Prince Ed ward Is land (PEI) is a largely ru ral province with

Epidemiology of Problem Gambling in Prince Edward Island: A Canadian Microcosm? Ja son P Doiron, BA 1, Rich ard M Nicki, PhD 2 Ob jec tives: To gather in for ma tion that de scribes the ex tent of gam bling and prob lem gam bling in Prince Ed ward Is land (PEI), to rig or ously an a lyze the re la tion be tween gam bling ac tiv i ties and prob lem gam bling, to doc u ment cog ni tive and emo tional corre - lates of prob lem gam bling, and to iden tify an at-risk gam bling group. Method: We se lected a ran dom, strat i fied sam ple (n = 809) to rep re sent the adult pop u la tion of PEI. We ad min is tered both the South Oaks Gam bling Screen (SOGS) and an early version of the Ca na dian Prob lem Gam bling Index (CPGI) to par tic i pants who had gam bled. Re sults: The cur rent rate of prob lem gam bling was 3.1%. Prob lem gam blers were likely to be un der age 30 years, to be sin gle, and to re port cog ni tive, emo tional, and sub stance abuse cor re lates. Mul ti ple-re gres sion anal y sis iden ti fied a unique and sub stan - tial re la tion be tween prob lem gam bling and video lottery ter mi nal (VLT) use. We iden ti fied a group of at-risk gam blers (scor ing 1 or 2 on the SOGS), com pris ing 14% of the sam ple. Con clu sions: Gam bling and prob lem gam bling pat terns in PEI re sem ble those in most other prov inces. The re la tion found be - tween prob lem gam bling, VLT use, and cog ni tive, emo tional, and sub stance use cor re lates should ap ply to the greater population as well. (Can J Psy chi a try 2001;46:413 417) Key Words: pathological gambling, problem gambling, video lottery terminals, multiple regression, cognitive errors Prince Ed ward Is land (PEI) is a largely ru ral province with an adult pop u la tion of ap prox i mately 100 000. Prince Ed - ward Is land ers have var i ous gam bling ac tiv i ties avail able to them, in clud ing con ven tional lot ter ies, video lot tery ter mi - nals (VLTs), horse rac ing, bingo, and char i ta ble gam bling ac - tiv i ties. Al though PEI does not have a ca sino, one is avail able within driv ing dis tance, in Nova Sco tia. In 1999, the pro vin cial gov ern ment re ceived gam bling prof its to tal ling ap prox i mately 13 million dol lars, and PEI was sec - ond only to Quebec in per ca pita spend ing on gambling ($340) (1). Gov ern ment rev e nues sug gest that gam bling is a pop u lar ac tiv ity in PEI, and there has been con sid er able con - cern voiced from the pub lic, gov ern ment, and the med i cal com mu nity about gam bling s ad verse ef fects, es pe cially in re la tion to VLTs (2). Very few em pir i cal data have been re - ported, how ever, re gard ing the in ci dence of gam bling in the prov ince. Manuscript received August 2000 and accepted February 2001. Previously presented at the 11th International Conference on Gam bling and Risk Taking; June 12 16, 2000; Las Vegas, NV. 1 Doctoral graduate student, Department of Psychology, University of New Brunswick, Fredericton, New Brunswick. 2 Professor, Department of Psychology, University of New Brunswick, Fre - dericton, New Brunswick. Address for correspondence: Mr JP Doiron, Department of Psychology, Uni - versity of New Brunswick, Fredericton, NB, #45444, E3B 5A3. e-mail: r741n@unb.ca. In con trast, gam bling prev a lence stud ies for ex am ple, Rob - ert Ladouceur s work in the prov ince of Que bec (3,4) have been com pleted in most other prov inces, al though they have been re ported only rarely in the re search lit er a ture. Ta ble 1 sum ma rizes prev a lence rates of adult prob lem gam bling (less se vere prob lem rate plus prob a ble patho log i cal rate) ac cord - ing to re cent sur veys, all of which used the South Oaks Gam - bling Screen (SOGS) (5). Prob lem gam bling rates ap pear to be rel a tively stable across Can ada and are be low 5% in all prov inces. The pro file of prob lem gam blers is also rel a tively uni form. In gen eral, Ca - na di ans ex pe ri enc ing prob lems with gam bling are more likely to be male, sin gle, and youn ger than the gen eral pop u la tion. Missing from the Ca na dian reports is in for ma tion about the group of gam blers who score pos i tively on the SOGS but are be low the cut-off point for prob lem gam bling (a score of 3). Since the di ag nos tic threshold for gam bling pa thol ogy may be some what ar bi trary (13), SOGS scores of 1 or 2 could in di - cate an at-risk level of prob lem atic gam bling that might be tar geted for pri mary-pre ven tion in ter ven tions. Many of the Ca na dian stud ies have re ported par tic u larly strong links be tween prob lem gam bling and con tin u ous-play ac tiv i ties such as card games (not played in ca si nos), casino games, bingo, horse rac ing, VLTs, and in stant-win tick ets (6,7,9). Tech niques for es tab lish ing these links have usually in volved cal cu lat ing the prob lem gam bling rate among Can J Psychiatry, Vol 46, June 2001 413

414 The Canadian Journal of Psychiatry Vol 46, No 5 par tic i pants in each ac tiv ity or iden ti fy ing popular gambling ac tiv i ties among prob lem gam blers. Al though these ap - proaches sug gest which gam bling ac tiv i ties might be most re - lated to prob lem gam bling, they do not pro vide un equiv o cal ev i dence, be cause prob lem gam blers com monly en gage in more than one kind of gam bling ac tiv ity. To ad dress this lim i ta tion, the unique re la tion be tween in - volve ment (du ra tion of play, fre quency of play, time per play - ing ses sion, and monthly ex pen di ture) in each gambling ac tiv ity and scores on prob lem gam bling as sess ment de vices must be rig or ously as cer tained. Sta tis ti cal anal y sis in volv ing mul ti ple-re gres sion tech niques is well suited to this pur pose, with as sess ment de vice scores be ing the cri te rion vari able and sex and in volve ment in gam bling ac tiv i ties be ing pre dic - tor vari ables (to con trol for any re la tion be tween sex and prob lem gam bling). Cal cu la tion of the squared semipartial cor re la tion co ef fi cients (sr 2 ) readily pro vides the unique vari - abil ity in SOGS scores ac counted for by in volve ment in each gam bling ac tiv ity. The ques tion of how to mea sure prob lem gam bling in the gen - eral pop u la tion has been de bated in the re search community for some time. Al though the SOGS, a DSM-III-based as sess - ment de vice, has been the most com mon in stru ment with known psychometric prop er ties (5,14) for iden ti fy ing prob - lem gam blers, there have been sig nif i cant crit i cisms of it (15,16). Yet, the SOGS al lows for in ter na tional (and in ter pro - vin cial) com par i sons of prob lem gam bling rates an im por - tant ben e fit (16). Nev er the less, the prob lems iden ti fied with the SOGS un der line the im por tance of im prov ing prob lem gam bling as sess ment tools for use in gen eral pop u la tion sur veys. A new as sess ment de vice, based on DSM-IV (17) cri te ria for patho log i cal gam bling, is cur rently be ing de vel oped to ad - dress some of the prob lems with past prev a lence sur vey in - stru ments. Known as the Ca na dian Prob lem Gam bling In dex (CPGI) (18), it of fers sev eral im prove ments over pre vi ous as - sess ment de vices: it is being de signed for use in gen eral popu - la tion sur veys; it is mul ti di men sional and does not over em pha size cer tain in di ca tors of prob lem gam bling, for ex am ple, bor row ing money; it will in clude cor re late items that will fa cil i tate subtyping of prob lem gam blers; and it should pro duce more nor ma tive dis tri bu tions of scores (18). The pri mary pur pose of this study was to de ter mine the ex tent of rec re ational and prob lem gam bling in PEI. Three ad di - tional goals were to in ves ti gate high-risk gam bling ac tiv i ties us ing a more pow er ful sta tis ti cal anal y sis, to pres ent find ings from the CPGI about im por tant cog ni tive and emotional cor - re lates of prob lem gam bling and at the same time to take the first step to ward the val i da tion of an early ver sion of this new tool, and lastly, to identify and de scribe an at-risk group of gam blers. Method Par tic i pants The sam ple (n = 809) was se lected from the pro vin cial health care da ta base, which in cludes ev ery per son el i gi ble for pro - vin cial health care (that is, the en tire res i dent pop u la tion of PEI). First, a com puter pro gram ran domly se lected 3796 tele - phone num bers pro por tion ately stratified to rep re sent the adult pop u la tion of 5 geo graphic re gions of the prov ince. Use of this pro gram en sured that ev ery phone num ber in the da ta - base had an equal chance of being se lected. Next, pro fes - sional tele phone in ter view ers ran domly se lected and con tacted phone num bers within each re gion. The last birth - day method (19) was used to ran domly se lect par tic i pants in each house hold. The SOGS and CPGI were ad min is tered to all re spon dents who had gam bled at least once in the 12-month presurvey pe riod (n = 669). The re sponse rate for the sur vey was 43%, sim i lar to re sponse rates in other re cent stud ies con ducted in the re gion (7,8). Mea sures As noted above, the SOGS is a widely used DSM-III-based mea sure of prob lem atic gam bling ac tiv ity that has been found to be both re li able and valid (5,14). Scores of 0 to 2 are usu ally in ter preted as sug gest ing nonproblematic gam bling, scores of 3 to 4 as a less se vere gam bling prob lem, and scores of 5 and above as prob a ble patho log i cal gam bling. The early version of the CPGI, based on DSM-IV cri te ria for patho log i cal gam - bling, com prises 48 self-re port items that re late to 5 di men - sions of prob lem gam bling be hav iour: ex tent of in volve ment, spe cific be hav iours, con se quences, cog ni tive and emo tional cor re lates, and en vi ron men tal fac tors and cor re lates (18). Results Pat terns of Par tic i pa tion Data were an a lyzed us ing SPSS. Eighty-three per cent of the sur vey sam ple took part in at least 1 gam bling ac tiv ity in the 12-month pe riod prior to the sur vey. Over all, gam blers were sig nif i cantly more likely to be men (χ 2 = 4.47, P < 0.05) and to have an an nual house hold in come over $30 000 (χ 2 = 13.57, P < 0.001). Finally, 32% of gam blers gam bled 1 to 10 times yearly, 27% gam bled be tween 1 and 3 times monthly, and 41% gam bled at least once weekly. Par tic i pa tion rates for each ac tiv ity (per cent of the sam ple who played in the 12-month presurvey pe riod) were as fol - lows: char i ta ble wag er ing (57.1%), lot ter ies (55.4%), in - stant-win tick ets (48.6%), horse rac ing (15.2%), ca sino slots (14.4%), card games out side a ca sino (14.4%), VLTs (13%), bingo (9.1%), sports lot ter ies (7.6%), betting on games of skill (7.1%), ca sino ta ble games (6.5%), and dice games (0.8%). Men were more likely to par tic i pate in all gambling ac tiv i ties, with the ex cep tion of bingo, which was played more often by women.

June 2001 Epidemiology of Problem Gambling in Prince Edward Island: A Canadian Microcosm? 415 Prob lem and At-Risk Gam bling Prob lem gam bling af fected only a small pro por tion of the sam ple. Province Only 1.1% of re spon dents scored as hav ing less se vere gambling prob lems (scores of 3 or 4 on the SOGS), whereas 2% scored as be - ing prob a ble patho log i cal gam - blers (scores of 5 or more on the SOGS), re sult ing in a total prob lem gam bling rate of 3.1%. Prob lem gam blers were sig nif i cantly more likely than the gen eral population to be un der the age of 30 years (χ 2 = 4.08, P < 0.05) and sin gle (χ 2 = 8.63, P < 0.01). On the av er age, prob lem gam blers be gan gam bling at age 18 years. The 14% of re spon dents (n = 113) who scored 1 or 2 on the SOGS were con sid ered to be at-risk gam blers in stead of non-prob lem gam blers (that is, the usual clas si fi ca tion of SOGS scores). Re spon dents in the at-risk cat e gory were sig - nif i cantly more likely than the gen eral pop u la tion to be un der the age of 30 years (χ 2 = 11.93, P < 0.01), and single (χ 2 = 23.04, P < 0.001). At-risk gam blers, on av er age, be gan gam - bling at age 21 years. Table 1. Summary of Canadian problem gambling surveys Year Less se vere gambling prob lems (%) Prob able patho logi cal gam bling (%) Table 2. Unique relations between gambling activities and problem gambling Activity sr 2 P To tal problem gam bling rate (%) Quebec (4) 1999 2.4 2.1 4.5 Alberta (6) 1998 2.8 2.0 4.8 New Brunswick (7) 1996 1.9 2.2 4.1 Nova Scotia (8) 1996 2.8 1.1 3.9 British Columbia (9) 1996 2.8 1.1 3.9 Ontario (10) 1996 2.0 2.0 4.0 Manitoba (11) 1995 2.4 1.9 4.3 Saskatchewan (12) 1994 1.9 0.8 2.7 Video lottery terminals 0.41 < 0.001 Horse races 0.05 < 0.001 Casino slots 0.02 < 0.01 Bingo 0.01 < 0.001 Casino table games 0.01 < 0.01 Instant-win tickets 0.01 < 0.01 Cards 0.004 not significant Prob lem Gam bling CPGI Cor re lates In ter nal-con sis tency anal y sis of the early ver sion of the CPGI used in this study re vealed a Cronbach al pha co ef fi cient value of 0.69. Chi-square anal y ses in di cated that prob lem gam blers were sig nif i cantly more likely than gam blers in gen eral to en - dorse a num ber of emo tional, sub stance abuse, and cog ni tive er ror cor re late items on the CPGI. First, prob lem gam blers were more likely to re port hav ing been treated for stress-re - lated ill nesses (χ 2 = 5.36, P < 0.05), to have felt se ri ously de - pressed (χ 2 = 6.06, P < 0.05), and to have se ri ously con sid ered or at tempted suicide (χ 2 = 7.17, P < 0.01). Second, con cern - ing drug abuse, prob lem gam blers were more likely to re port hav ing a drug or al co hol prob lem (χ 2 = 8.24, P < 0.01), hav ing used drugs or al co hol while gam bling ( χ 2 = 16.16, P < 0.001), and hav ing gam bled while in tox i cated or high (χ 2 = 38.50, P < 0.001). Finally, prob lem gam blers were more likely to en - dorse 2 cognitive er ror items on the CPGI: they were more likely to be lieve that sys tems or strat e gies were helpful (χ 2 = 33.48, P < 0.001) and that a se ries of losses makes a win more prob a ble (χ 2 = 26.19, P < 0.001). Ac tiv ities Closely Linked to Prob lem Gam bling The de gree to which gam bling ac tiv i ties (that is, char i ta ble wag er ing, bingo, VLTs, or card play ing) were linked to prob - lem gam bling, as mea sured by the SOGS, was iden ti fied us - ing a mul ti ple-re gres sion anal y sis. De gree-of-in volve ment vari ables for each gam bling ac tiv ity, along with a vari able for Games of skill 0.003 not significant Sports lotteries 0.003 not significant Charity 0.002 not significant Traditional lotteries 0.001 not significant sex, were si mul ta neously en tered as pre dic tors of the cri te rion vari able, SOGS scores. Ta ble 2 lists the per cent age of SOGS score vari ance (sr 2 ) uniquely as so ci ated with in volve ment in each gam bling ac tiv ity (that is, vari ance that does not over lap with other ac tiv i ties and sex). F-tests were used to de ter mine which vari ances were sig nif i cantly greater than 0; P val ues for F-tests are also pro vided in Ta ble 2. Six of the 12 ac tiv i ties were found to have sig nif i cant and unique re la tions to SOGS scores. All of these ac tiv i ties were con tin u ous-play ac tiv i ties (ac tiv i ties with a short pe riod of time be tween wager and out come). Note wor thy is the finding that in volve ment in VLT play ing had by far the larg est unique re la tion to SOGS scores. VLT Gam blers: Playing Pat terns, De mo graph ics, Cor re lates In the 12 months prior to the sur vey, 105 out of the total of 809 re spon dents had played VLTs at least once. Of these, 64.6% played VLTs be tween 1 and 10 times yearly, 20.5% played 1 to 3 times monthly, and 14.9% played VLTs weekly.

416 The Canadian Journal of Psychiatry Vol 46, No 5 De mo graphically, VLT play ers dif fered from the gen eral pop u la tion in sev eral im por tant ways. They were more likely to be men (χ 2 = 19.87, P < 0.001), un der the age of 30 years (χ 2 = 16.91, P < 0.001), sin gle (χ 2 = 4.46, P < 0.05), and to have an in come over $30 000 (χ 2 = 7.38, P < 0.01). On the CPGI cor re late items, chi-square anal y ses re vealed that VLT play ers were more likely than the gen eral pop u la tion to re port a past drug or al co hol prob lem (χ 2 = 7.82, P < 0.01), that they had used al co hol or drugs while gam bling (χ 2 = 61.49, P < 0.001), and that they had gam bled while in tox i cated or high (χ 2 = 45.36, P < 0.001). VLT play ers did not dif fer from the gen eral pop u la tion in terms of ed u ca tion, em ploy ment, and men tal health (that is, stress-re lated ill nesses, de pres sion, and sui cide). Dis cus sion As in other Ca na dian prov inces, gam bling in PEI was found to be a pop u lar ac tiv ity, with only a small per cent age of the adult pop u la tion be ing clas si fied as prob lem gam blers. At the same time, some in ter est ing in ter pro vin cial dif fer ences were found. First, gam blers in PEI ap pear to have higher in comes than the pop u la tion in general, which is not true for gam blers in other prov inces. Sec ond, the rate of prob lem gam bling ap - pears to be lower in PEI than in all other prov inces, ex cept for Sas katch e wan; the ab sence of ca si nos might con trib ute to this lower rate. Third, the pro por tion of male prob lem gam blers in PEI did not exceed the pro por tion of men in the gen eral popu - la tion, which dif fers from find ings in other prov inces. Last, the pro por tion of gam blers in PEI with more se vere prob lems (that is, prob a ble patho log i cal gam bling) ap pears to be greater than the pro por tion with less se vere prob lems. The only other prov ince with this pattern of prob lem gam bling is the neigh bor ing prov ince of New Brunswick. One rea son for this pat tern might be the fact that VLTs were in tro duced in the At lan tic re gion be fore be com ing available in other Ca na dian ju ris dic tions. It is pos si ble that the higher level of more se vere gam bling prob lems is re lated to the ear lier avail abil ity of an ac tiv ity that, as this study sug gests, might be par tic u larly ad dic tive. In some Ca na dian sur veys, men tal health and sub stance use prob lems have been found to be as so ci ated with prob lem gam bling (6,8); this was also the case for PEI. In ad di tion, this study iden ti fied 2 im por tant cognitive cor re lates of prob lem gam bling. That is, prob lem gam blers in PEI were more likely than gam blers in gen eral to be lieve that play ing strat e gies were es sen tial to win ning and to be lieve that a se ries of losses makes win ning more likely. This re port iden ti fies the pres - ence of such cognitive errors in prob lem gam blers in the gen - eral pop u la tion and rep re sents an im por tant ex trap o la tion of lab o ra tory re sults in the re search lit er a ture (20,21): it im plies that such er rors in thinking among prob lem gam blers are rel a - tively com mon, which should have ob vi ous con se quences for the treatment of prob lem gam bling be hav iour. Mul ti ple re gres sion, an in no va tive ap proach to the anal y sis of the unique as so ci a tion be tween var i ous gam bling ac tiv i ties and prob lem gam bling, re vealed sig nif i cant links be tween in - volve ment in con tin u ous-play gam bling ac tiv i ties and SOGS scores. Only VLT play ing, how ever, had a unique relation to prob lem gam bling that was sub stan tial and there fore espe - cially im por tant. VLT play ers in PEI were more likely than the general pop u la - tion to be young, sin gle men, and to be rel a tively af flu ent. These find ings sug gest that VLT play ers in PEI have suf fi - cient means to use these ex pen sive machines and have much to lose if their in volve ment be comes prob lem atic. They were also shown to have a greater his tory of al co hol and drug abuse, to have used al co hol and drugs while gam bling, and to have gam bled while in tox i cated or high more than the gen eral pop u la tion. These find ings are im por tant in light of the re cent trend in PEI and other Ca na dian prov inces of gov ern ments that restrict VLT use to bars and other es tab lish ments in volv - ing al co hol con sump tion. The at-risk group of gamblers, iden ti fied as those gam blers scor ing 1 or 2 on the SOGS, might be vul ner a ble to the de vel op ment of prob lem gambling in these settings. In sum, the re sults of this study pro vided a sub stan tial amount of in for ma tion re gard ing gam bling ac tiv ity in PEI, a prov ince where such ac tiv ity had not been pre vi ously as sessed. Im por - tant cog ni tive and emo tional cor re lates of prob lem gambling were iden ti fied in the gen eral pop u la tion. A more pow er ful sta tis ti cal ap proach, not pre vi ously used in the lit er a ture in - ves ti gat ing the as so ci a tion be tween in volve ment in gambling ac tiv i ties and prob lem gam bling, re vealed a strong link be - tween VLT use and prob lem gam bling. Finally, an at-risk group that con sisted of a rel a tively large part of the sam ple was iden ti fied. On the other hand, these find ings have sev eral limitations. Our study was vul ner a ble to the prob lems with tele phone sur - vey re search out lined by Lesieur (22): gam blers might have dis con nected lines be cause phone bills were un paid, might not an swer the phones to avoid bill col lec tors, and might not want to dis cuss such per sonal mat ters with a stranger. Sec - ond, al though mea sures were taken to pro mote a high re - sponse rate (for ex am ple, a me dia re lease in the community prior to the sur vey), the ob tained rate, though ad e quate, was lower than the rate found in many stud ies in the field. Third, as with many stud ies in the field, the low base rate of prob lem and, prob a ble patho log i cal, gam bling in the gen eral pop u la - tion re sulted in only a small group of these gam blers being iden ti fied, which would ad versely af fect the power of some of the sta tis ti cal tests used. In con clu sion, the find ings of this study have pro vided the ba - sis for both pri mary and sec ond ary pre ven tion in ter ven tions in PEI. Further studies will be needed to as sess the success of such in ter ven tions, which will be rel e vant for all provinces.

June 2001 Epidemiology of Problem Gambling in Prince Edward Island: A Canadian Microcosm? 417 Acknowledgements The authors thank the Department of Health and So cial Ser vices, Prince Ed ward Island, for sup port in this project, and Dr Barry Spin - ner for his helpful sta tis ti cal advice. References 1. Mandal VP, Vanderdoelen C. Chasing light ning: gam bling in Can ada. Etobicoke (ON): United Church Pub lishing House; 1999. 2. Sullivan P. PEI s video-gam bling ma chines cre at ing an ad dic tion prob lem, is land MDs warn. CMAJ 1993;148:257 9. 3. Ladouceur R. Prev a lence es ti mates of patho log i cal gam bling in Que bec. Can J Psy - chi a try 1991;12:732 4. 4. Ladouceur R, Jacques C, Ferland F, Giroux I. Prev a lence of problem gam bling: a rep li ca tion study 7 years later. Can J Psy chi a try 1999;44:802 4. 5. Lesieur HR, Blume SB. The South Oaks Gam bling Sreen (SOGS): a new in stru - ment for the iden ti fi ca tion of patho log i cal gam blers. Am J Psy chi a try 1987; 44:1184 8. 6. Wynne Re sources. Gam bling and problem gam bling in Al berta. Ed mon ton (AB): Al berta Al co hol and Drug Abuse Com mis sion; 1998. 7. Baseline Mar ket Re search. Fi nal re port prev a lence study: problem gam bling wave 2. Fred er ic ton (NB): De part ment of Fi nance; 1996. 8. Baseline Mar ket Re search. Fi nal re port 1996 prev a lence study on prob lem gam - bling in Nova Sco tia. Hal i fax (NS): De part ment of Health; 1996. 9. An gus Reid Group. Brit ish Co lum bia prob lem gam bling sur vey. Vic to ria (BC): Brit ish Co lum bia Lot tery Cor po ra tion; 1994. 10. Ferris J, Stripe T, Ialomiteanu A. Gam bling in On tario: a re port from a gen eral pop - u la tion sur vey on gam bling-re lated prob lems and opin ions, re vised, sum mary re - port. To ronto (ON): Ad dic tion Re search Foun da tion; 1996. 11. Cri te rion Re search Cor po ra tion. Problem gam bling study: fi nal re port. Win ni peg (MB): Manitoba Lot teries Foun da tion; 1995. 12. Volberg RA. Gam bling and problem gam bling in Sas katch e wan. Re gina (SK): Min is ter s Ad vi sory Com mit tee on So cial Im pacts of Gam bling; 1994. 13. Beaudoin CM, Cox BJ. Char ac ter is tics of prob lem gambling in a Ca na dian con text: a pre lim i nary study using a DSM-IV based ques tion naire. Can J Psy chi a try 1999;44:483 87. 14. Volberg RA, Banks SM. A re view of two mea sures of patho log i cal gam bling in the United States. Jour nal of Gam bling Studies 1990;6:153 63. Clinical Im pli ca tion These findings have provided the basis for primary and secondary prevention interventions in PEI (and elsewhere) that could tar get cognitive and emotional vulnerabilities among problem gamblers. Limi ta tions Telephone sur veys could bias the sample. The response rate was lower than desired. The low rate of problem gambling reduced the power of some statis - tical tests. 15. Culleton RP. The prev a lence rates of patho log i cal gam bling: a look at methods. Jour nal of Gam bling Studies 1989;5:22 41. 16. Walker MB, Dickerson MG. The prev a lence of prob lem and patho log i cal gam - bling: a crit i cal anal y sis. Jour nal of Gam bling Studies 1996;12:233 48. 17. Amer i can Psy chi at ric As so ci a tion. Di ag nos tic and sta tis ti cal man ual of men tal dis - or ders. 4th ed. Wash ing ton (DC): Amer i can Psy chi at ric As so ci a tion; 1994. 18. Ferris J, Wynne H, Sin gle E. Mea suring problem gam bling in Can ada: draft fi nal re port. Sub mitted for the Ca na dian Centre on Sub stance Abuse; 1999. 19. Weisberg HF, Krosnick JA, Bowen BD. An in tro duc tion to sur vey re search, poll - ing, and data anal y sis. 3rd ed. Thou sand Oaks (CA): Sage Pub li ca tions; 1996. 20. Langer EJ. The il lu sion of con trol. Jour nal of Ap plied and So cial Psy chol ogy 1975;32:311 28. 21. Toneatto T, Blitz-Miller T, Calderwood K, Dragonetti R, Tsanos A. Cog ni tive dis - tor tions in heavy gam bling. Jour nal of Gam bling Studies 1997;13:253 66. 22. Lesieur HR. Ep i de mi o log i cal sur veys of patho log i cal gam bling: cri tique and sug - ges tions for mod i fi ca tion. Jour nal of Gam bling Studies 1994;10:385 98. Résumé Epidémiologie du jeu compulsif à l Île-du-Prince-Édouard : un microcosme canadien? Ob jec tifs : Re cueil lir de l in for ma tion qui dé crit l é ten due des jeux de hasard et du jeu com pul sif à l Île- du- Prince- Édouard (Î-P-É), ana lyser rig oure use ment la re la tion entre les activi tés de jeu et le jeu com pul sif, docu menter les corré lats cog ni tifs et émo tion nels du jeu com pul sif et iden ti fier un groupe à risque de jeu com pul sif. Méth ode : Nous avons sé lec tionné un échan til lon aléa toire strati fié (n = 809) pour re pré senter la popu la tion adulte de l Î-P-É Nous avons ad min is tré le South Oaks Gam bling Screen (SOGS) et une version pré limi naire de l In dice ca na dien des pro blèmes de jeu (ICPJ) aux par tici pants qui avaient joué. Résul tats : Le taux ac tuel de jeu com pul sif était de 3,1 %. Les joueurs com pul sifs étaient sus cep ti bles d a voir moins de 30 ans, d être cé li bataires, et de dé clarer des corré lats cog ni tifs, émo tion nels et de toxi co manie. Une analyse de ré gres sion mul ti ple a dé - ter miné une re la tion unique et sub stan tielle en tre le jeu com pul sif et l u tili sa tion des ap pa reils de lo te rie vidéo (ALV). Nous avons iden ti fié un groupe de joueurs à risque (score de 1 ou 2 au SOGS), englo bant 14 % de l é chan til lon. Con clu sions : Les modèles de jeux de hasard et de jeu com pul sif à l Î-P-É res sem blent à ceux de la plupart des prov inces. La re - la tion con statée en tre le jeu com pul sif, l u tili sa tion des ALV, et les corré lats cog ni tifs, émo tion nels et de toxi co manie devrait égale ment s ap pliquer à l en sem ble de la popu la tion.