Psychological Repom, 1997.80.

83-88 O l'sychological Reports 1997

THE LIEBET QUESTIONNAIRE FOR SCREENING

PATHOLOGICAL GAMBLERS '

EDWARD E. JOHNSON AND ROBERT HAMER

Robert Wood lohnson Medical School
University of Medicine and Dentistry of New lersey

RENA M. NORA
Deparfment of Veterans Affairs Medical Center
Lar Vegas, Nevada

BENITO TAN, NORMAN EISENSTEIN, AND CHARLES ENGELHART

Department of Veterans Affairs Medical Center
Lyons, New Jersey

Summnary.-A 2-item questionnaire was derived €rom 10 DSM-IV criteria €or

pathological gambling. Subjects were 362 men, 191 classified as pathological gamblers
and 171 as nonproblem-gambling controls. The two items were significant in sensitiv-
ity and negative predictive value and significant in specificity and positive predictive
value.

Gambling currently is widespread and growing, with American, Canadi-

an, and British studies showing over 8 % of adults approving of gambling
practices and two out of three adults participating regularly in this form of
recreation (Ladouceur, Dube, & Bujold, 1994). In 1974, 61% of Americans
gambled, spending $17.4 billion (Kallick, Suits, Dielman, & Hybels, 1979);
by 1992 the amount had reached $329.9 bilhon dollars, a 19-fold increase
(Christiansen, 1993). Murray (1993) has stated in a review that estimates of
the number of pathological gamblers in the United States vary between 1.1
and six m&on. Wolkowitz, Roy, and Doran (1985) have noted that even the
conservative estimates of addictive gambling exceed the prevalence of schizo-
phrenia. Lesieur (1994) has pointed out that most epidemiological surveys
probably have seriously underestimated the extent of the problem and of pa-
thological gambling. The increased access to various sophisticated forms of
g a m b h g and access to incredible sums of money and credit have contribut-
ed to pathological g a m b h g as a serious public health problem which cannot
be ignored.
Three well-known tools for screening gambling behavior are the 16-item

'Request reprints from E. E. Johnson, Ph.D. at the Ps chiatry Department, University of Medi-
cine and Dentistry of New Jersey-Robert Wood J o L s o n Medical School, Piscataway, New
Jersey 08854.
84 E. E. JOHNSON, ET AL.

South Oaks Gambling Screen (Lesieur & Blum, 1987), the Gamblers Anon-
ymous Twenty Questions (1984). and the Diagnostic and Statistical Manual
bf Mental ~ i s o r d e r s ,Fourth Editlo" criteria-of the American Psychiatric
Association (1994). Volberg (1994) and Lester (1994) have reported data in-
dicating that, when the availabhty of gambling is increased, the prevalence
of gambling-related ~roblemsin the general population is increased. The rise
in availability of opportunities to gamble can be expected to be reflected in
growing numbers of pathological gamblers; likewise, the demand for more
rapid identification also can be expected to increase. Thus a screening tool
more brief than any of the foregoing inventories would seem useful. It is irn-
portant, however, that the search for a more rapid tool for screening be
achieved without any significant loss of vahdty or rehability. The aim of this
study was to assess the minimum number of items from a 12-statement ques-
tionnaire based upon the 10 DSM-IV criteria for pathological gambling that
would differentiate dependably between gamblers and nonprob-
lem-gambling controls.

Subjects
Participating were 362 men, 191 classified as pathological gamblers and
171 as nonproblem-gambhg controls. The pathological gamblers were re-
cruited from Gamblers Anonymous groups throughout the country, while
the controls were recruited 'from Department of Veterans Affairs employees
who were neither pathological nor problem gamblers. Criteria for inclusion
in the pathological gambler group were membership in Gamblers Anony-
mous and no more than seven negative responses to the Gamblers Anony-
mous Twenty Questions.
Each prospective participant had to read a description of the study
which included a Department of Veterans Affairs consent form, the aim of
the study, and the method of gathering data. All subjects participated in the
study voluntarily.
Questionnaires
All participants completed the following questionnaires, a Demographic
Information Form, the Gamblers Anonymous Twenty Questions, and a 12-
item Gambling Questionnaire adapted from the &agnostic criteria for patho-
logical g a m b h g listed in DSM-IV. A copy of the Gambling Questionnaire
may be obtained on request from the corresponding author.

Demographic Characteristics
Pathological gamblers and control subjects were closely matched in
 SCREENING PATHOLOGICW GAMBLERS 85

mean age, with pathological gamblers averaging 44.3 (range 21 to 74 years)

and the controls 42.8 (range 17 to 82 years). A t test yielded no significant
dkference in mean age between the groups ( r = -.95, p < 3 5 ) so it does not
appear that age influences g a m b h g behavior, at least in this population.
Average years of m1Litary senrice were 4.9 for pathological gamblers and 7.2
for nonpathological-gambling controls, with the difference in mean years be-
ing nonsignificant. The percentage married was 58.4 for gamblers (111) and
46.4 for controls (79), with gamblers significantly more k e l y to be married
than controls (Fisher Exact p < .O3). Classified as gamblers were 70.3% (26)
of Jewish subjects, 60.4% (81) of Catholics, 43.3% (39) of Protestants, and
44% (42) of Others. The percentages of gamblers in the sample by race
were Caucasian 59.8% (177). African American 24% (6), fispanic 10% (11,
and other 20.7% (6), respectively.
Among gamblers 10.5% (20) had less than high school education,
53.7% (102) finished high school, 21.6% (41) graduated from college, and
14.2% (27) had advanced degrees. Comparable percentages for controls
were 6.4% ( l l ) , 32.3% (55), 40.9% (70), and 20.5% (351, respectively. In
general, gamblers had higher incomes than controls with 37.1% (71), earn-
ing $50,000.00 or more per year, while only 19.7% (34) of controls exceed-
ed $50,000.00 per year.
DETECTING PATHOLOGICAL GAMBLERS
Less than 5 % (7) of the control subjects answered yes to any of the 12
G a m b h g Questionnaire items, while more than 75% (146) of the problem
gamblers answered yes to most of the questions. Although the groups dif-
fered in their pattern of responses to almost all questions, we were inter-
ested in discovering a small subset of the questions and a cut-off rule that
would dfferentiate the problem gamblers from the controls. Since we had
sufficient subjects, one analysis was done by randomly splitting the sample
into two subsamples and performing a step-wise logistic regression (Hosmer
& Lemeshow, 1989) on each subsample. Although the groups dlffered some-
what on some of the demographic measures, we chose not to enter these
into the models because we wanted the brief questionnaire we developed to
be independent of demographic information. In both stepwise logistic re-
gressions
- G a m b h g Questionnaire Item 3 (Have you ever felt the need to
bet more and more money?) and Item 6 (Have you ever had to lie to people
important to you about how much you gambled?) emerged as the two first
predictors, and, given that they were in the model, selection of the next
highest pre&ctor produced failure to converge. Thus, in both subsamples,
the same two items emerged as the best in a joint logistic equa-
tion. When the model was refitted with the full sample and the same analy-
sis was performed, the same results occurred, i.e., again Questions 3 and 6
86 E. E. JOHNSON, ET AL.

emerged as the two first predictors, and the model did not converge when
selecting the next predictor. The lack of convergence was due to a lack of
new information in any of the other questions, given that Questions 3 and 6
were in the model.
Using the full sample, Questions 3 and 6 were combined to form the
sum of the number of yes responses. Given that there are two questions,
there can be 0, 1, or 2 responses. The results of placing a cut-off at any non-
zero response are reported in Table 1.
TABLE 1
FREQUENCIESUSEDIN COMPUTING
OPERATING
CHARACTERLSTICS
OF T H E L~E/BET
QU~ONNAIRE

Test Disease State

Positive Negative
Positive TP:190 FP:16
Negative FN: 1 TN: 155

Ln Table 1 "Disease State" is group, either gamblers (positive) or con-

trols (negative); "Test" is positive if either Question 3 or Question 6 was
answered affirmatively and negative otherwise; TP indicates true positive, FP
false positive, FN false negative, and TN true negative. A true positive is a
~roblerngambler whom the test identified as a ~ r o b l e mgambler; a true neg-
ative is a nonproblem gambler whom the test identified as a nonproblem
gambler; a false positive is a nonproblem gambler whom the test classified as
a problem gambler; and a false negative is a problem gambler whom the test
classified as a nonproblem gambler.
Although it was impossible to administer the questions on multiple oc-
casions and so to assess test-retest reliability, we did examine the agreement
between the two items used in the screen, Items 3 and 6. Kappa (a chance-
corrected measure of agreement) was .811, with an asymptotic standard er-
ror of .028. A 95% confidence interval on Kappa is the interval from .756 to
,866. This indicates substantial agreement between the two items.
Sensitivity and specificity are common ways of reporting the efficacy of
a diagnostic test. They have the advantage of being uninfluenced by the base
rate. Sensitivity can be thought of as the probabllity that a test result wdl be
positive, given that the person has the disease. A high value for sensitivity is
indcative of a test that might be useful for screening. It wdl catch many of
the true positives, while producing few false negatives. Specificity can be
thought of as the probabllity that a test result wdl be negative, given that the
person does not have the disease. Tests with high specificity generally wdl
produce positive results almost only when the person has the dsease and
not otherwise.
From the table above,
 SCREENING PATHOLOGICAL GAMBLERS

Sensitivity = TP/TP + FN = 190/191= .99

Specificity= TN/TN + FP= 155/171= .91.
Thus this test. has very high sensitivity and fairly high specificity.
The Positive Predictive Value (PPV) of a test is the predictive value of
a positive result defined as the proportion of subjects with a positive test
who have the lsease. The Negative Predictive Value (NPV) of a test is the
predictive value of a negative result defined as the proportion of subjects
with a negative test who are disease-free. Again from the table above,
Positive Prelctive Value (PPV) = TP/(TP + FP) = 190/206 = .92, and
Negative Predictive Value (NPV) = TN/(TN + FN) = 155/156 = .99.
PPV and NPV are useful in deciding what the result on the test means. In
this case, the Positive Prechctive Value of .92 means that approximately
92.23% of the people who achieve a positive result on this test are, in fact,
problem gamblers and that 7 or 8% of the people whom the test classifies
as problem gamblers w d , in fact, not be problem gamblers. The Negative
Prechctive Value of .99 means that approximately 99.36% of the people who
achieve a negative result on this test are, in fact, not problem gamblers. The
implication is that, if the test says that someone is not a problem gambler,
one is pretty certain that he is not, which is just how a screening test should
perform.
A lunitation of research in this area is that there is no totally accurate
way of deciding which participant may be denying problem gambling. It is
conceivable, also, that the point of demarcation chosen for the operational
definition of pathological gambhg, i.e., no more than seven negative re-
sponses to the Gamblers Anonymous Twenty Questions, may have been set
too low, i.e., may be over-inclusive, or may have been set too high, i.e., may
be under-inclusive. Again, generali~abht~ may be limited to participants in
Gamblers Anonymous, who are &g to admit the need for help. Another
lunitation of this research is that all sibjects were men. A comparative study
of women is presently being conducted.
Conclusion
The significant Sensitivity and Negative Predctive Value and the signifi-
cant Specificity and Positive Predictive Value obtained in this study inlcate
that the Lie/Bet Questionnaire is promising for screening pathological garn-
blers.
REFERENCES
AMERICAN PSYCHLATRIC &SOCIATION. (1994) Diagnostic and sfatistical manual of mental disor-
ders. (4th ed., Rev.) Washington, DC: Author.
CHRISTIANSEN, E. M. (1993) Income 1992 gross annual wager of h e US.: Part 1. Gaming and
Wagering Business, 14, 7 , 12-35.
88 E. E. JOHNSON, ET AL.

HOSMER, D. W., &LEMESHOW, S. (1989) Applied logistic regrexsion. New York: Wiley.
KNLICK,M., SUITS,D., DIELMAN, T., &HYBELF, J. (1979) A sum of American gambling afti-
trrde and behavior. Ann Arbor, MI: Univer. of Iihchigan ?urvey Center, Institute for
Social Research.
LADOUCEUR, R., DUBE,D., & BUJOLD, A. (1994) Prevalence of pathological gamblin and re-
lated roblems among college srudents in the Quebec metropoliran area. ~ a n a j i a n]our-
nal o&chiatry, 144, 1184-1188.
LESIEUR, H. R. Epidemiological surveys of athological gambling critique and suggestions for
modification. Journal of Ganzbling .Sttrles, 10, 385-398.
LESIEUR, H. R.. & BLUM,S. (1987) The South Oaks Gambling Screen (SOGS) (1987) A new
instrument for the identification of pathological gamblers. American ]orrrnal of Psychiatry,
144, 1184-1188.
LESTER,D. (1994) Access to gambling opportunities and compulsive gambling. The Interna-
tional Jozrrrzal of the Addictions, 29, 161 1-1616.
MURRAY. J. (1993) Review of research on pathological gambling. Psychological Reports. 72, 791-
810.
VOLBERG, R. (1994) The prevalence and demo raphics of pathological gamblers: implications
for public health. American ]ourno1 of pubti Heal~h,84, 237-241.
WOLKOWTZ, 0..ROY.A., & DORAN, A. (1985) Pathologic gambling and other risk-taking pur-
suits. Psychiatric Clinics of North America, 8, 3 l 1-322.

Accepted December 2, 1996.