close

Вход

Забыли?

вход по аккаунту

?

jsa.1987.48.202

код для вставкиСкачать
Journal of Studies on Alcohol, Vol. 48, No. 3, 1987
Validity of the MacAndrew
Medical Population
LEO J. DAVIS, J?,., PH.D., ROBERT
KENNETH
P. OFFORD, M.S.
Scale in a General
C. COLLIGAN,
PH.D.,
ROBERT
M. MORSE,
M.D.
AND
Department of Psychiatry and Psychology, and Section of Medicine Research Statistics, Mayo Clinic and Mayo
Foundation,
200 First Street S.W.,
Rochester, Minnesota 55905
ABSTRACT. The Minnesota Multiphasic Personality Inventory
(MMPI) was scored for the MacAndrew (MAC) Scale in a series
of samplestotaling 14,789 subjects. These samplesincluded 1077
substance-dependentpatients (739 men), 7090 selected medical
outpatients (2853 men), 5000 unselectedmedical outpatients (2500
men), 214 psychiatricinpatients (92 men) and 1408 contemporary
normals (646 men). Overall, the MAC Scale correctly classified
only 70.7�of the men and 37.9�of the women alcoholics.
However, when samples of men and women alcoholics were
broken down into four age groups, the MAC Scale correctly
identified 90�of men alcoholics in the 18- to 24-year age group
FFICIENTandinexpensive
methods
areneeded
for the detection
of alcoholism
because it is a
major health-care problem (NIAAA, 1983). A study
by Myers et al. (1984) suggestedthat alcohol abuse/
dependence is the most prevalent psychiatric disorder
among men in the United States. In women, it is
among the four most frequent disorders.
The use of questionnaires for the detection of
alcoholism has merit. They are generally inexpensive
and are appropriate for the mass screeningof various
populations. Miller (1976) divided questionnaires into
direct and indirect measures, where direct questionnaires inquire specifically about drinking behavior
and attitudes, and indirect questionnaires do not.
Most of the indirect measures have been developed
from the Minnesota Multiphasic Personality Inventory
(MMPI).
The MacAndrew (MAC) Scale (MacAndrew, 1965)
has been the most thoroughly investigated of the
currently available MMPI-derived scales and the one
that clinicians believe has shown the most promise.
However, in reviewing the literature on the MAC
Scale, we found that sample sizes were frequently
inadequate, reliability was seldom reported and few
studies focused on women alcoholics. In addition,
the MAC Scale was developedto differentiate probReceived: 23 December 1985. Revision: 21 July 1986.
202
but only 41�of women alcoholics in the same age group. In
addition, the predicted rate of alcoholism in the samples that
were not substance-dependent,
using the MAC Scale, was approximately 1.5 to 4 times the estimated lifetime prevalencerate
(this was also true in the 18- to 24-year age group). MMPI item
215 ("I have used alcohol excessively")correctly identified 95O7o
of the men and 94� of the women
alcoholics.
It was concluded
that, except for men less than 24 years old, the MAC Scale is
not an appropriate screeningtest for substancedependencein a
predominantly middle-classmedical population. (J. Stud. Alcohol
48: 202-206, 1987)
lem drinkers from psychiatric outpatients. This last
point was of particular concern because the MAC
Scale may not be appropriate for our target population of medical outpatients. With these findings in
mind, we decided to investigate the usefulness of the
MAC Scale in various inpatient and outpatient medical samples and the scale's performance in a normal
sample.
Method
Subjects
The subject pool for this investigation consistedof five
samples (Table 1):
TABLE1. Mean (+ SD) ages for the five samples
Men
Sample
Substance-dependent
Alcohol only
Drugs only
Alcohol + drugs
N
Women
Age
N
Age
739
43.0
+
14.7
338
44.3
+
525
46.0
+
14.0
211
46.0
+ 14.6
14.5
71
39.4
+
13.5
70
43.1
+
12.7
143
34.0
+
13.8
57
39.5
+
15.3
2853
47.6
+
14.6
4237
48.3
+
14.7
2500
outpatients
Contemporary normals 646
92
Psychiatric inpatients
50.5
+
13.7
2500
49.7
+
13.9
47.5
+
16.9
762
46.1
+
18.1
37.5
+
15.1
122
41.8
+
15.3
Selected medical
outpatients
Unselected
medical
MAcANDREW
SCALE
AND
MEDICAL
was obtained
before admission.
emotional
factors
were
considered
the 49-item
to have
a role
and the 51-item
version
of the MAC
of alcoholism.
??timates of base rate
The prevalence of alcoholism in our contrast samples is
unknown, although a study by Swenson and Morse (1975)
suggested that the prevalence rates for men and women in
our medical patient population may be 16 and 6%, respectively. A study by Robins et al. (1984) assessedthe
lifetime prevalence of alcohol abuse and dependence in
three sites--New Haven, Connecticut; Baltimore, Maryland;
and St. Louis, Missouri. Their data suggest that lifetime
prevalence rates of 20�for men and 5�for women are
in the
development of current symptoms.
Unselected medical outpatients. From 1962 to 1965, the
MMPI
was administered to 50,000 consecutive medical
outpatients at the Mayo Clinic. This sample is an unbiased
sample of medical patients because the MMPI was administered not on physician request but as part of a research
protocol (Swenson et al., 1973). A 10�random sample
(2500 men and 2500 women) from this group was used
for the present study. In addition, this subsample was
included in the present study becausethe MMPI data were
collected during the same time period as the original data
used in the development of the MAC Scale.
Contemporary normals. During 1983, MMPI data on a
realistic.
These
estimates
are
closest
to
those
found
for
New Haven, an area that most approximates our own
socioeconomic area. We used these estimates in our analyses
of the MAC Scale. The use of lifetime prevalence rates
instead of short-term rates is supported by data that suggest
the MAC Scale is a stable measure reflecting an enduring
characteristic of persons (Schwartz and Graham, 1979).
Results
Classification accuracy
randomly selectedmidwesternnormative sample (N = 1408,
646 men) were obtained by Colligan et al. (1983). These
data provided a sample of persons who were not under
physician care for any physically or mentally handicapping
For those with only alcohol dependence, the MAC
Scale correctly identified only 70.7�of the men and
37.9� of the women (Table 2). In addition, for
those with only drug dependence, only 29.6�of the
men and 22.9� of the women were correctly identified. These last findings are noteworthy because the
condition.
Psychiatric inpatients. This sample consistedof 214 psychiatric patients (92 men) hospitalized in 1983. The diagnoses varied, but most of the patients were treated for
depression and psychosomatic disorders.
MAC
Scale
has
been
described
as
a
measure
Patients who appeared in both the outpatient and the
Classification accuracy of the MAC Scale
Cutting score of men
Sample
Substancedependent
Alcohol only
Drugs only
Alcohol + drugs
Selected medical outpatients
Unselected medical outpatients
Contemporary normals
Psychiatric inpatients
Cutting score of women
23 and below
24 and above
23 and below
24 and above
N
N
N
N
245
154
50
41
1901
1478
400
50
O7o
33.2
29.3
70.4
28.7
66.6
59.1
61.9
54.3
of
"addiction proneness" (Kranitz, 1972; Lachar et al.,
1976). In the remaining samples, the MAC Scale
overestimated the lifetime prevalence rates by a factor
Data analysis
TABLE2.
version
Scale. MacAndrew (1965) had originally developed a 51item scale, but subsequently omitted item 215 ("I have
used alcohol excessively") and item 460 ("I have used
alcohol moderately, or not at all") because these items
dealt directly with alcohol use. We used his 49-item scale
and his recommended cutting score of >_ 24 as indicative
The first and fourth
criteria were employed to eliminate MMPIs obtained while
the patient was undergoing withdrawal or was still acutely
intoxicated. Patients were classified into three groups on
the basis of DSM-III (American Psychiatric Association,
1980) criteria for (1) alcoholism, (2) alcoholism and drug
dependenceand (3) drug dependence.Before 1980, National
Council on Alcoholism (NCA) (1972) criteria were used.
Contemporary selected medical outpatients. This sample
consisted of 7090 medical outpatients (2853 men) to whom
the MMPI was administered in 1983 at the request of the
primary or consulting physician because psychological or
203
ADDU samples were deleted from the outpatient group
but retained in the ADDU sample. When all duplications
had been eliminated, the protocols were scored for both
Substance-dependentpatients. This sample consisted of
1077 patients (739 men) admitted for treatment to the
Alcoholism and Drug DependenceUnit (ADDU) from 1979
through 1983. The initial sample consisted of 1645 patients
but 568 were excluded because (1) no MMPI was obtained
after admission, (2) an invalid MMPI was obtained, (3)
duplicate admissions were excluded or (4) the existing
MMPI
OUTPATIENTS
494
371
21
102
952
1022
246
42
O7o
66.8
70.7
29.6
71.3
33.4
40.9
38.1
45.7
223
131
54
38
3455
2055
603
93
O7o
66.0
62.1
77.1
66.7
81.5
82.2
79.1
76.2
115
80
16
19
782
445
159
29
O7o
34.0
37.9
22.9
33.3
18.5
17.8
20.9
23.8
204
DAVIS,
COLLIGAN,
MORSE
that varied from 1.5 to 4 times greater than the
expected value.
We thought that the MAC Scale might be differentially discriminating at various ages, so we divided
our samples into four age categories (18-24, 25-44,
45-64 and 65+ years). The MAC Scale correctly
identified
90� of
the
men
alcoholics
and
88� of
the alcohol-plus-drug group in the 18- to 24-year age
group. Sensitivity decreased rapidly, however, in the
older age groups. We noted no other substantial
changes in classification in the other samples across
these age groups.
Although we used the cutting score of _> 24 originally recommended by MacAndrew, other cutting
scores have been suggestedin the literature. Therefore, we decided to use a cutting score that approximated a sensitivity rate of 90�for men and women
with alcohol dependenceonly. This cutting score was
_> 21
for
men
and
_> 18 for
cutting scores, the MAC
89� of
both
our
men
women.
With
these
Scale correctly identified
and
women
with
AND
OFFORD
a prevalence rate of 20� for men alcoholics, we
obtained an estimate of positive predictive value of
30�and negative predictive value of 89�.Using a
prevalence rate of 5� for women alcoholics, we
obtained an estimate of positive predictive value of
10070and negative predictive value of 96�. These
positive predictive values mean that, on the average,
only 30� of the men in an unselected medical
outpatient population with a score of _> 24 would
be true positive for alcoholism. Similarly, 10070of
the women in the same population with a score of
_>24 would be true positive for alcoholism.
Endorsement frequencies
We studied the endorsementfrequenciesof all the
items in the MAC Scale, including items 215 and
460. We found item 215 effective in identifying the
patients in our alcoholic sample. This was most
notable in our sample of women alcoholics, when
compared with the performance of the MAC Scale.
alcohol
Item 215 correctly identified 95.0�of men alcoholics
dependence only, but at the same time produced
lifetime prevalence estimates in the other four samples
and 93.8�of women alcoholics (Table 3) and provided estimatesof lifetime prevalence that are closer
that varied
to the expected values than those obtained with the
from
60 to 69� for men and 47 to 55�
for women. Thus lifetime prevalence rates were overestimated by a factor that varied from 3 to 10.5
times the expected value.
The high false-positive rate--that is, low specificity--is especially damaging when the MAC Scale is
viewed from the perspective of predictive value. As
Baldessarini et al. (1983) have noted, specificity is
crucial to the estimation of the positive predictive
value of a test when applied to populations with
different prevalence rates. Briefly stated, positive
predictive value is the percentage of true-positive
results as a portion of all positive results. Negative
predictive value is the percentage of true-negative
results as a portion of all negative results. Applying
the formulas presented by Baldessarini et al. to the
performance of the MAC Scale on our unselected
medical outpatients yields the following results. Using
TABLE3.
MAC
Scale.
Internal consistency reliability coefficients
Kuder-Richardson
no.
20
coefficients
were
com-
puted for all five samples, by sex (Table 4). The
coefficients were uniformly low, suggestingthat the
items that comprise the MAC Scale do not form a
unified core of homogeneousitems.
Discussion
The MAC Scale did not accurately identify our
known alcoholics and was especially poor in identifying women alcoholics. This latter finding is not
unexpected because the scale was not developed on
female samples. Depending on sex and age group,
the MAC Scale overestimateslifetime prevalence rates
Endorsement frequenciesof MMPI item 215 ("I have used alcohol excessively")
Men
Women
Not endorsed
Sample
Substance dependent
Alcohol only
Drugs only
Alcohol + drugs
Selectedmedical outpatients
Unselectedmedical outpatients
Contemporary normals
Psychiatric inpatients
N
%
82
26
45
11
2026
2000
426
54
11.1
5.0
63.4
7.7
71.0
80.0
65.9
58.7
Endorsed
N
657
499
26
132
827
500
220
38
Not endorsed
%
N
%
88.9
95.0
36.6
92.3
29.0
20.0
34.1
41.3
79
13
60
6
3814
2375
648
96
23.4
6.2
85.7
10.5
90.0
95.0
85.0
78.7
Endorsed
N
259
198
10
51
423
125
114
26
%
76.6
93.8
14.3
89.5
10.0
5.0
15.0
21.3
MAcANDREW
TABLE4.
SCALE
AND
Internal consistency
reliabilitycoefficients(Kuder-Richard-
son no. 20)
Sample
Substancedependent
Alcohol only
Drugs only
Alcohol + drugs
Selected medical outpatients
Unselected medical outpatients
Contemporary normals
Psychiatric inpatients
Men
Women
.53
.49
.52
.57
.59
.47
.51
.65
.34
.37
.28
.24
.39
.37
.38
.43
by a factor of 1.5 to 4. When a different cutting
score that optimizes sensitivity is used, the result is
an unacceptably high false-positive rate in our contrast samples.
One explanation for this failure in case detection
of alcoholics is that the MAC Scale was developed
on a sample representing an entirely different target
population. We know from prior research (Dietvorst
et al., 1978) on the Wechsler Adult Intelligence Scale
(WATS) that our substance-dependentpopulation is
skewed toward higher socioeconomic levels, with a
higher proportion of professional and managerial
patients than was used in the standardization of
WATS. Also, our results suggest that age among
alcoholic
men is a crucial
factor
because the MAC
Scale identified 90�of our alcoholic men in the age
range 18 through 24 years while still overestimating
lifetime prevalence rates by a factor of 1.2 to 2. In
older age groups, the discriminatory accuracydeclines
rapidly.
Another
reason for the lack of accurate identifi-
cation may be that proponents of indirect scalesof
alcoholism,such as the MAC Scale, presupposethe
existenceof a certain behavioral type peculiar to
alcoholismor other addictionsthat might be identified
by empiricallyderived descriptors--in other words, a
type of personalityor personalitystyle with a predispositionto alcoholism.Much effort has gone into
the investigationof this issuein the past. Although
some clinicians
continue
to view
alcoholism
as a
MEDICAL
OUTPATIENTS
205
MMPI item 215 ("I haveusedalcoholexcessively")
is much more accuratein identifyingboth men and
women alcoholics than is the MAC Scale. Evaluation
of item 215 comes closer to estimates of lifetime
prevalence. That it does better than the 51-item
versionof the MAC Scale(which includesitem 215)
is probablydue to the largevariabilityintroducedby
the bulk of the scale, which obscures the discriminatory capacity of item 215.
An alternative explanation of the effectivenessof
item 215 is that it is endorsedby alcoholicsalready
admitted to an inpatient treatment program. However, we are skeptical of this explanation because
most of our patientshave not acceptedthe diagnosis
of alcoholismat the time of testing and enter treatment reluctantlyor under outsidepressure.While not
acceptingthat they are alcoholics,these patients are
nonethelesssurprisingly candid in reporting their
drinking behavior. This last statement also underscoresour experiencethat indirect scales, such as the
MAC Scale,are insensitive,at leastin our population
of medical patients. Our own unpublisheddata on
other MMPT-derived
alcoholism tests and a direct
scale, the Self-AdministeredAlcoholism Screening
Test (SAAST), a derivative of the Michigan Alcoholism Screening Test (MAST), indicate that the
SAAST differentiates alcoholics from nonalcoholics
with far greatersensitivityand specificitythan any
indirect scale.(COLLIGAN,
R. C., DAVIS,L. J., JR.,
MORSE, R. M. AND OFFORD, K. P. The not-somagnificent seven:The accuracyof MMPI alcoholism
scales in screeningmedical patients. Unpublished
manuscript, Mayo Clinic, 1986; DAVIS, L. J., JR.,
HURT, R. D., MORSE,R. M. AND O'BRIEN, P. C.
Discriminant analysis of the self-administeredalco-
holismscreening
test. Unpublished
manuscript,
Mayo
Clinic, 1987).
Acknowledgments
The authors thank Linda J. Rings of the Section of Medical
Research Statistics and Ellen J. Lichty of the Department of
Psychiatryand Psychologyfor their help in this study.
symptom of an underlying personality disorder, such
ideas regarding the cause of alcoholism have devel-
oped through the study of those already afflicted.
Several prospectivestudies (e.g., Beardslee and Vaillant, 1984; Vaillant, 1980) have suggestedthat no
personality style is predictive of alcoholism.
The hypothesis that the MAC Scale measures ad-
diction proneness(Kranitz, 1972; Lachar et al., 1976)
is not supported by our data. The scale identified
between 23 and 30% of our drug-dependentonly
groups. Addiction proneness may not have been
identified for the same reasons as noted above for
alcoholism.
References
AMERICAN PSYCHIATRIC ASSOCIATIONTASK FORCE ON NOMENCLA-
TURE AND STATISTICS.Diagnostic and Statistical Manual
of
Mental Disorders (DSM-III), Washington, D.C., 1980.
BALDESSARINI,
R. J., FINKLESTEIN,
S. AND ARANA, G. W. The
predictivepower of diagnostictestsand the effect of prevalence
of illness.Arch. gen. Psychiat. 40: 569-573, 1983.
BEARDSLEE,
W. R. AND VAILLANT,G. E. Prospectiveprediction
of alcoholismand psychopathology.
J. Stud. Alcohol 45: 500503, 1984.
COLLIGAN,
R. C., OSBORNE,
D., SWENSON,
W. M. ANDOFFORD,
K. P. The MMPI: A ContemporaryNormative Study, New
York: Praeger Pubs., 1983.
206
DAVIS, COLLIGAN,
DIETVORST,T. F., SWENSON,W. M. AND MORSE, R. M. Intellectual
assessmentin a midwestern alcoholism treatment population. J.
clin. Psychol. 34: 244-249, 1978.
KRANITZ, L. Alcoholics, heroin addicts and nonaddicts: Comparisons on the MacAndrew alcoholism scale of the MMPI. Q. J.
Stud. Alcohol 33: 807-809,
1972.
LACHAR, D., BERMAN, W., GRISELL,J. L. AND SCHOOFF,K. The
MacAndrew alcoholism scale as a general measure of substance
misuse. J. Stud. Alcohol 37: 1609-1615, 1976.
MACANDREW, C. The differentiation of male alcoholic outpatients
from nonalcoholic psychiatric outpatients by means of the
MMPI. Q. J. Stud. Alcohol 26: 238-246, 1965.
MILLER, W. R. Alcoholism scales and objective assessmentmethods: A review. Psychol. Bull. 83: 649-674, 1976.
MYERS, J. K., WEISSMAN, M. M, TISCHLER, G. L., HOLZER, C.
E., III, LEAF, P. J., ORVASCHEL,H., ANTHONY, J. C., BOYD,
J. H., BURKE, J. D., JR., KRAMER, M. AND STOLTZMAN,R.
Six-month prevalence of psychiatric disorders in three communities. Arch. gen. Psychiat. 41: 959-967, 1984.
NATIONAL COUNCIL ON ALCOHOLISM.Criteria for the diagnosis of
alcoholism. Amer. J. Psychiat. 129: 127-135, 1972.
MORSE AND OFFORD
NATIONALINSTITUTEON ALCOHOLABUSEAND ALCOHOLISM
(NIAAA).
Alcohol and Health, Fifth Special Report to the U.S. Congress,
DHHS Publication No. (ADM) 84-1291, Washington: Government Printing Office, 1983.
ROBINS, L. N., HELZER, J. E., WEISSMAN, M. M., ORVASCHEL,
H., GRUENBERG,E., BURKE, J. D., JR. AND REGIER, D. A.
Lifetime prevalence of specific psychiatric disorders in three
sites. Arch. gen. Psychiat. 41: 949-958, 1984.
SCHWARTZ,M. F. AND GRAHAM, J. R. Construct validity of the
MacAndrew alcoholism scale. J. cons. clin. Psychol. 47: 10901095, 1979.
SWENSON,W. M. ANI) MORSE,R. M. The use of a self-administered
alcoholism screening test (SAAST) in a medical center. Mayo
Clin. Proc. 50: 204-208, 1975.
SWENSON,W. M., PEARSON,J. S. AND OSBORNE,D. An MMPI
Source Book: Basic Item, Scale, and Pattern Data on 50,000
Medical Patients, Minneapolis: University of Minnesota Press,
1973.
VAILLANT, G. E. Natural history of male psychological health:
VIII. Antecedents of alcoholism and "orality." Amer. J. Psychiat. 137: 181-186,
1980.
8th Cape Cod Institute
29 June-4 September 1987
Cape Cod, Massachusetts
Sponsored by the Department of Psychiatry of the Albert Einstein College of Medicine, the Institute
consistsof 27 week-long summer coursesfor mental health and health scienceprofessionals.Tuition is
$355 for one program and $190 for each additional program attended. A student rate is available.
For a catalog and other information, contact:
Cape Cod Institute
Albert Einstein College of Medicine
1312 Belfer Building
Bronx, New York 10461
(212) 430-2307
Документ
Категория
Без категории
Просмотров
5
Размер файла
603 Кб
Теги
202, jsa, 1987
1/--страниц
Пожаловаться на содержимое документа