The Brief Addiction Science Information Source (BASIS)

Op-Ed/Editorials: Detecting At-risk Internet Gambling Behavior

Howard J. Shaffer, Ph.D.
Debi A. LaPlante, Ph.D.
Richard A. LaBrie, Ph.D.
Sarah E. Nelson, Ph.D.

Land-based and online gaming companies are racing to create universal algorithms that will identify people who are at risk for developing problematic gambling behavior. Unfortunately, supporters of this effort are nowhere near the finish line for that race; perhaps worse, the competitors are quite possibly off-track. Researchers and practitioners have yet to agree upon the most important identifying characteristics of problematic gambling behavior. Land-based efforts are at a starting advantage, simply because the extant scientific gambling literature predominantly addresses land-based gambling. Consequently, innovators who are trying to facilitate the development of identification algorithms for land-based gambling venues have a stronger scientific foundation available to them than online gaming companies.

There is so little available empirical information about Internet gambling behavior and characteristics of problematic Internet gambling, that early efforts to develop appropriate algorithms are at high risk for error. Researchers cannot, at this time, say with confidence that any of the leading candidates for land-based identifying characteristics of problematic gambling behavior apply to online gambling behavior. The nature of Internet gambling is sufficiently distinct from land-based opportunities to suggest that although some factors might generalize fairly well, others will not, and still others unique to Internet gambling remain to be determined. And, we do not know yet if there are unique risk factors associated with Internet gambling.

Public information related to land-based identification algorithms is limited. In one exception, the Saskatchewan Gaming Corporation (SGC) has published information in a peer-reviewed journal about their algorithm-guided identification system (iCare) to identify at-risk gamblers; they also presented related peer-reviewed findings at international conferences. (1)  Unless the development of such algorithm-guided systems follows transparent validation procedures (e.g., surveying and comparing validation groups of patrons), the sensitivity (i.e., likelihood of accurate identification of individuals who have a problem), specificity (i.e., likelihood of accurate identification of individuals who do not have a problem), and predictive validity (i.e., the ability to accurately predict future gambling problems) of its identification of “problem gambling” is uncertain. Adapting problematic gambling characteristics gleaned from land-based research and/or relying on professional or conventional wisdom about problematic gambling characteristics is unlikely to result in behavioral profiles that can withstand traditional scientific scrutiny.

The risks and hazards associated with prematurely bringing an identification algorithm to the market are considerable. A product without acceptable specificity, sensitivity, and predictive validity places a company and its consumers at unnecessary risk. To illustrate, problematic gambling behavior is a low base-rate phenomenon. This means that an algorithm can claim a success rate of 90% due to its specificity (i.e., likelihood of accurately identifying individuals who do not have gambling problems), but yield minimal to no sensitivity (i.e., likelihood of accurately identifying individuals who do have problems). Last week we posted an illustration of this phenomenon in the WAGER, The Importance of Sensitivity: In Models and for Interpreting What You Read. Finally, the algorithm must have predictive validity. Absent predictive validity, the best that an algorithm can offer is a transient identification that might not be accurate the next day or beyond. The likelihood of sensitivity errors (i.e., falsely identifying someone who has a problem as being problem-free) and the absence of predictive validity place companies at great risk for litigation and players at risk for ongoing harm that can be avoided with careful and systematic research and planning.

The best next steps in the development of responsible gambling for Internet gambling companies are to identify the actual public health risks of online gambling and to develop a comprehensive profile of characteristics that empirically can distinguish problematic Internet gambling behavior from recreational Internet gambling. Specifics related to the best means of accomplishing these tasks are beyond the scope of this paper. Until the next investigative steps are accomplished, identification algorithms for online gaming consumers will be of dubious value.

Notes

1. Davies, B. (2007). iCare: Integrating Responsible Gaming Into Casino Operation. International Journal of Mental Health & Addiction, 5(4), 307-310.

The DRAM, 3(11) - Childhood Sexual Abuse, Age of First Drink, and Onset of Alcohol Dependence

Studies linking childhood sexual abuse (CSA) to higher likelihood of alcohol dependence (AD) later in life (i.e., Dinwiddie et al., 2000; Fergusson, Horwood, & Lynskey, 1996; Kendler et al., 2000; Kilpatrick et al., 2000; Molnair, Buka, & Kessler, 2001; Nelson et al., 2002) might lack adequate control for environmental and genetic factors that influence the relationship between CSA and AD.  These studies are also inconclusive because an earlier age of first drink is associated with later AD regardless of CSA status (DeWit, 2000).  This week’s DRAM reviews a study that controls for the influence of environmental and genetic characteristics, as well as the influence of drinking at an early age using a sample of twins. 

Sartor, Lynskey, Bucholz, McCutcheon, Nelson, Waldron, et al. (2007) conducted telephone surveys with 2061 (response rate 78%) 18-29 year old female monozygotic or dizygotic twin sets participating in the Missouri Adolescent Female Twin Study (Heath et al., 2002).  The survey included the Semi-Structured Assessment for the Genetics of Alcoholism (SSAGA; Bucholz, 1994; Hesselbrock, 1999), age at first drink, age of onset of AD, and Traumatic Events and Early Childhood Experiences.  Participants met researchers’ criteria for CSA if they endorsed having experienced rape, sexual molestation, and/or forced sexual contact before age 16, or sexual contact by someone five or more years older before they reached age 16.  The authors measured speed of progression to AD by measuring age at first drink and age at onset of AD. 

Table 1.  Alcohol use measures adjusted for environmental, genetic, or early age of first drink correlates among those with CSA history.

   

* Significant
** OR = Odds ratio, HR = Hazard ratio
Adapted from Sartor, et al. (2007)

The sample was 86% Caucasian and 14% African American.  Almost 12% endorsed a history of CSA and 7.9% qualified for AD.  The authors adjusted for the genetic influence expressed by the higher risk for AD if her twin had AD, especially if they were monozygotic twins.  Table 1 shows that, after adjusting for genetic factors, CSA was associated with a greater likelihood of lifetime alcohol use, having a first drink at an early age, and AD.  However, the effect of CSA on AD was mediated by age at first drink. After adjusting for the genetic influence on AD and the CSA-related age at first drink, CSA history did not increase the speed of AD onset from 1st drink.

The study was based on self-report and defines CSA narrowly; these factors might lead to under reporting CSA.  Rather than compare sets of twins where only one female experienced CSA, researchers group all twins together, and then statistically controlled for differences between twin pairs.  This yields a larger sample size and, therefore, statistical significance for smaller differences.  The study did not report the rate of co-occurrence in twin pairs of CSA and AD, and congruence across twins. Adjusting for AD co-occurrence might adjust for CSA co-occurrence if congruence is common.  The results might not generalize well to the general population because a very small subset of the overall sample had both a history of CSA and AD; the entire sample was female, and most of the sample was Caucasian.

Contrary to expectation, CSA did not predict time to AD after adjusting for age at first drink.  Drinking at an early age predicted a more rapid pace for AD development; however, both twins having AD was the best predictor of rapid onset of AD after their first drink, suggesting genetics and environment are greater risk factors for this effect than CSA.  Future research could examine an early intervention’s effectiveness for reducing rates of later AD among early adolescents reporting CSA history. 

What do you think?  Comments can be addressed to Leslie Bosworth.

References

DeWit, D., Adlaf, E.M., Offord, D.R., Ogborne, A.C. (2000). Age at first use: a risk factor for the development of alcohol use disorders. American Journal of Psychiatry, 157, 745-750.

Dinwiddie, S., Heath, A. C., Dunne, M. P., Bucholz, K. K., Madden, P. A. F., Slutske, W. S., et al. (2000). Early sexual abuse and lifetime psychopathology: a co-twin control study. Psychological Medicine, 30, 41-52.

Fergusson, D. M., Horwood, L. J., & Lynskey, M. T. (1996). Childhood sexual abuse and psychiatric disorders in early adulthood.  II.  Psychiatric outcomes of childhood abuse. Journal of the American Academy of Childhood and Adolescent Psychiatry, 35, 1365-1374.

Heath, A. C., Howells, W., Bucholz, K. K., Glowinski, A. L., Nelson, E. C., & Madden, P. A. (2002). Ascertainment of a mid-western U.S. female aldolescent twin cohort for alcohol studies: assessment of sample representativeness using brither record data. Twin Research, 5, 107-112.

Kendler, K. S., Bulik, C. M., Silberg, J., Hettema, J. M., Myers, J., & Prescott, C. A. (2000). Childhood sexual abuse and adult psychiatric and substance use disorders in women.  An epidemiological and co-twin control analysis. Archives of General Psychiatry, 57, 953-959.

Kilpatrick, D. G., Acierno, R., Saunders, B., Resnick, H. S., Best, C. L., & Schnurr, P. P. (2000). Risk factors for adolescent substance abuse and dependence: data from a national sample. Journal of Consulting and Clinical Psychology, 68, 19-30.

Molnair, B. E., Buka, S. L., & Kessler, R. C. (2001). Childhood sexual abuse and the subsequent psychopathology: results from the National Comorbidity Study. American Journal of Public Health, 91, 753-760.

Nelson, E. C., Heath, A. C., Madden, P. A. F., Cooper, M. L., Dinwiddie, S. H., Bucholz, K. K., et al. (2002). Association between self-reported childhood sexual abuse and adverse psychosocial outcomes.  Results from a twin study. Archives of General Psychiatry, 59, 139-145.

Sartor, C. E., Lynskey, M. T., Bucholz, K. K., McCutcheon, V. V., Nelson, E. C., Waldron, M., et al. (2007). Childhood sexual abuse and the course of alcohol dependence development: Findings from a female twin sample. Drug and Alcohol Dependence, 89, 139-144.

The WAGER Vol. 12(11): The Importance of Sensitivity: In Models and for Interpreting What You Read

Have you noticed lately that several different cellular phone companies are running ads claiming to be America’s leading network? Have you wondered how they can all claim to be number one? One company claims to have the fewest dropped calls of any network. This sounds impressive until you begin to wonder how many calls actually connect in the first place. Companies can’t drop calls if they don’t complete them in the first place.

To understand and evaluate the wealth of information that advertising and press releases provide us, it is important to be an informed consumer.  Knowing what questions to ask when reading claims in the media is crucial to becoming an active processor, rather than a passive absorber, of this information. To know what questions to ask, a basic understanding of statistics and research methodology is necessary. We attempt to provide some of that understanding in the BASIS every week. Although the research we review is important, the ability to understand and evaluate the research you come across in other settings is just as important.

I recently read a press release touting a technology able to identify players at risk for gambling problems “with a precision of more than 90%.” A very impressive claim at first glance. But let’s investigate this claim more thoroughly. What does 90% precision mean in this case? The study behind this claim is not publicly available – no surprise – so we have to speculate. Given what we know about gambling disorders and problems, it is interesting to explore what data could lead to such a claim.

Here is one way to achieve 90% precision:

In this example, there are 7 actual gamblers with problems, and 93 actual recreational gamblers without problems, represented by the columns. The rows represent how the given model (e.g., technology, diagnostic test, etc.) classifies those people. In this case, the predictive model correctly classifies 90% of the sample (i.e., the 87 who are both predicted and are actual recreational gamblers and the 3 who are both predicted and are actual gamblers with problems). This represents the 90% precision claim. However, disordered gambling is a “low” base rate event: less than 2% of the population typically qualifies as pathological gamblers and approximately 3-5% more as problem gamblers. As a result, most people classified by any model will fall into the recreational category, both in terms of their actual and predicted symptoms.

In the example, 93 do not have gambling problems, and the model correctly classifies 94% of those 93. (1) This relationship accounts for the bulk of the model’s classification accuracy. If instead, we consider the model’s capacity to accurately classify actual gamblers with problems, a different picture emerges. There are 7 actual gamblers with problems in the example. The model correctly classifies 3 of them: 43%. (2) Of the 9 people that the model classifies as gamblers with problems, only 33% are actual gamblers with problems. (3)

There are two key points here: First, statistics can sound impressive, but without the appropriate context, statistics easily can be misinterpreted. Second, specific to a disorder or problem with a low occurrence in the general population, it is relatively easy to achieve very good overall classification accuracy. To truly evaluate a model’s ability to classify, as in this case, it is important to know how well it classifies actual cases (e.g., gamblers with problems), not just non-cases.

Protecting the public health is best served by developing screening procedures that have a high sensitivity (i.e., ability to identify those people who have problems). Assuming that the procedures themselves don’t have any extraordinary risks, it is generally optimal to include false positives (i.e., people identified as having problems who do not actually have problems) rather than to miss identifying the cases with problems. Informed consumers should demand to know all the parameters of a model’s accuracy, and in the case of infrequently occurring problems, particularly the sensitivity.

What do you think? Comments should be addressed to Sarah Nelson.

Notes

1  This (i.e., correctly classifying 94% of those without problems) is the specificity of the predictive model.
2  This (i.e., correctly classifying 43% of those with problems) is the sensitivity of the predictive model.
3  This (i.e., 3 of the 9 classified as having problem actually having problems) is the positive predictive value of the predictive model.

STASH Vol. 3(10) Anabolic steroid use and other risky behaviors: Not just a man’s issue.

During recent years, Centers for Disease Control and Prevention (CDC) researchers have observed an increase in prevalence rates of adolescent females’ self-reports of anabolic steroid use and other risky behaviors. These risky behaviors include, binge drinking, drug use, sexual behavior, and violence in school (Grunbaum et al., 2004).  For decades, researchers, health advocates, and the general public have associated these issues as being a “male problem” (Bahrke, Yesalis, & Wright, 1990; Egan, 2002; Potteiger & Stilger, 1994; "SCOUTING; School Lesson About Steroids", 1985) and targeted interventions for these activities primarily to males (See STASH Vol. 2(5)).  In this weeks’ STASH, we examine the anabolic steroid use of female high school students and its effect on various health behaviors.

Elliot et al. (2007) conducted an analysis of the CDC’s 2003 Youth Risk Behavior Surveillance System (YRBSS) dataset comparing health outcomes of steroid users to non-users among girls (N=7,447) in grades nine through twelve.  The YRBSS sampled participants using a three-stage cluster design (U.S. Region x School x Classroom) drawing a nationally representative sample of students in grades 9-12.  Participants completed a 95 question anonymous self-report survey regarding general health, substance use, violent behavior, and mental health.  Researchers conducted a logistic regression to estimate the association between anabolic steroid use and risky health behaviors; covariates in this analysis included race/ethnicity, grade level, and team sports participation.

About 5% (5.3%) of female U.S. high school students reported using anabolic steroids during their lifetime.  As Table 1 shows, adolescent female anabolic steroid users are more likely than female non-anabolic steroid users to also abuse other substances, engage in risky sexual behavior, attempt to lose weight in unhealthy ways (e.g., not eating for more than 24 hours), and suffer from recent depressive episodes and suicide attempts.  Researchers found participating in team sports to be a protective factor against anabolic steroid use, meaning those who participated in team sports were significantly less likely to use anabolic steroids (OR, 0.52; 95% CI,0.34-0.80).

Table 1: Comparing prevalence estimates of health-related behaviors for female U.S. high school students according to anabolic steroid use (Adapted from Elliot et al., 2007).

This study is not without limitations.  Researchers utilized a cross-sectional design.  A cross-sectional study allows researchers to ask questions about behavior and health outcomes during one point in time, does not allow examination of temporal relationships, and, therefore, limits our ability to understand causal influences.  The YRBSS data is based upon self-report, potentially leading to under-reporting of risky health behaviors and outcomes.  In addition, researchers only included participants who were present in school in their analysis.  Adolescents who do not attend school might engage in the most risky behaviors. If so, then this data evidences conservative estimates of risky health behaviors.

Despite these limitations, the prevalence of risky health behaviors among female students reporting anabolic steroid use raises important pubic health concerns.  These results suggest three things. (1) Female students using steroids might suffer from underlying psychiatric problems, requiring a more extensive treatment program such as counseling.  (2) Those who participated in team sports were less likely to use anabolic steroids, potentially indicating that women’s sports empower girls to make better health choices. These healthy choices might be mediated by increased self-esteem.  (3) It is possible that these results might be due in part to a greater awareness of the dangers of anabolic steroid use among female student athletes.  The benefits related to girls’ participation in team sports deserves further study.  In the meantime, school administrators and health advocates should consider implementing anabolic steroid treatment and prevention programs specific to adolescent females.

What do you think?  Comments can be addressed to John H. Kleschinsky.

Reference

Bahrke, M. S., Yesalis, C. E., 3rd, & Wright, J. E. (1990). Psychological and behavioural effects of endogenous testosterone levels and anabolic-androgenic steroids among males. A review. Sports Medicine, 10(5), 303-337.

Egan, T. (2002). Body-Conscious Boys Adopt Athletes' Taste for Steroids [Electronic Version]. The New York Times. Retrieved 12/5/2007 from http://query.nytimes.com/gst/fullpage.html?res=9E00E3DF1239F931A15752C1A9649C8B63&sec=&spon=&pagewanted=all#.

Elliot, D. L., Cheong, J., Moe, E. L., & Goldberg, L. (2007). Cross-sectional study of female students reporting anabolic steroid use. Archives of Pediatrics & Adolescent Medicine, 161(6), 572-577.

Grunbaum, J. A., Kann, L., Kinchen, S., Ross, J., Hawkins, J., Lowry, R., et al. (2004). Youth risk behavior surveillance--United States, 2003. MMWR Surveill Summ, 53(2), 1-96.

Potteiger, J. A., & Stilger, V. G. (1994). Anabolic Steroid Use in the Adolescent Athlete. Journal of Athletic Training, 29(1), 60-64.

SCOUTING; School Lesson About Steroids. (1985).   Retrieved December 5, 2007, from http://query.nytimes.com/gst/fullpage.html?sec=health&res=9A0DE3D91430F936A35755C0A960948260&n=Top/News/Health/Diseases,%20Conditions,%20and%20Health%20Topics/Steroids

Addiction & the Humanities Vol. 3(10) - Supply & Demand: Looking at addiction through drug trafficking

The guiding principle of supply and demand plays a key role in the indirect exchange between drug suppliers and drug users. In addition to local drug traffickers, there are also international drug rings whose successful operations make them essential collaborators in the drug business. One leader of such a drug ring was Khun Sa, whose life and death highlights differences in international government practices related to drug control. In this issue of Addiction and the Humanities, we use Khun Sa as a lens to examine the role of drug trafficking in the supply and demand that feeds drug addiction.

From 1974 to 1994, Khun Sa had unrivalled success and control over the growth and transport of opium, and its derivative, heroin, in the Shan State of Myanmar (formerly Burma) (AlJazeera.net, October 30, 2007; Economist, November 8, 2007). His rise to control came from a combination of confused reasoning and a government whose wish for economic and military aid clouded its judgment on how to deal with drug trafficking. Although never formally educated, Khun Sa received military training from the Chinese Nationalist forces. This experience no doubt proved valuable in commanding a local militia, and running his part of the already present opium business.

After moving his life and business to Myanmar (formerly Burma) (Fuller, November 5, 2007), Khun Sa maintained a love-hate relationship with both the Shan people and the Burmese Government; he argued that providing economic stability for the Shan people motivated him to engage in drug trafficking. In addition, although it publicly condemned him, the Burmese Government tolerated his control of Myanmar’s most economically rewarding export, and his arrangements with local police, soldiers, and generals. In 1996, the government reached a tacit agreement with Khun Sa ensuring that he would avoid extradition, receive government protection, and could retire comfortably; in return he had to surrender his home, weapons, ammunition, soldiers, refineries, and crops to the Burmese Government.

In 1977, Khun Sa offered to sell his entire opium crop to the United States (Johnson, March 1, 1982); this proposal suggested that buying his opium crop was the only way for the US to take it off the market and minimize the amount of opium traveling from Asia to the US. Khun Sa knew that his products, business, and influence stretched internationally, and with the Burmese Government’s help, he was able to avoid the consequences (e.g., indictment, extradition, trials) of international illegal activity. With the majority of demand for illegal drugs originating in the West, more specifically in the United States, foreign government cooperation with drug traffickers poses a threat to international attempts to put a stop to drug business’s power. In Khun Sa’s case, the Burmese Government enabled him in multiple ways. While Khun Sa controlled a personal army, the government accepted the idea of opium as a profitable crop and chose to ignore the evident negative effects on international relations. In fact, Khun Sa’s January 7, 1996 surrender to the Burmese Government placed a significant amount of weaponry and property in their hands.

Myanmar’s exit from the opium (heroin) business (Liu, November 6, 2007) made room for other Eastern countries' re-entry. Afghanistan, for example, has regained its position as the number one producer of heroin in the East; 92% of the world’s heroin is now produced in Afghanistan (Liu, November 6, 2007). The situation in Afghanistan is similar to that in Myanmar: without sufficient government regulations, drug traffickers can impose their power and deceive nations into thinking that their drug transactions function like all other import/export businesses. The difference in philosophy between Eastern and Western government ideas explains why Khun Sa’s proposal to the United States was met with an indictment. In the US, illegal drug activity is punished on both ends of the supply and demand scale; however, similar to the Shan state, currently some Eastern nations have less demand and can focus on their role as suppliers to bring them financial stability.

What do you think? Comments can be addressed to Ingrid Maurice.

References

AlJazeera.net. (October 30, 2007). Golden Triangle drugs kingpin dies.   Retrieved November 26, 2007, from http://english.aljazeera.net/NR/exeres/FA91ECD1-59CA-4963-83EE-34DCB6557F71.htm

Economist, T. (November 8, 2007). Obituaries: Khun Sa (Chang Chi-fu), master of the heroin trade, died on October 26th, aged 73.   Retrieved November 19, 2007, from http://www.economist.com/obituary/displaystory.cfm?story_id=10097596

Fuller, T. (November 5, 2007). Khun Sa, Golden Triangle Drug King, Dies at 73. New York Times, Asia Pacific Section   Retrieved November 19, 2007, from http://www.nytimes.com/2007/11/05/world/asia/05khunsa.html

Johnson, M. (March 1, 1982). The Great Opium War. Time Magazine   Retrieved November 26, 2007, from http://www.time.com/time/magazine/article/0,9171,922799,00.html

Liu, M. (November 6, 2007). Death of the Golden Triangle's most powerful druglord. Newsweek   Retrieved November 19, 2007, from http://www.blog.newsweek.com/blogs/ov/trackback.aspx?PostID=65236