Zombie stat: Prescribed drugs are the third leading cause of death...

[Reposted from Zombiestatistics. Original date 1/3/2014]
Perpetrator: Richard Smith, most recently.1 All over Google, in various formats, wherever it feeds some hyperbole...
Let's look at this a bit more closely. Where did such an alarming statistic originate? Well, go to PubMed and type "iatrogenic deaths" – you'll find loads of stuff. However, the most cited in popular articles (i.e. "on the internet") is an article by Barbara Starfield, published in JAMA,

is entitled "Is US health really the best in the world?"2 This article looks at the deficiencies of the US healthcare system in general, but cites some earlier studies detailing the causes of death in US hospitals. It correctly cites an earlier study,3 which identified an alarming 105 000 deaths/year from "non-error adverse effects of medications". Not a great statistic – sure, we all know there are risks associated with taking medication and if you're in hospital, you're probably more likely to be taking medication with a higher risk:benefit ratio than a paracetamol, but still, 105 000 people...
To put this number in to context, there were 2 337 256 deaths reported in the USA in 1998.4 So my maths says 105 000/2 337 256 is 4.5%. The third leading cause of death in the US in 1998, according to the CDC National Vital Statistics report, was cerebrovascular disease (after cancer and heart disease), at a whopping 6.8%.4
But, my statistician friends would be tutting at me at this point. The 105 000 deaths occurred in a US hospitalized population, the 2 337 256 deaths occurred across the whole country, regardless of setting. So maybe we're narrowing it down to "Prescribed drugs are the third leading cause of death in hospitalized patients in the US..." ? Not such a great headline, but perhaps more accurate.
Popping back to the 1998 mortality statistics in the US, there were 16 926 people who died from drug-induced causes, which includes both legal and illegal drug use, as well as poisoning from medically prescribed and other drugs. If you assumed all those deaths were as a result of prescription drugs (which they weren't, by the way), then 16 926/2 337 256 gives us 0.72% dropping down to about 14th on the list – pardon me, the US list. Unfortunately, the WHO mortality data isn't that easily interrogated for deaths related to prescription drugs, which I think would give a much better balanced view.

References
1. Smith R, Gotzsche PC, Groves T. Should journals stop publishing research funded by the drug industry? Bmj. 2014;348(jan14 3):g171–g171. doi:10.1136/bmj.g171
2. Starfield B. Is US health really the best in the world? JAMA. 2000;284(4):483–5. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10904513. Accessed February 7, 2014
3. Lazarou J, Pomeranz BH, Corey PN. Incidence of adverse drug reactions in hospitalized patients: a meta-analysis of prospective studies. JAMA. 1998;279(15):1200–5. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9555760. Accessed February 4, 2014.
4. Murphy SL. Deaths: final data for 1998. Natl Vital Stat Reports. 2000;48(11):1–106.