Fictional Epidemiology

August 17th, 2006 by Ben Goldacre in bad science, statistics | 47 Comments »

Hands up, it’s a pretty cheap shot, but nevertheless I thought you might enjoy this bit of amateur epidemiology from some Christians who’ve looked at cause of death and early mortality in “rock stars”, and then done some sums.

I particularly enjoy the way they take the average age at death of the normal population, and compare that against the average age at death of… some rock stars who died young. Rock stars who die young do indeed, on average, die young. You may be able to spot some other deliberate methodological errors.

Anyway, the actual breakdown on cause of death is actually quite interesting: I had a quick scout about and to my disappointment I couldn’t find anyone else doing the same research more robustly. It seems that the British research effort in this field has focused on mortality and morbidity in UK soap stars, as described in this seminal paper:

It’s a dangerous industry, showbusiness. As the Christian chums say at the end of their study on rock star mortality:



If you like what I do, and you want me to do more, you can: buy my books Bad Science and Bad Pharma, give them to your friends, put them on your reading list, employ me to do a talk, or tweet this article to your friends. Thanks! ++++++++++++++++++++++++++++++++++++++++++

47 Responses

  1. Filias Cupio said,

    August 17, 2006 at 10:55 am

    So how would we do this properly?

    First we need a suitable definition of “rock star”. A good start would be all recordings chosen according to some criteria from the the weekly record sales charts of nation-of-your-choice since 1950. (perhaps anything that peaked within top 5, or was in top 20 for at least 10 weeks, or whatever.)

    Then we need some criteria for it to be ‘rock’. I’d suggest getting hold of the play-lists of some rock radio stations. There is a bias problem here in that stations which play 1950s rock might be more eclectic in their playlist than a station which plays only current top 40.

    Now with our list of “rock hits”, we need to determine the musicians. I’d suggest a limit of 6 per hit, so you don’t include entire 1950s Big Bands or the occasional backing orchestra. Take the top 6 as listed in the record’s credits – this should mostly get the real starts and leave out many of the nobodies. Perhaps you might only take people who have a minimum number (say 5) qualifying recordings.

    Now we have the list of rock stars, we need to find out when they were born, and if/when they died. This bit is easy – make a graduate student do it.

    Finally, we need a random comparison sample matched for year of birth, sex and perhaps ethnicity and/or country of residence. (Or possibly we trim our rock stars list to a single country of residence.)

  2. pseudomonas said,

    August 17, 2006 at 6:20 am

    Hmm. They say they haven’t rigged the list and they’ve used all the rock stars they could find – but missed the ones that are still alive. And given that rock is a fairly young genre…

    Anyway, the dubious geology and sensationalist HTML in is great.

  3. jeremymiles said,

    August 17, 2006 at 6:52 am

    Similar problems to all the studies that show that left handed people died younger, as satisfactorily debunked in this paper:

  4. zonknz said,

    August 17, 2006 at 8:07 am

    i want to make my avatar it’s just so amazing.

  5. coracle said,

    August 17, 2006 at 8:41 am

    I like the idea of amateur epidemiology though. It would be great if there were sufficient data released by the ONS for us all to have a bash.

  6. kim said,

    August 17, 2006 at 8:59 am

    Who’d have thought it, hey? A lifestyle of drugs, alcohol, fast cars and loose women (or should that be loose cars and fast women?) isn’t as conducive to long life as one of eating brown bread and watching Songs of Praise.

    Excuse me while I lie down and recover.

  7. ceec said,

    August 17, 2006 at 9:10 am

    What about Mick Jagger? He’s going to live forever! And Keith Richards! He can even survive “falling out of a coconut tree” (is this a euphemism?)

  8. Tom P said,

    August 17, 2006 at 9:12 am

    I don’t see “mysterious gardening accident” on that list.

  9. JohnD said,

    August 17, 2006 at 9:27 am

    The ‘Notes & Queries’ column in the Guardian answered the question, “Who are the deadest rock group?”, in the sense of the group still being in existence.
    The answers listed cause-of-death for twenty one members of said groups: Liver failure (2 – 10%), Drug overdose (4 – 20%), Trauma (8 – 38%), Cancer(4 – 20%), Stroke (1), Heart attack (1 – 10%)), respiratory failure (?!) (1), “natural causes (?!?) (1), “cocktail cherry “(perhaps should be put in with ‘respiratory failure’?) (1)

    The UK Registrar General’s report on causes of death for 2004 (I know, the groups existed long ago, but anyway) lists 5271 deaths in the age (not rock) groups 20-34 years, 226 of which were due to liver disease (4%), narcotic drug overdose 228 (4%), 820 to transport accidents (15%), 935 cancer (17%), ischaemic heart disease 141(3%). Comparision fails as the group sample is far too small, but the distribution is essentially the same as the general population.

    I have to borrow Andrew Muir’s comment that the worst job in rock, as six died while still in the group, is drummer for Spinal Tap.


  10. germslayer said,

    August 17, 2006 at 9:31 am

    Errr… How come they included John Lennon, but missed out George Harrison?

  11. JohnD said,

    August 17, 2006 at 9:37 am

    Did tnhe same for that website, after I did the RG’s list.
    Liver (as ‘alcohol’) 3%
    Drug o/d 12%
    Trauma 18%
    Cancer 8%
    heart attack 13%

    Same caveats.
    With the exception of cancer. same pattern.

    ‘Young people’ die of similar causes, general population or hell-bound hedonists.

  12. tomh said,

    August 17, 2006 at 10:39 am

    They made a mistake, Elvis isn’t dead, everyone knows that he is still alive.
    Elvis Presley 77-08-16 42 Drug Overdose
    (couldn’t help it 😉

  13. Daniel Rutter said,

    August 17, 2006 at 10:41 am

    Not only no Mick and Keef, but also no Paul McCartney, no James Brown, no Jerry Lee Lewis, no Chuck Berry, no Bo Diddley…

    And they even mention Ozzy Osbourne’s name, in the Randy Rhoades entry – but they still don’t include him in the list!

    They can approximate the form, but they just don’t get the content.

  14. Evil Kao Chiu said,

    August 17, 2006 at 10:52 am

    Having ‘chosen death’ above, I’m pleased to see that they claim Jesus Christ ‘did’ for me. Perhaps that was supposed to read ‘died’ or perhaps, more sinister, Big J has set in train a series of events that will eventually result in my demise.

  15. Ben Goldacre said,

    August 17, 2006 at 10:58 am

    So how would we do this properly?

    i guess if we wanted to be preposterously visionary, someone could make the list fairly easily, then we could divvy it up between 26 groups each doing one letter of the alphabet, then collate the results. we’d probably be able to publish it somewhere – somewhere very obscure – but it would be the world’s first ad hoc blog publication. any volunteers to coordinate? i reckon it might also be quite a good school project, although the follow up on all those rock stars could get a bit dull after a while. i can sort out some maths people to do the stats/SMR‘s afterwards (jeez can you believe nobody has written an SMR entry on wiki yet?).

  16. Evil Kao Chiu said,

    August 17, 2006 at 11:03 am

    Let’s not forget that ‘years lived’ needs to be adjusted to take account of ‘fun had’ – some sort of multiplier should be applied for the number of hotel rooms snorted, cocaine slept with and supermodels trashed. Or something like that. I was never very rock ‘n’ roll.

  17. ceec said,

    August 17, 2006 at 11:06 am

    you’d have a job working out the denominator. Maybe you could do a case-control study.

  18. Ben Goldacre said,

    August 17, 2006 at 11:09 am

    well case control is the obvious easy way out, but we’d get the denominator alright if we got all the rock stars for a given time frame and followed them up, as filias suggested? (this might well be wrong, i’ve been up all night working, and, er, posting blog entries at 4 am BST).

  19. regordane said,

    August 17, 2006 at 11:10 am

    I seem to remember that Jesus is believed to have been about 33 when he died.

  20. Evil Kao Chiu said,

    August 17, 2006 at 11:17 am

    Good point!

    3.9 years less than the average dead rock star.*

    *who died young.

  21. sockatume said,

    August 17, 2006 at 11:37 am

    Does that mean that Jesus Was F’ing Metal?

  22. PJA said,

    August 17, 2006 at 12:00 pm

    Does that mean that Jesus Was F’ing Metal?

    What with his long hair and spiky headgear? How could he be anything but.

    On doing it properly, I’d have guessed that there are actuarial tables with comparative data for the population at large. But the real point of the exercise is in the brilliant pages you get to when you click on the I choose death bit at the bottom of the page.

    “For their rock is not like our Rock”. Too right.

  23. Childeric said,

    August 17, 2006 at 12:11 pm

    Tom P: in fact they do list Jeff Porcaro, the drummer from Toto who actually *did* die in a bizarre gardening accident (he seems to have had an allergic reaction to a pesticide he was using), but disappointingly they’ve only got him down as ‘heart attack’.

  24. jj_hankinson said,

    August 17, 2006 at 12:19 pm

    There is also the issue that the cause of death is often vastly oversimplified on death certificates and official registers (‘heart attack’ covers a multitude of sins). You’d have to work out what source of cause of death to trust when following up the cases…

  25. kim said,

    August 17, 2006 at 12:21 pm

    You know, it’s as if they haven’t noticed that “live fast and die young” or “hope I die before I get old” are part of the rock philosophy, if that isn’t too grandiose a term.

    They also haven’t picked up on the disproportionately high number of rock stars who died age 27. No doubt some statistician will pick me up on this…

  26. kim said,

    August 17, 2006 at 12:54 pm

    Just to expand on that point, I don’t know if anyone saw John Sutherland’s piece in the graun this week, which was an interview with an epidemiologist who was carrying out a longitudinal study of the drinking habits of civil servants. The gist of her findings were that civil servants were largely dull people who tended to drink in moderation. Sutherland seemed desperate to get her to say something interesting but she didn’t oblige.

  27. ayupmeduck said,

    August 17, 2006 at 12:57 pm

    This is an awful list. Where is my hero Joe Strummer (50)? What about Elliott Smith (34) – with his lifestyle he was lucky to last as long. Arthur Lee (61) may be excused as it’s recent. Malcolm Owen died at 26 not 24.

    There is a North American bias, as I noted off the top of my head that Ian Dury (57), George Harrison (58) & John McGeoch (48) and Syd Barrett (60) are not included. It seems reasonable to me to assume that Jesus just doesn’t like American rock stars (who does?), so we can hope that we’ll soon be rid of Peter Frampton now that he’s changed his citizenship (sorry Peter, only joking).

    In any case, without even doing the stats, I’m sure the theory is incorrect, otherwise how does it explain that all the members of Motörhead are still alive?

  28. acedia said,

    August 17, 2006 at 1:07 pm

    They also appear to have mysteriously overlooked “rock stars” (heck, if the term can encompass Karen “Goddess of Thunder” Carpenter, it can stretch to a few old bluesmen/women) who died old like John Lee Hooker (who died at 83), Ray Charles (73), Muddy Waters (68), Nina Simone (70), Howlin’ Wolf (66), Albert King (69)… Need I go on? And the blues is the devil’s music, after all.

  29. John A said,

    August 17, 2006 at 1:35 pm

    So how would we do this properly?

    I’d like to see a Christian Rock control group. Since many of them are born again they would be expected to have a similar background to the other rocker group.

  30. ceec said,

    August 17, 2006 at 2:18 pm

    John A: That is an excellent suggestion. Does the rock cancel out the christianity though? I think we need three arms, actually:

    1. Christian rockers
    2. Sinners (i.e. other rockers)
    3. Non-rock orientated people choosing “life” (none of you cynics choosing “death” – you are obviously veering too much towards category 2.)

  31. Ben Goldacre said,

    August 17, 2006 at 2:22 pm

    i think christian rockers could be the key to this. maybe we should work out the average age at death of the ones who died young…

  32. ceec said,

    August 17, 2006 at 3:52 pm

    The internet is not being helpful in defining “christian rock”. This is what I found. They’re not giving much away here, but I’m getting the distinct impression that it’s not quite de rigeur to use the word “rock” in this connection…

    “Contemporary Christian Music (or CCM) is a classification of Christian music as well as popular music in general. It is not a musical style or genre, as it refers to several types of music. Instead, it is called “contemporary” because it is distinct from traditional and southern gospel music. It is called “Christian” “on account of a perceived connection to what [self-identified fans of CCM] regard as Christianity” “(Encyclopedia of Contemporary Christian Music, 2002).

  33. Ben Goldacre said,

    August 17, 2006 at 3:59 pm

    no way dude, i’ve been slam dancing at my desk to christian rock all afternoon

    much of it is what chris moyles would describe as “ghey”.

  34. pseudomonas said,

    August 17, 2006 at 4:08 pm

    Some kind soul has already compiled

  35. Ben Goldacre said,

    August 17, 2006 at 4:18 pm

    i know wiki is famously unreliable, but we could get a figure pretty easily by going through all the biogs under “hard rock”. given the crazed obsessiveness of music fans i’d say rock musicians biographies are probably among the more accurate info on wiki, at least for the most basic data, if not rat munching antics.

  36. jj_hankinson said,

    August 18, 2006 at 12:17 pm

    Should Chris Rock get his own (confusing) category too?

  37. Gleamhound said,

    August 18, 2006 at 12:22 pm

    I am surprised and disappointed that they didi not use the venerable internet list of “Dead Porn Stars”. Obvious sinners the lot of them. I have been using this list for years as a brilliant example of bad epidemiology.

  38. CDavis said,

    August 18, 2006 at 4:24 pm

    Have seen no existing mention, I feel I should point out the hugely apposite contribution to this debate made by the excellent Mitch Benn:


  39. Bob O'H said,

    August 18, 2006 at 4:55 pm

    Even with case control, it’s not easy: the controls have to be people who have lived up to the point where they could become famous rock stars. I would recommend talking to your tame statistricians first: getting the right data will make everything else a lot easier.

    I wonder if the way to approach the analysis is to view it as a counting process: death rates in the general population can be calculated, and then one can simply see if the death rates of rock stars are higher or lower than predicted (I’m, um, simplifying a bit). The following paper might be useful:
    at least as a starting point. This is still an active area of research, but not one I follow too closely.


  40. amoebic vodka said,

    August 18, 2006 at 10:01 pm

    Well, you can always use cddb, or the free alternative freedb as they categorise CDs to genres.

    We’re not entirely sure if it can be queried by genre, but it would save having to classify what counts as rock. As we doubt they were classified with this experiment in mind, then that will stop bias creeping in from deciding that anyone under 30 isn’t a rock star.

  41. Chris Coldham said,

    August 19, 2006 at 10:49 pm

    “Occupation” is recorded on death certificates. You could select out “rock musician” or equivalent and compare to a control group.

  42. Robert Carnegie said,

    August 19, 2006 at 11:42 pm

    Anecdotally, Billie Piper seems to have given up pop music for a far more dangerous career. (Fill in your own punch-line. Twiddly-dum, twiddly-dum, twiddly-dum, oo-ee-oo… I cross the void beyond the mind…)

  43. Robert Carnegie said,

    August 19, 2006 at 11:44 pm

    But wait, what about jazz musicians?

    I remember Lenny Henry pointing out that American popular music forms whose names -don’t- refer to the sexual act are relatively rare…

  44. Nurn said,

    August 21, 2006 at 9:16 pm

    Does Mozart count as a dead rock star? dead at 26 (I believe), he must drag the average down a bit…

  45. Littleshim said,

    August 21, 2006 at 10:19 pm

    Definitely some sort of control is needed for the varying “rock” quotients of the musicians. Maybe someone could come up with that?
    That being said, rock musicians seem a less-than-ideal choice for this study. As people have said, it hasn’t been around that long, and rockers aren’t the only hedonists around. What about actors? Cinema’s been around a lot longer, and there’s plenty of drink, sex and drugs in theatre. There’s a much bigger subject base in there too.
    You could probably do something with the opium and absinthe-glugging Romantic poets as well. They were pretty wild. And generally died young, I believe. Then again, so did everyone else back then.

  46. ceec said,

    August 23, 2006 at 11:35 am

    I’m sure there are other forms of sin than rock, but let’s face it, rock is the worst. I’m pretty sure that once you get to a certain level of rock you will burn in hell, i.e. it is a binary condition. If you insist, though, you could calibrate rock quotients on a scale using:

    Length of hair (for boys)
    Largness of hair (for girls)
    Tightness of trouser
    Ever claimed to enjoy long guitar solo
    Lifetime consumption of amphetamines
    Plans to grow a very big beard later in life (if positive for this condition, likely to burn in hell anyway so not a failsafe indicator)

  47. jiangjiang said,

    December 8, 2009 at 2:53 am

    ed hardy ed hardy
    ed hardy clothing ed hardy clothing
    ed hardy shop ed hardy shop
    christian audigier christian audigier
    ed hardy cheap ed hardy cheap
    ed hardy outlet ed hardy outlet
    ed hardy sale ed hardy sale
    ed hardy store ed hardy store
    ed hardy mens ed hardy mens
    ed hardy womens ed hardy womens
    ed hardy kids ed hardy kids ed hardy kids