MEASURING DRINKING PATTERNS AND PROBLEMS:
THE EXPERIENCE OF THE LAST HALF CENTURY[1]
Robin Room
Centre for Social Research on Alcohol and Drugs
Sveaplan
Stockholm University
S-106 91 Stockholm, Sweden
(Robin.Room@sorad.su.se)
Surveys on drinking behaviour are
not a phenomenon only of the last half century. To pick a couple of earlier
examples, as part of the investigations of the Committee of Fifty to
Investigate the Liquor Problem, a survey of “drinking among brain workers of
the United States” was carried out by John Billings (1903); and a survey of
drinking by children in Bochnia, Poland was carried out in 1913 (Swiecicki,
1972).
Two main features differentiate the
modern tradition of surveys of drinking behaviour from earlier efforts. Before the study by Straus and Bacon (1953)
of drinking among U.S. college students, drinking surveys focused only on the
fact of drinking at all, or on the frequency of drinking. Surveys of illicit drug use today still
conventionally ask only about the fact of use or frequency of use. Straus and Bacon’s broadening of the scope of
questioning and analysis about drinking thus marked a decisive break with the
focus in and after the temperance era on alcohol use per se as the
problem. Straus and Bacon’s extension of
the scope was in two main ways: by using a typology of drinking patterns, with
attention to the amount of use per occasion; and by going beyond questions
about drinking behaviour to ask also about problems related to drinking.
The second feature which marks the
modern tradition of drinking surveys was the shift to probability sampling
methods. Fully probabilistic sampling
methods made their way into U.S. public opinion polling gradually, since they
raised the expense of conducting in-person household surveys (U.S. response
rates with mail surveys are unacceptably low, and too few houses had telephones
until recent decades). The process was
helped along by debacles such as the Literary Digest poll’s prediction that
Roosevelt would lose the 1936 election, which he won by a huge margin (probed
in an early publication by Don Cahalan -- Cahalan and Meier, 1939). Earlier adult general population surveys such
as those of Mulford and Miller (1960a) in Iowa used “modified quota samples” and other methods
short of full probability sampling, and such methods can still be found in use
in some U.S. alcohol surveys in the early 1970s (Harris, 1971; Johnson et al.,
1977). The use of sampling quotas,
typically to replace the last link in the chain of area-probability sampling,
tended to overrepresent those in the population more likely to be found at
home, and thus, for instance, to overrepresent abstainers. In the U.S., full probability sampling
methods were first applied in a community drinking practices survey by
Genevieve Knupfer and associates (Knupfer et al., 1963), and in a U.S. national
sample by Don Cahalan and associates (Cahalan, Cisin and Crossley, 1969).
By the mid-1970s, surveys of adult
drinking behaviour had been carried out and reported in a number of
societies. In addition to surveys in the
U.S. -- some already mentioned -- there had been surveys in Finland (Mäkelä,
1971) and more generally in the Nordic countries (Jonsson and Nilsson, 1969),
in Britain (Edwards et al., 1972; Dight, 1976), France (Sadoun et al., 1965),
Canada (Cutler and Storm, 1973), Australia (Encel et al., 1971), the
Netherlands (Gadourek, 1963), and Switzerland (Wütrich and Hausheer, 1979).
There was some cross-fertilization
of ideas between the research groups working in different countries from the
start, in the form of joint publications, correspondence, and in some cases
study visits. The Drinking and Drug Practices Surveyor, initiated in
1970, provided a venue for methodological work and discussions. Even more, the advent of the annual Alcohol
Epidemiology meetings in 1975, initially as a section of the International
Council on Alcohol and Addictions, provided a meeting-place for exchanges on
ideas and methods.
DEVELOPMENTS
IN ASKING ABOUT DRINKING
Despite the cross-fertilization,
substantial differences have persisted in national traditions of asking about
drinking behaviour. A major division has been between the tradition of asking
about recent drinking occasions, on the one hand, and asking for the respondent’s
summary of his or her customary drinking behaviour, on the other. The former tradition has been more often used
in European studies, and the latter in north American studies (Alanko, 1984,
Room, 1990). Recently, there seems to
have been some convergence of thinking about ways of asking about amounts of
drinking (e.g., Dawson, 1998, Stockwell et al., 1999). First, it can be agreed that there is no
single best way of asking about drinking behaviour -- the optimum method will
depend on the purpose for which the data is being collected. Second, if a recent-occasions method is being
used, an adequate characterization of an individual’s drinking requires
information stretching over several drinking occasions. Thus, unless frequent
drinking is characteristic in the cultural situation, collecting data only on
drinking occasions in the last week will result in considerable misestimation
and misclassification, particularly for less-frequent drinkers. Third, if a respondent’s-summary method is
being used, “usual quantity” is not sufficient as the method of asking about
quantities consumed on an occasion (Rehm et al., 1999). In particular, for most
purposes questions need to be asked about drinking larger amounts on an
occasion, even if they are not a “usual” amount.
There are several factors which have
kept the national traditions, once established, divergent. Analysts are used to working with a
particular set of questions, and there may also be a substantial investment in
computer code for converting responses to summary measures. More importantly, keeping questions
comparable with earlier surveys allows for analysis of trends over time, an
aspect of the literature that has become increasingly important as survey datapoints
cumulate. A particular set of questions
may also be attuned to the particular drinking customs of the culture. This
point was underlined by the experience of the WHO 3-country Community Response
Study in the 1970s: the last week’s drinking was an adequate way of asking
about drinking in Scotland, where drinking occasions are relatively frequent,
but was very unsuited to the infrequent drinking which is common in Mexico
(Rootman and Moser, 1985). While these
factors make it likely that the divergent questions in the national survey
series will continue to be asked, it may be possible to bridge the gaps between
national traditions with a short selection of cross-culturally comparable
questions to be asked in addition.
DEVELOPMENTS
IN SUMMARIZING DRINKING PATTERNS
Once the questions on drinking
behaviour have been asked, the researcher faces the task of aggregating and
summarizing them in the analysis.
Earlier surveys of drinking in the United States summarized the
patterning of drinking in terms of typologies, combining frequency of drinking
and summary dimensions of quantity per occasion in various ways. Although they
were composed from two and sometimes three dimensions of drinking behaviour,
the typologies were often treated in analysis as a single ordered dimension,
one which implicitly gave special weight to heavier drinking occasions. However, an alternative tradition, starting
with Knupfer et al. (1963) and continuing with the Volume-Variability measure
of Cahalan et al. (1969, Appendix 1) and later typologies of frequency of
drinking and of drinking 5+ drinks (Room, 1990), clearly distinguished a
dimension of frequency (or in which frequency was important) from a dimension
centred on whether or not the respondent sometimes drank larger quantities. A methodological article by Knupfer (1966)
emphasized how different the correlates of frequent light drinking and less
frequent heavy drinking were, although the two patterns might result in about
the same volume of drinking. As she wrote in a 1965 letter, “I have been
carrying the torch for the importance of ‘quantity’.... The essence of the
point might be put this way: we want a index that is more related to the blood
alcohol level of the drinker than to the profit level of the alcoholic
beverages literature” (quoted in Room, 1990).
In line with this theme, an article a few years later by Kettil Bruun
(1969) used a measure of frequency of intoxication based on calculations of
times respondents were above a threshold blood-alcohol level.
For a considerable period, this
emphasis was swamped by a shift in the literature to a single-minded focus on
summarizing drinking behaviour in terms of overall volume of drinking, in terms
of an average amount consumed per unit of time.
There were probably several reasons for this shift in focus. First, the literature became self-conscious
about issues of validity (e.g., Pernanen, 1974), and in this context the
proportion of alcohol sold which could be accounted for a survey became the
“gold standard” for the validity of drinking measures; volume of drinking was
the most direct equivalent of the sales figures, expressed as units of ethanol
per member of the drinking-age population.
Second, the Ledermann model became important in the literature. Since it
focused on hypotheses about the distribution of volume of drinking in a
population, in the course of controversies over its validity much energy was
spent on computing distributions of drinking volumes in different populations
(e.g., Bruun et al., 1975, p. 33). In
this light, complaints were voiced at Alcohol Epidemiology sessions about the
habit in U.S. surveys of using categorical typologies which could not be
converted into the volume dimension.
Third, volume of drinking was in principle a single continuous and
quantifiable dimension, and because of this was readily used in a variety of
multivariate statistical techniques which assumed a dependent variable of this
type. The fact that all the
respondent’s drinking was included in a volume measure made it seem like the
obvious choice as a summary measure of amount of drinking. Last, as medical epidemiological studies
began to pay more attention to drinking, they used a volume of drinking measure
as a matter of course (Edwards et al., 1994, p. 45), by analogy with other
measures of diet and behaviour; the influence of this prestigious literature
filtered back into the drinking patterns literature.
While the tradition of describing
the drinking pattern in terms of two or more dimensions never entirely
disappeared, the main emphasis for some years has been on volume of
drinking. Recently, however, there has
been a new emphasis on the importance of the patterning of drinking,
accompanied by critiques of the adequacy of approaches focused only on volume
of consumption (e.g., Single and Leino, 1998).
While this emphasis has sometimes been regarded as a new departure, in
reality it represents a return of a perspective which would have seemed
self-evident to such researchers in the earlier days of drinking surveys as
Genevieve Knupfer or Kettil Bruun.
With the return to thinking of and
characterizing drinking patterns in terms of two or more dimensions, however,
the field has still not reached any consensus on methods of summarization, and
particularly on how to handle drinking patterns in multivariate analyses. Categorical typologies remain awkward to use
in such analyses. On the other hand, the
problem of collinearity may hinder the solution of introducing two or more
dimensions of drinking patterns as separate variables. Introducing interaction terms can result in
difficulties in interpreting the results.
It is time for the field to get beyond conceptual arguments
re-emphasizing the importance of patterns in characterizing drinking, and on to
concrete discussions about ways of characterizing and summarizing patterns,
particularly in the context of multivariate analysis.
AREAS
FOR DEVELOPMENT IN CHARACTERIZING DRINKING PATTERNS
Driven in part by the medical
epidemiological literature’s focus on volume of drinking, a major
methodological focus has been on developing questions which are convertible
more and more exactly into an absolute metric of grams of ethanol consumed per
time period. But one cannot get
meaningful responses by asking respondents how many grams of ethanol they drink
on an average day. Most surveys ask
instead about something like “drinks” -- the units in which alcoholic beverages
are customarily consumed. Obviously, for a conversion from a “drink” to grams
of ethanol, one needs to know the strength of the alcoholic beverage, and how
much of the alcoholic beverage was poured into the drink. Both of these may vary from one occasion to
another. Methodological analyses showing how much “a drink” may vary by occasion,
by respondent, and by society (e.g., Turner, 1990) have fueled a search for an
international “standard drink”.
The survey researcher’s quandary
about these matters of measurement has sometimes reflected back into social
policy and programming. Since it was
developed by Susan Dight for a Scottish survey (Dight, 1976), the “standard
unit” has been a feature of British surveys.
The unit was a researcher’s construction to deal with the problem that
the predominant Scottish beverage, beer, was sold primarily in two different
drink-sizes, a half-pint and a pint.
Dight chose the smaller size as the “standard unit”, although an
ordinary male drinker in Scotland would think of “a drink” in terms of a pint
-- two standard units. When British
governments then moved to promoting “sensible limits” on drinking, the Dight
unit took on a new role as the metric for stating these limits. Not surprisingly, in view of the ordinary
drinker’s definition of “a drink”, the British “sensible limits” are often
misinterpreted.
From the point of view of survey
methods, the emphasis on standard units or drinks seems to me misplaced. Our basic job in asking respondents about
their drinking is to attune the questions to ways the respondent can
comfortably answer, not to try to impose some standard unit on them. The problem, of course, is that in
respondent’s-summary approaches the analyst would like respondents to summarize
in terms of equal levels of ethanol intake.
But this problem is probably best solved by allowing the respondent to
answer in terms of the respondent’s preferred units, but with the level
specified in terms of those units.
From the point of view of
understanding drinking behaviour itself, and also of understanding relations
between drinking and social and health harm, I would put first priority not on
a more exact calibration of grams of ethanol intake, but rather on paying more
attention to aspects of the cultural definition and social meaning of
drinking. Whether one drinks at all,
whether one takes a drink on a particular occasion, whether one gets drunk and
how drunk one gets are all structured by and hold implications for how people
think of and define themselves and others with respect to drinking. But the survey research literature has only
fitfully visited this territory of the social meaning of drinking -- most
commonly, probably, by asking questions about “reasons for drinking”. As Bacon
noted about American Drinking Practices (Bacon, 1969), the most common
mode of analysis has been basically a “demographic analysis”, which “does not describe the styles, procedures and
qualities of the drinking activity,... and only considers the sociocultural
settings in broad, almost abstract categories”.
Since American Drinking Practices,
indeed, the field may have somewhat regressed in terms of what Bacon was
looking for. For instance, Cahalan and
his coworkers (1969) did pay attention to type of beverage consumed, but it is
only recently has the literature returned to paying sustained attention to the
rather different profiles of harm associated with beer, wine and spirits
drinking (Room, 1976). A couple of
factors have turned our attention away from such matters. In the first place, drinking surveys have
most commonly been done in cultures where for some, at least, drinking at all,
and particularly drinking more than a little, has been a morally questionable
activity. To ask about socially defined
categories, researchers feared, was to invite responses oriented to social
desirability. And in fact, cross
cultural surveys have found evidence of such an effect. In societies with a
temperance tradition, a substantial fraction of people define themselves as
abstainers even though they had taken a drink at least once in the last year;
conversely, in very “wet” societies, some people who had not taken a drink in
the last year nevertheless define themselves as drinkers (Lindgren, 1973;
Nelker, 1973).
Second, meanings and definitions of
drinking are diverse. Excursions into
classifying drinking into socially meaningful categories have often ended up
with many types varying on many dimensions (e.g., Martin et al., 1992). Once these typologies have been described,
analysts do not find them easy to use further in multivariate analysis.
Third, the methodological
individualism of most survey research means that our attention has mostly
focused on the individual’s drinking patterns, rather than on the drinking
occasion as a collective social context (but see Simpura, 1991). The collective pattern of drinking in a
particular type of occasion may well have a stronger relation to harm or other
outcomes than the individual drinking patterns participants bring to the
occasion.
Measuring amount of ethanol consumed
on an occasion, or as a drinking pattern, is obviously important in
characterizing an individual’s drinking and understanding its relation to
potential harms. But it is far from the
full picture. Other parts of the picture include how the respondent and
bystanders defined the drinking, both in terms of the occasion and in terms of
patterns. Did the respondent consider
him/herself drunk, and did others? Was
the consumption and comportment while and after drinking expected, allowed or
disapproved of in the situation? These
matters of the social definitions surrounding drinking and intoxication need to
be measured alongside the grams of ethanol.
DEVELOPMENTS
IN ASKING ABOUT DRINKING-RELATED PROBLEMS
Asking general-population
respondents about adverse consequences of their drinking, like asking them
detailed questions about drinking patterns, really begins with Straus and
Bacon’s study of college students (1953), and has continued for the intervening
half century. But issues of measurement
and aggregation in this area have received less sustained collective attention
by researchers than the area of drinking practices.
Studies have varied greatly in the
number of items asked concerning drinking problems, but not very much in terms
of the areas about which questions are asked.
Usually, respondents are asked about adverse reactions of others to
their drinking. The occurrence of
casualties and physical health problems related to drinking are commonly
included in the list of questions.
Respondents are also often asked about problematic drinking comportment:
arguments or fights while drinking, drinking-driving, going to work with a
hangover. Also asked in one survey or
another are a wide variety of drinking-related behaviours or occurrences
considered symptomatic of addiction: such items as use of alcohol for coping,
drinking to relieve withdrawal, gulping drinks when noone is looking, and drinking
longer or more than intended. Many of
these items date back to the Grapevine survey constructed by members of
Alcoholics Anonymous and analyzed by Jellinek (1946) in terms of phases in the
natural history of alcoholism. While
surveys with problem items across this kind of range have perhaps been most
common in the United States, similar lists of “types of experiences related to
drinking” can be found in studies elsewhere, for instance, in Nordic surveys
(Mäkelä, 1981).
A new kind of item entered into use
as psychiatric epidemiology became more involved in measuring alcohol
problems. In constructing survey items,
survey researchers usually try to keep items as simple as possible, and avoid
“double-barrelled” questions. The
criteria for the psychiatric diagnoses in the field, however, often
deliberately combine different conceptual areas into the same criterion. A criterion like “continued drinking despite
knowledge of adverse consequences”, for instance, combines behaviour (continued
drinking), cognition (knowledge of...) and the occurrence of adverse
consequences. In seeking to
operationalize such criteria, those constructing questionnaires have felt
forced to construct items which are difficult to understand and to answer, and to which responses are
difficult to interpret.
Questions about drinking problems in
early drinking surveys were often phrased in terms of lifetime occurrence --
“did this ever” occur? Phrasing the
question in such terms obviously has the greatest chance of picking up positive
responses. Asking questions on a lifetime basis was also encouraged by the
clinical tradition of regarding alcoholism, along with other psychiatric
conditions, as lifelong once incurred.
Those of us engaged in longitudinal studies quickly realized that such
questions greatly hindered studies of changes in drinking problem status over
time -- with lifetime question, respondents could never get better, they could
only become invalid. In early studies,
the time period specified for “current problems” varied, from 6 months, in studies based on the DIS, to as
long as three years (Cahalan and Room, 1974).
The sporadic nature of many problems discouraged short time-periods; in
the end, the literature has settled down to 12 months as the usual time-period
for “current” problems.
This often raises problems for
analyses of the relation between drinking patterns and drinking problems. It would usually be desirable to have the two
domains measured on the time-period, but some drinking-patterns measures have been
based on shorter periods -- the last seven days or two weeks or 30 days. On the other hand, measuring drinking
patterns on a twelve-month base raises the issue of whether and how to measure
and analyze variability in patterns within the period. A variety of expedients have been used to
deal with this issue, but there has been no agreement on a particular solution.
It should be noted that the alcohol
survey tradition operates on a quite different epistemology from general
medical epidemiology in terms of the relation of alcohol consumption to social
and health problems (Edwards et al., 1994, pp. 48-50). Whereas the classic problem in medical
epidemiology is to demonstrate causation by correlating two conceptually
unrelated phenomena, in the alcohol survey tradition the causal connection is
built into the question a-priori. Often
the respondent him/herself is asked whether there is a problem and to make the
causal connection (“did your drinking have a harmful effect on your marriage or
home life”). In other questions, the
respondent is being asked about problematization by others (“a friend’s
feelings about your drinking threatened to break up your relationship”). In a third type of question, the
problematization comes from the researcher.
On its face, “I have often taken a drink first thing when I get up in
the morning” or “I find I have to drink more now to get the same effect as
before” do not describe problems; they become problematized only in terms of
the researcher’s interpretation of the behaviour. (The researcher’s interpretation does reflect
general clinical and cultural interpretations, raising the complication that
the respondent, too, is likely to know s/he is giving an answer that will be
seen as signalling a problem.)
These issues of imputation of cause
and of problem-ness deserve wider discussion in the international alcohol
research community.
DEVELOPMENTS
IN SUMMARIZING DRINKING PROBLEMS
From the first, U.S.
general-population surveys sought to aggregate drinking-problems items into one
or more summary scales. Mulford and Miller (1960b), for instance, constructed
one scale for “troubles due to drinking”, and another on “preoccupation with
alcohol”; in later work Mulford came to see the latter as more or less an
operational measure of alcoholism concepts.
Genevieve Knupfer, trained both in psychiatry and in sociology, took a
pragmatic and eclectic view of what should be measured under the rubric of “problem drinking” in the general population
when she turned to this issue in the mid-1960s (Knupfer, 1967). Knupfer’s approach was to identify different
conceptual areas of “problems from drinking”, and construct subscales in each
area. Some of these areas reflected
interpersonal problems, e.g., job problems, spouse problems, and problems with
the police. Along with physical health
problems from drinking, this group of problems were sometimes called “tangible
consequences”. Other problem-areas
described aspects of drinking behaviour which were defined by the analyst as
problematic -- e.g., “binge drinking”, “use of alcohol for coping”, “symptomatic drinking”, “loss of control”. A
“serious problem”, a “moderate problem” and a “no problem” level was defined in
each problem-area, either by a-prior decision (e.g., job loss was defined as
more serious than complaints at work), or by the number of positive responses
given to items in the area.
This basic system of about a dozen
problem-area scores was used in a series of publications by members of the
Berkeley group. Analysts differed, however,
in how the problem-area scores were presented in analysis. While Knupfer (1967) and Cahalan (1970)
presented prevalence rates for the individual problem-area scores, their main
attention tended to be on an “overall problems score” which added together
scores from all the problem-areas. Clark
(1966), on the other hand, kept the problem-area scores separate, focussing on
the extent of overlap between a positive score in one problem-area and a
positive score in another -- an approach Room (1977) also applied to problems
from opiate use. A third approach, used
by Cahalan and Room (1974), used a typology distinguishing “tangible
consequences” from binge drinking and other problematic consumption.
In later work in the same tradition,
Hilton (1991) primarily analyzed drinking problems in terms of two domains, one
identified as “dependence” and the other as “consequences”. A similar division between “personal” and
“social” consequences was used in analyzing a set of items in the WHO study of
Community Response to Alcohol Problems (Rootman and Moser, 1985).
Psychiatric epidemiology’s entry in
the alcohol epidemiology field affected summarizations of alcohol problems in a
number of ways. In the first place, the
tradition’s orientation to psychiatric nosology meant that drinking-problem
items were now to be aggregated in terms of the dichotomy of “making” or “not
making” a diagnosis for the particular respondent. Initially, the questionnaires and analyses
were oriented to DSM-III, a classification with two main diagnoses, “alcohol
dependence” and “alcohol abuse”. Given
the fact that the latter diagnosis could only be made in the absence of the
former, and that the two diagnoses were not conceptually very distinct, most
publications in the initial wave reported only a combined dichotomy of those
qualifying or not qualifying for either alcohol dependence or abuse.
Currently, psychiatric
epidemiological studies in the alcohol field usually measure whether a
respondent qualifies for a diagnosis on four main diagnoses: “alcohol
dependence” and “alcohol abuse” in DSM-IV, and “alcohol dependence syndrome”
and “harmful use of alcohol” in ICD-10.
The two alcohol dependence measures are close but not identical. In principle, ICD-10 “harmful use” (and, in
view of overlap in criteria, ICD-10 dependence) are supposed to include harm to
physical and psychological health, but not to include social and interactional
consequences of drinking. On the other
hand, DSM-IV “alcohol abuse” is unambiguously a measure of legal and other
external and social consequences of drinking.
Given its orientation and
epistemology, the psychiatric epidemiology tradition has been little interested
in the issues of causal relationship and conceptual clusterings that have
concerned the social epidemiological tradition of measuring alcohol problems. On the other hand, the psychiatric
epidemiology tradition has been much more oriented to psychometric traditions
of measurement and of establishing the scientific respectability of measures
with test-retest reliability studies (Kirk and Kutchins, 1992). Studies in both traditions tended to find
that, applying large assortments of “problem” items to a non-clinical
population, a strong general factor tended to emerge in principal components
factor analysis. The psychiatric
epidemiology tradition has tended to regard this as evidence for the validity
of a single generalized dependence concept (e.g., Hasin et al., 1994). The social epidemiology tradition has tended
to take a more limited view of the significance of this finding, regarding the
underlying commonality indicated by the factor as simply a willingness to get
quite drunk (or to acknowledge getting quite drunk).
The psychiatric epidemiology
tradition, on the other hand, is having considerable trouble
fitting
the findings for “alcohol abuse” or “harmful use” into its paradigm. Again, both traditions report the same
findings: items from abuse/harmful use load into a single common factor with
dependence items in factor analyses (e.g., Hasin et al., 1994), but the
commonality among items tends to be least for indicators of social reactions
and other problems related to drinking.
From the point of view of psychiatric epidemiology’s psychometric
traditions, this has led to questioning of the viability of a separate arena of
harmful use or abuse (i.e., tangible consequences of drinking).
In my view, these developments
suggest the limitations of the standard psychometric paradigm as a guide to
conceptualization and aggregation in measuring alcohol problems. The general factor underlying the items can
be interpreted in terms other than a unified dependence syndrome. On the other
hand, the fact that two items do not have strong positive correlation does not
indicate much about their conceptual relationship. If they are conceptualized as alternative
manifestations of the same phenomenon, they might even have a strong negative
correlation, and still belong in the same measure. I suspect also that the psychometric paradigm
has driven the adoption of conceptually mixed criteria in the nosologies, noted
above as resulting in a nedd to create technically problematic survey
items. It may be hard to know what
responses to the resulting “portmanteau” or double-barreled items indicate, but
it certainly tends to increase the alphas and other scale-construction
statistics if the different scale items all reach across component conceptual
domains.
In recent years, a further tradition
has strengthened its position in the field of drinking problems measurement,
with the development and application of brief screening instruments in
nonclinical populations. Given its
pragmatic purposes, a screening instrument makes no claims to be measuring
diagnoses, or about the conceptual status of its component items. The criterion for including items in a
screening instrument are firstly the extent to which, as scored together, they
approximate an underlying condition which is of clinical interest, and
secondly, that “false negatives” be kept to a minimum. In this context, conceptual clarity is
irrelevant, and screening measures often combine items across a range of
conceptual domains, frequently asked on a lifetime basis. The two screening measures which are probably
now most widely used in population surveys, CAGE and AUDIT, thus both include
items on drinking behaviours, on cognitions about drinking, and on the
reactions of others; responses across these different domains are simply summed
to yield an overall score on the measure.
WHERE
DO WE GO IN MEASURING DRINKING PROBLEMS?
The field is currently in a confused
state with respect to the measurement of drinking problems. On one side of the literature, the
development in psychiatric epidemiology has culminated in very lengthy sets of
questions, designed to map as exactly as possible the DSM-IV and ICD-10
diagnostic criteria and specifications.
The AUDADIS questionnaire developed by Bridget Grant and her coworkers
(Stinson et al., 1998) may represent the furthest likely elaboration of this
tradition, with the number of questions needed for the diagnostic algorithms
threatening to take over the entire interviewing time. In terms of the standards of the psychiatric
epidemiology literature, such a questionnaire is undoubtedly state-of-the-art,
both in terms of the detailed coverage of the diagnostic specifications and in
terms of the impressive psychometric underpinnings such as cross-cultural
reliability testing (Chatterji et al., 1997).
But instruments like AUDADIS are beyond the scope of a multipurpose or
monitoring survey.
At another boundary of the
literature is the kind of analysis represented by studies by Mäkelä and
Mustonen (1988) of the relation of drinking problems to alcohol intake, in
which each of a number of drinking problem items is analyzed separately, with
no aggregation at all. Such an analysis
avoids entering the psychometric entanglements of aggregating across drinking
problems (although the question of the validity of responses of course
remains). This strategy also has the
advantage of relating more immediately to contextual and environmental
approaches to preventing drinking problems, since the contextual and
environmental issues tend to vary from one kind of problem to another.
A third direction in the literature
is the search for relatively limited lists of items which can measure a fair
representation of alcohol problems, and
usable summary measures, in multiple-purpose questionnaires. A common recourse for this purpose at the
moment is a screening measure such as the AUDIT, which has the advantage of a
considerable track-record of use and of psychometric testing. For purposes such as a general tracking
measure to be used in repeated surveys in a national population, a measures
like the AUDIT may indeed be suitable.
But if the study’s purposes include
a better understanding of interactions -- for instance, of the relation between
drinking patterns and alcohol-related problems -- a measure like the AUDIT,
which reaches across these dimensions, is useless unless dissolved into its
component parts. Here what are needed
are usable measures with a clear separation of domains of meaning -- at a
minimum, there is a need to return to a separation between drinking behaviour,
cognitions about drinking, and adverse consequences of drinking. Alternatively, a DSM-IV-based split between
drinking behaviour, “alcohol dependence”, and “alcohol abuse” would be
serviceable, although with a recognition that both drinking-behaviour and
consequences (“abuse”) elements are hidden within the dependence
construct.
At the moment, the market is fairly
open for relatively short measures which cover such domains of meaning. In the area of cognitive experiences of
craving and impairment of control, scales developed and initially subjected to
psychometric testing in clinical environments, such as the 25-item Alcohol
Dependence Scale, are probably serviceable.
An alternative, with less psychometric testing so far, would be a short
summary scale measuring the criteria of ICD-10 dependence, such as has been
used at the Alcohol Research Group in Berkeley and ARF in Toronto, and for
marijuana at the Sydney centre (Swift et al., 1998).
In the area of tangible consequences
of drinking -- legal, social, interactional and health problems -- there may be
a need to start again. Discussion is
needed about the question of causal attribution -- whose attribution we should
be depending on, for what analytical purposes.
Work is needed on building and testing new measures in each specific
area of tangible consequences of drinking.
To a considerable degree, the task is to undertake an updating of the
kind of thinking done by Genevieve Knupfer in building her problem-area scales,
with attention to defending the results in the light of the current
psychometric literature. An area which
needs particular attention is family, relationship and other interactional
problems. This area bulks large in
social concerns about drinking, but we have almost no alcohol-specific social
statistics for this area, and monitoring population levels and trends in problems
in this area will depend on developing adequate survey measurements.
Also needed is thinking and
collective discussion about the bases for and methods of aggregation across
problem-areas in terms of such constructs as “tangible consequences” of
drinking. Here current thinking about revising
the International Classification of Impairments, Disabilities and Handicaps
(WHO, 1997; see http://www.who.ch/icidh), and associated work on developing
disablement assessment instruments, may provide some useful leads.
CONCLUSION
In the light of the foregoing, the
conclusion to this discussion will be no surprise. Much has been done and learned in the last
half-century in measuring and analyzing drinking patterns and problems. But, particularly with respect to drinking
problems, we have reached the point of seeing that there are substantial
problems with all the approaches common in the literature. We are far from reaching the stage of mature
science. There is thus plenty of
thinking and research about these issues to be done in the new millenium.
REFERENCES
Alanko, Timo, An
overview of techniques and problems in the measurement of alcohol consumption, Research
Advances in Alcohol and Drug Problems 8:209-226, 1984.
Bacon, Selden D.,
Introduction, pp. xv-xxvi in Don Cahalan, Ira H. Cisin and Helen M. Crossley, American
Drinking Practices: A National Study of Drinking Behavior and Attitudes,
New Brunswick, NJ: Rutgers Center of Alcohol Studies, Monograph No. 6, 1969.
Billings, John S. Data
relating to the use of alcoholic drinks among brain workers in the United
States, pp. 307-338 in J.S. Billings, ed., Physiological Aspects of the
Liquor Problem, vol. 1. Boston: Houghton Mifflin, 1903.
Bruun, Kettil, The
actual and the registered frequency of drunkenness in Helsinki, British
Journal of Addiction 64:3-8, 1969.
Bruun, Kettil, et al.,
Alcohol Control Policies in Public Health Perspective. Helsinki: Finnish
Foundation for Alcohol Studies, Volume 25, 1975.
Cahalan, Don, Ira
Cisin and Helen Crossley, American Drinking Practices: A National Study of
Drinking Behavior and Attitudes. New Brunswick, NJ: Rutgers Center of
Alcohol Studies, Monograph No. 6, 1969.
Cahalan,
Don and N.C. Meier, The validity of mail-ballot polls. Psychol. Rev.
3:1-12, 1939.
Cahalan, Don and Robin
Room, Problem Drinking among American Men. New Brunswick, NJ: Rutgers
Center of Alcohol Studies, Monograph No. 7, 1974.
Chatterji, Somnath,
John B. Saunders, Radu Vrasti, Bridget F. Grant, Deborah Hasin and D. Mayer,
Reliability of the alcohol and drug modules of the Alcohol Use Disorder and
Associated Disabilities Interview Schedule -- Alcohol/Drug-Revised
(AUDADIS-ADR): an international comparison, Drug and Alcohol Dependence
47:171-185, 1997.
Clark, Walter,
Operational definition of drinking problems and associated prevalence rates, Quarterly
Journal of Studies on Alcohol 27:648-668, 1966.
Cutler, Ron and Thomas
Storm, Drinking Practices in Three British Columbia Cities: I. General
Population Survey. Vancouver: Alcoholism Foundation of British Columbia,
1973.
Dawson, Deborah, ed.,
Proceedings: International Workshop on Consumption Measures and Models for Use
in Policy Development and Evaluation, Alcoholism: Clinical and Experimental
Research 22(2) Supplement:1S-81S, 1998.
Dight, Susan, Scottish
Drinking Habits. London: HMSO, 1976.
Edwards, Griffith, J.
Chandler and Celia Hensman, Drinking in a London suburb: I. correlates of
normal drinking, Quarterly Journal of Studies on Alcohol, Supplement
6:69-93, 1972.
Edwards, Griffith, et
al., Alcohol Policy and the Public Good.
Oxford, etc.: Oxford University Press, 1994.
Encel,
S., K.C. Kotowicz and H.E. Resler, Drinking practices and attitudes in an
Australian
metropolitan area, Social
Science and Medicine 6:469-482, 1971.
Gadourek, I. Riskante gewoonten en zorg voor eigen
welzijn. Groningen: Wolters, 1963.
Harris, Louis and
associates, American Attitudes toward Alcohol and Alcoholics, study
2138, for NIAAA. Louis Harris and Associates, 1971.
Hasin, Deborah B.,
Bengt Muthén, K.S. Wisnicki and Bridget Grant, Validity of the biaxial
dependence concept: a test in the U.S. general population, Addiction
89:573-579, §994.
Hilton, Michael,
Demographic characteristics and the frequency of heavy drinking as predictors
of self-reported drinking problems, pp. 194-212 in Walter B. Clark and Michael
E. Hilton, eds., Alcohol in America: Drinking Practices and Problems.
Albany: State University of New York Press, 1991.
Jellinek, E.M. Phases
in the drinking history of alcoholics: analysis of a survey conducted by the
official organ of Alcoholics Anonymous, Quarterly Journal of Studies on
Alcohol 7:1-88, 1946.
Johnson, Paula, David
Armor, Michael Polich and Harriet Stambul, U.S. Adult Drinking Practices:
Time Trends, Social Correlates and Sex Roles, Report to NIAAA. Santa
Monica, CA: RAND Corp., 1977.
Jonsson, Erland and
Tom Nilsson, Samnordisk undersokning av vuxna mans alkoholvanor (A joint
Nordic study of the drinking habits of adult males). Stockholm, mimeographed,
1969.
Kirk, Stuart A. and
Herb Kutchins, The Selling of DSM: The Rhetoric of Science in Psychiatry.
Hawthorne, NY: Aldine de
Gruyter, 1992.
Knupfer, Genevieve,
Some methodological problems in the epidemiology of alcoholic beverage usage:
definition of amount of intake, American Journal of Public Health
56:237-242, 1966.
Knupfer, Genevieve,
The epidemiology of problem drinking, American Journal of Public Health
57:973-986, 1967.
Knupfer, Genevieve,
Raymond Fink, Walter B. Clark and Angelica S. Goffman, Factors Related to
Amount of Drinking in an Urban Community, Drinking Practices Study Report
No. 6. Berkeley, CA: State Department of Public Health, 1963.
Lindgren, Åke, Some
results from an international series of drinking surveys, Drinking and Drug
Practices Surveyor 8:34-45, 1973.
Mäkelä, Klaus, Measuring
the Consumption of Alcohol in the 1968-1969 Alcohol Consumption Study,
Report No. 2. Helsinki: Social Research Institute of Alcohol Studies, 1971.
Mäkelä, Klaus, Scandinavian
Drinking Survey: Construction of Composite Indices of Drinking Attitudes and
Personal Experiences Related to Drinking. Oslo: National Institute for
Alcohol Research, SIFA mimeograph No. 47, 1981.
Mäkelä, Klaus and Heli
Mustonen, Positive and negative experiences related to drinking as a function
of annual alcohol intake, Biritish Journal of Addiction 83:403-408,
1988.
Martin, C., Alan
Wyllie and Sally Casswell, Types of New Zealand drinkers and their associated
alcohol-related problems, Journal of Drug Issues 22:773-796, 1992.
Mulford, Harold A. and
Donald E. Miller, Drinking in Iowa: II. The extent of drinking and selected
sociocultural categories, Quarterly Journal of Studies on Alcohol
21:26-39, 1960a.
Mulford, Harold A. and
Donald E. Miller, Drinking in Iowa. IV. Preoccupation with Alcohol and
Definitions of Alcohol, Heavy Drinking and Trouble due to Drinking, Quarterly
Journal of Studies on Alcohol 21:279-291, 1960b.
Nelker, Gunnar, Total
abstention -- as an attitude and as a behavior -- some further notes, Drinking
and Drug Practices Surveyor 8:45-47, 1973.
Pernanen, Kai, Validity
of survey data on alcohol use, Research Advances in Alcohol and Drug
Problems 1:355-374, 1974.
Rehm, Jürgen, Thomas
K. Greenfield, Gordon Walsh, Xiaodi Xie, Linda Robson and Eric Single,
Assessment methods for alcohol consumption, prevalence of high risk drinking
and harm: a sensitivity analysis, International Journal of Epidemiology
28:219-224, 1999.
Room, Robin, Beverage
type and drinking problems in a national sample of men, Drinking and Drug
Practices Surveyor 12:29-30, 1976.
Room, Robin, Notes on
the spectrum of opiate use, pp. 40-50 in Joan D. Rittenhouse, ed., The
Epidemiology of Heroin and Other Narcotics, DHEW Publication No. (ADM)
78-559. Washington: National Institute
on Drug Abuse, Research Monograph No. 16, 1977.
Room, Robin, Measuring
alcohol consumption in the United States: methods and rationales, Research
Advances in Alcohol and Drug Problems, 10:39-80, 1990.
Rootman, Irving and
Joy Moser, Community Response to Alcohol-Related Problems. Washington, DC: USGPO, 1985.
Sadoun, R., G. Lolli
and M.M. Silverman, Drinking in French Culture. New Brunswick, NJ:
Rutgers Center of Alcohol Studies, 1965.
Simpura, Jussi,
Studying norms and contexts of drinking, Contemporary Drug Problems
18:477-498, 1991.
Single, Eric and
Victor Leino, Levels, patterns and consequences of drinking, pp. 7-24 in Marcus
Grant and Jorge Litvak, eds., Drinking Patterns and Their Consequences.
Washington, DC: Taylor and Francis, 1998.
Stockwell, Tim,
Deborah Dawson, Harold Holder, David Jernigan, Maria Elena Medina Mora and Eric
Single, International Guidelines for Monitoring Alcohol Consumption and Harm,
draft March 1999.
Straus, Robert and
Selden D. Bacon, Drinking in College. New Haven: Yale University Press,
1953.
Swiecicki, Andrzej. Teenagers and drinking
in Poland, Drinking and Drug Practices Surveyor 6:4-6, 1972.
Swift, Wendy, Jan
Copeland and Wayne Hall, Choosing a diagnostic cut-off for cannabis dependence,
Addiction 93:1681-1692, 1998.
Turner, Charlotte, How
much alcohol is in a “standard drink”? An analysis of 125 studies, British
Journal of Addiction 85:1171-1175, 1990.
World Health
Organization, ICIDH-2: International Classification of Impairments,
Activities, and Participation: A Manual of Dimensions of Disablement and
Functioning: Beta-1 Draft for Trials, June 1997. Document
WHO/MSA/MNH/EAC/97.2. Geneva: World Health Organization, 1997.
Wütrich, Peter and H.
Hausheer, Alkohol in der Schweiz. Frauenfeld: Huber, 1979.