Tuesday, November 20, 2012

Jane Brody and the New York Times: Counting Smokers Badly

Jane Brody falsely reported in her November 12 New York Times “Personal Health” column that, according to the National Survey on Drug Use and Health (NSDUH), the smoking prevalence among young adults (age 18-25 years) is 40% (here).  The error was picked up by the American Council on Science and Health (here), and the Times later posted a correction, noting that the rate is 34%. (The original number was the prevalence of all tobacco use, rather than the prevalence of smoking.)  

But is 34% the accurate figure?  Perhaps; it depends on which federal survey is used. 

Anti-tobacco forces have for years selectively used two sets of numbers based on two federal surveys in a manner that recalls the schoolyard taunt, “Are you bragging or complaining?”  When zealots brag about how regulation, legislation and litigation have driven down smoking rates, they use the CDC-supported National Health Interview Survey (NHIS).  According to the 2010 NHIS, the prevalence of smoking among young adults (18-24 years) was 20% (here).  But when they complain that smoking rates are too high, and demand more regulation, legislation and litigation, they use the NSDUH, which produces higher estimates than NHIS – in this case 34%.

Why does this matter?  Using Jane Brody’s corrected percentage from NSDUH, the number of young adult smokers (18-24 years) in the U.S. is 10,489,854; using NHIS, the number is 6,165,090. 

With respect to variability, this difference of 4.5 million smokers is huge, and here’s why. 
NSDUH estimates are higher because it counts as a smoker anyone who answers “yes” to this question: "During the past 30 days, have you smoked part or all of a cigarette?”  The comparable question in the NHIS is, "Do you now smoke cigarettes every day or some days?"  

I published a formal study of this problem in 2009 (abstract here), and I wrote a blog post about it (here).  At that time, I calculated that there were 9 million more smokers of all ages nationally in NSDUH compared with NHIS.  

How can the federal government develop effective policies to help American smokers when the government can’t even develop and employ consistent estimates of how many Americans smoke? 

I am grateful to Jane Brody for bringing attention to this egregious problem. 

1 comment:

Nick said...

This is an ongoing issue that still hasn't grabbed the attention of researchers and practitioners. What are seemingly minute discrepancies in questioning across surveys clearly leads to big differences - in the case of NSDUH vs. NHIS, 14% points! I wonder what the prevalence rates would look like if the NSDUH used a combination of two questions like the CDC - ironically, the NSDUH actually includes the same question as the NHIS that asks whether a respondent has smoked 100 cigarettes in their lifetime, yet it is not used to define current smoking. Like your 2009 study, Ryan et al. (http://www.hindawi.com/journals/jeph/2012/918368/) also explore prevalence rates using 2008 NSDUH and NHIS data, noting that certain subpopulation estimates become comparable when using a modified NSDUH smoking definition. Why SAMHSA does not adopt this more comparable definition remains a mystery.

Another frequently unmentioned issue is what I call the prevalence smokescreen: although prevalence rates have dropped over the past several decades, this is actually due to population growth. In fact, the absolute number of smokers has remained relatively stable, but unfortunately, many folks in public health still attribute this reduced prevalence to the success of anti-smoking campaigns and programs. Not to say these sorts of efforts are frivolous, but it's clear that there's more than meets the eye.