Editorial Credits/ Crédits éditorials

Subscription Rates /Prix d'abonnements

Advertising Rates / Tarifs publicitaires (PDF)

Guest Editorial
Gambling: The Hidden Addiction
Robert Ladouceur
(PDF)

In Review
The Road Less Travelled: Moving From Distribution to Determinants in the Study of Gambling Epidemiology
Howard J Shaffer, Richard A LaBrie, Debi A LaPlante, Sarah E Nelson, Michael V Stanton
(PDF)

Assessing and Treating Problem Gambling: Empirical Status and Promising Trends
Tony Toneatto, Goldie Millar
(PDF)

Review Paper Preventing Postpartum Depression Part II: A Critical Review of Nonbiological Interventions
Cindy-Lee E Dennis
(PDF)

Continuity of Care in Mental Health Services: Toward Clarifying the Construct
Anthony S Joyce, T Cameron Wild, Carol E Adair, Gerald M McDougall, Alan Gordon, Norman Costigan, Anora Beckie, Laura Kowalsky, Gloria Pasmeny, Fran Barnes
(PDF)

The Boundary Between Borderline Personality Disorder and Bipolar Disorder: Current Concepts and Challenges
Chandra A Magill
(PDF)

Selective Serotonin Reuptake Inhibitor and Venlafaxine Use in Children and Adolescents With Major Depressive Disorder: A Systematic Review of Published Randomized Controlled Trials
Darren B Courtney
(PDF)

Original Research Twenty-Year Course of Schizophrenia: The Madras Longitudinal Study
R Thara
(PDF)

Book Reviews
(PDF)

Evidence and Experience in Psychiatry. Volume 6. Eating Disorders
Review by
Hany Bissada

Integrated Treatment for Mood and Substance Use Disorders
Review by
Nady el-Guebaly

Letters to the Editor
(PDF)

Attention-Deficit Hyperactivity Disorder in a Sample of Omani Schoolboys

Effective Use of Olanzapine for Obsessive–Compulsive Symptoms in a Patient With Bipolar Disorder

Monoamine Oxidase Inhibitors and Subarachnoid Hemorrhage

Beyond Haloperidol: Teaching Emergency Medicine Residents to Manage Acute Agitation and Aggression in the Emergency Department
Hypothalamic–Pituitary–Adrenal Function and Preventing Major Depressive Episodes

A Romanian Adoptee’s Journey From Latency Age to Adolescence

Stéatose hépatique non alcoolique secondaire à la clozapine

Re: A Case–Control Study on Psychological Symptoms in Sleep Apnea-Hypopnea Syndrome (SAHS)

Reply: A Case–Control Study on Psychological Symptoms in Sleep Apnea-Hypopnea Syndrome (SAHS)

In Review

The Road Less Travelled: Moving From Distribution to Determinants
in the Study of Gambling Epidemiology

Howard J Shaffer, PhD¹, Richard A LaBrie, EdD², Debi A LaPlante, PhD³, Sarah E Nelson, PhD³, Michael V Stanton, BA⁴

By primarily focusing on the distribution of gambling and gambling disorders, the epidemiologic study of gambling has progressed along the same initial path as most psychiatric epidemiology. Now, this field faces divergent roads. Down one is the extension of the well-travelled path of prevalence studies in geopolitical areas (for example, nations, states, and provinces) and targets of opportunity (for example, schools, organizations, and local groups). Down the other—the road less travelled—lies a turn toward studies of gambling determinants.

A vast assortment of games of chance have existed throughout recorded history; however, the 20th century has seen remarkable growth in gambling’s visibility, availability, and accessibility through the proliferation of legalized casinos and gambling houses (1,2). In addition to the potential loss of money, research links gambling with various physical and mental health problems (3–5). As gambling’s availability and popularity have grown, so has the extent of gambling-related problems, leading to increased interest in the study of gambling and its consequences.

Gambling studies grew rapidly over the last 30 years. Most gambling-related research examined epidemiology (6)—the distribution and determinants of gambling—and most of that research focused on distribution. Measuring the distribution of a phenomenon (for example, gambling) across population segments involves identifying prevalence (that is, the proportion of people with the disorder at a specific time) and incidence (that is, the proportion of people who acquire the disorder during a specific period). Measuring the determinants of a phenomenon (for example, disordered gambling) involves identifying the causes and risk factors associated with the problem of interest; incidence studies provide ideal designs to permit the identification of these factors. Currently, there are more than 200 existing studies of gambling prevalence; however, there are few true incidence studies (2,7–9). Further, there are few studies of the determinants of gambling and disordered gambling (5).

As this relatively young field continues to develop, extant evidence suggests that the current course of broad, population-based prevalence studies is nearing its end and that the field now needs to develop its measurement tools and turn to the study of determinants as its central strategy. Identifying determinants requires an understanding of both the incidence and nature of disordered gambling. Researchers can accomplish this by focusing on factors that influence population segments to 1) initiate gambling, 2) persist in gambling despite suffering a gambling-related problem or disorder, 3) stop or reduce intemperate gambling, and 4) relapse to disordered gambling states after a period of remission. Each area of understanding is central to advancing prevention and treatment efforts. If the field is to move successfully toward the study and understanding of the determinants of gambling-related problems, researchers need to improve both the theory identifying the conceptual research targets and the instruments used to measure these constructs.

This article proposes that the field of gambling studies is at a transitional point in its development: it is poised between initial descriptions of the distribution of gambling-related problems and theory-driven investigations of the proximal and distal risk factors that influence the developmental course of those problems. To this end, after giving readers a brief history of the psychiatric epidemiology of gambling studies, we will review the current assessment tools and concepts used, summarize pertinent epidemiologic findings, consider the theoretical issues facing the field as it advances toward a new era of research, and discuss the potential directions for this new wave of investigation.

Laying the Road: Defining and Identifying Gambling Disorders

Defining Intemperate Gambling as a Psychiatric Disorder
In 1980, the diagnosis of “pathological” gambling joined pyromania, kleptomania, and intermittent and isolated explosive disorders as an impulse disorder in the DSM-III of the American Psychiatric Association (APA) (10). According to the APA, clinicians identify the presence of a gambling disorder by confirming at least 5 of 10 DSM criteria; each criterion currently carries equal weight, despite the likelihood that each differentially reflects the extent of the disorder. In addition, clinicians must determine that the pattern of excessive gambling is not caused by a manic disorder. Table 1 summarizes the most current DSM-IV criteria (11). Taken together, the 10 diagnostic criteria derive from 3 conceptual domains often associated with addictive disorders: compulsion or craving; loss of control; and continuing behaviour, despite the presence of adverse consequences. Although the DSM-IV currently classifies pathological gambling as an impulse control disorder, many clinicians consider it to be an addiction, a label not yet included in the APA nomenclature (11). Indeed, the DSM criteria for gambling disorders closely parallel the signs and symptoms of substance use disorders, which are commonly considered addictions (10,11).

Assessing the Prevalence of Gambling and Gambling-Related Problems
During the mid-1970s, on behalf of the US Commission on a National Policy Toward Gambling, Kallick and colleagues first described the nature and scope of gambling activities in the US (12). Lacking an instrument for measuring disordered gambling, they created the Institute for Social Research (ISR) scale, an 18-item scale with items derived from related concepts and research on heavy gamblers (12). The ISR was the first instrument to measure disordered gambling and helped stimulate large-scale gambling-related epidemiologic research.

In 1987, Lesieur and Blume developed the South Oaks Gambling Screen (SOGS) to create “a consistent, quantifiable, structured instrument that could be administered easily by nonprofessional as well as professional interviewers” (13, p 1184). These investigators used the DSM-III-R criteria (14) to guide both the development and validation of the SOGS, which rapidly became the instrument of choice among researchers estimating disordered gambling prevalence and remains the screen most frequently used to assess the presence of gambling-related problems (15).

There are over 27 instruments for identifying disordered gambling, with many more in development; however, there is considerable debate about these instruments and what each purports to measure (15–17). Although various instruments are available to assess the prevalence of disordered gambling, Shaffer and colleagues suggest that each instrument is best understood by considering the context of its origin, its driving motivation, its relation to funding, and its inherent strengths and weaknesses (15). By fully appreciating the multidimensional characteristics of each instrument, investigators can better assess the validity of the estimates they produce. Although researchers constantly strive for a perfect psychiatric nosology capable of classifying every disorder accurately, it is important to remember that a map is not the territory (18) and a diagnosis is not the same as the disease (19).

Cutpoints and Continua
Although many diagnostic schemes (for example, the DSM-IV) recognize only the presence or absence of a clinical disorder, evidence suggests that subclinical instances of most disorders are more prevalent and cause more public harm because of their greater numbers (20–22). Consistently, gambling-related problems reside on a behavioural continuum that can range from none to a great deal. Over time, researchers and clinicians have transformed this quantitative continuum into discrete categories, using different and often confusing labels (for example, at-risk, problem, subclinical, pathological, probable pathological, extremely pathological, and compulsive).

To simplify and organize this jumble of categories, researchers have developed more inclusive categorical standards for measuring the continuum of disordered gambling behaviour. Shaffer and Hall proposed a universal system that categorizes gambling along a continuum (23). This public health strategy allows clinical workers to separate and classify many levels of gambling-related problems by assigning a value to each level of gambling behaviour, ranging from nongambler to severely disordered gambler. In Shaffer and Hall’s system, level 0 represents nongamblers, level 1 represents gamblers who do not report any gambling-related signs or symptoms, level 2 represents gamblers who are experiencing subclinical levels of gambling-related problems, and level 3 represents gamblers who meet diagnostic criteria for having a gambling disorder (Note 1). Note that gamblers can move in 2 directions within this system, either to a more disordered state (for example, from level 2 to level 3) or to a less disordered state (for example, from level 2 to level 1). This system more accurately classifies the range of gambling-related problems, more finely measures change than do dichotomous strategies (for example, the DSM-IV), and provides clinical workers with a comprehensive tool to guide the allocation of resources to gambling-related needs. Determining the point at which subclinical becomes diagnostic depends on the characteristics of each screening instrument; in addition, alternate cutpoints have been suggested for existing measures (for example, the SOGS). Because the level system includes cutpoints to determine levels for various instruments, it is descriptive rather than corrective.

The Elusive Gold Standard
Although worldwide prevalence estimates of disordered gambling seemingly reflect a relatively stable and robust phenomenon, investigators have not yet established independent markers that can distinguish gambling disorders as a unique psychiatric construct. As with most psychiatric diagnoses, the lack of a diagnostic gold standard (that is, an independent criterion that can validate the presence or absence of a disordered state) constrains the development and planning of both treatment strategies and public policy initiatives designed to ameliorate or regulate gambling-related problems. The absence of a gold standard does not suggest that disordered gambling is invalid as an underlying construct; it merely suggests that diagnostic classification—whether clinician- or instrument-based—needs to be refined (5,24,25). Fortunately, new research holds the potential for more precise physiological, neurologic, and psychological markers that will lead to greater diagnostic accuracy and validity.

Two Roads Diverge: Gambling Epidemiology Studies

Overall, population prevalence rates provide a useful measure of what proportion of people at risk for developing a problem have that problem at a given time, how the extent of the problem may have changed over time in that population, and how it compares to the rate of the target disorder in other segments of the population and in other populations. Despite the various methodologies and measures used by researchers in epidemiologic gambling studies, prevalence estimates of level 3 disordered gambling within the general adult population remain fairly stable from study to study, time to time, and place to place. Given the stability of these estimates, researchers now need to study more specific population segments and learn more about the disorder’s onset and etiology. Calculating prevalence rates within population subgroups begins to clarify who is vulnerable to gambling-related problems and allows research to focus on the incidence and course of disordered gambling among these groups: improved understanding of the variation of risk and protective factors across population segments can in turn increase our ability to prevent and treat gambling disorders. The following 2 sections review research on the prevalence of disordered gambling in general adult populations and the prevalence of disordered gambling in specific population segments. The third section addresses the lack of research on the incidence of disordered gambling.

The Road Well Travelled: Population Prevalence Rates
Prevalence research has identified between 0.2% and 2.1% of the world population as past-year pathological gamblers (that is, level 3, according to Shaffer and Hall’s classification system). Despite diverse social settings, these prevalence rates are remarkably similar from study to study and method to method, suggesting that the estimates are robust (26).

US and Canadian Prevalence Rates. In 2 metaanalyses, Shaffer and colleagues analyzed studies of disordered gambling prevalence in the US and Canada, classifying each study’s prevalence estimates by gambling level (that is, level 1 = no gambling-related problems, level 2 = subclincial gambling problems, and level 3 = clinical gambling disorder). Shaffer and colleagues also grouped these estimates as either past-year or lifetime (7,15). By 1999, they observed that the average lifetime prevalence of level 3 gambling was 1.92% and average past-year prevalence was 1.46% (Table 2). Shaffer and colleagues also estimated prevalence rates using other measures of central tendency less sensitive to outliers (for example, the median and a trimmed mean); these estimates were similar to the unweighted means (that is, 1.785% to 1.80% lifetime and 1.20% to 1.27% past-year).

Chronological Changes. Research has begun to address speculation about the impact of the 20th-century expansion of gambling in the US and Canada on the prevalence of gambling-related problems. Opportunities to gamble have certainly increased since the first gambling prevalence study was published in 1979 (12): states have adopted lotteries, legalized casinos, and adapted new electronic technology for gambling. Despite this growth in gambling opportunity, national gambling surveys have measured only a small increase in past-year gambling prevalence, from 61% in 1975 to 63% in 1998 (27,28). Since prevalence studies began, the rate of disordered gambling has increased in the general population (26). The first metaanalysis of prevalence studies, spanning 1975 through mid-1997, identified a positive statistically significant correlation between the year a study was conducted and past-year level 3 gambling prevalence (r = 0.56; 25). An update of Shaffer and colleagues’ original meta-analysis (26) extends and supports this trend through 1999, revealing that general adult population prevalence rates estimates showed a positive statistically significant correlation between the prevalence estimates and the year in which the study was conducted (r = 0.31) (7).

Although a possible interpretation of the increased disordered gambling prevalence rates across time is to conclude a causal relation between gambling exposure and gambling-related problems (29), this is not the only plausible explanation. The effect of gambling exposure on prevalence may hold only for certain segments of the population, may vary nonlinearly over time (owing to processes such as adaptation), or may not be causal at all (5).

Changes in Response to Exposure. Some research has attempted to examine changes in prevalence rates as a response to the introduction of a casino or lottery into a community. Two studies of the effect of Britain’s national lottery on gambling problems found limited evidence for a relation: gambling expenditure and endorsed disordered gambling symptoms increased, but the pathological gambling rate did not change (30,31). In Canada, studies of the impact of casino openings have also produced mixed results: one found increased reported gambling problems after the opening of a Niagara Falls casino (32), one found no such increase after the opening of a casino in Windsor (33), and one found increased reports of knowing someone with a gambling problem but no change in disordered gambling prevalence rates after a casino opened in the Hull area (34).

Shaffer, Vander Bilt, and Hall found that casino employees, a population with direct exposure to gambling, did have higher rates of level 2 and level 3 gambling than the general population (35). Shaffer, LaBrie, and LaPlante similarly showed that Nevada counties with high exposure to gambling had higher rates of problem and pathological gambling than Nevada counties with low exposure (36).

Recent research, however, indicates that any exposure effect that does exist may relate to novelty and therefore be time-limited. For example, Volberg recently noted that several states have reported decreased gambling prevalence and stabilization of problem and pathological gambling rates since casinos were introduced (37). Phase 2 of New Zealand’s national prevalence survey, first conducted in 1991 in response to increased gambling opportunities, found a significant decrease in disordered gambling from 1991 to 1999 (38). Shaffer and Hall also found that the elevated rates of problem and pathological gambling they observed among casino employees actually decreased over time (8). In fact, the same analysis that showed higher disordered gambling rates in more exposed Nevada counties noted that prevalence rates were higher among new Nevada residents and short-term casino employees than in residents and employees who had presumably experienced gambling exposure for longer (36,39). These studies of exposure have mixed findings, but that very mix may signal the beginning of a trend toward adaptation—a nonlinear response to exposure. Studies using more systematic measures of gambling exposure are needed to test the role of exposure and adaptation in determining prevalence and incidence (see 36).

International Prevalence Rates. Recently, researchers have published prevalence estimates of disordered gambling in countries other than the US and Canada. Table 3 summarizes international prevalence estimates for level 2 (that is, subclinical) and level 3 (that is, meeting clinical diagnostic criteria) gambling. In general, it appears that gambling disorders have slightly lower prevalence rates in European countries than in the US. Nevertheless, the rates are remarkably similar, given the range of methods and measures. This consistency across settings establishes level 3 disordered gambling as a unique, detectable phenomenon. Consequently, we encourage the field to study more closely the proportion of the population (that is, 0.2% to 2.1%) who are level 3 gamblers.

The Road Less Frequently Travelled: Population-Segment Prevalence Rates
The era that emphasized general population prevalence rates is drawing to a close. When research shows stable rates of population prevalence for a particular disorder (that is, rates within a range), epidemiologists typically begin to focus on population segments with higher or lower rates of the disorder than the general population (5,40,41). For example, other successful studies of psychiatric epidemiology have taken a similar turn toward the road less travelled:

After identifying illness base rates in the population, the ECA [Epidemiologic Catchment Area] study was to identify high-risk subgroups within the population, those with unusually high rates of illness as well as groups with unusually low rates. In addition to providing profiles of individuals at unusually high risk for developing an illness, such a strategy allows testing of causal hypotheses. While it is unlikely that a single cause of mental disorders will be found, just as no single cause explains cancer or heart disease, the eventual aim of epidemiological research is to identify specific components in a causal chain of factors that produce an illness. For factors, that are amenable to change, direct interventions may be designed to reduce rates of illness in a population (42, p 3).

This practice gives investigators the opportunity to identify high-risk (that is, vulnerable) and low-risk (that is, resilient) population segments. As Table 4 illustrates, epidemiologic research reveals that disordered gambling prevalence varies along many demographic dimensions. In the following discussion, we describe these primary dimensions in more detail.

Age
Adolescence. Many aspects of problem behaviours emerge during adolescence. Compared with adults or those younger, adolescents are more likely to take drugs (43), act delinquently (44), and commit serious crimes (45) (see 46 for a review of this literature). Jessor and Jessor have posited a “problem behaviour syndrome” to explain this multifaceted increase (47). Some researchers (for example, 48) suspect that gambling may be another facet of this syndrome, implying that the prevalence of gambling-related problems in adolescents may be similarly inflated. Indeed, prevalence studies have consistently shown that adolescents evidence higher rates of problem and pathological gambling than adults. Across 32 studies of adolescent problem gambling prevalence, Shaffer and Hall found pathological and problem gambling rates to be almost twice those of adults (7) (Table 4).

Despite higher overall prevalence, during the second half of the 20th century prevalence rates of adolescent gambling disorders have not shown the increase observed in adult rates (7,49). It may be that individuals who engage in underage gambling are not affected by the increased availability of legal gambling because they are engaged in less mainstream forms.

College. College students are another population segment of interest. Alcohol consumption has reached and maintained alarming levels on college campuses—almost one-half of college students can be classified as binge drinkers (50)—and has been a public concern for some time. More recently, public policy and epidemiologic attention has turned toward college gambling. Whether the alcohol phenomenon is an artifact of college culture or an extension of adolescent problem behaviour, many researchers have hypothesized that as gambling opportunities become more available, college students’ gambling habits may mirror their excessive drinking habits (for example, 51). The 19 college studies included in Shaffer and Hall’s metaanalysis confirm these suspicions (see Table 4). Aggregated, these studies yield a 5.56% prevalence rate of level 3 gambling and a 10.88% rate of level 2 gambling among college students—estimates that are even higher than those found among adolescents. More recently, Oster and Knapp observed similarly elevated rates of level 3 gambling in their college student samples (52).

In contrast to these elevated estimates of gambling prevalence, a recent longitudinal study of Missouri college students found that fewer than 1.0% could be classified as pathological gamblers (9). The discrepancies among the rates observed in these studies might be owing in part to the convenience sampling strategy used in most college studies and in part to different time frames (for example, past year or lifetime), instruments, and diagnostic cutpoints. The single study of gambling prevalence among a scientifically selected national sample of college campuses found gambling involvement to be similar to that of adults (53).

The Elderly. Older adults are a population with lower-than-average prevalence of gambling-related problems. They were not considered at risk for gambling problems until the recent expansion of opportunities for older adults to get involved in activities such as casino trips, bingo, and lotteries. This development has led to speculation that the elderly may be a population segment newly vulnerable to gambling problems; however, representative prevalence studies have not yet been conducted. Nevertheless, research has shown that elderly involvement in gambling increased by 45% between 1975 and 1998 (27) and that gambling trips are some of the most frequented activities by senior citizens in retirement centres (54). Research also shows that elderly people who participate in such outings report greater social support in their lives than those who do not (55). More studies are needed to determine the consequences of increased access to gambling opportunities for the elderly.

Sex
Sex differences in both gambling and gambling problems have been well-documented. In general, men gamble more than women and are more likely to have gambling problems (12,29,37) (see Table 4). Across the years, however, prevalence rates of gambling and disordered gambling among women have increased (28), and some researchers have found that men and women have comparable rates of gambling problems (56,57). In addition, women who do gamble appear to develop gambling problems more quickly than men (58). Research has begun to investigate the mechanisms behind these differences, studying the kinds of games men and women play (58,59) and their motivation for playing (57). However, more work is needed to understand the trajectories that lead men and women to gambling-related problems.

Ethnicity
Several research studies have found that ethnic minority groups have a higher prevalence of gambling-related problems and are at greater risk of gambling problems than whites. In a review of studies across 5 states, Volberg found that 80% of the general population sampled were white, but only 64% of the identified pathological gamblers were white (29). Native Americans, specifically, are a population vulnerable to gambling-related problems. In a review of studies of Native American gambling, Wardman and colleagues observed prevalence rates of problem and pathological gambling 5 to 15 times greater than rates in the general population (60) (see Table 4). Further, recent research by Petry and colleagues has found elevated rates of gambling disorders in a sample of 96 Southeast Asian refugees—a level 3 prevalence of 59% (61). Australian research in a convenience sample from a Chinese- speaking community found that 7.8% scored 5 or higher on the SOGS (meeting diagnostic criteria for pathological gambling on that measure) (62). Although the studies do not provide national prevalence estimates for these ethnic groups, the elevated rates in these 2 populations deserve further investigation. It may be that gambling’s integral role in Native American and Asian cultures contributes to the increased prevalence in these groups. By addressing such a hypothesis, researchers can learn more about which groups are at risk and also about the mechanisms behind that risk.

Socioeconomic Status
Studies have identified higher rates of gambling-related problems among samples with lower socioeconomic status (SES). Although individuals from this population segment tend to gamble less money than those in higher SES brackets, their gambling involves a higher proportion of their personal income (63). This disproportionate spending can have important adverse consequences. Welte and colleagues found that 5.6% of participants in the lowest SES bracket of their study could be classified as pathological gamblers, compared with only 1.6% of participants in the highest SES bracket (28). Individuals suffering from disordered gambling are more likely than the general population to be high school dropouts (29) and unemployed (64). Samples of homeless people seeking substance abuse treatment (65) and people receiving community services (66) show higher prevalence rates of disordered gambling than the general population.

Mental Health
Although pathological gambling is listed in the DSM as a distinct disorder, it often cooccurs with other mental disorders. In a seminal review of comorbid conditions and gambling, Crockford and el-Guebaly observed that gamblers in general, and those suffering from disordered gambling in particular, had elevated rates of mental illness (3). Pathological gamblers were 6 times more likely to have antisocial personality disorder than the general population; between 25% and 63% reported having had a substance use disorder at some point (3). Depression and attention deficit disorder are also more common in those suffering from disordered gambling than in the general population (3,67). The relations between disordered gambling and these other disorders suggest common underlying mechanisms (see 68–70) or multiple trajectories by which people can develop disordered gambling.

Taken together, this body of epidemiologic research on special populations reveals that various population segments evidence different prevalence rates of disordered gambling or other gambling-related problems. Why these rates differ is not yet clear, nor are the causal factors that underlie these differences. Conceptual advances are necessary to identify the proximal and distal determinants influencing different prevalence rates across population segments. As new models and theories explain the causal relations among the possible set of determinants, it will become increasingly possible to match prevention and treatment efforts to specific population segments.

The Road Not (Yet) Taken: Incidence Studies
Although prevalence estimates provide measures of the distribution of gambling-related problems within the population, between different populations, and within population subgroups, these estimates reveal little about the onset of problem gambling or how exposure to gambling opportunities influences that onset. Although changes in prevalence rates can and often do reflect changes in incidence, we show in the next section that this is not a necessary relation. Despite the importance of studying incidence as a source of information about the development of gambling problems, few studies include incidence estimates. Those that do exist are all prospective studies; these have found evidence that pathological gambling is an intermittent problem and not the chronic, progressive disease often portrayed (8,9,71). For example, in 1999, Phase 2 of New Zealand’s national gaming survey reinterviewed some participants from the 1991 survey. Three-quarters of 1991 past-year pathological gamblers no longer met the criterion in 1999, and one-half did not even qualify as past-year problem gamblers (38). More prospective studies are needed to provide information about the course and determinants of disordered gambling.

Road Hazards: Current Issues for the Epidemiology of Gambling

Epidemiologic prevalence studies, as described above and elsewhere, have contributed to a fundamental understanding of gambling and gambling-related disorders. These estimates are only the beginning of the story, however. After establishing population-based prevalence estimates for illness, researchers must begin to identify the factors that determine the development of gambling-related problems. Once this more complex research emerges, new and improved primary and secondary prevention efforts can commence. Researchers must also more closely scrutinize the methods and procedures used to derive existing prevalence and risk estimates. This undertaking will allow researchers and treatment providers to identify more accurately the nature and extent of gambling disorders and related problems for specific population segments and geographic regions. In the following section, we review 3 central issues that limit the usefulness of many existing prevalence estimates: the stability of gambling over the lifetime, screening consistency, and diagnostic immaturity.

Individual Variability vs Aggregate Stability
As described previously, regardless of instrumentation, national and international estimates of past-year and lifetime problem and pathological gambling are more often similar than not. This is true even within such special populations as the elderly and youth. At first glance, this suggests that the disorder is chronic, that is, relatively unchanging over time. Closer scrutiny reveals that this is not necessarily accurate. Since Shaffer and colleagues first proposed that gamblers with symptoms were moving in 2 directions—both toward and away from more disordered states—additional evidence has emerged to support this view (15,72). The first prospective, longitudinal research studies demonstrated that people with gambling-related problems move in and out of disordered states much more than was indicated by cross-sectional research (8,9,73). For example, a 3-year follow-up study of casino employees demonstrated that, over time, more employees improved than worsened their gambling status (that is, they showed a less severe level of disorder) (8). Similarly, throughout the 11 years of a longitudinal study with 4 assessment periods, the overall prevalence of participants with at least 1 past-year problem remained steady at 2% to 3%, but different subjects contributed to these rates during the study (9). Within the study sample, young individuals who experienced and subsequently conquered a gambling problem were highly unlikely to report a problem at subsequent follow-ups. In a 1-year follow-up study, Wiebe and colleagues found that more than 50% of Ontario participants initially classified by the Canadian Problem Gambling Index as at-risk gamblers and 26% of participants initially classified as moderate problem gamblers screened as nonproblem gamblers or non- gamblers during follow-up (71).

The stability of prevalence rates results from people moving in and out of disordered gambling levels at approximately the same rate. In other words, aggregate prevalence estimates obscure individual disorder trajectories. This phenomenon unfortunately disguises important changes and trends among individual gamblers and misleads observers into thinking that the disorder is more chronic than it is.

Instrumentation
In addition to obscuring individual trajectories, aggregate prevalence rates obscure the differences among screening instruments. An issue widely ignored by investigators is that different instruments tend to identify different people as having a gambling disorder. This suggests that these measures evaluate different dimensions associated with gambling involvement. Because different screens will yield different diagnoses, researchers, clinicians, and diagnosticians need to understand the various dimensions of disordered gambling that each instrument measures. Lack of clarity can lead to mismatched treatment and, therefore, less-than-optimal treatment outcomes. Instrumentation discrepancy is also a concern for epidemiologists trying to accurately estimate the prevalence of gambling disorders. Case-identification disagreements only become visible when multiple screening instruments are applied to the same group of people.

Interscreen Agreement
Generally, when investigators apply screening instruments to gambling populations with clinical disorders, the different instruments yield high total scores, regardless of the dimensions measured, because this population segment tends to score positively on all or most of the disorder’s dimensions. Consequently, the lack of classification agreement between instruments occurs when scientists apply them to the general population. Stinchfield, for example, compared the SOGS and the DSM-IV diagnostic criteria and reported “good classification accuracy in the gambling treatment sample” but “poorer classification accuracy in the general population . . . with a high false positive rate of 0.50” (17, p 12–13). In a study of the Nevada general population (39), Volberg administered both the SOGS and the National Opinion Research Center (NORC) DSM-IV Screen for Gambling Problems (NODS; 27) and reported a low correlation between total SOGS and NODS scores for the current year time frame (r = 0.47).

Inspection of the cross-classification of Nevada residents by the current-year SOGS and the lifetime NODS (Note 2) suggests that the existing statistical agreement (for example, correlation) between instruments results from the similar classification of people with no problems. Of the 567 survey participants who endorsed no problems on one or the other instrument, both instruments classified two-thirds (65%) as having no problems. Agreement between the instruments was lower for the other reported categories of problem severity: 21% agreement for people with 1 or 2 problems; 11% agreement for those with 3 or 4 problems; and as expected, given the better accuracy of recognizing severe levels of problems, 33% agreement for people with 5 or more problems.

To date, researchers have responded strategically to these problems by simply developing more instruments. More than 30 screening tools are now available to assess gambling-related problems. So far, this proliferation of instruments has done more to confirm the problem than to rectify it. Researchers ought instead to consider defining the dimensional structure of their screens so that psychometricians can develop a comprehensive universal screen.

Dimensions of Disordered Gambling: An Illustration
Inconsistency among measured dimensions is problematic for many reasons and likely contributes to interscreen discrepancy. Although a comprehensive illustration of this problem is beyond the scope of this paper, we will digress briefly to present a hypothetical example. Let us start by assuming that the concept of disordered gambling is multidimensional and accept the following premises.

Premise 1. Disordered Gambling Has 5 Dimensions (that is, A, B, C, D, and E). When researchers constructed their screening instruments, they identified and consequently measured different dimensions. Given that the number of dimensions is not infinite and researchers were studying roughly the same phenomenon, we also assume that there is some commonality in the dimensions that different screens measure.

For this discussion, consider 2 instruments that 1) measure 1 common dimension and 2) evaluate 2 dimensions unique to each instrument. For this assumption, let us accept the following premise.

Premise 2. Test 1 Measures Dimensions A, B, and C; Test 2 Measures Dimensions A, D, and E. Let us also assume that each dimension is independent and equally representative of a gambling disorder (Note 3). Since most measurement devices score by summing the number of items endorsed and use cut-offs for determining the extent of problems, we will use a hypothetical scoring level of multiple items endorsed from 2 or more dimensions (that is, positive responses to 2 or more items within each of 2 dimensions) to indicate the presence of an underlying disorder (specifically, level 3 gambling).

For Test 1 to identify an individual as having level 3 problems, the individual would need to exhibit one of the following endorsement patterns: ABC, AB, AC, or BC. For Test 2 to identify an individual as having level 3 problems, the individual would need to show one of the following endorsement patterns: ADE, AD, AE, or DE. There are 10 opportunities for common identification across the 2 screening tests. Of these 10, 5 combinations include the common dimension A and would classify the same subjects into the highest level of gambling severity (Table 5). Consequently, the likelihood of randomly assigning the same people with gambling problems to level 3 is about 50% under the assumptions of independence and equal probability of symptom occurrence.

The Nevada study confirms this illustration. In it, 40% of the participants identified by the NODS or SOGS as having gambling problems were classified as level 3 by both instruments (39).

This exercise can be repeated for any level of gambling disorder. Even less congruency is obtained at lower levels of gambling disorder because fewer endorsed items are needed. Because the levels of gambling disorder also have different probabilities of being observed in the public, the extent of congruent classification depends on the relative frequency of the levels. For example, adjusting the random agreement in level 3 classification (P = 0.50) by the prevalence of level 3 gambling in gamblers who endorse at least 1 problem (that is, 15%, as recently provided in the SOGS estimates in the Nevada household population study; 39) yields an agreement of 8%.

Consequences of Incomplete Dimension Coverage
The hypothetical scenario described above reasonably replicates the results of comparative studies of existing gambling screening instruments, because each of these tests incompletely covers the dimensions underlying gambling disorders. This lack of completeness might be traced to the populations (for example, gamblers in treatment, members of Gamblers Anonymous, or general household populations) and purposes (for example, screening for risk, diagnosis, and prevalence) influencing the development of these instruments; it might also be traced to limited awareness of the dimensions that a screen actually assesses. As long as clinicians and epidemiologic screens fail to identify the full set of dimensions underlying gambling disorders, incongruent classification will continue to create diagnostic problems and compromise the accuracy of population-prevalence estimates.

The scenario above also replicates the results of existing psychometric research: the unidimensional additive scoring of screening instruments is inadequate to represent a multi- dimensional latent state. The method of summing endorsed characteristics assumes that all dimensions exist on the same additive continuum and that all dimensions equally predict gambling disorders (Note 4). This means that certain patterns of gambling (for example, chasing), consequences of gambling (for example, stealing to pay off gambling debts), classic symptoms (for example, withdrawal), and even relatively benign experiences (for example, losing more than intended) all apply psychometric force in the same direction. This equivalence is highly unlikely and misleading.

Diagnostic Immaturity
The instrumentation problem suggests that different instruments ignore multiple dimensions of gambling disorders, because each instrument tends to identify preferred aspects of the disorder and to ignore other characteristics. However, gambling instrumentation faces other problems as well. Most are due to the recent development of many screens and the scientific youth of the field in general. The following discussion considers 3 fundamental instrumentation problems associated with estimating gambling disorder prevalence: the arbitrariness of the dimensions that screening instruments measure, the utility of different self-report time frames, and general problems associated with self-report.

To start, the dimensions different screens measure and the ways in which they measure them (for example, by assessing behaviour or attitudes or cognitions) are often arbitrary. However, little to no research has examined the relative importance of different dimensions and the validity of different methods of assessing those dimensions. This situation contributes to interscreen discrepancy for reasons similar to those described above.

Different screens use different time frames for self-report. The most common time frames are the past 6 months, past year, and lifetime. Lifetime frames are problematic, because assessment screens do not consider in their calculations the time period for which individuals confirm diagnostic criteria; self-report of the past 6 months and past year probably do not run this risk. In the SOGS, for example, persons aged 60 years might confirm 5 items over the course of their life and be categorized as probable pathological gamblers; however, each confirmed criterion could technically have occurred 10 years apart. Contrast this with the individual who confirms 5 criteria in the past month. The extent of the first individual’s gambling problem is clearly an overestimate, because time is not weighted in any way. As currently used in gambling studies, lifetime self-reports are not necessarily stable. For example, Abbott found that only one-quarter of a New Zealand sample that met criteria for a lifetime diagnosis of pathological gambling met those same criteria 7 years later (38). The host of problems to which lifetime reports are vulnerable, such as memory fatigue and social desirability, reduces the importance of lifetime estimates. As Walker and Dickerson note, “If lifetime prevalence has a valid meaning, then it refers to the occurrence of some characteristic that, if present at all, is present for life” (74, p 239).

Finally, all gambling screens suffer from the common burdens of self-report (such as self-presentation bias or faulty memory). The accuracy of self-reported behaviours is questionable; as well, self-report is scientifically and clinically undesirable, because it forces individuals to virtually self-diagnose. Individual self-reports of problems frequently contradict assessments gleaned from psychiatric measures (75,76). Further, the self-report of past behaviours and attitudes is problematic because the strategy uses the disorder’s consequences to identify its existence. This circular reasoning can lead to overestimating gambling disorders: not everyone with a problem has an underlying psychological disorder. More time and research should be spent on developing advanced, unobtrusive diagnostic tools that do not rely on self-reported behaviours and attitudes. Some of these tools might include functional magnetic resonance imaging, event-related potential, neurochemical analysis, psycho- physiological assessment, and reaction time measures.

The Road Ahead: Conclusions and Future Directions

The epidemiologic study of gambling has reached a crossroads. Globally, research on the distribution of gambling has revealed an apparently robust and reliable phenomenon among adult populations. In this article, we suggest that prevalence studies of the general population based on existing instrumentation have matured sufficiently for this strategy to be no longer fruitful. General-population incidence studies remain rare and should provide an important source of information about the nature of gambling-related problems, as will general-population prevalence studies that use more refined measures sensitive to the influence of new cases, relapses, and recoveries on prevalence rate fluctuations. However, as with more mature areas of epidemiologic inquiry (for example, those that have an incidence literature), the next path for the epidemiologic study of gambling should include segments of the population vulnerable to the development of gambling disorders. This turn in the road represents a strategic shift away from studies of gambling distribution and toward studies of gambling determinants. To adequately study the determinants of disordered gambling, scientists must improve theoretical models that identify causal paths and clarify the dimensions that underlie gambling-related problems. These advances will permit—perhaps even encourage—the development of better psychometric tools. Taken together, these changes will require the implementation of prospective incidence studies of vulnerable population segments. The road ahead is demanding and filled with difficult hazards. However, journeying the road less travelled encourages the epidemiology of gambling disorders to become more meaningful for prevention and treatment efforts. Finally, once scientists and clinicians distinguish and implement influential prevention and treatment protocols for vulnerable populations, the road less travelled and the road well travelled will converge. These 2 strategic paths should inform one another and encourage a recursive and integrated public health research strategy (41). As both roads become well travelled by repeated investigative cycles of whole populations and vulnerable population segments, scientists and clinicians will gain a more precise definition and comprehensive understanding of the development of gambling, as well as its potential positive and negative consequences.

Notes

1. Shaffer and Hall’s system also defines level 4 gamblers as those who seek treatment for gambling-related problems (23). Typically, treatment seekers emerge from level 3 gamblers; however, level 2, and even level 1, gamblers will occasionally seek assistance. Level 1 gamblers seeking treatment can be thought of as having pseudoaddiction (77).

2. Volberg (39) did not report the cross-classification results between current NODS and current SOGS; this finding emerged as part of a secondary data analysis performed for this review.

3. Although it is now the screening convention to have every dimension reflect equal weight, this is of course unreasonable. For example, it is highly unlikely that anyone could be diagnosed as a pathological gambler if they were not preoccupied with gambling. The current DSM classification system, however, permits such a circumstance.

4. One exception to this blind scoring method is the Massachusetts Adolescent Gambling Screen (72).

Funding and Support

This work was supported in part by funding from the National Center for Responsible Gaming and the Iowa Department of Public Health.

Acknowledgements

The authors extend special thanks to Gabriel Caro, Tony Donato, Rachel Kidman, David Korn, Christine Reilly, and Christine Thurmond for their contributions and advice regarding this manuscript.

References

1. National Gambling Impact Study Commission. National Gambling Impact Study Commission report. Washington (DC): National Gambling Impact Study Commission; 1999.

2. National Research Council. Pathological gambling: a critical review. Washington (DC): National Academy Press; 1999.

3. Crockford DN, el-Guebaly N. Psychiatric comorbidity in pathological gambling: a critical review. Can J Psychiatry 1998;43:43–50.

4. Cunningham-Williams RM, Cottler LB, Compton WM, Spitznagel EL, Ben-Abdallah A. Problem gambling and comorbid psychiatric and substance use disorders among drug users recruited from drug treatment and community settings. J Gambl Stud 2000;16:347–76.

5. Shaffer HJ, Korn DA. Gambling and related mental disorders: a public health analysis. Annu Rev Public Health 2002;23:171–212.

6. Eber GB, Shaffer HJ. Trends in bio-behavioral gambling studies research: quantifying citations. J Gambl Stud 2000;16:461–7.

7. Shaffer HJ, Hall MN. Updating and refining prevalence estimates of disordered gambling behaviour in the United States and Canada. Can J Public Health 2001;92:68–172.

8. Shaffer HJ, Hall MN. The natural history of gambling and drinking problems among casino employees. J Soc Psychol 2002;142:405–24.

9. Slutske WS, Jackson KM, Sher KJ. The natural history of problem gambling from age 18 to 29. J Abnorm Psychol 2003;112:263–74.

10. American Psychiatric Association. DSM-III: diagnostic and statistical manual of mental disorders. 3rd ed. Washington (DC): American Psychiatric Association; 1980.

11. American Psychiatric Association. DSM-IV: diagnostic and statistical manual of mental disorders. 4th ed. Text revision. Washington (DC): American Psychiatric Association; 2000.

12. Kallick M, Suits D, Dielman T, Hybels J. A survey of American gambling attitudes and behavior (Research report series, Survey Research Center, Institute for Social Research). Ann Arbor (MI): University of Michigan Press; 1979.

13. Lesieur HR, Blume, SB. The South Oaks Gambling Screen (SOGS): a new instrument for the identification of pathological gamblers. Am J Psychiatry 1987;144:1184–8.

14. American Psychiatric Association. DSM III-R: diagnostic and statistical manual of mental disorders. 3rd ed. Revised. Washington (DC): American Psychiatric Association; 1987.

15. Shaffer HJ, Hall MN, Vander Bilt J. Estimating the prevalence of disordered gambling behavior in the United States and Canada: a meta-analysis. Boston (MA): Presidents and Fellows of Harvard College; 1997.

16. Derevensky JL, Gupta R. Prevalence estimates of adolescent gambling: a comparison of the SOGS-RA, DSM-IV-J, and the GA 20 questions. J Gambl Stud 2000;16:227–51.

17. Stinchfield R. Reliability, validity, and classification accuracy of the South Oaks Gambling Screen (SOGS). Addictive Behaviors 2002;27:1–19.

18. Shaffer HJ, Robbins M. Manufacturing multiple meanings of addiction: time-limited realities. Contemporary Family Therapy 1991;13:387–404.

19. Szasz T. Diagnoses are not diseases. Lancet 1991;338:1574–6.

20. Brownson RC, Newschaffer CJ, Ali-Abarghoui F. Policy research for disease prevention: challenges and practical recommendations. Am J Public Health 1997;87:735–9.

21. Ratey JJ, Johnson C. Shadow syndromes. New York: Pantheon Books; 1997.

22. Slutske WS, Eisen S, True WR, Lyons MJ, Goldberg J, Tsuang M. Common genetic vulnerability for pathological gambling and alcohol dependence in men. Arch Gen Psychiatry 2000;57:666–73.

23. Shaffer HJ, Hall MN. Estimating the prevalence of adolescent gambling disorders: a quantitative synthesis and guide toward standard gambling nomenclature. J Gambl Stud 1996;12:193–214.

24. Kleinman A. Culture and clinical reality: commentary on culture-bound syndromes and international disease classifications. Cult Med Psychiatry 1987;11:49–52.

25. Shaffer HJ. Assessment of addictive disorders: the use of clinical reflection and hypotheses testing. Psychiatr Clin North Am 1986;9:385–98.

26. Shaffer HJ, Hall MN, Vander Bilt J. Estimating the prevalence of disordered gambling behavior in the United States and Canada: a research synthesis. Am J Public Health 1999;89:1369–76.

27. Gerstein D, Murphy S, Toce M, Hoffmann J, Palmer A, Johnson R, and others. Gambling impact and behavior study: Report to the National Gambling Impact Study Commission. Chicago (IL): National Opinion Research Center; 1999.

28. Welte JW, Barnes GM, Wieczorek WF, Tidwell M-C, Parke, J. Gambling participation in the US results from a national survey. J Gambl Stud 2002;18:313–37.

29. Volberg RA. The prevalence and demographics of pathological gamblers: implications for public health. Am J Public Health 1994;84:237–41.

30. Grun L, McKeigue P. Prevalence of excessive gambling before and after introduction of a national lottery in the United Kingdom: another example of the single distribution theory. Addiction 2000;95:959–66.

31. Shepherd R-M, Ghodse H, London M. A pilot study examining gambling behaviour before and after the launch of the National Lottery and scratch cards in the UK. Addiction Research 1998;6:5–12.

32. Room R, Turner NE, Ialomiteanu A. Community effects of the opening of the Niagara casino. Addiction 1999;94:1449–66.

33. Govoni R, Frisch G, Rupcich N, Getty H. First year impacts of casino gambling in a community. J Gambl Stud 1998;12:347–58.

34. Jacques C, Ladouceur R, Ferland F. Impact of availability on gambling: a longitudinal study. Can J Psychiatry 2000;45:810–5.

35. Shaffer HJ, Vander Bilt J, Hall MN. Gambling, drinking, smoking and other health risk activities among casino employees. Am J Ind Med 1999;36:365–78.

36. Shaffer HJ, LaBrie RA, LaPlante DA. Laying the foundation for quantifying regional exposure to social phenomena: considering the case of legalized gambling as a public health toxin. Psychol Addict Behav 2004;18:40–8.

37. Volberg RA. The epidemiology of pathological gambling. Psychiatr Ann 2002;32:171–8.

38. Abbott MW. What do we know about gambling and problem gambling in New Zealand? New Zealand Gaming Survey Report Number 7. Wellington (NZ): New Zealand Department of Internal Affairs; 2001.

39. Volberg RA. Gambling and problem gambling in Nevada. Northampton (MA): Gemini Research Ltd; 2002.

40. Fisher S. Measuring the prevalence of sector-specific problem gambling: a study of casino patrons. J Gambl Stud 2000;16:25–51.

41. Shaffer HJ, Kidman R. Conceptualizing gambling as a public health issue. In: Grant JE, Potenza MN, editors. Understanding and treating pathological gambling. Washington (DC): American Psychiatric Publishing. Forthcoming.

42. Regier DA, Robins LN. Introduction. In: Regier DA, Robins LN, editors. Psychiatric disorders in America: the Epidemiologic Catchment Area Study. New York: The Free Press; 1991. p 1–10.

43. Johnston LD, O’Malley PM, Bachman JG. National survey results on drug use from the Monitoring the Future Study, 1975 1992. Rockville (MD): National Institute on Drug Abuse; 1993.

44. Moffitt TE. Adolescence-limited and life-course-persistent antisocial behavior: a developmental taxonomy. Psychol Rev 1993;100:674–701.

45. Federal Bureau of Investigation. Uniform crime reports for the United States. Washington (DC): US Department of Justice; 2001.

46. Siegel AW, Scovill LC. Problem behavior: the double symptom of adolescence. Dev Psychopathol 2000;12:763–93.

47. Jessor R, Jessor SL. Problem behavior and psychosocial development: a longitudinal study of youth. New York: Academic Press; 1977.

48. Vitaro F, Brendgen M, Ladouceur R, Tremblay RE. Gambling, delinquency, and drug use during adolescence: mutual influences and common risk factors. J Gambl Stud 2001;17:171–90.

49. Stinchfield R. Youth gambling: how big a problem? Psychiatr Ann 2002;32:197–202.

50. Wechsler H, Davenport A, Dowdall G, Moeykens B, Castillo S. Health and behavioral consequences of binge drinking in college: a national survey of students at 140 campuses. JAMA 1994;272:1672–7.

51. Winters KC, Stinchfield R, Botzet A, Anderson N. A prospective study of youth gambling behaviors. Psychol Addict Behav 2002;16:3–9.

52. Oster SL, Knapp TJ. Underage and pathological gambling by college students: emerging problem on campus? Psychology and Education: An Interdisciplinary Journal 2001;38:15–9.

53. LaBrie R, Shaffer HJ, LaPlante DA, Wechsler H. Correlates of college student gambling in the United States. J Am College Health 2003;52:53–62.

54. McNeilly DP, Burke WJ. Disposable time and disposable income: problem casino gambling behavior in older adults. Journal of Clinical Geropsychology 2002;8:75–85.

55. Vander Bilt J, Dodge H, Pandav R, Shaffer HJ, Ganguli M. Gambling participation and social support among older adults: a longitudinal community study. J Gambl Stud. Forthcoming.

56. Hing N, Breen H. Profiling Lady Luck: an empirical study of gambling and problem gambling amongst female club members. J Gambl Stud 2001;17:47–69.

57. Hraba J, Lee G. Gender, gambling and problem gambling. J Gambl Stud 1996;12:83–101.

58. Potenza MN, Steinberg MA, McLaughlin SD, Wu R, Rounsaville BJ, O’Malley SS. Gender-related differences in the characteristics of problem gamblers using a gambling helpline. Am J Psychiatry 2001;158:1500–5.

59. Ladd GT, Petry NM. Gender differences among pathological gamblers seeking treatment. Exp Clin Psychopharmacology, 2002;10:302–9.

60. Wardman D, el-Guebaly N, Hodgins D. Problem and pathological gambling in North American Aboriginal populations: a review of the empirical literature. J Gambl Stud 2001;17:81–100.

61. Petry NM, Armentano C, Kuoch T, Norinth, T, Smith L. Gambling participation and problems among South East Asian refugees to the United States. Psychiatr Serv 2003;54:1142–8.

62. Blaszczynski A, Huynh S, Dumlao V, Farrell E. Problem gambling within a Chinese speaking community. J Gambl Stud 1998;14:359–80.

63. Shaffer HJ. A public health perspective on gambling: the four principles. AGA Responsible Gaming Lecture Series. American Gaming Association; 2003;2:1–27. Available: www.americangaming.org/programs/responsiblegaming/rg_lectures.cfm.

64. Cox BJ, Kwong J, Michaud V, Enns MW. Problem and probable pathological gambling: considerations from a community survey. Can J Psychiatry 2000;45:548–53.

65. Shaffer HJ, Freed CR, Healea D. Gambling disorders among homeless persons with substance use disorders seeking treatment at a community center. Psychiatr Serv 2002;55:1112–7.

66. Lepage C, Ladouceur R, Jacques C. Prevalence of problem gambling among community service users. Community Ment Health J 2000;36:597–601.

67. Rugle L, Melamed L. Neuropsychological assessment of attention problems in pathological gamblers. J Nerv Ment Dis 1993;18:107–12.

68. Bergh C, Eklund T, Södersten P, Nordin C. Altered dopamine function in pathological gambling. Psychol Med 1997;27:473–5.

69. Blum K, Cull JG, Braverman ER, Comings DE. Reward deficiency syndrome. American Scientist 1996;84:132–45.

70. Breiter HC, Aharon I, Kahneman D, Dale A, Shizgal P. Functional imaging of neural responses to expectancy and experience of monetary gains and losses. Neuron 2001;30:619–39.

71. Wiebe J, Single E, Falkowski-Ham A. Exploring the evolution of problem gambling: a one year follow-up study. Toronto: Responsible Gambling Council; 2003. Available: www.responsiblegambling.org. Accessed December 1, 2003.

72. Shaffer HJ, LaBrie R, Scanlan KM, Cummings TN. Pathological gambling among adolescents: Massachusetts Gambling Screen (MAGS). J Gambl Stud 1994;10:339–62.

73. Stinchfield R, Winters KC. Outcome of Minnesota’s gambling treatment programs. J Gambl Stud 2001;17:217–45.

74. Walker MB, Dickerson MG. The prevalence of problem and pathological gambling: a critical analysis. J Gambl Stud 1996;12:233–49.

75. Hardoon K, Derevensky JL, Gupta R. Empirical measures vs. perceived gambling severity among youth: why adolescent problem gamblers fail to seek treatment. Addictive Behaviors 2003;28:933–46.

76. Narrow WE, Rae DS, Robins LN, Regier DA. Revised prevalence estimates of mental disorders in the United States. Arch Gen Psychiatry 2002;59:115–23.

77. Zinberg NE, Lewis DC. Narcotic usage: I. A spectrum of a difficult medical problem. N Engl J Med 1964;270:989–93.

78. Volberg RA, Abbott MW, Roennberg S, Munck IM. Prevalence and risks of pathological gambling in Sweden. Acta Psychiatr Scand 2001;104:250–6.

79. Bondolfi G, Osiek C, Ferrero F. Pathological gambling: an increasing and underestimated disorder. Schweizer Arch Neurol Psychiatrie 2002;153:116–22.

80. Abbott MW. Problem and non-problem gamblers in New Zealand; a report on phase two of the 1999 National Prevalence Survey. New Zealand Gaming Survey Report Number 6. Wellington (NZ): New Zealand Department of Internal Affairs; 2001.

81. Sproston K, Erens B, Orford J. Gambling behaviour in Britain: results from the British Gambling Prevalence Survey. London (UK): National Centre for Social Research; 2000.

82. Collins P, Barr G. Gambling and problem gambling in South Africa: a national study. National Center for the Study of Gambling; 2001. Available: http://www.responsiblegaming.co.za/gamblingreport_pretext%20design.pdf. Accessed 2003 Dec 1.

83. Wong ILK, So EMT. Prevalence estimates of problem and pathological gambling in Hong Kong. Am J Psychiatry 2003;160:1353–4.

84. Becona E. Prevalence surveys of problem and pathological gambling in Europe: the cases of Germany, Holland and Spain. J Gambl Stud 1996;12:179–92.

85. Gotestam KG, Johansson A. Characteristics of gambling and problematic gambling in the Norwegian context: a DSM-IV based telephone interview study. Addictive Behaviors 2003;28:189–97.

86. Productivity Commission. Australia’s gambling industries: final report (number 10). Canberra: AusInfo; 1999.

87. Volberg RA, Abbott M. Ethnicity and gambling: gambling and problem gambling among indigenous peoples. Subst Use Misuse 1997;32:1525–38.

Author(s)

Manuscript received and accepted February 2004.

1. Associate Professor of Psychology, Department of Psychiatry, The Cambridge Health Alliance; Director, Division on Addictions, Harvard Medical School, Boston, Massachusetts.

2. Associate Director for Research and Data Analysis, Division on Addictions, Harvard Medical School, Boston, Massachusetts.

3. Instructor in Psychiatry, The Cambridge Health Alliance and the Division on Addictions, Harvard Medical School, Boston, Massachusetts.

4. Technical Research Assistant, Division on Addictions, Harvard Medical School, Boston, Massachusetts.

Address for correspondence: Dr HJ Shaffer, Division on Addictions, Harvard Medical School, The Landmark Center, 401 Park Drive, Second Floor East, Boston, MA 02215

e-mail: Howard_Shaffer@hms.harvard.edu

1 | 2