Journal ID (publisher-id): jgi
Publisher: Centre for Addiction and Mental Health
The Centre for Addiction and Mental Health
Received Day: 16 Month: October Year: 2012
Accepted Day: 16 Month: March Year: 2014
Publication date: 2015
First Page: 140 Last Page: 173
Publisher Id: jgi.2015.30.7
|Construct Development for the FocaL Adult Gambling Screen (FLAGS): A Risk Measurement for Gambling Harm and Problem Gambling Associated with Electronic Gambling Machines|
|1Dalhousie University, Halifax, Nova Scotia, Canada
|2Focal Research Consultants Limited, Halifax, Nova Scotia, Canada
|This article was peer-reviewed. All URLs were available at the time of submission.
For correspondence: Tony Schellinck, PhD, Focal Research Consultants Limited, 7071 Bayers Rd., Suite 326, Halifax, , NS, , Canada B3L 2C2., e-mail: email@example.com, , Website address: http://www.focalresearch.com
Competing interests: None declared.Ethics approval: The Ontario Institutional Review Board (ON IRB). Final protocol approval was obtained for “Preliminary Development of a Self Administered Gambling Risk Assessment Instrument for Slots” on June 23, 2008.Funding: Funding was received from the Victoria Gambling Research Panel under the leadership of Dr. Linda Hancock, to develop the Victoria Self-Administered Problem Gambling Scale (SAPGS) in 2003. Further funding was received from the Victoria Department of Justice, Melbourne Victoria in 2006 to test the validity and value of the Victoria SAPGS. In 2008, the Ontario Problem Gambling Research Centre provided funding for further development of what became known as the FLAGS-EGM and its then Director, Rob Simpson, provided guidance in expanding measures used in the instrument.Contributors: TSchellinck planned the document. TSchellinck and HS drafted and wrote the manuscript with editorial contributions from TSchrans and MB. HS and TSchrans conducted the gambling-related literature review. TSchrans and TSschellinck conceptualized the research design and conducted the focus group and survey studies. TSchellinck and MB assessed the current analytical literature and designed the analysis approach. TSchellinck conducted the analysis and finalized the design of the constructsDr. Tony Schellinck is an Adjunct Professor in the Faculty of Graduate Studies and the Rowe School of Business at Dalhousie University, Canada, as well as CEO of Focal Research Consultants Limited. From 1996 to 2013 he was the F. C. Manning Chair in Economics and Business at Dalhousie University. Since 1989 he has conducted research into gambling behaviour for industry, government, public health and regulatory agencies. This work included a ten-year large-scale monthly tracking study of gambling behaviour, over 300 focus group sessions with gamblers, the 1998 Nova Scotia Video Lottery Study, two large scale studies into the value of responsible gambling features on VLT machines, and the Nova Scotia Adolescent Gambling Exploratory Research: Identification of Risk and Gambling Harms Among Youth. Dr. Schellinck worked on creating the first algorithms deployed in casinos that identified using player loyalty data high-risk gamblers.
Ms. Tracy Schrans is Principal and President of Focal Research Consultants an independent research firm in Halifax, NS. Over the last twenty years Tracy has conducted numerous government, public health, and industry-sponsored research projects on a wide range of issues, with a particular emphasis on gambling- and alcohol-related issues. She consults internationally in responsible gambling and corporate social responsibility, social policy, player tracking and loyalty data analysis. Schrans is one of the developers of new instruments for measuring pre-harm risk for gambling among adults (FLAGS-EGM and FLAGS General) and adolescents (FYGRS) for prevention applications. She continues to work at the forefront of gambling behavior analytics, assisting gambling stakeholders in using system data, measurement, and technology to help identify, manage and prevent gambling risk and harm among their customers.
Dr. Heather Schellinck, PhD, is an Adjunct Professor in the Faculty of Graduate Studies and Department of Psychology and Neuroscience at Dalhousie University. Her research is primarily focused on learning and memory in animal models of neurodegenerative disease.
Dr. Michael Bliemel is an associate professor of Management Information Systems at Dalhousie University in Halifax, NS. He completed his PhD at McMaster University in Management Science/Systems, specializing in the quantitative modeling of consumer behaviour with health information systems. His current research interests include the strategic management of information systems and innovation in organizations, and business intelligence applications.
This is the first of two papers describing the development of the FocaL Adult Gambling Screen for Electronic Gambling Machine players (FLAGS-EGM). FLAGS-EGM is a measurement approach for identifying gambling risk, a tool that incorporates separate reflective and formative constructs into a single instrument. A set of statements was developed that captured ten constructs associated with gambling risk or which were considered components of problem gambling. Following completion of focus groups with regular slot players, a survey with the reduced set of statements was then administered to a sample of 374 casino slot players in Ontario, Canada. Nine of the proposed constructs passed tests for reliability and validity (Risky Cognitions Beliefs, Risky Cognitions Motives, Preoccupation Desire, Risky Practices Earlier, Risky Practices Later, Impaired Control Continue a Session, Impaired Control Begin a Session, Negative Consequences, and Persistence). A tenth construct (Preoccupation Obsession) requires further development through the addition of improved statements.
Voici le premier de deux articles décrivant la mise au point d’un instrument appelé le FocaL Adult Gambling Screen for Electronic Gambling Machine players (FLAGS-EGM). Il s’agit d’une méthode d’évaluation du risque de dépendance au jeu qui réunit deux volets distincts en un seul instrument, l’un réflexif et l’autre formatif. Nous avons formulé un ensemble d’énoncés traduisant dix constructs associés au risque ou considérés comme des éléments constitutifs des problèmes de jeu. Après avoir mené des groupes de discussion avec des joueurs qui s’adonnent régulièrement aux machines à sous, un questionnaire formulé à partir d’un ensemble limité d’énoncés a été administré à 374 joueurs en Ontario (Canada). Neuf constructs sur dix ont réussi les tests de fiabilité et de validité (croyances cognitives risquées, motivations cognitives risquées, préoccupation relative au désir, pratiques risquées antécédentes, manque de contrôle–continue la séance, manque de contrôle–entame une séance, pratiques risquées ultérieures, conséquences négatives et persistence). Un dixième construct (préoccupation relative à l’obsession) nécessitera une mise au point grâce à l’ajout d’énoncés améliorés.
With the current emphasis on preventing gamblers from self-harm prior to the development of gambling problems, an instrument that clearly identifies those at risk is urgently required. Originally a number of screens, such as the South Oaks Gambling Screen (SOGS) (Lesieur & Blume, 1987 1993) and the DSM-IV-TR (4th ed., text rev.; DSM–IV–TR; American Psychiatric Association, 2000), were designed to identify problem gamblers among treatment populations. Debate surrounding the identification of pathological versus problem gamblers and the utility of the existing gambling screens for general population use (Dickerson, 1993; Lesieur & Blume, 1993; Volberg, Dickerson, Ladouceur, & Abbott, 1996; Walker & Dickerson, 1996) led to the development of other measures, including the National Opinion Research Center DSM–IV-based Screen for Gambling Problems (NODS) (Gerstein et al., 1999), the Canadian Problem Gambling Index (Ferris & Wynne, 2001), the Victoria Gambling Screen ( Tolchard & Battersby 2010), the Gamblers Beliefs Questionnaire (Steenbergh, Meyers, May, & Whelan, 2002) and the Gambling Related Cognitions Scale, developed by Raylu and Oei (2004). Several of these newer screens also included risk estimates as a component of identifying problem gambling, yet none of the screens incorporated the use of unique constructs specifically to identify gambling risk as a component separate from problem gambling.
Presently, the Problem Gambling Severity Index (PGSI) component of the Canadian Problem Gambling Index appears to be used most often to assess risk as a component of identifying problem gambling severity. With this instrument, a score of 0 is labelled “no risk,” scores of 1–2 are labelled “low risk,” 3–7 are labelled “moderate risk,” and individuals scoring 8 or higher are classified as problem gamblers (Ferris & Wynne, 2001). As such, the PGSI views risk as part of a single concept, and includes classification criteria for two risk categories. It assumes that risk is captured as a lower score whereas higher scores represent problem gambling. Although no evidence exists to suggest that (1) lower scores reflect lower risk as opposed to lower certainty that someone is a problem gambler, or (2) the severity of the problem gambling is indeed lower, it nevertheless seems likely that some at-risk gamblers are in fact identified through this instrument. Nonetheless, no published research has demonstrated that the PGSI or any other instrument that identifies or categorizes gamblers actually predicts risk due to problem gambling.
Currie, Hodgins, and Casey (2013) compared the characteristics of gamblers in the four PGSI categories, on variables previous research had found were associated with pathological gambling, to determine if there were significant differences among the gamblers in each category. The reason for doing so was as follows: if the gamblers were significantly different on these dimensions, then the gamblers could in turn be considered as belonging in valid and distinct groups. Currie et al. found not only that the no-risk group was distinctively different from the low- and moderate-risk groups, as well as the problem gambling groups, but also that few differences existed between the low- and moderate-risk groups. This finding suggests the PGSI low- and moderate- risk groups do not in fact comprise distinct categories of gamblers. Currie et al. did improve the distinctiveness of the low and moderate risk groups by redefining the PGSI score thresholds (from 1–3 for low risk to 1–4) but this modification did not deal with the fact that the PGSI design, by relying on a continuum, has consequent limitations. This problem suggests in turn the need for an instrument that is better able to form distinct groups of gamblers with differing risk profiles.
Thomas, Jackson, and Blaszczynski (2003) have emphasized the need for a tool to independently determine risk as distinctive from problem or pathological gambling. Shaffer, LaBrie, LaPlante, Nelson, and Stanton (2004) also pointed out the pressing need to investigate risk and protective factors that influence the onset of gambling disorders. Given the limitations of existing problem gambling screens in identifying risk, the FocaL Adult Gambling Screen (FLAGS-EGM) was developed specifically to address this measurement gap. Our instrument is designed to work similarly to screens the medical community has established to identify factors for specific high-risk conditions (e.g., Naghavi, Falk, Hecht, & Shah, 2006 ).
Using item response theory and statistical modeling, with the detailed play behaviour and attitudinal data gathered for regular machine gamblers during the 1998 Nova Scotia Video Lottery Players Study, Schellinck and Schrans (1998) developed the first hierarchical model of the antecedents of problem gambling for EGMs. Three principal considerations underlay underlied the creation of FLAGS-EGM:
As the first step in achieving these objectives, an operational definition for gambling harm and problem gambling, as well as of risk, was in all three cases required. In 1994, the American Psychiatric Association (APA) defined pathological gambling as the “persistent and recurrent maladaptive gambling behaviour that disrupts personal, family and vocational pursuits” (4th ed., text rev.; DSM–IV–TR; APA, 2000). Individuals so defined are preoccupied with gambling, may be unable to control their gambling, and both chase losses and suffer negative consequences as a result of these problems. Many subsequent definitions of problem and pathological gambling have been advanced. Most of these definitions have concerned themselves in particular with continued excessive involvement in gambling despite associated negative consequences for the individual. More recently they have also dealt with negative outcomes for the gambler’s family, community or society at large (Neal, Delfabbro, & O’Neil, 2005). These authors also noted the need for an instrument that would differentiate definitively among individuals at different levels of risk.
We have operationally defined problem gambling as the co-occurrence of two conditions composed of “negative consequences” (outcomes) and “persistence” (behaviour). Problem gamblers are characterized as those persons who have experienced negative consequences directly related to gambling in the past 12 months and who persist in gambling despite the occurrence of these negative consequences. Other characteristics such as those noted in the DSM-IV (4th ed., text rev.; DSM–IV–TR; APA, 2000), such as loss of control, chasing losses, and preoccupation, are conceptualized as risk factors leading to problem gambling, and are used in the FLAGS-EGM model as precursors to gambling harm and problem gambling. A key objective of the FLAGS-EGM measure is the development of constructs to capture fully the dimensions of risk, harm and problem gambling.
In creating this instrument, we included both reflective and formative constructs. Previously, researchers have developed and assessed most gambling screens based only upon reflective constructs. Reflective constructs presuppose that an underlying latent construct causes the observed variation in the measures (Nunnally, 1978). As items within a reflective construct are all indicative of the underlying latent variable, high correlation among the items comprisingthe measure should result: in theory, a gambler should endorse either all or none of the items being flagged through the construct. This method is a highly desirable for conceptualizing and measuring a single homogeneous factor, or specific concept, such as preoccupation or persistence. Either an individual meets the conditions for identification on this dimension or that person does not. For example, a construct such as impaired control is best designed as a reflective construct. It is designed this way to capture the specific nature of a gambler’s tendency to find it difficult to stop gambling once engaged in play.
In practice, those researchers developing or interpreting problem gambling screens may assume that the number of items endorsed for a reflective construct represents a continuum. It is, however, incorrect to presuppose that the higher the number of items selected, the greater the impact or severity. Indeed, research in the area of construct development has challenged assumptions that constructs should always be reflective (Diamantopolos & Siguaw, 2006; MacKenzie, Podsakoff, & Jarvis, 2005), and the relative merits of using reflective and formative measures for theory development are still being debated. General consensus argues that formative measures are suitable for prediction of given outcomes when used with structural equation modeling (SEM) (Diamantopoulos, Riefler, & Roth, 2008; Diamontaopoulos & Sigaw, 2006; Freeze & Raschke, 2007; Howell, Breivik, & Wilcox, 2007; Wilcox, Howell, & Breivik, 2008).
In contrast to a reflective measure, a formative construct is said to predict the latent variable (Bollen & Lennox, 1991; Gefen, Straub, & Boudreau, 2000). The items comprising formative constructs represent different, often uncorrelated dimensions of the latent variable. Endorsement is additive such that the more items endorsed the greater the severity of impact. This characteristic is a desirable one for an instrument intended to identify levels of risk or harm: the more harm components an individual endorses, the greater the severity of gambling problems, or, in the case of risk components, the higher one’s risk for problem gambling. To represent adequately the scope of a variable, such as risky beliefs, would require a formative construct that included various diverse concepts. Examples of those concepts include beliefs that game outcomes could be influenced, that chances of winning improved with continued play, or that outcomes could be predicted. These beliefs may not all be held by the same persons but all are associated with risk.
A problem may arise when a screen is used as a multi-purpose measure to assess more than one dimension of gambling, e.g., problem gambling and an individual’s risk for developing problem gambling. Although characteristics of risk for becoming a problem gambler and the characteristics of being a problem gambler are not necessarily the same, the items retained in most instruments are all highly correlated with each other. As a result, those gamblers who are at various stages of risk may be undetected or misclassified. To resolve this problem, we included separate constructs to measure individual risk elements.
The analyses undertaken to develop and test each of the constructs for the FLAGS-EGM instrument was extensive and exceeded the scope of this paper. Consequently, we have divided the work into two parts: this paper describes the first phase of the instrument design process, including (1) selection of the constructs, (2) a description of the statement process, followed by the assignment of the statement to the constructs, and (3) testing of the constructs’ validity and reliability. A companion paper describes the development of the FLAGS-EGM instrument, using Partial Least Squares (PLS) modeling to establish relationships among the constructs in a path leading to problem gambling. The results of the PLS analysis in the second paper have implications for the design and testing of the constructs we refer to in the current article and the reader should consult that publication for further information.
Many of the general characteristics associated with problem gambling have already been thoroughly described (Johansson, Grant, Kim, Odlaug, & Götestam, 2009; Turner, Jain, Spence, & Zangeneh, 2008). To avoid the possibility of creating false positives and to limit, in the development of our instrument, the size of FLAGS-EGM, we included only gambling-specific constructs, each comprised of statements that referred to the respondent’s gambling cognitions, behaviours or experiences. Based upon our original research in this area (Schellinck & Schrans, 1998) and a comprehensive literature review, we created the reflective and formative constructs described below.
An individual who is preoccupied with thoughts of gambling has been defined as “having a fixation on gambling, is continually reliving past experiences, planning the next outing and thinking about how to get money for such an excursion” (4th ed., text rev.; DSM–IV–TR; APA, 2000). These criteria have been found to be reliable and valid predictors of problem gambling (Hodgins, 2004; Lakey, Goodie, Lance, Stinchfield, & Winters, 2007; Stinchfield, Govoni, & Frisch, 2005; Toce-Gerstein, Gerstein, & Volberg, 2009; Wickwire, Burke, Brown, Parker, & May, 2008). Gamblers who are obsessed with their play think constantly about their gambling. This fact in itself may not be considered a negative consequence of gambling; however, these thoughts can become so prevalent that they are deemed harmful, insofar as the person is tormented by these thoughts, is unable to function normally, or both. In these circumstances, gambling is seen to have harmed the gambler.
Merely whiling away one’s time with thoughts of gambling (Preoccupation Desire) is considered a risk indicator that occurs in the absence of harm. We have defined Preoccupation Desire as “having a strong desire to gamble frequently or as often as possible”; other researchers have referred to this characteristic as “craving” (Ashrafioun & Rosenberg, 2012; Young & Wohl, 2009). Wanting to gamble frequently may be a common characteristic of everyone who enjoys gambling; however, a strong desire that leads to increased or more extreme gambling activity could be an effective indicator of elevated risk. This characteristic could also be particularly relevant in subsequent modeling analysis, if it is shown to be a precursor to Impaired Control and Risky Practices.
Impaired control has been characterized as “repeated unsuccessful attempts to resist the urge to gamble in the context of genuine desire to cease” (Blaszczynski & Nower, 2002) as well as an inability to resist opportunities to gamble (e.g., begin a session) and cease the activity once engaged (e.g., continue a session) (Dickerson & O’Connor, 2006). We operationally defined the reflective construct, Impaired Control, as a gambling-specific construct based upon an individual’s personal experience.
Impaired control has been considered a cause of problem gambling (Ladouceur, Cantinotti, & Tavares, 2007) and thus has the potential as a risk indicator. Gamblers may attempt to justify their losses by perceiving them as a product of a lack of control, whether or not this is actually the case (Dickerson & O’Connor, 2006). Regardless of whether the action is perceived or factual, if a gambler recognizes that he is acting contrary to his intentions, this perception may be an indicator of risk. In creating our construct we modified several statements from the Scale of Gambling Choices (Baron, Dickerson, & Blaszczynski, 1995). We also used the two behaviours described by Dickerson and O’Connor (2006). In doing so we considered impaired control to be a two-dimensional construct such that an individual could suffer from one or both of the following problems: (1) Impaired Control Continue a session, defined as an inability to cease gambling once engaged, or (2) Impaired Control Begin a session, defined as an inability to resist starting a session .
Persistence is generally referred to as playing for a long time within a session, particularly when losing (Dickerson, 1993), and as excessive play in addition to continuing to play in attempts to recover previous losses (Breen & Zuckerman, 1999). The term has also been used in the context of assessing the number of trials at play (Kassinove, 1999) or sessions gambled per month (Ladouceur & Sévigny, 2005). Whether or not the player suffers harm from playing over an extended period of time is not generally addressed. Given our interest in using the concept of persistence to characterize problem gambling, we defined the reflective construct as the engagement of risky practices over an extended period despite that behaviour leading to negative consequences. We did not include “chasing losses” in our Negative Consequences construct, and instead positioned it as an element of the formative construct Risky Practices, where it could serve as a risk indicator or precursor to negative outcomes.
Formative constructs are basically lists of items that as exhaustively as possible capture and thus define the latent variable being measured. A principal goal of this literature review was to identify a broad range of items and, subsequently, to select those items that defined unique elements of the construct. Thus, each item retained could contribute to the identification of individuals who have indications of the latent variable.
Many gamblers believe irrationally they can use skill to influence the outcome of games that have completely random results. For example, certain gamblers may think that pressing the buttons quickly on a gambling machine will increase their odds of winning. Moreover, many players mistakenly believe that the probability of winning is greater than is actually the case. This “illusion of control” concept advanced by Ladouceur and Walker (1996) and further defined by Toneatto (1999) provided the background for many of the scales developed to investigate the role of erroneous cognition in maintaining problem gambling. The PGSI includes two statements regarding faulty cognitions (Ferris & Wynne, 2001), and items assessing similar concepts have been used to discriminate successfully between pathological/problem and non-pathological/non-problem gamblers (Källmén, Andersson, & Andren, 2008; Raylu & Oei, 2004; Steenbergh et al., 2002; van Holtz, van den Brink, Veltman, & Goudriann, 2010; Xian, Shah, Phillips, Scherrer, Volberg, & Eisen, 2008). We created the formative construct Risky Cognitions Beliefs. The construct comprises belief statements consistently found in the literature as being associated with risk for electronic gambling.
An individual’s motivation is considered to be a reflection of the internal and external forces that direct him to take action. For example, a person could be internally motivated to gamble for the feeling of excitement or externally motivated to gamble to win money (Lee, Chae, Lee, & Kim, 2007). The Motivation Towards Gambling Scale incorporated the following motivations: reward seeking; self-imposed pressures, such as the need for recognition; and goals, such as socialization, knowledge, accomplishment and stimulation (Chantal, Vallerand, & Vallières, 1995). Subsequent work by Clarke (2004, 2005, 2008) and Pantalon, Maciejewski, Desai, and Potenza (2008) confirmed the role of these factors in shaping play behaviours. The motivation “to escape one’s problems” is also highly correlated with both frequency of gambling and progression towards pathological gambling (Clarke, 2008; Nelson, Gebauer, Labrie, & Shaffer, 2009; Nower & Blaszczynski, 2010; Thomas, Allen, & Phillips, 2009). Our construct Risky Cognitions Motives only contains those factors found to be associated with problem gambling and does not include all the motivators for gambling described in the literature.
Problem gamblers have been found to engage in behaviours such as chasing losses, and participating in illegal activities to finance gambling and lying about the extent of their gambling (4th ed., text rev.; DSM–IV–TR; APA, 2000). Such risky practices may also occur during play. For example, an individual may attempt to borrow money from another player (Schellinck & Schrans, 1998). As playing behaviour begins to cause problems, gamblers engage in a variety of risky practices that may escalate in severity (Campbell-Meiklejohn, Woolrich, Passingham, & Rogers, 2008; Hong, Sacco, & Cunningham-Williams, 2009; Sumitra & Miller, 2005). Fewer gamblers tend to engage in the more risky practices (Schellinck & Schrans, 1998; Schellinck, Schrans, & Walsh, 2000). Although players may frequently use maximum bet options, or use a bank card to obtain additional cash during a session of play, they less commonly borrow money on their credit cards to keep playing. If the endorsement rates varied substantially among the risky behaviours analysed, we would consider creating two formative constructs, “Risky Practices Earlier” and “Risky Practices Later.”
Several problem gambling screens (Ferris & Wynne, 2001; Lesieur & Blume, 1987; Toce-Gerstein et al., 2009) include statements regarding harmful outcomes. Our measure included characteristics described by Thomas et al. (2009), such as negative impacts on work and financial well-being, problems with health, interpersonal relationships, and deceptive behaviour. We also incorporated statements assessing the impact on the individual’s sense of self-worth as described by Suurvali, Cordingley, Hodgins, and Cunningham (2009), in this formative construct. We have excluded more extreme consequences, such as engaging in criminal behaviours and having suicidal tendencies, as these questions were considered to be too threatening for a self-administered survey. As well, the less severe consequences would be experienced first so the instrument’s ability to identify problem gamblers would not be compromised by their exclusion.
We selected 190 prototype statements for assessment in the constructs. Each item selected for testing was hypothesized to be associated with gambling risk or harm as variously supported by the literature, the OPGRC risk framework (Simpson, Goodstadt, Wynne, & Williams, 2008) and our own research examining determinants of problem gambling (Schellinck & Schrans, 1998; Schellinck et al., 2000). The instrument was evaluated in a two phase process with regular slot machine gamblers. First, six qualitative focus groups (n=63) were completed to conduct a beta-test of the items. Second, based upon the outcome of the focus groups, a quantitative telephone survey (n=374) with a reduced instrument was completed. The research was subject to independent ethics approval by the Institutional Review Board Services.
Regular slot machine gamblers, i.e., those gamblers who on average played once a month or more over the past 12 months, were approached at the Slots at Western Fair in London Ontario over a one-week period (n≈650), prescreened and invited to join a confidential research panel. The original panel sample was generated through on-site recruitment over a five-day period including three weekdays (Tuesday, Wednesday and Thursday) and the weekend (Saturday and Sunday) over day and evening shifts covering periods from 8 a.m. to 10 p.m. Whereas the resulting sample constituted a convenience sample it was nonetheless fairly representative of the population of frequent gamblers, with high cooperation rates during the recruiting process (refusal rates < 20%).
Panel members were then re-contacted to take part in a series of focus groups to assess a beta version of the instrument. We did this to assist in determining the clarity of the items. Participants were recruited and grouped depending upon their risk score on the PGSI, (i.e., 0, 1– 4, or 5 or more) and how long they had been gambling on a regular basis of once a month or more (less than 2 years, or 2 or more years). An equal number of men and women took part in the sessions, with ages ranging from 23 to 74 years. All participants arrived 30 minutes before the session to self-complete a beta version of the instrument. During the discussion that followed, participants referred to a blank copy of the items to preserve the confidentiality of their personal responses. All sessions were audiotaped and the tapes transcribed by independent support personnel. An independent observer kept detailed notes during the sessions for use in thematic analysis of the statements. The discussion groups lasted approximately two hours and participants received an honorarium of $60.00 for taking part in the sessions.
If the participants had several interpretations of the wording, the statements were either revised or discarded as unsuitable. Many of the statements were intentionally similar in wording; those examples found to align most closely to our original intended meaning were retained, with less definitive versions eliminated. Based upon group responses and subsequent assessment of the 63 surveys, 132 dichotomous response statements were generated for testing in a larger quantitative sample.
The remaining panel members were contacted by telephone and asked to complete the reduced instrument. Informed consent was obtained from respondents before data collection took place. Individuals currently receiving assistance for substance use, gambling, or a mental health issue were excluded, as were those persons who worked for (1) the media, (2) a political or lobby group, (3) Addiction Services, or (4) Ontario Lottery Gaming or an affiliate. Participants who wished to do so had their names entered into a draw for one of four $100 grocery gift certificates (with no cash value) at a retail grocer of their choice. Of 422 eligible panel members, 374 met the selection criteria, i.e., played slots more than once a month, were not a member of any of the excluded groups, and responded to the survey. This process resulted in a completion rate of 69.2 %. Prior to analyses, 18% of those persons who responded to the survey were re-contacted by the field staff supervisor and key questions were repeated to ensure consistency and accuracy in responses to the survey. This process ensured that (1) the respondents were answering honestly and that they were not having problems remembering key estimates, and (2) that the interviewers were administering the survey correctly. No surveys needed to be deleted upon completion of this process.
This sample was comprised of 150 males and 224 females; the median age was 63 with ages ranging from 23 to 89 years. Over half (53.5%) of the sample played the slots weekly, 0.8% played daily and 45.7% played less than weekly on the slots. According to the PGSI 54.8% of the respondents were no risk gamblers, 19.3% were low risk gamblers, 20.3% were medium risk gamblers and 5.6% were problem gamblers. The test instrument and classification questions took approximately 26 minutes to administer (range 20–46). Statements were randomized for each participant to reduce the risk of common method bias (Bliemel & Hassanein, 2007).
Following analysis, the 132 statements were reduced to 53 items across the 10 constructs (see Appendix). As described below, the criteria for selecting the statements for the proposed constructs differed depending on whether the constructs were reflective or formative. Three steps were used to select the statements for the reflective constructs. First, exploratory principal components analysis (PCA) was performed on the three sets of statements designed to measure the original constructs—impaired control, preoccupation and persistence. Those statements having loadings greater than 0.5 on the resulting varimax rotated constructs were retained (Table 1). Second, statements comprising the resulting five constructs that had substantially different endorsement rates from the mode endorsement rate for each construct were dropped. Third, if a construct still had more than five statements those statements with the lowest loadings on the component were dropped so that the maximum number of statements in a construct to be tested for reliability and validity was five.
The formative construct statement selection process itself had three steps. First, PCA analysis of those statements designed for a construct was conducted, though for formative constructs it was expected that many components would be formed, and that one or two statements would be selected from each component to ensure all dimensions in the construct were captured. Second, the degree in overlap among the statements was then tested for by examining the Variance Inflation Factor (VIF) score of the statements when tested together, and statements with too much overlap with the other statements (i.e., VIF > 10.0) were dropped. Third, endorsement rates within the risky practices construct were compared to determine if certain statements might indicate earlier or later levels of risk. The large disparity in endorsement rates (4.0% – 34.8%) led to the splitting of the construct into two (one indicating early risky practices, the other later risky practices).
The statements created for each reflective construct were expected to reflect a single underlying latent variable. Several criteria, outlined below, were used to decide which statements would be assigned to the reflective constructs, which statements would be dropped, and whether any statements might need modification of wording to capture better the latent variable being measured.
Analyses for validity and reliability were conducted for the formative and reflective constructs using methods appropriate to the respective form of construct.
The construct reliability for reflective constructs is shown in Table 3. An outcome of the statement selection process was that the Preoccupation Obsession construct had only two statements. This fact meant that reliability tests could not be performed in this instance. We left this two-statement construct in the analyses when testing the other constructs. The four constructs consisting of three or more statements all had component reliability above the recommended level of 0.70 (Nunnally, 1978), indicating sufficient internal consistency. The convergent validity was evaluated using the average variance extracted (AVE) and all five reflective constructs were found to perform above the guideline of 0.5 as recommended by Fornell and Larcker (1981).
The discriminant validity of the five reflective constructs was evaluated using two approaches. The first approach compared the square root of the Average Variance Extracted (AVE) to the correlations of the other constructs. Adequate discriminant validity was indicated if the square root of the construct’s AVE was greater than its correlations with the other constructs (Compeau, Higgins, & Huff, 1999). Table 4 presents the square root of the AVE in the diagonal and the correlations in the off diagonal. All five reflective constructs passed the test for discriminant validity. For formative constructs, it is inappropriate to report AVE and not applicable (n.a.) has therefore been entered in the diagonal for these constructs.
The second approach (Gefen & Straub, 2005) compared the correlations between the individual items and the PLS calculated latent variable scores generated by conducting confirmatory factor analysis (Table 5). For the construct to have discriminant validity the item loadings for the reflective construct must be greater by 0.10 than the construct’s correlations with the other items. The five reflective constructs passed the test for discriminant validity. Again, this test does not apply to formative constructs (Diamantopoulos & Winklhofer, 2001) as the items comprising such a construct are not expected to be correlated with each other. The formative constructs were retained in the table so that correlations with the tested reflective constructs can be evaluated.
We adopted the four methods recommended by Henseler, Ringle, and Sinkovics (2009, p. 309) for assessing the validity of formative constructs. 1. Nomological validity: The relationships between the formative index and other constructs in the path model, which are sufficiently well-known through prior research, should be strong and significant. 2. External validity: The formative index should explain the variance of an alternative reflective measure of the focal construct. 3. Significance of weights: Estimated weights of formative measurement models should be significant. 4. Multicollinearity: Manifest variables in a formative block should be tested for multicollinearity. The variance inflation factor (VIF) can be used for such tests. As noted previously, a VIF greater than 10.0 indicates the presence of harmful collinearity (Diamantopoulos & Siguaw, 2006).
In the companion article to this paper, we created a Structural Equation Model using PLS (SEM PLS); we have reserved discussion of PLS and our hypotheses for that article. To describe the nomological validity of the constructs, however, we have reported principal results in Table 6. All five formative constructs predicted another construct as hypothesized, i.e., each had significant SEM PLS coefficients and four formative constructs were sequentially connected to preceding constructs as hypothesised.
To determine if the formative construct could explain a significant portion of the variance in an alternative but similar reflective measure of the construct, we undertook such a comparison for Negative Consequences and Persistence, the latter being a reflective measure that has negative consequences as part of its makeup. The weight for this connection in the PLS model was 0.569 (t=9.21), supporting its validity. This test was not viable for the other formative constructs as equivalent reflective constructs were not available.
The t-scores of the construct item weights for each of the formative constructs are found in Table 2. As recommended by Henseler et al. (2009), we retained certain non-significant indicators if they had sufficiently low VIF scores and were judged to be conceptually correct.
The statement selection process for formative constructs eliminated statements in each construct until all statements in the construct had VIF scores below 10.0 (Diamantopoulos & Siguaw, 2006; Henseler et al., 2009). We reported here the resulting range of VIF scores for each formative construct. The highest VIF score for Risky Cognitions Beliefs was 1.11 and for Risky Cognitions Motives 1.32. These scores indicate that multicollinearity was not an issue for either of these constructs. Both Risky Practices Earlier with a maximum VIF of 1.58 and Risky Practices Later with a maximum VIF of 1.84 had some degree of multicollinearity, but they were still well below the accepted threshold of 10 for inclusion in a formative construct. Similarly, Negative Consequences, with a maximumVIF score of 2.79, had some multicollinearity but again still met the criterion.
The data were examined for Common Method Bias using the method recommended by Podsakoff, MacKenzie, Lee, & Podsakoff (2003) for Harmon’s one-factor test. PCA was performed on all 53 indicators chosen for inclusion in the constructs, and the unrotated solution was assessed to determine the number of components with an eigenvalue greater than one. Strong method bias is present if the analysis produces a single component; less bias is present when more components are produced. In the current analysis, eleven components emerged with eigenvalues greater than one with the first component accounting for 33.6% of the variance and, collectively, the 11 other components accounting for 66.3% of the original variance. The results supported the conclusion that the amount of variance because of common method bias was not sufficient to explain our findings.
The results of testing and analysis produced an instrument with 53 statements comprising five formative and four reflective constructs designed to identify individuals experiencing risk of harm from gambling. A fifth reflective construct, Preoccupation Obsession, needs further development and testing to be an accurate measure of risk. The five formative constructs—Risky Cognitions Beliefs, Risky Cognitions Motives, Risky Practices Earlier, Risky Practices Later and Negative Consequences—passed the tests for nomological validity and external validity. The majority of the items had significant regression weights and any other statements retained were theoretically valid. All of the items passed the test for multicollinearity with VIF scores significantly below 10.0 and none greater than 2.8. The hypothesized relationships found in the PLS analysis are described in the companion paper as well as the potential applications for the FLAGS-EGM instrument.
The formative construct Risky Practices was divided into two constructs as some behaviours were found to be quite commonly endorsed by more gamblers whereas others were less frequent and more likely to be exhibited by those at higher risk, i.e., located closer to problem gambling status. As a result, the construct was subdivided to identify those exhibiting behaviour either early or late in the process of becoming a problem gambler. Using two constructs to define operationally early and late risk practices should provide a superior model of risk factors associated with problem gambling, in particular in terms of positioning, such constructs in the hierarchy of risk for problem gambling, e.g., earlier versus later risk factors.
Because FLAGS-EGM is composed of separate distinctive constructs we were able to focus on creating highly accurate statements to capture the essence of each specific variable. As a result, when designing the Negative Consequence construct, we could separate cause from effect, an important temporal consideration when designing an instrument for prevention applications. For example, borrowing money to gamble is not a negative consequence of gambling as consumers borrow money regularly for many purchases. Not being able to repay the loans or having to sacrifice other needed resources to pay for these loans, however, would be a negative consequence that could result from borrowing.
Some investigators consider obsession with gambling as a negative outcome that should be included as part of the definition of problem gambling rather than as a risk factor. We understand that obsession is closely related to problem gambling, but question whether on its own obsession is sufficient to define problem gambling. Many individuals can be obsessed with activities but in and of itself, this characteristic may in fact not be linked to negative consequences for an individual. For example, if an individual is obsessed with playing golf there may in fact be no negative impacts unless that person begins to miss work or neglect family responsibilities. We believe that if gamblers have become obsessed with gambling but have yet to suffer negative consequences, such a situation may accordingly be an excellent indicator of high-risk due to gambling. If further investigation indicates that only problem gamblers exhibit obsession with gambling, then we would reject the construct as a useful risk predictor.
The use of formative constructs should provide users of FLAGS-EGM added insight into the nature of the harms or risk factors gamblers face. For example, the gambler who self-administers the instrument can see which specific behaviours may be contributing to their risk and may be able to curtail or self-manage these behaviours. Public health, regulatory bodies and treatment providers will all be able to judge the degree to which problem gamblers suffer financial, relationship, or psychological harm and in turn, influence policy and practices. Even more importantly from a prevention and harm reduction perspective, it will be possible to assess the respective impacts of specific actions undertaken to resolve or mitigate risk for problem gambling, a means of measurement and evaluation that is not possible using existing screens.
Factor analysis and the literature supported the hypothesis that two of the reflective constructs, Preoccupation and Impaired Control, should each be subdivided into two separate constructs, Preoccupation Desire and Preoccupation Obsession, and Impaired Control Begin and Impaired Control Continue, respectively. Subsequent testing for discriminant validity confirmed that they are four distinct constructs. Four of the five reflective measures: Impaired Control Continue, Impaired Control Begin, Persistence, and Preoccupation Desire were each shown to hold sufficient internal consistency with component reliabilities above 0.70 and convergent validity with AVE scores above 0.5. Moreover, they each had discriminant validity, all having a greater square root of the Average Variance Extracted compared to the correlations with the other constructs. Moreover, the PLS-calculated latent variable scores for the reflective constructs were greater by 0.10 than the construct’s correlations with the other items.
The constructs were designed using a sample of regular Ontario slot gamblers. It could be argued that this restriction accordingly limits the generalizability of the results; however, we believe this limitation could apply to any set of constructs based on a single sample representing a specific population. The original items comprising FLAGS-EGM were developed and tested with video lottery players in Nova Scotia, Canada, and ‘pokie’ machine gamblers in Victoria, Australia. We did not use samples of university students, self-selected samples, or volunteers recruited through advertisements. Individuals diagnosed with comorbid disorders, treatment populations or prison inmates were not surveyed. Consequently, we are confident that these results are reasonably reflective of a general population of gamblers.
The item scales are dichotomous (yes/no) to facilitate understanding and ease of answering the items, a desirable characteristic for an instrument that is designed to be self-administered and which is composed of a large number of items. Examination of other screens suggests that it is difficult to create multi-item scales without adding considerable method bias as a result of the misfit between the statement and the scale anchors. Moreover, we built frequency distinctions into the statements to account for differences in the occurrence of certain behaviours, beliefs or outcomes. Certain statistics, particularly factor analysis, may be somewhat inaccurate when applied to dichotomous data. For example, on average, the loading may vary ± 0.08 compared to statistics derived using other techniques (Maguire 2001). Although we used PCA analysis to nominate statements for inclusion in the constructs, we relied on other means to actually test the reliability and validity of the constructs. As a result, we believe the use of factor analysis in this context does not reduce the validity of the constructs.
The current process did not yield a usable construct to measure Preoccupation Obsession. In the next phase of the research, additional statements will be included and tested to capture better this latent variable.
FLAGS-EGM holds strong potential. This approach to risk measurement moves beyond simple identification of those persons who have problems to active risk tracking for prevention and evaluation purposes. We will be conducting further research to develop this instrument by applying it to other larger, random samples of EGM players in different markets, testing new candidate statements for the preoccupation construct and retesting the constructs for reliability and validity.
The next steps, as detailed in the companion paper in this issue, required the use of SEM PLS to determine the causal paths of the relationships among the constructs. The placement and grouping of the constructs as indicators of risk will then determine the appropriate number of risk levels to create. FLAGS-EGM is composed of separate construct measures of risk, and thus when compared with current instruments, it should more accurate assess an individual’s risk due to gambling as well as identify those.
Risky Cognitions Beliefs
Risky Cognitions Motives
Risky Practices Earlier
Risky Practices Later
Impaired Control Continue
|American Psychiatric Association ( 2000). Diagnostic and statistical manual of mental disorders (4th ed., text revision). WashingtonDC: Author.|
|Ashrafioun, L.. Rosenberg, H. . ( 2012). Methods of assessing craving to gamble: A narrative review. Psychology of Addictive Behaviors, 26(3), 536–49.
|Baron, E.. Dickerson, M.. Blaszczynski, A. . ( 1995). The scale of gambling choices: Preliminary development of an instrument to measure impaired control of gambling behaviour. In O’Connor, J. V.
(Ed.), High stakes in the nineties: 6th National Conference of the National Association for Gambling Studies, Fremantle, Western Australia (
pp.153–168)Sydney, AU:The Association.
|Blaszczynski, A.. Nower, L. . ( 2002). A pathways model of problem and pathological gambling. Addiction, 97(5), 487–499.
|Bliemel, M.. Hassanein, K. . ( 2007). Consumer satisfaction with online health information retrieval: A model and an empirical study. e-Service Journal, 5(2), 53–83.|
|Bollen, K.. Lennox, R. . ( 1991). Conventional wisdom on measurement: A structural equation perspective. Psychological Bulletin, 110(2), 305–314.|
|Breen, R. B.. Zuckerman, M. . ( 1999). Chasing in gambling behavior. Personality and cognitive determinants.. Personality and Individual Differences, 27(6), 1097–1111.
|Campbell-Meiklejohn, D. K.. Woolrich, M. W.. Passingham, R. E.. Rogers, R. D. . ( 2008). Knowing when to stop: The brain mechanisms of chasing losses. Biological Psychiatry, 63(3), 293–300.
|Chantal, Y.. Vallerand, R. J.. Vallières, E. F. . ( 1995). Motivation and gambling involvement. The Journal of Social Psychology, 135(6), 755–763.
|Clarke, D. . ( 2004). Impulsiveness, locus of control, motivation and problem gambling. Journal of Gambling Studies, 20(4), 319–345.
|Clarke, D. . ( 2005). Motivational differences between slot machine and lottery players. Psychological Reports, 96(3:1), 843–848.
|Clarke, D. . ( 2008). Older adults’ gambling motivation and problem gambling: A comparative study. Journal of Gambling Studies, 24(2), 175–192.
doi: 10.1007/s10899-008-9090-z. Epub 2008 Feb 14.
|Compeau, D. R.. Higgins, C. A.. Huff, S. . ( 1999). Social cognitive theory and individual reactions to computing technology: A longitudinal study. MIS Quarterly, 23(2), 145–158.
|Currie, S. R.. Hodgins, D. C.. Casey, D. M. . ( 2013). Validity of the problem gambling severity index interpretive categories. Journal of Gambling Studies, 29(2), 311–327.
|Diamantopoulos, A.. Riefler, P.. Roth, K. P. . ( 2008). Advancing formative measurement models. Journal of Business Research, 61(12), 1203–1218.
|Diamantopoulos, A.. Siguaw, J. A. . ( 2006). Formative versus reflective indicators in organization measure development: A comparison and empirical illustration. British Journal of Management, 17(4), 263–282.
|Diamantopoulos, A.. Winklhofer, H. M. . ( 2001). Index construction with formative indicators: An alternative to scale development. Journal of Marketing Research, 38(2), 269–277.
|Dickerson, M. G. . ( 1993). Internal and external determinants of persistent gambling. In Heather, N. Miller, W. M. Greeley, J.
(Eds.), Self-control and the addictive behaviours (pp. 317–338). Sydney, AU: Maxwell Macmillan.
|Dickerson, M.. O’Connor, J. . ( 2006). Gambling as an addictive behaviour: Impaired control, harm minimisation, treatment and prevention. Cambridge, UK: Cambridge UP.|
|Ferris, J.. Wynne, H. . ( 2001). The Canadian problem gambling index: Final report. Ottawa, ON: Canadian Centre on Substance Abuse.|
|Fornell, C.. Larcker, D. F. . ( 1981). Evaluating structural equation models with unobservable variables and measurement error. Journal of Marketing Research, 18(1), 39–50.|
|Freeze, R. D.. Raschke, R. L. . ( 2007, June7–9). An assessment of formative and reflective constructs in IS research. Paper presented at the European Conference on Information
Systems (ECIS) 2007 Conference.
Retrieved from http://aisel.aisnet.org/ecis2007/171.
|Gefen, D.. Straub, D. . ( 2005). A practical guide to factorial validity using PLS-Graph: Tutorial and annotated example. Communications of the Association for Information Systems, 16, 91–109.|
|Gefen, D.. Straub, D. W.. Boudreau, M.-C. . ( 2000). Structural equation modeling and regression: Guidelines for research practice. Communications of the Association for Information Systems, 7, 1–78.|
|Gerstein, D.. Hoffman, J.. Larison, C.. Engelman, L.. Murphy, S.. Palmer, A.. … Sinclair, S. . ( 1999). Gambling impact and behavior study: Report to the National Gambling Impact Study Commission. Chicago, IL: National Opinion Research Center at the University of Chicago.|
|Henseler, J.. Ringle, C. M.. Sinkovics, R. R. . ( 2009). The use of partial least squares path modeling in international marketing. In Zou, S..
(Series Ed.)and Ghauri, P.N.. Sinkovics, R.R. .
(Vol. Eds.), Advances in international marketing.: Vol. New challenges to international marketing (
pp.277–319). Bingley, UK: Emerald.
|Hodgins, D. C. . ( 2004). Using the NORC DSM Screen for Gambling Problems as an outcome measure for pathological gambling: psychometric evaluation. Addictive Behavior, 29(8), 1685–1690.|
|Hong, S. I.. Sacco, P.. Cunningham-Williams, R. M. . ( 2009). An empirical typology of lifetime and current gambling behaviors: Association with health status of older adults. Aging and Mental Health, 13(2), 265–73.
|Howell, R. D.. Breivik, E.. Wilcox, J. B. . ( 2007). Reconsidering formative measurement. Psychological Methods, 12(2), 205–218.
|Jarvis, C. B.. MacKenzie, S. B.. Podsakoff, P. M. . ( 2003). A critical review of construct indicators and measurement model misspecification in marketing and consumer research. Journal of Consumer Research, 30(2), 199–218.
|Johansson, A.. Grant, J. E.. Kim, S. W.. Odlaug, B. L.. Götestam, K. G. . ( 2009). Risk factors for problematic gambling: A critical literature review. Journal of Gambling Studies, 25(1), 67–92.
|Källmén, H.. Andersson, P.. Andren, A. . ( 2008). Are irrational beliefs and depressive mood more common among problem gamblers than non-gamblers A survey study of Swedish
problem gamblers and controls.. Journal of Gambling Studies, 24(4), 441–450.
|Kassinove, J. . ( 1999). Effects of the “near miss” and the “big win” on persistence at slot machine gambling (Doctoral dissertation).
Retrieved from Dissertations & Theses : Full Text. (AAT 9933933).
|Ladouceur, R.. Cantinotti, M.. Tavares, H. . ( 2007). Impaired control: A look at the laying brick of pathological gambling. Revista Brasileira de Psiquiatria, 29(3), 203–204.
|Ladouceur, R.. Sévigny, S. . ( 2005). Structural characteristics of video lotteries: Effects of a stopping device on illusion of control and gambling persistence. Journal of Gambling Studies, 21(2), 117–131.
|Ladouceur, R.. Walker, M. . ( 1996). A cognitive perspective on gambling. In Salkovskis, P. M.
(Ed.), Trends in cognitive therapy (pp. 89–120). Oxford, UK: Wiley.
|Lakey, C. E.. Goodie, A. S. Lance, C. E.. Stinchfield, R.. Winters, K. C. . ( 2007). Examining DSM-IV criteria for pathological gambling: Psychometric properties and evidence from cognitive biases. Journal of Gambling Studies, 23(4), 479–498.
|Lee, H.-P.. Chae, P. K.. Lee, H.-S.. Kim, Y.-K. . ( 2007). The five-factor gambling motivation model. Psychiatry Research, 150(1), 21–32.
|Lesieur, H. R.. Blume, S. B. . ( 1987). The South Oaks Gambling Screen (SOGS): A new instrument for the identification of pathological gamblers. American Journal of Psychiatry, 144(9), 1184–1188.|
|Lesieur, H.R.. Blume, S.B. . ( 1993). Revising the South Oaks Gambling Screen in different settings. Journal of Gambling Studies, 9(3), 213–223.
|MacKenzie, S. B. . ( 2003). The dangers of poor construct conceptualisation. Journal of the Academy of Marketing Science, 31(3), 323–326.|
|MacKenzie, S. B.. Podsakoff, P. M.. Jarvis, C. B. . ( 2005). The problem of measurement model misspecification in behavioral and organizational research and some recommended solutions. Journal of Applied Psychology, 90(4), 710–730.
|Maguire, E. R. . ( 2001). Software review of MicroFACT 1.1, a microcomputer factor analysis program for dichotomous and ordered polytomous data and
mainframe sized problems. Structural Equation Modeling: A Multidisciplinary Journal, 8(1), 150–156.
|Naghavi, M.. Falk, E.. Hecht, H. S.. Shah, P. K. . ( 2006). The first SHAPE (Screening for Heart Attack and Education) guideline. Critical Pathways in Cardiology. 5(4), 187–190.
|Neal, P.. Delfabbro, P.. O’Neil, M. . ( 2005). Problem gambling and harm: Towards a national definition: Final report (ABN 0 975119 4 1). Melbourne, AU: Victorian Government Department of Justice.
Retrieved from http://www.adelaide.edu.au/saces/gambling/publications/ProblemGamblingAndHarmTowardNationalDefinition.pdf.
|Nelson, S. E.. Gebauer, L.. Labrie, R. A.. Shaffer, H. J. . ( 2009). Gambling problem symptom patterns and stability across individual and timeframe. Psychology of Addictive Behavior, 23(3), 523–533.
|Nower, L.. Blaszczynski, A. . ( 2010). Gambling motivations, money-limiting strategies, and precommitment preferences of problem versus non-problem gamblers. Journal of Gambling Studies, 26(3), 361–372.
|Nunnally, J. C. . ( 1978). Psychometric theory
(2nd ed.). New York, NY: McGraw Hill.
|Pantalon, M. V.. Maciejewski, P. K.. Desai, R. A.. Potenza, M. N. . ( 2008). Excitement seeking gambling in a nationally representative sample of recreational gamblers. Journal of Gambling Studies, 24(1), 63–78.
|Podsakoff, P. M.. MacKenzie, S. B.. Lee, J. Y.. Podsakoff, N. P. . ( 2003). Common method biases in behavioral research: A critical review of the literature and recommended remedies. Journal of Applied Psychology, 88(5), 879–903.
|Raylu, N.. Oei, T. P. S. . ( 2004). The Gambling Related Cognitions Scale (GRCS): Development and confirmatory factor validation and psychometric properties. Addiction, 99(6), 757–769.
|Schellinck, T.. Schrans, T. . ( 1998). The 1998 Nova Scotia video lottery survey. Halifax, NS: Nova Scotia Department of Health.|
|Schellinck, T.. Schrans, T.. Walsh, G. . ( 2000). Nova Scotia Department of Health regular VL players follow-up study: A comparative analysis of problem development and resolution. Halifax, NS: Nova Scotia Department of Health.|
|Shaffer, H. J.. LaBrie, R. A.. LaPlante, D. A.. Nelson, S. E.. Stanton, M. V. . ( 2004). The road less traveled: Moving from distribution to determinants in the study of gambling epidemiology. Canadian Journal of Psychiatry, 49(8), 504–516.|
|Steenbergh, T. A.. Meyers, A. W.. May, R. K.. Whelan, J.P. . ( 2002). Development and validation of the Gamblers Beliefs Questionnaire.. Psychology of Addictive Behaviors, 16(2), 143–149.
|Stinchfield, R.. Govoni, R.. Frisch, G. R. . ( 2005). DSM-IV diagnostic criteria for pathological gambling: reliability, validity, and classification accuracy. American Journal of Addiction, 14(1), 73–82.
|Sumitra, L. M.. Miller, S. C. . ( 2005). Pathologic gambling disorder. How to help patients curb risky behavior when the future is at stake. Postgraduate Medicine, 118(1), 31–37.|
|Suurvali, H.. Cordingley, J.. Hodgins, D. C.. Cunningham, J. . ( 2009). Barriers to seeking help for gambling problems: A review of the empirical literature. Journal of Gambling Studies, 25(3), 407–424.
|Tolchard, B.. Battersby, M. W. . ( 2010). The Victorian Gambling Screen: Reliability and Validation in a Clinical Population. Journal of Gambling Studies, 26(4),623-638.|
|Thomas, A. C.. Allen, F. C.. Phillips, J. . ( 2009). Electronic gaming machine gambling: Measuring motivation. Journal of Gambling Studies, 25(3), 343–355.
|Thomas, S.A.. Jackson, A.C.. Blaszczynski, A. . ( 2003). Measuring problem gambling: Evaluation of the Victorian Gambling Screen. Melbourne: Gambling Research Panel.
|Toce-Gerstein, M. . Gerstein, D.R. ., & Volberg, R.A. . ( 2009). The NODS-CLiP: A rapid screen for adult pathological and problem gambling. Journal of Gambling Studies, 25(4), 541–549.
|Toneatto, T. . ( 1999). Cognitive psychopathology of problem gambling. Substance Use and Misuse, 34(11), 1593–1604.
|Turner, N. E.. Jain, U.. Spence, W.. Zangeneh, M. . ( 2008). Pathways to pathological gambling: Component analysis of variables related to pathological gambling. International Gambling Studies, 8(3), 281–298.
|van Holtz, R. J.. van den Brink, W.. Veltman, D. J.. Goudriann, A. E. . ( 2010). Why gamblers fail to win: A review of cognitive and neuroimaging findings in pathological gambling. Neuroscience and Biobehavioral Reviews, 34(1), 87–107.
|Volberg, R. A.. Dickerson, M. G.. Ladouceur, R.. Abbott, M. W. . ( 1996). Prevalence studies and the development of services for problem gamblers and their families. Journal of Gambling Studies, 12(2), 215–231.
|Walker, M. B.. Dickerson, M. G. . ( 1996). The prevalence of problem and pathological gambling: A critical analysis. Journal of Gambling Studies, 12(2), 233–49.
|Wickwire, E. M. .,Burke, R. S.. Brown, S. A.. Parker, J. D.. May, R. K. . ( 2008). Psychometric evaluation of the National Opinion Research Center DSM-IV Screen for Gambling Problems (NODS). American Journal of Addiction, 17(5), 392–395.
|Wilcox, J. B.. Howell, R. D.. Breivik, E. . ( 2008). Questions about formative measurement. Journal of Business Research, 63(12), 1219–1228.
|Xian, H.. Shah, K. R.. Phillips, S. M.. Scherrer, J. F.. Volberg, R.. Eisen, S. A. . ( 2008). Association of cognitive distortions with problem and pathological gambling in adult male twins. Psychiatry Research, 160(3), 300–307.
|Young, M. M.. Wohl, M. J. . ( 2009). The Gambling Craving Scale: Psychometric validation and behavioral outcomes. Psychology of Addictive Behaviors, 23(3), 512–522.