Thursday, November 10, 2011

Data Show No Effect of Smoking Ban on Heart Attack Admissions in North Carolina, But That Doesn't Stop Anti-Smoking Groups from Claiming a 21% Decline


Frank Increase in Heart Attack Admissions Among Women Makes Anti-Smoking Groups' Conclusions Completely Untenable


A new report out of the North Carolina Department of Health concludes that the statewide smoking ban, implemented on January 1, 2010, caused a 21% decline in heart attack admissions during 2010.

The report assessed age-adjusted rates for weekly heart attack admissions during the two years prior to the smoking ban (2008 and 2009) and one year after the ban (2010). The effect of the ban was estimated using "generalized estimating equation (GEE) Poisson regression models with an autoregressive correlation matrix to adjust for short-term autocorrelations within clusters of gender, age, and county" (i.e., using a fancy statistical model).

The study results were as follows: "Comparison of QIC statistics suggested that the model that provided the best fit to the data was that which only adjusted for the weekly average temperature and for county. This model indicated that the rate of ED visits for AMI was 21 percent lower following the implementation of the smoke-free restaurant and bars legislation. (Rate Ratio [RR] Estimate was 0.79 ; 95% CI: 0.75-0.83)."

The study goes further. It then calculates the health care savings due to the smoking ban's reduction of heart attacks. A North Carolina Department of Health and Human Services press release boasts that: "Emergency room visits by North Carolinians experiencing heart attacks have declined by 21 percent since the January 2010 start of the state’s Smoke-Free Restaurants and Bars Law. ... the decline in heart attacks in North Carolina in 2010 represents an estimated $3.3 to $4.8 million in health care cost savings."

An immediate announcement by Dr. Stan Glantz, following the release of the report, boasted that heart attacks in North Carolina are "down 21 percent in the first year after the North Carolina smokefree restuarant and bar law took effect."

The Rest of the Story

The rest of the story is the actual data, which show quite clearly that the rate of decline in heart attack admissions in North Carolina has slowed since the smoking ban went into effect and that among women, the number of heart attacks has increased.

Moreover, the health department's report acknowledges both of these findings!

The combination of these two facts leads me to question whether the conclusion of the report is attributable to bias on the part of the anti-smoking groups involved in the research, which (like me) desperately want to see the health effects of their policies.

Here are the actual data, which show the rate of decline in heart attack admissions observed in North Carolina during the year before and the year after the smoking ban:

Year Before the Smoking Ban (change from 2008 to 2009): 10.5% decline in heart attacks
Year After Smoking Ban (change from 2009 to 2010): 5.5% decline in heart attacks

Thus, the actual data show that the rate of decline in heart attacks slowed in the first year after the smoking ban went into effect.

The study acknowledges this finding, admitting that: "Interestingly, the rates appear to have consistently declined between the year 2008 and 2009; after that period the rates leveled off at a consistently lower level in the year 2010 (Figure 1)."

In other words, the report readily admits that the data demonstrate an increase in the expected number of heart attacks following the smoking ban. Based on the existing secular trend, one would have expected to see about a 10% decline in heart attack admissions. However, the actual decline was only 5.5%. As the report itself acknowledges, the heart attack rates, which were declining prior to the smoking ban, "leveled off" in 2010.

Now let's examine what the so-called 21% decline in heart attack admissions meant for the number of heart attack admissions among women in North Carolina:

Question: From 2008 to 2009 (the year prior to the smoking ban), the number of heart attack admissions among women in North Carolina decreased by 661, from 4182 to 3521, representing a decline of 15.8%. Given the anti-smoking groups' conclusion that the smoking ban led to a 21% reduction in heart attacks, which of the following is the observed reduction in heart attack admissions among women in North Carolina from 2009 to 2010 (the first year of the smoking ban)?

a. From 3521 to 2678, a decline of 23.9%
b. From 3521 to 2790, a decline of 20.8%
c. From 3521 to 3012, a decline of 14.5%
d. From 3521 to 3245, a decline of 7.8%
e. From 3521 to 3456, a decline of 1.8%

The answer is:

...

...

f. None of the above.

The number of heart attack admissions among women increased from 3521 to 4273, an increase of 21.4%.

Do you see how ridiculous it is to conclude that the smoking ban in North Carolina led to a 21% decline in heart attack admissions when the actual data show that heart attack admissions among women increased by 21%?

Moreover, the report readily acknowledges this finding: "the number of ED visits for AMI increased slightly from 2009 to 2010 in younger and older women."

Now you can see why I cannot help but wonder whether there is a bias taking place here: researchers and policy makers are eager to find an effect of their interventions. I, too, share this bias, as I have testified in favor of smoking bans in many states and cities, and would like nothing more than to be able to document a substantial reduction in heart attacks due to these policies.

However, if the data don't show it, then the data don't show it and we shouldn't be disseminating these ridiculous, unsupported conclusions.

Frankly, the actual data come much closer to showing a 21% increase in heart attack admissions in North Carolina during the first year of the smoking ban than they do to showing a 21% decline in admissions.

By the way, if you don't believe me, look at the data yourself. In particular, look at Figure 1, which shows the raw data. You'll see that there was a decline in heart attack admissions prior to the smoking ban but that the decline leveled off somewhat after the ban went into effect. Even the authors of the report do not dispute this.

How then, does the report come up with the conclusion that there was a 21% decline in heart attacks due to the smoking ban? It appears to me that this is the result of a flaw in the methodology: the analysis fails to control for the secular trend in heart attacks. In other words, the analysis simply compares the number of heart attacks in 2010 to the number in 2008 and 2009 and finds that it is about 21% lower. If you don't control for the baseline trend in heart attacks, you are going to create an effect that doesn't actually exist.

The rest of the story is that instead of a 21% decline in heart attacks due to the smoking ban, there was actually a 21% increase in heart attack admissions among women and a decrease in the rate of decline in heart attacks overall from 10.5% to 5.5%. These data are not consistent with the conclusion in the report, in the press release, or in the announcement being disseminated by Dr. Glantz.

My readers understand that I support smoking bans and that this support does not hinge on demonstrating an immediate effect of such bans on heart attacks. The inability to demonstrate such an immediate effect is not a reason to stop protecting workers from secondhand smoke exposure. However, by disseminating these unsupported conclusions and adopting shoddy science to support our policies, I believe that the tobacco control movement is risking its credibility. And if that is lost or undermined, then we will lose our ability to continue to pass laws that protect workers from the dangers of secondhand smoke exposure. And that's the real shame with the rest of the story.

No comments: