Robin --

But didn't YOU write up the prevalence comedy in one of your own "science
meets alcohol" pieces?  I thought it was in your chapter in Roman's (1991)
book, but when I looked it wasn't there.

In your reply to Nancy and me you wrote:

"There is a paper by Popham from the 1960s which argues for the validity of
the Jellinek formula because look, all these other estimates agreed with
it.  My argument against that (I'm not sure if and where I put it in
writing) was that the Jellinek estimate tended to yield figures which were
about the geometric mean of the politically believable range."

That's, I believe, Popham's 1970 chapter, "Indirect Methods of Alcoholism
Prevalence Estimation: A Critical Evaluation," in Popham, R.E. (ed.),
_Alcohol & Alcoholism_, Toronto, ARF.  Wrote Popham, "The most important
evidence to consider in an evaluation of the Jellinek method is the extent
to which estimates based upon it agree with estimates derived by entirely
independent methods" (p. 299).  This was also the paper in which Popham
published his table showing how cirrhosis mortality and per capita
consumption correlated so nicely together in so many lands and time periods
-- save for that one pesky anomaly in the UK from 1931-58.  It was that
anomaly, in turn, that in due course launched Skog's celebrated 1980 and
1984 papers attempting to use time lag to return the UK anomaly to the
fold.  It struck me as curious and odd that Popham -- up at ARF, where the
focus on per capita consumption was already a decade old by 1970, dating
back to Seeley's (1960) paper -- would trouble himself sit down and write a
paper defending Jellinek's highly suspect alcoholism prevalence formula.
The answer to that puzzle didn't become clear until my colleagues and I
were sorting through the argument for the "Overlooking Terris" paper we
presented at the KBS meetings in Florence in 1998.  The ARF need the
Jellinek formula as much as the Yale/Rutgers gang needed it -- because it
linked the focus of their research attentions, cirrhosis mortality, to
something with a lot more pizzazz in contemporary alcohol studies, i.e.,
the then hot problem of alcoholism.  I would have bet cash money that your
response to Popham's convergent validity argument would appear in your
Ewing & Rouse chapter, but a quick look at least does not turn it up.  A
couple of authors later further elevated this bogus convergent validity
theme in a series of papers (Celentano, D.D. and McQueen, D.V., "Comparison
of alcoholism prevalence rates obtained by survey and indirect estimators,"
J. Stud. Alcohol 39:420-434, 1978; "Reliability and validity of estimators
of alcoholism prevalence," J. Stud. Alcohol 39:869-878, 1978; "Extent of
problem drinking:  A community based validation and assessment of
measurement strategies," Jun 1978. 238 P. (ETOH reference).






----------
From: Robin Room <[log in to unmask]>
To: [log in to unmask]
Subject: Re: How many alcoholics?
Date: Saturday, February 05, 2000 6:46 AM

Nancy and Ron --
    First, let me add Ron's thanks for this account.
    There are at least two other published accounts of how estimates of the
number of alcoholics were arrived at in the late 1960s and early 1970s,
which view the elephant from different legs.  One is pp. 55-60 in Joseph
Gusfield, The Culture of Public Problems: Drinking-Driving and the Symbolic
Order,  Chicago: University of Chicago Press, 1981.  I believe there is
another in Carolyn Weiner,  The Politics of Alcoholism, Transaction Books,
1981 -- I can't find a copy of that with me here.   And there is a third
short discussion of the issue in Don Cahalan, Understanding America's
Drinking Problem, San Francisco: Jossey-Bass, 1987, p. 17.  All three of
these accounts depend to some extent on recollections from Don Cahalan or
myself.
    As I remember hearing it from Don at the time (slightly different from
Gusfield's account), the original "nine million alcoholics" came from when
Jack Mendelson, then the head of the Center in NIMH which immediately
preceded NIAAA, was testifying to Congress, presumably in 1969 or 1970.
Someone on the hill asked him the one question of any political interest in
those days: how many alcoholics were there in the U.S.?  Nate Rosenberg,
our project officer at NIMH, was told to call Don Cahalan and ask him the
question.  Don demurred, going into a long song-and-dance about how it
depended how you defined it, and anyway we weren't measuring "alcoholism"
in our surveys.  Eventually Nate gave up.  But this was not an acceptable
answer.  Shortly after, Nate called back to say that NIMH was funding us to
do epidemiology, and we had to provide an answer.  After further demurrers,
Don eventually said something like: well, if you want to provide a number,
here is where you should look in our publications.  My memory don't have
the report in front of me, but I believe it distinguishes between
"alcoholics and "problem drinkers", with 5 million for the former and 9
million if you included the latter.
    For the second report to Congress (1974),  Tom Harford from NIAAA was
deputed to ask me to write a paragraph about how many alcoholics there are
in the US.   The sentence which finally appeared in the report read as
follows:
      "The number of Americans whose lives alcohol has adversely affected
depends on definition: those under active treatment for alcoholism in
public or private agencies are probably in the upper hundreds of thousands,
but there may be as many as 10 million people whose drinking has created
some problem for themselevs or their families or friends or employers, or
with the police, in the past year."
    I had written a further sentence which said something like: "The number
whose drinking has created such problems sometime in their life may be as
many as 20 million".  Apparently this number was too high for the political
process to stand, and the sentence was dropped.
    The "10 million" was chosen as a way of expressing that we were talking
about orders of magnitude, rather than hard figures (see Ron's post on the
problem that we are dealing with a continuum, or maybe more than one).
    But I was told that someone got up in Congress and, pointing to the 10
milllion figure, said something like: Look how bad things are.  We've been
fighting alcoholism for three years and the numbers have gone up by one
million!
    Note that "whose lives alcohol has adversely affected" in those far-off
days referred only to the drinkers themselves -- the focus on and estimates
of the numbers other than the drinkers whose lives were affected by the
drinkers lay in the future.
    There is a paper by Popham from the 1960s which argues for the validity
of the Jellinek formula because look, all these other estimates agreed with
it.  My argument against that (I'm not sure if and where I put it in
writing) was that the Jellinek estimate tended to yield figures which were
about the geometric mean of the politically believable range. Noone would
believe a figure below 1 million.  But neither would anyone believe a
figure above 13-15 million.  So any estimate which did not fall in this
range would be discarded -- probably not published in the first place. This
way, there would be fair agreement with any estimate which ended up with
figures around 5 million or so.
    The puzzling thing about all this is that the numbers in Nancy's
recollections don't look at all the same as the numbers in the tradition
coming from Alcohol Research Group recollections.  I wonder how the stwo
histories can be put together.
    -- Robin

-----Original Message-----
From: Ron Roizen <[log in to unmask]>
To: [log in to unmask] <[log in to unmask]>
Date: den 4 februari 2000 19:57
Subject: How many alcoholics?


>Nancy,
>
>Accidental or not, thanks for posting this to the whole list!
>
>I'd like to take issue with one element of your account, though.
>
>There were, and there are, no epidemiological studies that did or could
>confirm your 13+ million figure regarding the prevalence of alcoholism in
>the U.S.  The central problem was that survey measures of "drinking
>problems" resulted in continuous rather than bi-modal distributions on
>their drinking problems measures.  Hence, the results did not lend
>themselves to drawing an easy cutting point between alcoholic and
>subalcoholic drinking and drinking problems.  (The problem was first
>articulated in Walter Clark's classic paper -- "Operational Definitions of
>Drinking Problems and Associated Prevalence Rates," Quart. J. Stud.
Alcohol
>27:648-668, 1966 -- and then echoed and ramified through countless
>publications at the Alcohol Research Group in years following.)  A
>secondary problem was that general population surveys would tend to miss
>what's called the "nonhousehold population" of alcoholics -- i.e.,
residing
>in one or another institutional or "group quarters" or even homeless
>circumstances.
>
>Of course, this did not stop the movement on behalf of the expansion of
>alcoholism treatment in the U.S. from pressuring survey researchers to
>produce alcoholism prevalence estimates regardless.  And the meeting of
>these two forces -- survey researchers saying they couldn't do it and
>political-social advocates saying "Give us a number anyhow!" -- produced
>the sort of comedy and absurdity you described in your own experience in
>Congress.  In due course the folly of the scientific situation was
reported
>to a wider scientific public -- see, Barnes, Deborah M., "Drugs: Running
>the Numbers," Science 240:1729-1731, (24 June) 1988.  Robin Room has
>somewhere recounted the situation surrounding prevalence estimation, too,
>but at the moment I can't remember where.  Also see Hilton, M.E., "What I
>would most like to know: How many alcoholics are there in the United
>States?". British Journal of
>Addiction, 84(5):459-460, 1989.
>
>There may have been 13+ alcoholics in the U.S.; your guess was as good as
>the epidemiologists'!