Lynch WJ, Maciejewski PK, Potenza MN. Psychiatric Correlates of Gambling in Adolescents and Young AdultsGrouped by Age at Gambling Onset. Arch Gen Psychiatry. 2004;61(11):1116-1122. doi:10.1001/archpsyc.61.11.1116
Gambling is a prevalent behavior, yet few studies have investigated
its mental health correlates. Although early-onset engagement in behaviors
with addictive potential has generally been associated with more severe problems,
direct investigation of a nationally representative sample of gamblers grouped
by age at onset of gambling has not been performed.
To identify differences in psychiatric correlates of gambling and gambling-related
attitudes and behaviors in adolescents (aged 16-17 years) and in young adults
(aged 18-29 years) with early-onset (before age 18 years) and adult-onset
Logistic regression analysis.
Public access data set derived from random-digit-dialing telephone surveys.
The study analyzed data from adolescent (n = 235), early-onset
adult (n = 151), and adult-onset (n = 204) past-year gamblers
and adolescent (n = 299) and adult (n = 187) nongamblers
in the Gambling Impact and Behavior Study.
Main Outcome Measures
Gamblers and nongamblers were compared within each group on measures
of sociodemographics and psychiatric health. Adolescent, early-onset adult,
and adult-onset past-year gamblers were compared on measures of gambling attitudes
Adolescent gamblers were more likely than adolescent nongamblers to
report alcohol and drug use and abuse/dependence and depression. Elevated
rates of alcohol and drug use and abuse/dependence were observed in early-onset
adult gamblers vs adult nongamblers, and only elevated rates of alcohol use
were observed in adult-onset gamblers vs adult nongamblers. Substantial differences
in reasons for and patterns of gambling were observed among the 3 groups of
Adolescent-onset gambling is associated with more severe psychiatric
problems, particularly substance use disorders, in adolescents and young adults.
More research is needed to investigate the relationships and inform prevention
and treatment strategies.
Access to gambling has increased dramatically in the past several decades.1 Approximately 68% of the US adult population has gambledlegally in the past year.2Although most adultsgamble without apparent problems, approximately 9 million are classified asproblem gamblers and another 3 million as pathological gamblers.3 Adultproblem and pathological gambling are associated with substance use problems,depression, psychiatric treatment, poor subjective general health, arrest,and incarceration.1,4
Data suggest that adolescents may be more vulnerable than adults togambling-related problems. Although gambling among adolescents is largelyillegal, 50% to 90% of youths aged 12 to 17 years report gambling within thepast year.5 These high rates underscore theneed to investigate the psychiatric correlates and social consequences ofadolescent gambling. Surveys have consistently found high rates of at-risk,problem, and pathological gambling in adolescents.6 Regionalsurveys7- 12 haverevealed that heavy or disordered gambling in adolescents is associated withadverse health and social consequences, including substance use/abuse anddepression.
Adverse consequences associated with adolescent gambling may persistinto adulthood. Retrospective studies12,13 ofadult problem and pathological gamblers reveal that age at gambling onsetis typically before adulthood. The extent to which early onset of gamblingis associated with more severe gambling problems, as has been suggested forsubstance use disorders, has not been adequately studied (eg, in longitudinalinvestigations), although existing data suggest that early onset of gamblingis associated with more severe gambling problems.2,13 Moststudies of adult gamblers do not assess differences related to age at gamblingonset. Given that there exist age-at-onset and treatment-related differencesin the characteristics of individuals with alcohol dependence,14 adisorder frequently co-occurring with pathological gambling15 andlinked at a genetic level,16 further examinationof the effect of age at gambling onset on adult gambling and substance usebehaviors is warranted.
Few studies have directly compared adolescent gamblers with adult gamblers,particularly on a national level. Criteria used to define disordered gamblingfrequently vary between studies, and they are generally less stringent instudies of adolescents, complicating comparisons of gambling in adolescentsand adults.5,6 The 1998 GamblingImpact and Behavior Study (GIBS) surveyed adults and adolescents to examinetheir gambling attitudes and behaviors. The present study sought to investigatethe characteristics of adolescent (aged 16-17 years) and young adult (aged18-29 years) past-year gamblers on measures of psychiatric health and gambling.The young adult group of gamblers was separated into early-onset (gamblingonset at age <18 years) and adult-onset (gambling onset at age ≥18 years)groups. We hypothesized that (1) each group of gamblers would report morepsychiatric problems than nongamblers, (2) adolescent and early-onset gamblingwould show stronger associations with adverse measures of psychiatric healthcompared with adult-onset gambling, and (3) adolescent, early-onset adult,and adult-onset gamblers would differ on reasons for and patterns of gambling,with adolescent and early-onset adult gamblers demonstrating greater gamblingseverity (higher quantity/frequency) than adult-onset gamblers.
Data analyzed were obtained from the GIBS, a national US civilian householdsurvey conducted by the National Opinion Research Center (NORC). The goalof the GIBS was to examine the impact of gambling in the United States.17,18 Participants were interviewed throughthe use of random-digit-dialing (RDD) surveys ( 534 adolescents and 2417adults) and face-to-face interviews at selected gambling venues (530 adultsonly). Because face-to-face interviews were not conducted with adolescentgamblers and different questionnaires were used for the gambling venue andRDD adult samples, only data from the RDD surveys were analyzed in the presentstudy.
Data for the adult RDD survey were obtained using a list-assisted approachand one-plus sampling (selecting from banks of 100 telephone numbers withat least 1 listed telephone number), as is typical for the NORC, and theyare described elsewhere.17 Telephone numberspurchased from Survey Sampling International (Fairfield, Conn) were stratifiedby state lottery status, and working residential telephone numbers were identifiedin part through screening by Survey Sampling International.17 Theindividual interviewed from the contacted household was determined via a variantof the Troldahl/Carter/Bryant method.17
The adolescent population screened in the GIBS consisted of 16- and17-year-olds, a population representing less than 7% of US households. Inan effort to increase access to this population, adolescents were recruitedusing not only the RDD telephone lists like those used for adults17- 19 but also “enriched”lists that had a higher probability of providing access to adolescents.17,18 Screening of adolescents requiredconsent from the adolescent’s parent or guardian and the adolescent.
The adolescent and adult RDD samples were weighted to be representativeof the US population for 16- and 17-year-old adolescents and adults 18 yearsand older, respectively. The weighting procedure has been described previouslyfor adolescent17,18 and adult17- 19 populations (see http://cloud9.norc.uchicago.edu/dlib/ngis.htm). Sample weights for theadolescent and adult RDD samples were scaled to sum to the actual sample sizesfor these surveys ( 534 adolescents and 2417 adults), and samplesizes reported refer to sums of sample weights.
Studies involving analysis of publicly accessible GIBS data were presentedto the Yale Human Investigations Committee and were exempted from review.Only participants aged 16 to 29 years, inclusive, were studied to minimizerecall bias with respect to age at gambling onset and confounding of cohorteffects. Participants were categorized by age (16-17 vs 18-29 years), past-yeargambling status (yes vs no), and age at gambling onset (<18 years vs ≥18years) to form 5 groups: adolescent gamblers (n = 235), adolescentnongamblers (n = 299), early-onset adult gamblers (n = 151),adult-onset gamblers (n = 204), and adult nongamblers (n = 187).Past-year gambling was defined as “placing a bet [during the past 12months] on the outcome of a race or game of skill or chance, or playing agame—including for charity—in which one might win or lose money.”17 Data from individuals missing data on gambling status(n = 13 adults) and adults missing data on age at gambling onset(n = 7 adults) were excluded from analyses. Sample weights wereadjusted for individuals included in the analyses.
Variables were derived from questions in the GIBS,17,18 asdescribed previously.20 Alcohol and drug useand abuse/dependence measures were based on DSM-IV criteriaas implemented in the National Household Survey on Drug Abuse.17 Past-yearalcohol use was defined as use of alcohol at least once or twice per monthfor a minimum of 12 days in the past year. Past-year drug use was definedas use of an illicit drug for a minimum of 5 days in the past year. Respondentsneeded to meet a threshold criterion of use of a substance (not includingnicotine or caffeine) for nonmedical purposes on at least 5 days in the previousyear to be asked follow-up questions that assess DSM-IV criteria for abuse/dependence (eg, questions about tolerance, withdrawal,and adverse physical or social effects of use). Substance abuse/dependencevariables assessed abuse/dependence of alcohol; marijuana and hashish; cocaineand crack; stimulants such as methamphetamine, amphetamines, and speed fornonmedical reasons; and tranquilizers such as diazepam and alprazolam fornonmedical reasons.
Two screening questions from the Diagnostic Interview Schedule measureddepression, assessing a lifetime history of 2 weeks when the respondent eitherfelt sad, empty, and depressed all the time or lost interest in most thingspreviously found enjoyable. As with drug and alcohol use measures, screeningquestions were designed to capture most respondents with a history of majordepression while minimizing respondent burden.
Most gambling measures (reasons for gambling, age at gambling onset,and quantity/frequency measures) were adapted directly from the GIBS.17,19 At-risk or problem gambling was definedas a lifetime NORC Diagnostic Screen score of 1 or more DSM-IV inclusionary criteria for pathological gambling. Although large-scale,population-based psychometric tests of the NORC Diagnostic Screen have notbeen published, existing data suggest that the instrument has strong internalconsistency and test-retest reliability (lifetime and past-year version teststatistics of r = 0.99 and r = 0.98, respectively) and strong validity in the identificationof pathological gamblers.17 The threshold ofacknowledgment of at least 1 inclusionary criterion is identical to that usedin the GIBS to define at-risk (1-2 criteria), problem (3-4 criteria), andpathological (≥5 criteria) gambling.17 Typesof gambling were grouped into categories of strategic, nonstrategic, machine,casino, and noncasino, as described previously.19,20
Adolescent, early-onset adult, and adult-onset past-year gamblers werecompared with similarly aged nongamblers on sociodemographic and psychiatricmeasures. Some sociodemographic variables were not included in the analysisfor adolescent gamblers because they were judged to be of limited heterogeneity(age, marital status, and employment) or reliability (household income). Alogistic regression procedure implemented in SAS (SAS Institute Inc, Cary,NC) was used to determine odds ratios (ORs) adjusted for between-group differencesas follows. Odds ratios associated with past-year alcohol use and abuse/dependence,past-year drug use and abuse/dependence, any past-year substance abuse, andlifetime depression were adjusted for sociodemographic factors. Adolescent,early-onset adult, and adult-onset past-year gamblers were compared on gamblingmeasures. Odds ratios corresponding to gambling measures were adjusted forsociodemographic variables, past-year substance abuse/dependence, and lifetimedepression.
Compared with adolescent nongamblers, adolescent past-year gamblerswere more likely to be Hispanic and less likely to be girls (Table 1). Compared with adult nongamblers, early-onset adult past-yeargamblers were more likely to never have been married and less likely to bewomen, to earn less than $24 000 annually, and to acknowledge nonwhiterace/ethnicity. Adult-onset gamblers were less likely than adult nongamblersto be African American, to report less than full-time employment, and to earnless than $24 000 annually.
All groups of gamblers were more likely than nongamblers to report past-yearalcohol use (Table 2). Adolescent gamblersand early-onset adult gamblers compared with adolescent and adult nongamblers,respectively, were more likely to report past-year alcohol abuse/dependence,drug use, drug abuse/dependence, and substance abuse/dependence. Adolescentgamblers were more likely than adolescent nongamblers to report lifetime depression.Relationships between psychiatric health measures and gambling status foradolescent gambling compared with those for adult-onset gambling (ie, [adolescentgamblers vs adolescent nongamblers] vs [adult-onset gamblers vs adult nongamblers])were significant for alcohol abuse/dependence (adjusted OR, 4.39; P< .001), drug abuse/dependence (adjusted OR, 4.96; P = .01), and substance abuse/dependence (adjusted OR, 3.01; P = .01). Relationships between psychiatric healthmeasures and gambling status for early-onset adult gambling compared withthose for adult-onset gambling (ie, [early-onset adult gamblers vs adult nongamblers]vs [adult-onset gamblers vs adult nongamblers]) were significant for alcoholuse (adjusted OR, 1.73; P = .01), alcoholabuse/dependence (adjusted OR, 3.43; P< .001),drug use (adjusted OR, 4.02; P< .001), drugabuse/dependence (adjusted OR, 4.39; P< .001),and substance abuse/dependence (adjusted OR, 3.25; P< .001).There were no such differences in relationships between psychiatric healthmeasures and gambling status for adolescent gambling compared with those forearly-onset adult gambling.
Compared with both groups of adult gamblers, adolescent gamblers weremore likely to report gambling for social reasons and less likely to reportgambling for the personal services received (ie, friendly or respectful treatmentfrom staff) or to win money (Table 3).Early-onset adult gamblers were more likely than adult-onset gamblers to gamblefor excitement.
Compared with both groups of adult gamblers, adolescent gamblers weremore likely to usually gamble with someone and less likely to experience largewins or losses (Table 3). Adolescent gamblers were also less likely than early-onsetadult gamblers to gamble weekly or daily and more likely than adult-onsetgamblers to report at-risk or problem gambling. Early-onset gamblers weremore likely than adult-onset gamblers to report usually gambling with someoneand largest wins of $100 or more.
Adolescent gamblers were less likely than both groups of adult gamblersto report past-year nonstrategic, machine, and casino gambling (Table 3).Adolescent gamblers were more likely than adult-onset gamblers to report past-yearstrategic gambling. Compared with adult-onset gamblers, early-onset gamblerswere more likely to report past-year strategic and noncasino gambling andless likely to report past-year casino gambling.
This study examined psychiatric health and gambling-related attitudesand behaviors in adolescent, early-onset adult, and adult-onset past-yeargamblers using data from younger age groups in a recent national survey. Althoughgambling was associated with psychiatric problems (eg, substance abuse/dependence)in adolescents and young adults, these relationships were largely confinedto the groups reporting an early age at gambling onset. Between-group differencesin reasons for and patterns of gambling were observed, with some measuressuggesting that early-onset adult gamblers display heavier patterns of gamblingthan do adult-onset gamblers. Taken together, these results suggest that gamblingduring adolescence may influence adult psychiatric functioning.
The finding that adolescent and early-onset adult past-year gamblerswere more likely than adolescent and adult nongamblers, respectively, to bemale is consistent with findings from other studies21,23 ofadolescent and young adult gambling. A study23 of21 297 students in grades 8 to 12 in Vermont found male sex to be associatedwith past-year gambling and past-year problem gambling. Data suggest thatof adult problem gamblers, women begin gambling later in life, but once theybegin gambling, they develop gambling problems more rapidly than men.20,24 Thus, the findings suggest that gamblingprevention interventions that target youth groups should preferentially focuson boys. However, given the sex-related differences in adult gambling andproblematic gambling behaviors,20,24 moreresearch is needed to investigate sex-related differences in the characteristicsof adolescent gamblers to inform youth gambling and adult problem gamblingprevention efforts.
Strong associations with alcohol and other drug use and abuse/dependencewere observed with gambling in adolescent and early-onset adult gamblers,and these associations were, except for alcohol use, not observed in the adult-onsetgamblers. In part, these findings may reflect aspects of adolescent risk taking,decision making, and experiential learning.25 Studies26 support the notion that risky behaviors (eg, druguse) often begin in adolescence, peak in early adulthood, and diminish overtime. Moreover, risky behaviors often co-occur during adolescence.27 For example, past-year gambling has been shown tobe associated with alcohol consumption, tobacco smoking, drug use, seat beltnonuse, aggressive behavior, and sexual activity in adolescents.25,27 Thesefindings not only suggest that gambling and other risky behaviors be consideredwithin a neurodevelopmental framework,28 butthey also highlight the importance of age-appropriate public health guidelinesfor gambling.29 The findings also suggest thatscreening and treatment efforts for a variety of adolescent risk behaviors,including those involving gambling and substance use, should be integrated.However, the present finding of a similar pattern of differences in psychiatricmeasures in the adolescent and early-onset adult gamblers suggests that theseeffects are not due exclusively to a simple effect of age but rather are relatedto the onset of gambling during or before adolescence.
Adolescent gamblers were more likely than adolescent nongamblers toreport lifetime depression. Depression has been reported15,29 inassociation with recreational and problem or pathological gambling in adults.As in studies of adults, the nature of the association between depressed moodand gambling in adolescents cannot be easily discerned from association studiessuch as the present investigation. The extent to which depressed mood leadsto increased gambling (eg, to escape), gambling leads to depressed mood (eg,through financial losses or other adverse consequences), or a mutual factorleads to both (common environmental or genetic factors) requires direct examination.
Adolescent and adult gamblers reported differences in motivations togamble. Compared with adult gamblers, adolescent gamblers more frequentlyacknowledged gambling for social activity and less frequently for personalservices and to win money, and these differences may, in part, reflect thetypes of gambling performed. Compared with adult gamblers, few adolescentgamblers reported casino gambling, a finding presumably reflecting legal effortsto restrict adolescent casino gambling or decreased availability (eg, restrictedor limited access to automobile transportation). Previous research22 suggests that adolescent gambling is socially oriented,similar to substance use behaviors in adolescents. Adolescents more frequentlythan adults acknowledge past-year gambling in private settings (28.5% vs 11.0%)and report comparable or lower rates of past-year gambling on other formsof gambling.17 The importance adolescents placeon socializing during risk behaviors has implications for teen gambling preventionstrategies, for example, money may not represent the major motivating factorfor adolescent gambling. As such, prevention efforts that focus on increasingthe availability of nongambling social activities warrant consideration.
Compared with adults, adolescents less frequently acknowledge largewins or losses. These differences may largely reflect the financial situationof adolescents, who generally earn much less than adults. The high rates ofat-risk and problem gambling in adolescents despite low quantity/frequencymeasures warrant notice. Because the most widely used screen for adult problemgambling, the South Oaks Gambling Screen,31 heavilyscores items related to finances (ie, debt), more research is needed intothe suitability of specific measures for identifying adolescents with gamblingproblems.
Early-onset adult and adolescent gamblers were similar in several importantways. Specifically, both groups were more likely than adult-onset gamblersto report usually gambling with someone and engaging in strategic forms ofgambling and less likely to participate in casino gambling. Thus, for boththe adolescent and early-onset adult groups of gamblers, an element of competitiverisk taking seems to exert a stronger influence on the reasons for gamblingand the types of gambling performed. Adult-onset gamblers were least likelyto report usually gambling with someone, suggesting that the social elementpresent in adolescent gambling persists for adult gamblers who began gamblingduring adolescence. As with the alcohol and drug abuse/dependence measures,the most pronounced differences in rates of at-risk and problem gambling andstrategic and casino gambling variables were observed in the comparisons betweenadolescent and adult-onset gamblers. Early-onset adult gamblers differed fromadult-onset gamblers on the measure of large wins, which suggests that earlyonset may be associated with “heavier gambling,” similar to previousresearch2,13 demonstrating anassociation between age at gambling onset and gambling severity. Consistentwith this notion, early-onset gamblers were more likely than adult-onset gamblersto report high-frequency gambling (at least weekly; OR, 1.72), although thisdifference did not reach statistical significance.
This study is the first, to our knowledge, to compare adolescent, early-onsetadult, and adult-onset gamblers systematically on measures of psychiatrichealth and gambling-related attitudes and behaviors. The strengths of thestudy include a representative sample of the US population acquired throughan RDD method using a structured interview. One limitation involves the psychiatrichealth measures. These measures were limited to depression and alcohol anddrug use and abuse/dependence and were assessed by a limited number of questions.These questions did not include complete diagnostic criteria, a limitationin part mitigated by incorporation of diagnostic criteria into the surveyquestions, following the precedent of other national surveys.17,32 Anotherlimitation involves survey differences, particularly in the screening processfor gambling pathology. The adolescent and adult surveys were conducted separately,introducing the possibility of selection bias, and the screening measuresused differed. All adolescents who had ever gambled were asked the NORC DiagnosticScreen; in contrast, only adults who acknowledged ever losing $100 or morein a year were surveyed on the NORC Diagnostic Screen. This difference maycontribute to lower levels of at-risk and problem gambling in the adult population.The use of the age of 18 years to distinguish adolescents and adults has prosand cons because states vary regarding ages at which specific forms of gamblingbecome legal. However, 18 years is generally the youngest age at which individualsmay gamble legally. Other limitations include use of a small sample size relativeto statewide surveys and limitations inherent to cross-sectional surveys ingeneral, including the potential for recall bias, cohort effects, differencesin respondents’ interpretations of questions, and the inability to discerncause and effect in the observed associations. Attempts were made to minimizethe effect of recall bias and cohort effects by studying a narrowly definedage group of young adults (aged 18-29 years). The findings of differencesin adolescent and early-onset adult gambling compared with adult-onset gamblingin relation to measures of psychiatric health, and of differences in gamblingattitudes and behaviors among groups of adolescent, early-onset adult, andadult-onset gamblers, hold multiple implications for youth gambling healthguidelines and prevention and treatment strategies. The observation that theassociations of gambling with substance use and abuse/dependence measureswere most pronounced in groups reporting early-onset gambling suggests thatgambling during adolescence may substantially impact adult function. Thesefindings highlight the need for longitudinal studies to examine the effectof gambling on younger age groups, particularly as the availability and socialacceptance of legalized gambling increases. Information gathered from suchstudies could be of great value in generating health guidelines for gamblingin adolescents and adults.
Submitted for Publication: June 2, 2003; finalrevision received April 12, 2004; accepted April 22, 2004.
Correspondence: Wendy J. Lynch, PhD, Departmentof Psychiatry, Yale University School of Medicine, 34 Park St, New Haven,CT 06516 (firstname.lastname@example.org).
Funding/Support: This work was supported bygrant K12-DA114038 from the National Institutes of Health Office of Researchon Women's Health/National Institute on Drug Abuse, Bethesda, Md (DrLynch); grant K12-DA00366 from the National Institute on Drug Abuse/AmericanPsychiatric Association (Dr Potenza); grant K12-DA00167 from the NationalInstitute on Drug Abuse (Dr Potenza); and Women’s Health Research atYale.
Acknowledgment: We thank the Yale InterdisciplinaryWomen’s Health Research Scholar Program on Women and Drug Abuse andthe Women and Addictive Disorders Core.