On 24 April 2014 I sent to the Director for Policy, Development and External Affairs of the NYS Gaming Commission the following e-note with two attachments:
Attached are two documents I earnestly hope the Chairman and all the Commissioners will read carefully and discuss with the GFLB. Both are about “problem gambling,” the subject of the April 9 forum convened by the Gaming Commission. Watching the videotape and reading the transcript (everyone should thank the GC for providing these so fast) I saw that “problem gambling” was an elusive term. The extreme importance to the casino economy of net losses from problem gamblers was nowhere mentioned except when the speaker from Caesars deflected the issue. Yet around the “central statistic of casino revenues,” on which I have written to the Commission, is the “central dilemma” of regulation: the better the regulation is at preventing problem gambling, the lower is the casinos’ profit margin.
Selected prevalence statistics were presented as if they are the be-all and the end-all of gambling behavior studies. They are about all we have, but a poor stand-in for what we really want to know about time trends in social impacts, i.e. incidence and duration. Under the placid surface of what looks like stable prevalence, much new damage continues; as problem gamblers recover or die, new ones must be recruited to take their places.
As I have offered before, I am[ready] almost any time to meet with the Commissioners and staff to explain the critiques in more detail and to talk about “the central statistic.”
Thank you for your attention.
Stephen Q. Shafer, MD, MA, MPH Chairperson, Coalition Against Gambling in New York 917 453 7371 http://cagnyinf.org
To no one’s surprise, the Commission has not responded to my unsolicited comments. Does that mean the Commissioners have all accepted the ” adaptation hypothesis” [summarized in the next paragraph] that gambling expansion has little long-term population impact? If they have, ” this was a grievous fault,” but unless NYS media call them out sharply on it, the Commission will not have to answer it grievously. If, as I hope, they have not bought it, one sign will be that the Facilities Location Board questions searchingly all applicants for a casino license on what they will really do that will really stop new cases of problem gambling.
Gentlemen of the Gaming Commission, how say you?
[Scientists from the Division on Addictions of the Cambridge Health Alliance have proposed an “adaptation hypothesis,” which acknowledges that new gambling opportunities may lead to a temporary increase in prevalence of problem gambling for the surrounding population. Then, so goes the hypothesis, novelty fades, individuals become more “responsible” in their gambling behaviors and the crest subsides.]
One of the two attachments, slightly revised a month later, is below. It begins with Summary. It concerns the public health profile of “problem gambling.” The other, related to an operational definition of “problem gamblers” and to their fiscal significance, will soon be posted on the CAGNY web site.
Summary: “Problem gambling” is not a fixed uniform term. In his introduction to the April 9 forum on Problem Gambling, Dr. Gearan seems to take it, as I do, to mean both strata of gambling disorder combined, not just the less severe stratum often referred to as “problem gambling.” Statistics on the prevalence of “problem gambling” or its subgroup “pathological gambling” are often used to reassure policy-makers that gambling expansion has not worsened problem or pathological gambling. These statistics can be challenged on several grounds, but even if prevalence as a proportion of the population is truly unchanged in the long term, there are hundreds of thousands of new cases nationally hidden in it, millions of individuals affected. To keep prevalence stable there have to be new pathological gamblers brought on to take the place of those who died or entered recovery. Opponents of big tobacco use the term “replacement smokers.” We extend this concept to “replacement” problem gamblers. Tobacco companies need replacement smokers; society does not. Casinos need replacement problem gamblers; society does not. Quite the reverse.
Introduction In the April 9th 2014 Problem Gambling forum hosted by the New York State Gaming Commission two speakers, Dr. Sarah Nelson and Ms Christine Reilly, used lack of change over time in “the prevalence of problem gambling” to argue that expanded gambling opportunities do not increase endemic harms to public health. I will challenge that premise as an epidemiologist by digging into the population dynamics of “prevalence.” With problem gambling, as with most chronic conditions, prevalence is more easily measured than the rate at which new cases develop. That does not mean it’s a perfectly valid marker of causation. It hides new cases, the replacement problem gamblers that predatory gambling must endlessly cultivate to keep profits high. To speak of the prevalence of “problem gambling” requires first defining the latter term. In the second and third sections of this narrative I will talk about the pitfalls of prevalence to assess “problem gambling” in a population.
Section 1. Unofficial glossary of problem gambling
In the April 9th forum, speakers were not unified on the definition of “Problem Gambling.” Commission Chairman Gearan said in his introduction “The act [Upstate Gaming Act of 2013] requires the state and commercial casinos to develop programs to combat compulsive – this is from the legislation — to combat compulsive and problem gambling. That is a quote from the act.” The breaking-out of “compulsive” [or a near-equivalent like “pathological”] from “problem” is widespread usage. It can cause confusion, however, when the two words “problem gambling” appear alone as they do in paragraphs just above and below Dr. Gearan’s citation from the Upstate Gaming Act, e.g. “the complex issues of problem gambling.” Dr. Gearan cannot mean that issues of “compulsive gambling” are not “complex.” He has to be, like almost everyone else in the field (including this writer) combining a term like “compulsive gambling” with “problem gambling” into one larger category also called (to universal confusion) “problem gambling.”
Thus we have “problem gambling” as a subset of “problem gambling.” Unsatisfactory, but no one party is to blame. This commentary will not resolve it. The terms have been in flux for several decades. I highlight the confusion to alert the commissioners to it. “Problem gambling” is a shape-shifter throughout this forum and throughout the literature. The next few paragraphs lay out the terms such as “problem gambler” in common use now to categorize those gambling behaviors that in a small minority of the population harm the individual (and persons around him or her).
Below is a draft field guide to terms used in studies of gambling behavior. The prevalence estimates are rounded off. They will be disputed by some but make clear the relative size of the two subsets that comprise what we think forum speakers generally meant by “problem gambling, ” viz. what was formerly known as “disordered gambling.” That is the combination of “compulsive and problem gambling” or, in other terms, the combination of level 3 plus level 2. For the most damaging level of gambling (aka level 3) we at Coalition Against Gambling in New York (CAGNY) prefer “gambling addiction” to “disordered gambling” but are willing to make a gradual transition to the latter term.
Older terms ~ Pop. prevalence DSM V term CAGNY term
Level 2 2-3% none At risk gambling
aka Sub-clinical pathological
Level 3 1% Disordered prefer “Addicted”
aka Pathological can use Disordered
Disordered 3-4% none Problem
We believe that “problem gambling” should be used in future for the super-category that combines what used to be called “level 2 and level 3” or what Chairman Gearan termed “compulsive and problem gambling.” Analysts of gambling socio-economics right now lack a universally-recognized term for the whole spectrum of adverse effects of gambling. “Problem gambling” is the best choice as long as it subsumes what is now called “disordered gambling” (also known as pathological, compulsive, addicted or level 3).
The next four paragraphs are background to the table above and can be skipped once the table is in mind.
Most American adults gamble not at all (20-30%) or so little they get no damage from it. The huge quantifiable socio-economic costs of legalized gambling, estimated by Grinols at 60 billion dollars (2012$) move almost entirely through current “pathological gamblers” (near-synonyms are “addicted,” “compulsive” or “level 3”) with their families, employers, employees and associates drawn in, and through current “problem gamblers.” Gamblers in the latter category, sometimes denoted “level 2,” are more common but impose each a smaller quantifiable socioeconomic cost than those classified as “pathological” or like terms. “Pathological gambling” has recently been re-termed by the American Psychiatric Association “disordered gambling.” I don’t know how readily this will be adopted by researchers on gambling.
Neither Diagnostic and Statistical Manual IV nor DSM V (2013) have a term for what used to be “level 2” (also known as “problem gambling”). My colleague Arnold Lieber, M.D. has suggested “at risk gambling, ” though little is known about the degree of risk for moving from there into the more severe category now called “disordered.” Coalition Against Gambling in New York proposes the “at risk” term for lack of a better. We believe that “problem gambling” should be used in future for the super-category that combines what used to be called “level 2 and level 3” or in the Upstate Gaming Act of 2013 is termed “compulsive and problem gambling.” Analysts of gambling socio-economics right now lack a universally-recognized term for the whole spectrum of adverse effects of gambling. “Problem gambling” is the best choice as long as it subsumes what is now called “disordered gambling” (also known as pathological, compulsive, addicted or level 3). “Disordered gambling” for the combined group had adherents, but can no longer work now that this adjective has been transferred to the erstwhile level 3. “Level 2” and “level 3” were terse and distinct but too bland to work in 21st-century nosology.
“Disordered gambling” has in Diagnostic and Statistical Manual V replaced “Pathological gambling” (formerly equivalent to Level III). Regrettably, DSM V does not have a term for what Dr. Howard Shaffer and colleagues used to call “sub-clinical pathological gambling (i.e. problem gambling.” ) Shaffer, H. & Martin, R., Disordered Gambling: Etiology, Trajectory, and Clinical Considerations, Annual Review of Clinical Psychology, Vol. 7, pp. 483-510 (2011). This creates a serious danger that in the near future “problem gambling” will be equated with “disordered gambling,” a less common but even more serious condition. This mistake would greatly set back efforts to get gamblers into lasting recovery before they have reached the nadir of what is to be called henceforth (euphemistically) “disordered gambling.” Not all gamblers in what used to be called level 2 will get worse and enter level 3, but the risk is appreciable.
Pathological (“disordered”) gamblers make up conservatively 1.14% of adults in North America; problem gamblers (“at risk of disordered”), at least 2.8%. 1 In the U.S. that would be now 2.6 million pathological gamblers (“disordered”) and 6.5 million persons at risk of moving into “disordered.” Summed, that rounds to nine million problem gamblers. Lower estimates for prevalence can be found. 2,3 Coalition Against Gambling in New York uses 4% for the two categories combined based on the Shaffer et al 1999 meta-analysis, 4 which drew on findings from scores of reports studies.
Section 2 What prevalence tells and doesn’t tell
Prevalence is usually given as the proportion (can be %) of a population that has the condition of interest (in this case a certain level of gambling) at a given moment or at some point during a time interval (e.g. past twelve months, past one month, on interview day). Prevalence can also be a count. It is not the same as incidence. Incidence is a rate, the number of new cases in a time interval divided by the number at risk. It too can be correctly given as a count of new cases but still must have a denominator per unit of time.
Prevalence is not a good marker of the rate at which new cases occur. If “the prevalence of compulsive (aka pathological or disordered or level 3 or addicted) gambling” seems stable over time, that does not mean that people are no longer becoming gamblers in his category. Likewise for “problem gamblers.” To maintain the prevalence of current pathological gambling, replacement pathological gamblers must be developed to take the place of those who have recovered or died or otherwise disappeared from the population. The goal of public health is to prevent new cases. Prevalence tells little about progress on that front.
Prevalence depends on both incidence and duration. Not all cases prevalent at a certain time or during a certain year will be the same persons as at a different time. Individuals leave the current or “active” (in gambling studies usually = “past-year”) prevalence pool by recovery, moving far away, death, incarceration, or disabling illness. They can also move from the prevalence pool of one category into another, for better or for worse. New cases enter the pool in appropriate categories. Even if prevalence in a test-retest pair of extraordinarily accurate studies was at a plateau over (say) ten years, that stability required replacement pathological gamblers in the stead of those who during that decade died; or recovered (one estimate is that 1/3 recover 5); or went to jail (more than half of pathological gamblers commit prosecutable crimes); or moved to another country.
Different from current (past year) prevalence is lifetime prevalence. In the context of, let’s say, problem gambling, “lifetime” is misleading. It connotes lifelong, but really means “at any point in life” [could be years ago]. The lifetime prevalence pool is depleted only by death or out-migration. Problem gamblers whose recovery began more than a year ago are no longer past-year problem gamblers but will always be “lifetime” in that category.
Example: using past year prevalence for levels 2 and 3 from the Shaffer et al meta-analysis of 1999 (see below) we picture in a population of 100,000 adults that there are 3940 (= 100000 * (0.028 + 0.0114)) past year problem gamblers. There are 5450 “lifetime” problem gamblers (= 100000 * (0.0385 + 0.016)). Thus 28% of the lifetime group are have not been active in the past year.
The longer the duration of a case (of any condition), the closer will be the “current” (e.g. past year) prevalence and the “lifetime” prevalence. If all cases are literally lifelong, if no one ever recovers, the current year and “lifetime” prevalences will be the same. If all cases recover inside a year (which does not happen in problem gambling) , the two figures will be very different, with “lifetime” much higher. In very broad terms, prevalence can be equated to incidence multiplied by duration. This inexact relationship is no use, however, in gambling behavior research. Here even less is known about two of the three variables — duration and incidence – than about prevalence.
The table below, (data from Shaffer et al 1) uses earlier terminology to compare lifetime and past year prevalences (95% confidence intervals in parentheses). Lifetime figures, curiously, are not far higher than past year, as one would expect if sustained recovery were common. Quite possibly “lifetime” figures are falsely low; otherwise we must conclude that two thirds of persons who had ever been in level 3 (pathological or disordered) up to the interview time were still there.
Prevalence of level 2 and of level 3 Gambling as % of Specified Population Sector, United States ( 95% confidence intervals in parentheses)
Adult general pop >18
Lifetime Past year
Level 2 (~= problem aka at risk of disordered) 3.85 2.8
Level 3 (~= pathological aka disordered) 1.6 (1.35, 1.85) 1.14 (0.9, 1.38)
From Shaffer et al meta-analysis 1999 ref 1
There are almost no figures to directly reckon incidence of pathological or problem gambling in adults. Incidence has to be inferred from changes in prevalence over a relatively short interval. Volberg, for example, found in Iowa adults a sharp increase in prevalence between 1989 and 1995, contemporaneous with a rapid expansion of gambling in that state. In 1989 0.05% (one twentieth per cent) were past-year pathological; in 1995, 1% were. This is a twenty-fold multiple and a 0.95% absolute increase. 6 It points to a high incidence, but cannot prove it.
What is the average duration of problem gambling in the USA no one knows. It can’t be less than a few years. Most cases of gambling addiction that make the newspapers have been running for at least a few years. To analyze duration would call for following a representative cohort of addicted (aka pathological aka”:disordered”) gamblers or of the larger class, problem gamblers. This would be expensive, administratively challenging, and, in my opinion, unethical. A researcher cannot simply follow the natural history of a condition known to be deleterious unless there is absolutely no treatment possible; for problem gambling there is treatment known to work.
Prevalence is a tool for planning on (say) “how many cases need we prepare to treat each year?” It is poorly suited for understanding causation, assessing risk factors. For those challenges, incidence is the key. In time series rapid climbs in past-year or even lifetime prevalence imply, but do not prove, relatively high incidence and not short duration. Rapid drops suggest high rates of recovery or death combined with low replacement incidence. Estimates of incidence of pathological gambling in North American adults must be drawn from changes in prevalence, full of pitfalls. Incidence of problem gambling ought to be estimated to reckon the marginal impact of expanding legalized gambling or, separately, the ongoing impact of continuing it at a certain level. We will likely never have incidence figures and will have to rely on distant second-best prevalence figures.
Throughout this discussion I have assumed that differences in prevalence between two times are unrelated to methodology and that all point estimates (e.g. 0.6%) are not only completely accurate but also very precise (e.g. 95% confidence intervals are very narrow). This is unrealistic, but in this narrative I’m talking about the limitations in preventive medicine of even optimal estimates of prevalence. Critiquing research methods would be a bigger task yet.
Section 3. Prevalence statistics should not reassure makers of health policy
Now we can turn at last to the several mentions in the forum about prevalence as a measure of societal impacts of “legalized gambling.” Dr. Nelson dwelt on this at length. She noted that the prevalence of lifetime “pathological gambling” in a nationwide study published 2008 (probably that of be Kessler, Hwang, LaBrie et al Psychol Med (2008) 38:1352-1360 ; the same study I think, is described in ref 3 below) was 0.6%, actually down slightly but not significantly different from the 0.7% found three decades earlier for the Commission on the Review of the National Policy Toward Gambling, 1976 . She, and others from the Division on Addictions, see this lack of change as reassurance that over three decades in which gambling opportunities expanded many-fold in so many states there was no corresponding change in impacts. “So nationally,” Dr. Nelson said at the forum, “there has been no significant increase in gambling disorders since 1975 despite the growth [of gambling opportunities].” Depending on what studies one looks at,however, a strong case can be made that there was in fact an increase during the 1990s.
The well-known meta-analysis (1999) that examined over 100 studies covering a similar span of years (mid-70s to 1997) came from the highly-respected group Dr. Nelson has worked with since 2003, headed by Dr. Howard Shaffer. Yet important estimates in it are not mentioned by members of that group when they declare no change over 30 years. In the meta-analysis the estimate for lifetime prevalence of “pathological gambling” (1.6%) was much higher than either of the “stable” figures cited by Dr. Nelson. In fact, the lower bound of its confidence interval was higher than twice the 0.6% found in the 2003-2004 replication. Ms Reilly did cite the 1.6%, only to pass over it to the more comforting 0.6%.
Indeed, the meta-analysis showed an increase in the prevalence of past-year level 3 (abbreviated here to PYL3) between the earlier years it covered (1977-1993) and the last three years (1994-1997). PYL3 in adults went from 0.84% to 1.29%, a statistically significant increase (p<.05). Comparing the same two epochs, lifetime level 2 in adults went from 2.93% to 4.88% (p < .05). The absolute increase in prevalence of PYL3 in adults was 0.45% , the relative increase 154%. When the authors merged all the studies however, the higher prevalence figure in the later ones (1.29%) was lost to view. This may have been a comparison suggested by the data, which could make the “p value” less impressive.
Two other observations by Dr. Shaffer’s group, however, point to rising prevalence in the 1990s. A 2001 paper by Shaffer and Hall 4supplemented the compendium of studies included in the meta-analysis published 1999 . The authors added ones published since 1997 and also some that had never been published, furnished by their authors. 139 studies with at least one prevalence estimate (some compared two instruments) were analyzed for the 2001 paper. For adults only Past Year Level 3 1.46% lifetime level 3 1.92%. There was a positive correlation (r = .313, p<.05) for later year and higher prevalence. The authors found fifteen geographic areas in which earlier and later estimates had been done by the same methods. PYL3 averaged 1.02% in the earlier look, 1.33% in the second (p < .05).
Scientists from the Division on Addictions of the Cambridge Health Alliance have proposed an “adaptation hypothesis,” which acknowledges that new gambling opportunities may lead to a temporary increase in prevalence for the surrounding population. Then, so goes the hypothesis, novelty fades and individuals become more “responsible” in their gambling behaviors and the crest subsides. I hold that proponents of the “adaptation hypothesis” who highlight the similarity in estimates of the lifetime prevalence of pathological gambling between the mid-70s co-morbidity study and its replication thirty years later are selecting data points to support their case while ignoring ones their own group reported that contradict the “no significant increase” conclusion.
Even if the two low estimates are completely valid for the US population, however, as prevalence figures they hide awful damage. See table below. Over 30 years, some (we don’t know how many) of the 1.134 million died or disappeared otherwise (e.g. left the country) from the sampling frame. Let’s guess, cautiously 200,000. In 2004 with its higher population the prevalent count was up by 180,000. Thus 380,000 new cases arose sometime in the thirty years.
Year US pop over age 18 Preval. lifetime pathological number lifetime pathological
1975 216 million 0.7% 1.134 million
2004 292 million 0.6% 1.314 million
The near-alike prevalence figures from mid-70s and early 2000s say little about the human toll, which was probably much greater than 380,000. Multiply that conservative estimate by the number of persons around a single pathological gambler whose lives are damaged or ruined; some use a factor of 4, some use 10. Lesieur estimated 17. Over a million lives badly damaged, to whose profit ? The operators of state-sanctioned predatory gambling.
The Gaming Commission should be commended for videotaping its events and providing a transcript right away.
The writer, Stephen Q. Shafer MD, MA, MPH retired in 2010 as Clinical Professor of Neurology at Harlem Hospital Center, Columbia University. He lives in Saugerties NY. He is Chairperson of Coalition Against Gambling in New York, a non-profit organization registered in Buffalo, for which he receives no pay. email@example.com or firstname.lastname@example.org 917 453 7371
- Shaffer HJ, Hall MN, Vander Bilt J Estimated Disordered Gambling Behavior in the United States and Canada Report to National Gambling Impact Study Final Report 1999 https://divisiononaddictions.org/html/publications/meta.pdf
2. Gerstein D et al Gambling Impact and Behavior Study. Research done by NORC for NGISC http://govinfo.library.unt.edu/ngisc/reports/gibstdy.pdf . Table 1 (just below) based on this report p. 25
Table 1. Prevalence of Problem and Pathological Gambling , by time period and by method of interviewing RDD: Random digit dialing survey Patron Intercept: face to face survey at gambling venues. Data line 1 Problem data line 2 Pathological
RDD Patron Intercept RDD + PI combined
Lifetime Past year Lifetime Past year Lifetime Past year
1.3 0.4 5.3 4.9 1.5 0.7
0.8 0.1 7.9 5.3 1.2 0.6
3. Petry N, Stinson FS, Grant B Comorbidity of DSM-IV Pathological Gamblers and Other Psychiatric Disorders. J Clin Psychiatry 2005. 66(5): 566-574
4. Shaffer HJ, Hall MN Updating and Refining Prevalence Estimates of Disordered Gambling Behaviour in the United States and Canada. Canadian J Pub Health 2001 92(3): 168-172
5. Slutske WS Natural Recovery and Treatment –Seeking in Pathological Gambling. Am J. Psychiatry 2006: 163:297-502 The researcher looked at the past-year experience of the 201 persons who had met criteria for lifetime pathological gambling in one of two surveys, found that about a third no longer met criteria for the preceding year.
6. Grinols Earl L Gambling in America: Costs and Benefits, Cambridge University Press, 2004 pp 178-79.
The image of a game table is from flickr creative commons. Original title = snooker-3151824312_431fc8b6f2.jpg
The opinions in this narrative are those of the writer and are not necessarily shared by any or all members of CAGNY. This piece may be reproduced in whole or in part as long as the permalink above is cited.