The road less travelled a longitudinal study Rosa Billi, Manager, Research European Association of Gambling Studies Helsinki, Finland, September 2014 Page 1
Victorian Gambling Study Where is Victoria?
Presentation Overview Background Study design Findings Challenges The future Acknowledgements Page 3
The road more travelled... Most gambling epidemiology studies are cross sectional and/or retrospective Assess current participation and problems (prevalence) Give an indication of past participation and problems Provide information on distribution and potential risk/protective factors Additional information, e.g. help-seeking Frequently methodologically compromised Limitations - temporal sequence, causal inference, risk/protective factors for problem onset and progression Prevalence study 'replications' provide estimates of change over time (snapshots at population level) Page 4
Travelling the road A wider variety of research methods, including longitudinal studies. would broaden our base of knowledge about gambling and problem gambling Abbott and Volberg (1996) The road less travelled moving from distribution to determinants in the study of gambling epidemiology. Shaffer et al (2004) The hope that this road will become more travelled and in the process help shift the focus of gambling studies from gambling distribution to gambling determinants. Abbott and Clarke (2007) Page 5
Early journeys along the road Early longitudinal studies - shortcomings: Usually Clinical populations Short time spans Mix of 'add-ons' and stand alone studies Mostly small/moderate samples not representative of general population Various methodological problems including high attrition Psychological focus Prospective (longitudinal) important: Page 6 Volberg (2010) el-guebaly et. al (2008) Relatively little knowledge re people who report problems at a particular point in time (prevalence) and whether these will resolve over time- stability of condition Provides insights into incidence or the number of new cases that develop over time Delfabbfro (2013)
Background Objectives to explore: Risks and vulnerabilities related to changes in gambling status Incidence Movements in and out of PGSI states Eleven hypotheses (including) Gamblers move in and out of PGSI states Problem gambling is transitory in nature Co morbidities are clustered together Chasing wins to cover losses is the biggest predictor of problem gambling Contextual factors contribute to problem gambling EGMs and other continuous forms of play are more likely to result in problem gambling than non continuous forms of play Gamblers with moderately high PGSI scores are more likely to transition to problem gambling. Page 7
Design Cross Sectional (W1) 15,000 RDD CATI survey 70/20/10 representative of Vic population Prospective cohort or longitudinal (W2, W3, W4) annual follow up x 3 Qualitative component face to face interviews (n=44) Page 8
Design Map of Victorian Government Regional Boundaries 2008
Data collection periods Wave One July 2008 - October 2008 Wave Two September 2009 - January 2010 Wave Three September 2010 - January 2011 Qualitative May 2011 - August 2011 Wave Four October 2011 - January 2012
Sample Page 11
Gambling participation questions Gambling participation in 12 activities: informal private betting; electronic gaming machines (EGMs); table games (e.g, blackjack, roulette, poker); horse or harness racing or greyhounds; sports and event results; Lotto, Powerball or the Pools; Keno; scratch tickets; bingo; telephone or SMS competitions; raffles, sweeps and other competitions; and speculative stock investments. Gambling behaviour using the Problem Gambling Screening Index (PGSI): Nine-item index with scores from 0 to 27 Non-gambler, non-problem gambler (PGSI=0), low-risk gambler (PGSI=1-2), moderate-risk gambler (PGSI=3-7), problem gambler (PGSI=8-27) Lifetime risk of gambling using NORC DSM-IV Screen for Gambling Problems Control, Lying and Preoccupation (NODS-CLiP2) scale: Lifetime non-problem gambler (NODS=0); lifetime at-risk gambler (NODS=1,2); lifetime problem gambler (NODS=3-4); lifetime pathological gambler (NODS 5) Page 12
PGSI and NODS CLiP2 PGSI Measures past year risk Non gambler 0 Non problem gambler 0 Low risk gambler 1-2 Moderate risk gambler 3-7 Problem gambler 8+ NODS CLiP2 Measures lifetime risk Lifetime non problem 0 Lifetime at risk1-2 Lifetime problem gambler 3-4 Lifetime pathological gambler 5+
Health and wellbeing questions Core non-gambling questions W1 W2 W3 W4 Health, K10, readiness to change, life events, recreation, smoking CAGE etc Additional contextual questions for specific waves Global Financial Crises (W2) Economic Stimulus Package (W2) Vic Bushfires (W2) Linked Jackpots (W3) Major sporting events (W3) Additional social capital (W4) Trauma and hardship (W1 and W4) Loneliness (W4) Page 14
Some findings Prevalence (total stock) Vic prevalence rate from wave one - 0.7% Incidence (new cases) 12 month incidence rate 0.36% (95% CI 0.21% - 0.57%) NODS CLiP2 0.12% (CI 0.03% - 0.25%) - (of 0.36%) new problem gamblers 0.24% (CI 0.13% - 0.41%) - (of 0.36%) previous history of path/problem gambling relapse Page 15
Co-morbidities Co-morbidity: a condition (or disorder) existing simultaneously but independently with another condition in a person, or a condition (or disorder) in a person that causes, is caused by, or is otherwise related to another condition in the same person. Valderas et al. (2009) Shaffer and Korn (2002) believe that the complex relationships between co-morbid disorders include the possibilities that: both disorders are independent of each other one disorder protects against the other one disorder causes the other both disorders share the same cause or are components of a more complex set of symptoms Page 16
Co-morbidities Co-morbidities (from various cross sectional studies) and problem gamblers: Depression: 37-71% Anxiety disorders: 41-60% Severe psychological distress: 25-30% Personality disorders: 61% (US) Nicotine dependence: 47-64% Alcohol abuse & dependence: 48-72% Drug dependence: 38% (US) Suicide ideation: 9-27% In Victoria more likely to report poor health, lung conditions, obesity and a disability affecting everyday life (wave one). Kessler et al. (2008) Petry et al. (2005) Productivity Commission (1999) Thomas & Jackson (2008) Dept of Justice (2009) Page 17
Co-morbidities conditions 70% 60% 50% 40% 30% 20% 10% 0% current smokers sign of alcohol abuse obesity anxiety depression troubles w ith w ork increased arguments w ith someone close non-problem gamblers low risk gamblers moderate risk gamblers problem gamblers Source: Victorian Gambling Study, 2008 (Sample =15,000, w eighted) unable to get help w hen needed Page 18
Co-morbidities number of co-occurring conditions 70% 60% 50% 40% 30% 20% 10% 0% 0 1 2 3 4 >=5 non-problem gamblers low risk gamblers moderate risk gamblers problem gamblers Source: Victorian Gambling Study, 2008 (Sample =15,000, w eighted) Page 19
Co-morbidities Yeah I ve been, I ve had mental problems since I was little, for social phobia and growing up I was anorexic for ten years, so I didn t go to school, I was hospitalised my whole teenage adolescence and the children s but yeah, it was just always depression, obsessive compulsive disorder and ahm, borderline personality disorder, so. female qualitative study, Victorian Gambling Study, I had the accident, I feel like, I can t do a lot and just you know you get a bit depressed, do you know what I mean? male qualitative study, Victorian Gambling Study But that s what happens. You get depressed, you go and blow your money and then you re depressed because you ve blown your money. So work that out. male qualitative study, Victorian Gambling Study Page 20
Psychological distress Page 21
Alcohol abuse Page 22
Life event triggers Page 23
Co-morbidities Chicken or egg? Little research that clarifies how the onset of problem gambling relates temporally to the onset of other disorders. Question 1 The relationship between onset (new cases) of high risk (MR/PG) gambling behaviour and co morbidities Question 2 The relationship between onset (new cases) of co morbidities and high risk (MR/PG) gambling behaviour Page 24
Co-morbidities The onset of co-morbidities Question 2 The significant variables were: being male (OR=2.0, CI 1.3-3.0, p=0.002) age (OR= 1.02, CI 1.00-1.03, p=0.008) disability (OR=2.1, CI 1.9-4.0, p=0.028), and PGSI problem gambling risk category (OR=4.2, CI 0.9-18.9, p= 0.061). Page 25
Co-morbidities Question 1 The onset of high risk gambling Scoring as an at-risk lifetime gambler (NODS CLiP2) was significantly associated with new onset of high risk gambling behaviour during the study period (OR=6.3, p=0.007, CI 1.7-23.9). Any health condition (OR=2.7, p=0.027, CI 1.1-6.7) Current smoker (OR=2.7, p=0.035, CI 1.1-6.8) Further analysis on any health condition Participants with anxiety were x 4 more likely to develop MRPG (OR=4.0 p=0.036, CI 1.1-14.6) [adjusted for NODS and smoking] Page 26
Transitions Page 27
PGSI Risk Group Transitions Wave Four risk groups from Waves One, Two and Three For example, problem gamblers and their risk group in previous waves 100 90 80 70 60 50 40 30 20 10 0 2008 2009 2010 2011 2008 2009 2010 2011 2008 2009 2010 2011 2008 2009 2010 2011 Page 28 ZR LRG MRG PG Zero risk Low Risk Gamblers Moderate Risk Gamblers Problem Gamblers
PGSI Risk Group Transitions PGSI Risk Group Transitions Wave One risk groups thro Waves Two, Three and Four For example, problem gamblers and their risk group in subsequent waves Page 29
Transitions MR greatest probability of transitioning to PG - 11% Most PG are likely to remain PG, regardless of gender - 71% 19% of PG likely to decrease to MR, and Probability that PG will cease gambling is close to zero - <1%. Markov chain to predict the probabilities of transitioning in and out of gambling risk states. Page 30
Lifetime gambling risk compared to PGSI category across the four waves Page 31
Frequency of EGM use W1 to W3 Page 32
Person time Wave One July 2008 - October 2008 What is person years? The time at risk for all persons in a population Each year a participant contributes to a study = one person year In our study 3686 participants completed all four waves = 14,744 person years Page 33
Person time Wave One July 2008 - October 2008 Most problem gamblers (71%) were likely to remain problem gamblers from one year to the next Approximately 22% of problem gamblers were likely to decrease to moderate risk The probability that problem gamblers were likely to cease gambling was close to zero (0.1%) Moderate risk (9%) had the greatest probability of becoming problem gamblers Non gamblers or non problem gamblers had a very low probability of becoming problem gamblers (0.1%) Page 34
Stability probability CPGSI Category Overall Stability n % % NG 2,148 14.57 48.86% NPG 11,225 76.13 82.51% LR 896 6.08 35.67% MR 345 2.34 43.34% PG 130 0.88 59.09% Total 14,744 100 Page 35
Challenges and learnings Ambitious project Multi supplier model (international, national) (Baseline + additional) funding Time - lack thereof (govt dept) Analysis, analysis, analysis (for example, response rate with priorities) Definitional changes Measurement (PGSI anchors, definitional changes etc) Page 36
Challenges and learnings Attrition (endeavour to counteract) Awareness of mortality (loss of research staff) Collaboration- international doh moments - what should have been asked Familiarity Spreading the word Replication Page 37
Where to from here? Short cohort follow up survey in 2015-2016? (another world first) - report changes over 7-8 years? Short, sharp three-monthly follow up of participants to see short term transitions? Examine remaining hypotheses Secondary analyses currently underway Use findings for targeted prevention (e.g. MR to PG) Collaboration with Sweden & New Zealand & USA (article in press) & data pooling Fact sheets underway Peer reviewed articles (methods paper submitted) Delivery of findings via presentations Page 38 (Caution: findings need to be confirmed via other studies)
Research team Max Abbott Rosa Billi Sarah Hare Damien Jolley Penny Marshall Paul Marden Jan McMillen Elmer Villanueva Rachel Volberg Christine Stone We would like to acknowledge and remember Damien Jolley