 |
 |
|
|
|
The rise of commercial gambling is an international phenomenon. In different countries, the number of exposed individuals has risen and so have prevalence rates of pathological gambling (PG) (1). Therefore, demand is growing for more efficient and more widely available treatment for PG. However, research on treatment for PG is only beginning and, when reviewing the topic, researchers face several problems. There are few randomized controlled trials (RCTs), and different criteria for assessing success rates hinder comparisons of treatment modalities (2). A review of psychosocial interventions for PG (3) identified only 4 RCTs: 2 Canadian studies (4,5), 1 from Spain (6), and 1 from Australia (7). The criteria for treatment success varied across the 4 studies. The outcome measures adopted by the Canadian studies were frequency of gambling, perception of control, perceived self-efficacy, desire to gamble, and number of DSM-III-R criteria (8) met by participants. The outcome measures for the Spanish and Australian studies were frequency and amount of money invested in gambling. However, in the Australian study, both abstinence and controlled gambling (defined as expenditures no greater than AUS$10.00 weekly) were considered favourable outcomes. Recently, authors have developed scales for assessing the treatment efficacy of pharmacologic trials for PG. These new scales have been used in conjunction with such general measures of clinical improvement as the Clinical Global Impression scale (CGI). DeCaria and colleagues developed the Yale-Brown Obsessive Compulsive Scale modified for pathological gambling (PG-YBOCS). It is based on perceived similarity between obsessive–compulsive disorder and pathological gambling. Reliability and validity of the PG-YBOCS have been presented but not published (9). Kim and colleagues developed the Gambling Symptom Assessment Scale (G-SAS), a 12-item instrument that evaluates improvement in treatment for pathological gambling (10). The Obsessive-Compulsive Drinking Scale modified for pathological gambling (OCDS-PG) has been used in one study (11). The PG-YBOCS, G-SAS, and OCDS-PG are craving-focused according to the main objective of the clinical trials for which they were developed. Data on validity and reliability have been published only for the G-SAS (10). Petry has recently validated a gambling section for the Addiction Severity Index (ASI) (12). The ASI is probably the most used instrument in the addiction field. It is a semistructured interview that assesses 7 variables considered crucial for treatment: medical condition, employment or support, alcohol use, drug use, illegal activity, family and social relationships, and psychiatric condition (13). However, the ASI requires considerable training, and its length might be an obstacle when working with large clinical services requiring fast and repeated measures over time. With this in mind, we designed the Gambling Follow-Up Scale (GFS), a free adaptation of the ASI, to follow up pathological gamblers under treatment. To allow change sensitivity, we gave preference to ASI items that are susceptible to change in relatively short periods of time. The first version of the GFS had 4 items: “frequency and time gambling,” “work status,” “family relationship,” and “leisure.” A fifth item, “enrolment in Gamblers Anonymous (GA),” was added. As Taber has pointed out, participation in GA may favour recovery, especially if associated with clinical assistance (14). The item choice, as in the ASI, reflects the concepts that ideal treatment provides suppression of the addictive behaviour coupled with better psychosocial functioning and that both objectives are interdependent. The GFS is meant for either self-report or semistructured interview, depending on resources available and the patient’s level of education. Its structure is purposefully simple so that it can be used along with other scales specific to the treatment modality under evaluation (for example, scales assessing cognitive distortions in cognitive-based treatments). The GFS goal is to provide researchers and clinicians with a simple set of reliable measures for follow-up and treatment outcome. This study provides preliminary data on the use of the GFS as a semistructured interview, its interjudge reliability, and its construct validity. As a secondary goal, we have also examined the impact of the other GFS items on the gambling item, as well as the impact of treatment on all GFS items. MethodsScale Development: Structure and Content We arranged the GFS items along a 5-point scale. For each point, we developed a short text describing the item status that corresponded to the given score (1 for the worst-case scenario and 5 for the best). We consulted 9 specialists to test the text ordering in each item. These individuals were psychiatrists and psychologists specializing in either addiction or psychometrics and selected for their clinical and research experience from among professionals in the 2 leading medical schools in Brazil. For each item, we omitted the original scores and shuffled the texts, asking the specialists to order them according to their judgment. We also asked whether the specialists considered the item relevant for assessing improvement among pathological gamblers under treatment. At the end of the questionnaire, an open-ended question asked for further suggestions. The specialists considered all 5 items relevant, although 2 suggested that gambling behaviour needs a more comprehensive assessment than just frequency and time. All of them reproduced the original ordering for the “frequency and time gambling” item; 8 out of 9 reproduced the original text ordering for “work status,” “leisure,” and “enrolment in GA.” The “family relationship” item raised some controversy, with 4 of 9 specialists disagreeing with the original text ordering. There were doubts about whether the option “living with relatives, but in isolation” was better or worse than “living with relatives with frequent verbal aggressions.” Commentaries further criticized the fact that the “family relationship” item simultaneously tapped into more than one concept, such as interpersonal relationships, aggression, and support. Overall, the original structure and content of the GFS was approved. There were few corrections to improve clarity. This last version then proceeded to reliability testing. Table 1 shows the English translation of the GFS items.
Sample Collection From January 1 to June 30, 2000, we collected the sample in 2 different treatment centres in the city of Sno Paulo (a highly industrialized and populated area having 12 million inhabitants): the Institute of Psychiatry of the University of Sno Paulo (USP) and the Gambling Outpatient Program of the Federal University of Sno Paulo (UNIFESP). Along with GA, these centres represent the only specialized and free-of-charge programs available for gambling in Brazil (15). Patients attending USP were receiving 1 of 3 possible treatment modalities: psychiatric treatment combined with individual psychotherapy, weekly GA meetings, or both. Patients attending UNIFESP received psychiatric treatment combined with group psychotherapy. Work with human subjects reported in this study complies with the guiding policies and principles for experimental procedures of the Helsinki Declaration (16). Reliability Testing All 3 authors were engaged in this phase of the scale development. The most experienced (HT) was chosen as reference for reliability testing. Time to perform the interview was recorded. A 3-phase process was designed. First, HT and VC interviewed 20 outpatients being treated for pathological gambling, with HT leading the first 10 interviews and VC leading the remaining 10. The same procedure was repeated for the next pair of interviewers (HT and DF). Then, HT retired, and VC and DF, alternating as leading interviewer, performed another set of 20 interviews . The intraclass correlation (ICC) test verified the concordance rate for each pair of interviewers, as well as the overall concordance rate for all 60 interviews. We used the Wilcoxon test (17) to compare mean values between raters. Construct Validity We compared the GFS items with each other and with length of treatment, using Pearson correlation tests. We included length of treatment to appraise the effect of the amount of treatment. In addition, we performed an exploratory factor analysis to investigate the scale’s structure. Gambling Behaviour and Treatment Impact To further investigate gambling behaviour, all variables that correlated with the item “frequency and time gambling” according to the Pearson tests (P = 0.10) entered a stepwise linear regression, with “frequency and time gambling” as the dependent variable. We classified type of treatment according to 3 possible scenarios: attending only GA, attending only clinical treatment, and attending both GA and clinical treatment. In setting this classification, we followed Taber’s suggestion that clinical assistance combined with GA may render better outcomes (14). Using analysis of covariance (ANCOVA), we compared the 3 types of treatment in relation to the GFS items, controlled for treatment length. We used the SPSS software package for the statistical analysis (18). ResultsSample Description We assessed 47 subjects: 16 attending only GA, 13 receiving only clinical treatment (all receiving psychiatric treatment, with 3 receiving individual therapy and 10 receiving group therapy), and 18 both attending GA and receiving psychiatric treatment (all in individual therapy). We assessed 13 patients twice, with a minimum interval of 6 months between assessments, performing in total 60 GFS evaluations. Patient age averaged 44.7 years, and patient education averaged 11.3 years. Of the patients, 37 (80%) were white, 5 (10%) were black, and 5 (10%) were Asian. Nineteen (40%) were women. We found no differences regarding the demographic profiles for patients attending different treatment centres. However, analysis of variance (ANOVA) revealed significant differences in treatment length. The mean (standard error) months under treatment were 13.2 (2.1) for the GA-only group, 3.7 (2.3) for the clinical-only group, and 9.6 (2.0) for subjects receiving both forms of treatment (F2,44 = 4.5, P = 0.016). Additional ANOVAs did not uncover sex differences in the GFS scores. GFS Reliability The GFS interviews took on average 6.0 minutes, SD 2.7 minutes. On all GFS items, interviewer concordance was high for the 3 pairs of interviewers, with concordance rates ranging from 82% to 95% (n = 60, ICC coefficients ranging from 0.85 to 0.99, P < 0.001). Wilcoxon tests found no differences between the mean scores of the 3 raters. Consistency and Construct Validity For this phase of the analysis, we used the GFS scores of the leading interviewer, and for those patients interviewed twice, we considered only the scores of the first interview (n = 47). Table 2 shows the outcome of the Pearson correlation analysis. In general, the items correlated well. The exception was the item “work status,” which correlated significantly only with “leisure.” The latter, in turn, presented the greatest correlation coefficients of all items. To investigate the GFS structure, we performed 2 factor analyses. In the first, we entered only the 5 GFS items. The outcome was a solution with only one factor (Eigenvalue = 2.4) accounting for 47.6% of the variance. The component matrix showed considerable factor loadings of the items “leisure,” “frequency and time gambling,” and “family relationship” (0.84; 0.71; and 0.71, respectively), and moderate contributions of the items “enrolment in GA” and “work status” (0.59 and 0.56, respectively). In the second factor analysis, we added treatment length to account for the effect of the amount of treatment on the GFS scores. The outcome was a 2-factor solution (Eigenvalues = 2.5 and 1.1; percentage of variance explained, 42.1% and 18.9%, respectively; cumulative variance explained, 61.0%). Factor 1 encompassed all GFS items, and factor 2 comprised only treatment length.
Prediction of Gambling Behaviour and Treatment Impact We set a stepwise linear regression model having the item “frequency and time gambling” as the dependent variable. Four independent variables entered the model in the following order: “leisure,” length of treatment, “family relationship,” and “enrolment in GA.” The final regression model significantly correlated “frequency and time gambling” (R2 = 0.356; F2,44 = 12.2; P < 0.001) with “leisure” (β = 0.495; P = 0.001) and length of treatment (β = 0.054; P = 0.016). Using an alternative approach, we entered the 4 variables and length of treatment simultaneously in the regression model. The outcome was similar, with more leisure and lengthier treatment significantly related to less gambling. Regarding the impact of different treatment modalities, ANCOVA did not uncover differences for “frequency and time gambling,” “work status,” or “family relationship.” However, different types of treatment did have an impact on the item “leisure” (F2,43 = 5.00, P = 0.011). Patients attending only GA (n = 16) had a mean (SD) score of 4.1 (1.3); patients who received psychiatric treatment and attended GA (n = 18) had a mean (SD) score of 3.9 (1.2); and patients who received only psychiatric treatment (n = 13) had a mean (SD) score of 2.5 (1.6). Bonferroni post hoc analysis revealed that the difference was located in the group of patients receiving only psychiatric treatment: their scores on the item “leisure” were significantly lower than the scores of patients who attended only GA (mean difference –1.6, standard error 0.50; P = 0.008) and the scores of patients attending GA plus psychiatric treatment (mean difference –1.4, standard error 0.49; P = 0.017). DiscussionThe high rates of concordance between independent raters for all items prove the reliability of the GFS. Its single-factor structure is compatible with the rationale that, for those who suffer from pathological gambling, gambling behaviour and psychosocial functioning are intertwined aspects that influence each other. Moreover, this unitary structure remained stable even when we accounted for different treatment lengths. Despite previous criticism from the consulted specialists, the item “family relationship” displayed a considerable loading on the factor analysis, together with relevant correlations to the other GFS items. Conversely, “work status” correlated only with “leisure.” Perhaps the item needs restructuring, or perhaps the observable impact of “frequency and time gambling” on “work status” was weakened by contingencies not related to gambling, such as the community’s economic status. In the stepwise regression procedure the items “enrolment in GA” and “leisure” seemed to covary, causing the exclusion of the former. Accordingly, although no treatment modality proved superior, the data indicate that “enrolment in GA” is particularly relevant for improvement in “leisure.” Considering the playful and entertaining properties of gambling, it makes sense that subjects able to abstain are those who have found alternatives to fill the void. The significant correlation of treatment length to gambling behaviour suggests that treatment has been effective in diminishing gambling involvement. However, the cross-sectional, uncontrolled nature of this study precludes our appraising efficacy of the treatments assessed, and it is beyond the scope of the current investigation. Overall, the GFS proved to be a valid, reliable, and quick instrument, potentially compatible with complementary and more specific treatment measures. The fact that GA was associated with better leisure but not with the formal treatment modalities underscores the GFS’s ability to localize particular assets of different treatments. Future studies must address sensitivity to change, cross-validity, and the inclusion of new variables while balancing these factors with the priority of keeping the scale short and versatile. It is important to underscore that the current results apply to the Portuguese version of the GFS. Future studies must determine whether the English translation presented in Table 1 retains the psychometric properties of its original. Ideally, new versions of the scale should encompass a small set of variables that reflect an international consensus regarding the variables that are important for assessing the efficacy of gambling treatment. Funding and SupportMs Castro, Mr Fuentes, and Dr Tavares were supported by the State of São Paulo Research Foundation, Brazil (FAPESP grants 99/01622-6, 99/02803-2, 98/015406-0). AcknowledgementsWe thank the staff of the gambling outpatient clinics at the state and federal universities and the following colleagues for their suggestions and feedback on the initial version of the GFS: Antonio Carlos Pacheco e Silva Neto, Arthur Guerra de Andrade, Clarice Gorenstein, David C Hodgins, Laura Andrade, Marcelo Fernandez, Marcos Costa Leite, Maria Lizak, Maria Paula MT de Oli veira, Monica Levit Zilberman, Nicole Peden, Richard Rosenthal, Ron Richard, and Sandra Scivoletto. References1. Shaffer HJ, Korn DA. Gambling and related mental disorders: a public health analysis. Annu Rev Public Health 2002;23:171–212. 2. Tavares H, Zilberman ML, el-Guebaly N. Are there cognitive and behavioural approaches specific to the treatment of pathological gambling? Can J Psychiatry 2003;48:22–7. 3. Oakley-Browne MA, Adams P, Mobberley PM. Interventions for pathological gambling. Cochrane Database Syst Rev 2000;2:CD001521. 4. Ladouceur R, Sylvain C, Boutin C, Lachance S, Doucet C, Leblond J, and others. Cognitive treatment of pathological gambling. J Nerv Ment Dis 2001;189:774–80. 5. Sylvain C, Ladouceur R, Boisvert JM. Cognitive and behavioral treatment of pathological gambling: a controlled study. J Consult Clin Psychol 1997;65:727–32. 6. Echeburúa E, Báez C, Fernández-Montalvo J. Comparative effectiveness of three therapeutic modalities in the psychological treatment of pathological gambling: long-term outcome. Behavioral and Cognitive Psychotherapy 1996;24:51–72. 7. McConaghy N, Blaszczynski A, Frankova A. Comparison of imaginal desensitisation with other behavioural treatments of pathological gambling. A two-to-nine-year follow-up. Br J Psychiatry 1991;159:390–3. 8. American Psychiatric Association. Diagnostic and statistical manual of mental disorders. 3rd ed. Revised. Washington (DC): APA; 1987. 9. DeCaria CM, Hollander E, Begaz T, Schmeidler J, Wong C, Cartwright C, and others. Reliability and validity of a pathological gambling modification of the Yale-Brown Obsessive-Compulsive Scale (PG-YBOCS): preliminary findings. Paper presented at the Third International Obsessive-Compulsive Disorders Conference; 1998; Madeira (Portugal). 10. Kim SW, Grant JE, Adson DE, Shin YC. Double-blind naltrexone and placebo comparison study in the treatment of pathological gambling. Biol Psychiatry 2001;49:914–21. 11. Zimmerman M, Breen RB, Posternak MA. An open-label study of citalopram in the treatment of pathological gambling. J Clin Psychiatry 2002;63:44–8. 12. Petry N. Validity of a gambling scale for the addiction severity index. J Nerv Ment Dis 2003;191:399–407. 13. McLellan AT, Kushner H, Metzger D, Peters R, Smith I, Grissom G, and others. The fifth edition of the Addiction Severity Index. J Subst Abuse Treat 1992;9:199–213. 14. Taber JI, McCormick RA, Russo AM. Follow-up of pathological gamblers after treatment. Am J Psychiatry 1987;144:757–61. 15. Tavares H, Martins SS, Zilberman ML, el-Guebaly N. Gamblers and treatment-seeking: why haven’t they come earlier? Addictive Disorders and Their Treatment 2002;1:65–70. 16. World Medical Association Declaration of Helsinki: ethical principles for medical research involving human subjects. J Postgrad Med 2002;48:206–8. 17. Bartko JJ, Carpenter WT Jr. On methods and theory of reliability. J Nerv Ment Dis 1976;163:307–17. 18. SPSS Inc. SPSS Statistical Package for Social Sciences. Version 10.0 for Windows. Chicago (IL): SPSS Inc; 1997. Author(s)Manuscript received November 2003, revised, and accepted March 2004. 1. Psychologist, The Gambling Outpatient Unit, Institute and Department of Psychiatry, University of Sno Paulo, Sno Paulo, Brazil. 2. Coordinator, The Gambling Outpatient Unit, Institute and Department of Psychiatry, University of Sno Paulo, Sno Paulo, Brazil. Address for correspondence: Dr H Tavares, The Gambling Outpatient Unit, Institute and Department of Psychiatry, University of Sno Paulo, Rua Purpurina 155, conj 126/128, Sno Paulo SP, Brasil, 05435-030 e-mail: hermanot@uol.com.br
1 | 2
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||