in 2004 and in the large 12.5% decline in heart attack admissions in South Carolina in 2004.
This is shoddy science. It's shoddy because there is simply too much underlying variability in the data to establish a definite trend from one year's data point and to eliminate the possibility that the single year change that is being observed is simply due to random variation, rather than to an effect of the smoking ban.
And here, the sample sizes are huge compared to any of the studies that are being relied upon to claim that smoking bans dramatically and immediately reduce heart attack rates. In Nebraska, there were about 1800 heart attack admissions per year, compared to only about 80 in Helena. The Helena conclusion was based on a reduction of about 16 heart attacks, while the Nebraska "conclusion" is based on a reduction of about 540.
In fact, an examination of the HCUP data reveal that the variation in heart attacks is rather strongly related to the sample size. In states with very large populations, there is little variability. For example, in the two largest states in the database (Florida and California), there was no year-to-year variation of greater than 5.5%. But in the smallest state (Hawaii), there was as much as a 15% year-to-year variation in heart attack admissions.
With small populations, it is going to take more than just five years of data to adequately establish baseline trends and to estimate the random variability in heart attack admissions. And it is going to take more than one year to establish that a definite change in the trend has occurred, rather than simply random variation.
Massachusetts is a great example of this. There was a 12.0% decline in heart attacks in Massachusetts in 2004. This is in sharp contrast to the average 9.2% increase in heart attacks in Massachusetts during the period 2001-2003. Does this mean that something which occurred in late 2003 or early 2004 caused the sharp decline?
We can't really tell yet. First of all, we would need to know the underlying variation in the data. Do we observe changes from year-to-year in Massachusetts of this magnitude?
Well, from 2001 to 2002, heart attack admissions in the state increased by 18.3%, but admissions were essentially stable from 2002 to 2003. Is this just random variation, or is there a real trend? You need to go back further than 2001 and farther ahead than 2004 to find out. After all, there were more heart attacks in Massachusetts in 2004 than in 2000. That certainly doesn't sound like a trend of declining heart attacks. The 2005 data point may help to clarify this. If heart attacks continue to fall, then it appears there may be a real trend. But if they are up a little, it will make it appear that there is just a fair amount of variability occurring.
The bottom line is that even in a large population with many heart attacks occurring, we cannot draw causal conclusions regarding the effect of smoking bans based on these data. It is hard to imagine drawing sweeping causal conclusions from similar data in a population that is exceedingly smaller. It's hard to believe that a change in heart attack admissions from 40 to 24 is being used as the basis of a national campaign to convince the public that acute secondhand smoke exposure causes heart attacks.
The rest of the story is that shoddy science is being used to promote smoking bans and the science is shoddy because one cannot validly conclude that the observed year-to-year change in one data point in the rather small populations of Helena, Pueblo, and Saskatoon were due to the smoking ban, rather than to baseline variation, secular changes, or some other factors.