Content uploaded by Matthew Browne
Author content
All content in this area was uploaded by Matthew Browne on Jul 25, 2016
Content may be subject to copyright.
ORIGINAL PAPER
Breaking Bad: Comparing Gambling Harms Among
Gamblers and Affected Others
En Li
1
•Matthew Browne
1
•Vijay Rawat
1
•Erika Langham
1
•
Matthew Rockloff
1
Springer Science+Business Media New York 2016
Abstract This article examines gambling harms from both gamblers and affected others’
perspectives. Participants (3076 gamblers and 2129 affected others) completed a retro-
spective survey that elicited information on harms they experienced from gambling across
their lifetime. Their responses were analyzed through testing measurement invariance,
estimating item-response theoretic parameters, calculating percentages, confidence inter-
vals, and correlations, as well as regressions. The results indicated large commonalities in
the experience of harms reported by gamblers and affected others. Further, gamblers
appeared to ‘export’ about half of the harms they experienced to those around them. The
findings also provided detailed profiles of evolving harms as problem gambling severity
varies.
Keywords Gambling harms Gamblers Affected others Gambling problems Item-
response theoretic (IRT) parameters
Introduction
When you have children, you always have family. They will always be your priority,
your responsibility. – Breaking Bad (Season 3, Episode 5) (polokill 2013)
Gambling can have negative impacts not only on gamblers themselves, but can also lead
to adverse consequences for those connected to them, in particular their family and close
friends. Children of problem gamblers could experience reduced material and non-material
well-being, as well as greater risk of developing unhealthy behaviors (Darbyshire et al.
2001; Jacobs et al. 1989). Spouses or partners of people who gambled were also reported to
&En Li
e.li@cqu.edu.au
1
Central Queensland University, Queensland 4701, Australia
123
J Gambl Stud
DOI 10.1007/s10899-016-9632-8
suffer from financial insecurities, health problems, psychological difficulties, or even
deteriorating relationships (Dickson-Swift et al. 2005; Holdsworth et al. 2013; Lorenz and
Shuttlesworth 1983; Lorenz and Yaffee 1988). Parents of people who experienced prob-
lems with gambling might face manipulation from their gambling children, and suffer from
emotional impacts such as fear, guilt, and anger (Kalischuk et al. 2006). In addition to
children, partners, and parents, grandparents, siblings, or close friends of problem gamblers
could also report increased mental health problems and other well-being impairments
(Salonen et al. 2014). A recent review by Velleman et al. (2015) further confirmed various
negative impacts (e.g., on finance, relation, or health) that might occur to gamblers’
affected others.
1
Given the complex nature of gambling harms experienced among affected others, there
has been increasing number of studies assessing these experiences or their consequences,
from either affected others or gamblers’ perspective. For example, Krishnan and Orford
(2002) utilized interviews and a coping questionnaire to examine coping strategies and
received support reported by partners, parents, and landlord of gamblers. In addition to the
coping questionnaire, Orford et al. (2005) also proposed a series of other measures for
testing harmful stress, ill-health experiences, and future hope among family members of
people with gambling (or drug/alcohol) problems. Recently, Dowling et al. (2015) assessed
problem gamblers’ opinions on the effects of their gambling on their own family, through
both qualitative measurement and a newly-developed family impact scale. They found over
30 % of their respondents, namely problem gamblers who sought gambling treatment, did
not admit the existence of family impact due to their gambling (Dowling et al. 2015). This
potential under-reporting through ‘gamblers’ voice’, along with previous studies focusing
on ‘affected others’ voice’ (e.g., Krishnan and Orford 2002; Orford et al. 2005), has
highlighted the importance of assessing gambling harms from both gamblers and affected
others’ perspectives. Moreover, understanding the individual level harms (rather than
societal harms) requires examining both harms occurring to gamblers, as well as harms
experienced by people closely connected to them. So, how do the gambling related harms
experienced by gamblers themselves compare—both qualitatively and by degree—to those
experienced by their family and friends (i.e., affected others)?
Two studies have provided some preliminary answers to this question. Ferland et al.
(2008) conducted an exploratory study to compare the perceptions of seven pathological
gamblers and their spouses. It was found that those spouses’ perceived consequences from
pathological gambling were more severe than those gamblers’ perceptions (Ferland et al.
2008). However, the very limited sample size of the study advises caution in interpreting
the results. In a more recent study, Langham et al. (2016) employed qualitative methods to
understand gambling harms to gamblers and their affected others; via focus groups
(n =35), individual interviews (n =25), and posts from online gambling help/support
forums (n =469). They found that harms could be organized into identical domains for
both gamblers and affected others: those being financial, relationship, emotional/psycho-
logical, health, work/study, cultural, criminal activity, and lifecourse and intergenerational
harms. The authors also identified a large number of specific harms within each domain,
such as ‘bankruptcy’ in the financial domain, ‘reduced performance due to tiredness or
distraction’ in the work/study domain, and ‘reduced engagement in cultural rituals’ in the
cultural domain (Langham et al. 2016). This provides a strong conceptual basis for
approaching harms from gambling, as well as a detailed enumeration of the specific types
1
Affected others is the general term we employ to describe any person with a significant relationship to a
gambler who is affected by the gambler’s behavior.
J Gambl Stud
123
of harms within reasonable categories. However, the qualitative methodology did not
permit comparisons between the two groups—e.g., whether harms to the affected others
had a characteristically different profile from harms to the gamblers themselves. Quanti-
tative work is required in order to make comparisons between gamblers and affected others
regarding the prevalence and degree of harms experienced as a result of gambling.
In addition to the much-needed comparison of gambling harms as they occur to gam-
blers and those around them, there has been a lack of research into the conceptual simi-
larities and differences between gambling problems (as defined by symptoms) and
gambling harms (which are negative outcomes from gambling). As one of the most popular
measures of gambling problems, the Problem Gambling Severity Index (PGSI) developed
by Ferris and Wynne (2001) has been utilized in numerous gambling studies since its
publication (e.g., Browne et al. 2015; Dixon et al. 2014; Li et al. 2016; Rockloff et al.
2014; Wohl et al. 2013). Conceptually the PGSI does not measure the ‘amount of harm
experienced’, but is rather a clinical screening instrument for gambling problems (Hodgins
et al. 2011). However, problem gambling is intimately connected to gambling related
harms, and the PGSI does probe a limited number of key harms as indicators of problem
gambling (it also includes some symptoms that are not harms; Currie et al. 2009; Svetieva
and Walker 2008). What is lacking is a systematic investigation into a broad range of
specific harms associated with different levels of problem gambling/gambling participation
(Blaszczynski 2009; Rodgers et al. 2009), and an understanding of their relative preva-
lences across PGSI categories. The present paper addresses this issue, by taking a com-
prehensive approach and casting a ‘wide net’ in terms of probing for specific harms that
may affect gamblers and those in their immediate social network.
The aims of the current study are threefold, including: examining experiences of
gambling harms from both gamblers and affected others’ perspectives; testing domains of
harms identified by previous qualitative research, through a large-scale retrospective sur-
vey; and assessing a comprehensive set of harms, along with their relationship to the PGSI.
By accomplishing these objectives, we intend to evaluate various harms occurring to
gamblers with different degrees of symptoms, as well as corresponding harms to their
closely related others. The outcome of the survey could also generate insights benefiting
detection of gambling related harms and problems, as well as targeted treatment and
support strategies for both gamblers and affected others.
Method
Development of Harm Checklists
As mentioned, prior qualitative research (Langham et al. 2016) identified a taxonomy of
specific harms based on extant literature, focus groups, interviews, and online forum posts.
These harms were developed into a set of personal statements, following a series of
criteria:
1. Covering the comprehensive set of harms identified within the taxonomy, while using
plain language and providing examples where appropriate.
2. Avoiding content overlap between items while making each item unitary in scope. For
example, a candidate harm ‘spent less time and got less enjoyment from spending time
with people I care about’ was broken into two more specific items. This enabled
respondents to respond definitively to each item.
J Gambl Stud
123
Table 1 Abbreviated and full item labels for potential harms arising from gambling
Item
abbreviation
Full item label
Financial harms
Bankrup Bankruptcy
Loss utilities Loss of supply of utilities (electricity, gas, etc.)
Emerg. Acc. Needed emergency or temporary accommodation
Add. Employ. Took on additional employment
Loss assets Loss of significant assets (e.g. car, home, business, superannuation)
Welfare Needed assistance from welfare organisations (foodbanks or emergency bill payments)
Sold items Sold personal items
Inc. CC. Debt Increased credit card debt
Red. Ben. Exp. Less spending on beneficial expenses such as insurances, education, car and home
maintenance
Red. Ess. Exp. Less spending on essential expenses such as medications, healthcare and food
Late bills Late payments on bills (e.g. utilities, rates)
Red. Rec. Exp. Less spending on recreational expenses such as eating out, going to movies or other
entertainment.
Red. Sav. Reduction of my savings
Red. Spend. Reduction of my available spending money
Work/study harms
Exc. Study Excluded from study
Lost job Lost my job
Conflict Conflict with my colleagues
Hin. Job. Seek Hindered my job-seeking efforts
Resources
a
Used my work or study resources to gamble
Lack Prog. Lack of progression in my job or study
Time
a
Used my work or study time to gamble
Absent Was absent from work or study
Late Was late for work or study
Red. Perf. Reduced performance at work or study (i.e. due to tiredness or distraction)
Health harms
Emerg. Treat.
a
Required emergency medical treatment for health issues caused or exacerbated by
gambling
Overeating Ate too much
Suicide Attempted suicide
Self-harm Committed acts of self harm
Living cond. Unhygienic living conditions (living rough, neglected or unclean housing, etc.)
Service
a
Increased use of health services due to health issues caused or exacerbated by my
gambling
Medical needs Neglected my medical needs (including taking prescribed medications)
Hygiene Neglected my hygiene and self-care
Alcohol Increased my consumption of alcohol
Malnutrition Didn’t eat as much or often as I should
Tobacco Increased my use of tobacco
Physical
activity
a
Reduced physical activity due to my gambling
J Gambl Stud
123
Table 1 continued
Item
abbreviation
Full item label
Stress problems Stress related health problems (e.g. high blood pressure, headaches)
Red. Sleep
gamb
a
Loss of sleep due to spending time gambling
Depression Increased experience of depression
Red. Sleep
Worry
a
Loss of sleep due to stress or worry about gambling or gambling-related problems
Emotional/psychological harms
Escape Thoughts of running away or escape
Worthless Felt worthless
Vulnerable Felt insecure or vulnerable
Ext. Distress Feelings of extreme distress
Failure Felt like a failure
Hopeless
a
Feelings of hopelessness about gambling
Distress
a
Felt distressed about my gambling
Anger
a
Felt angry about not controlling my gambling
Shame
a
Felt ashamed of my gambling
Regret
b
Had regrets that made me feel sorry about my gambling
Escape Thoughts of running away or escape
Relationship harms
Actual ending Actual separation or ending a relationship/s
Belittled Felt belittled in my relationships
Threat ending Threat of separation or ending a relationship/s
Isolation Social isolation (felt excluded or shut-off from others)
Red. Enjoyment Got less enjoyment from time spent with people I care about
Increased
conflict
Experienced greater conflict in my relationships (arguing, fighting, ultimatums)
Reduced events Spent less time attending social events (non gambling related)
Increased
tension
Experienced greater tension in my relationships (suspicion, lying, resentment, etc.)
Neglected Resp. Neglected my relationship responsibilities
Reduced time Spent less time with people I care about
Other harms
Children Unsup. Left children unsupervised
Arrested driving Arrested for unsafe driving
Shame culture
a
Felt that I had shamed my family name within my religious or cultural community
Violence Had experiences with violence (include family/domestic violence)
Theft
government
Petty theft or dishonesty in respect to government, businesses or other people (not
family/friends)
Children
neglected
Didn’t fully attend to needs of children
Red. Connec.
Cult.
Felt less connected to my religious or cultural community
Outcast
a
Outcast from religious or cultural community due to involvement with gambling
Red. Contrib.
Cult.
Reduced my contribution to religious or cultural practices
J Gambl Stud
123
3. Using phrasing that was appropriate regardless of whether the source of the harms was
one’s own gambling, or someone else’s gambling. This facilitated comparisons
between the two groups.
This process resulted in a set of 73 specific potential harms arising from gambling
(Table 1), organized within six of those broad domains adapted from Langham et al.
(2016). These six domains included financial, work/study, health, emotional/psychological,
relationship, and other harms (note that other harms covered both cultural and criminal
activity relevant harms). The large item set led us to adopt a checklist approach; in the
interest of maintaining a reasonable time for study participants to complete, and yielding
more interpretable results in terms of the relative prevalence of harms. Accordingly, for
each domain, the checklist involved participants reviewing the list, and checking each item
if they experienced that issue as a result of the gambling. A single 4-point Likert response
item that assessed the overall level of harm experienced in that domain followed each
domain checklist. For example, the financial domain concluded with the following item:
‘Overall, what level of impact did your gambling have upon your financial security during
this time?’
Survey Design
The key rationale behind our survey design was to understand the prevalence of harmful
outcomes, relative to different levels of gambling problems. Given the low expected
prevalence of currently existing gambling problems or harms in the general population, we
opted for a retrospective survey design in order to elicit information from across partici-
pants’ lifetime. The cost to this decision involved accepting the use of a PGSI modified
slightly to suit retrospective responding. The benefit was to greatly increase the amount of
useful data obtained, for a given sampling effort. We were interested in harms that accrue
to significant others around the gambler (‘affected others’), as well as the gambler
themselves. Participants were requested to focus on the 12-month period of their life when
gambling was causing the most problems. Throughout the survey, participants were
reminded to reflect on that 12-month period. This approach of reminding participants to
continue to focus on the relevant 12-month period was matched by an accompanying
retrospective version of the PGSI. The PGSI items themselves were not modified except
for the addition of the prefix ‘At this time…’ and the utilization of past tense (e.g., ‘At this
time, did gambling cause you any health problems, including stress or anxiety?’). For
Table 1 continued
Item
abbreviation
Full item label
Crime
a
Felt compelled or forced to commit a crime or steal to fund gambling or pay debts
Pay money Promised to pay back money without genuinely intending to do so
Took money Took money or items from friends or family without asking first
a
The core contents of these items remained the same in both questionnaires, however their phrasing was
slightly varied to suit either gamblers’ or affected others’ perspectives. For example, the full item for
‘‘Resources’’ read ‘‘Used my work or study resources to gamble’’ in the questionnaire for gamblers, and read
‘‘Used my work or study resources to assist with matters arising from their gambling’’ in the questionnaire
for affected others
b
This item was only asked in the questionnaire for gamblers
J Gambl Stud
123
affected others, the PGSI was completed second hand, from the perspective of the affected
person (e.g., ‘At this time, did gambling cause the person any health problems, including
stress or anxiety?’). However the harms were measured in both cases as a self-report from
the person who experienced them.
Recruitment of Participants
Our goal for recruitment was to obtain a stratified sample of harms across PGSI categories,
across participants’ lifetime. Recruitment for online completion of the survey was done
through an ISO-accredited Australian commercial panel provider in two stages. In the first
stage of recruitment, the criteria for participation were either: that the participant’s own
gambling had caused them problems, no matter how minor (directed toward the ques-
tionnaire for gamblers), or having had a close relationship with a person whose gambling
had caused them problems, no matter how minor (directed toward the questionnaire for
affected others). Participants completed only one questionnaire. In the case that a partic-
ipant fulfilled both criteria, they were directed to complete the questionnaire for gamblers.
Fig. 1, Panel a illustrates the recruitment process for this stage.
N = 5597 total
responses
N = 4136
completed
N = 1461
screened-
out/incomplete
N = 2458
gamblers
N = 1678 affected
others
N = 1294 total
responses
N = 1069
completed
N = 225 screened-
out/incomplete
N = 618 gamblers
N = 451 affected
others
Panel a:
Panel b:
Fig. 1 Overview of the recruitment process. Panel a: Stage 1 recruitment (inclusion criteria: lifetime
experience of gambling harms). Panel b: Stage 2 recruitment (inclusion criteria: lifetime experience of
frequent gambling)
J Gambl Stud
123
Of the 2458 gamblers and 1678 affected others who met the eligibility criteria and
completed the survey, 71 % of gamblers and 81 % of affected others reported gambling
problems in the most severe problem gambling category (i.e., reporting a PGSI score of 8
or more). These surprisingly high proportions were presumably due to participant inter-
pretation of the screening criteria—those participants who admitted having problems (in
the screen), also tended to be classified by the PGSI as having gambling problems. In order
to achieve a greater representation of participants in lower-risk PGSI categories, we ini-
tiated a second stage of recruitment from the remaining panel members (Fig. 1, Panel b).
The screening criteria were modified to indicate a time in the participant’s life when they
were gambling often (directed toward the questionnaire for gamblers), or had a close
relationship with someone who was gambling often (directed toward the questionnaire for
affected others). Thus, the criteria for inclusion made no reference to gambling problems,
only towards ‘gambling often’. A further 618 gamblers and 451 affected others were
recruited in the second stage, with 35 % of gamblers and 64 % affected others in this group
meeting the criteria for problem gambling based on the PGSI cut-off (i.e., reporting a PGSI
score of 8 or more).
Data Analyses
All analyses were conducted in the R statistical programming environment. Specifically,
we utilized the lavaan package (Rosseel 2012) for testing of measurement invariance of the
PGSI instrument, and the ltm package (Rizopoulos 2006) for estimating item-response
theoretic (IRT) parameters of the harm items, and general purpose functions for calculating
key item indicators such as percentages, confidence intervals, and correlations.
Our sample included a reasonably large proportion of individuals who reported having
experienced gambling problems and harms in the last 12 months. This allowed scope to
check for the validity of retrospective reporting by testing for measurement invariance of
the PGSI instrument between individuals reporting on current problems, and those who
were reflecting on a historical 12-month period in their lives. We also made comparisons in
PGSI responding between those reporting on their own problems, versus those reporting on
the problems of a significant other. This was done via standard measurement tests (Van-
denberg and Lance 2000), comparing group-variant and group-invariant confirmatory
factor analysis (CFA) models using the comparative fit index (CFI) and the root mean
square error of approximation (RMSEA). These are commonly employed in questionnaire
development to ascertain whether an instrument performs similarly across different groups.
In the present context, this allows us to test whether the PGSI instrument was functioning
similarly, as reported by gamblers and affected others. Both CFI and RMSEA incorporate
penalty terms for degrees of freedom, and therefore improved fit measures are possible for
more highly constrained models (in the present cases, involving a common model for both
gamblers and affected others), when the extra degrees of freedom are not supported by data
fit. Additionally, t-tests were employed to compare latent PGSI scale means between
groups of participants.
For each domain of gambling harms considered, a separate IRT model was applied. IRT
modeling assumes the existence of a latent dimensional construct (e.g. financial harm),
higher scores of which are manifested by a greater probability of observing positive scores
on a set of measurable indicators (i.e. the specific harms on our checklists). In a two-
parameter model, items can differ in terms of their severity (i.e., ’difficulty’) and dis-
crimination parameters. A higher item severity parameter means that the indicator tends to
be positive only when latent scores are relatively high. That is, sensitive for capturing
J Gambl Stud
123
differences between high degrees of gambling harm. Conversely, a lower severity
parameter indicates that the item is sensitive for capturing differences between lower levels
of harm. The discrimination parameter describes how reliably the indicator discriminates
individuals overall with respect to the latent construct (i.e., discriminates between higher
and lower levels of the latent construct). Because IRT does not make use of information
outside of the items being considered, it is related to item reliability rather than validity.
Therefore, correlations and cross-tabulations of checklist harms with the PGSI provide an
alternative measure of item functioning with respect to an external test. Prevalence and
95 % confidence intervals were calculated for all four PGSI categories. The items’ point-
biserial correlation coefficient was calculated with respect to (a) the PGSI, (b) the general
domain Likert item measuring ‘overall harm’ (within the respective domain), and (c) the
sum of positive answers in corresponding harm domain (item-total correlation, excluding
the current item). Lastly, we summarized the overall relationship between number of harms
reported and the PGSI. A series of linear and loess smoothed regressions were run for
gamblers and affected others separately. In these simple regression models, the PGSI score
was the predictor, and the number of specific harms reported for each harm domain, and
across all domains, were the response variables. The standardized beta coefficients for
these models provide a means to compare the degree to which each of the domains of
harms are related to increasing gambling problems, as measured by the PGSI.
Results
In total, 3076 (1364 female) complete responses were obtained from participants reporting
on harms arising from their own gambling (hereafter gamblers), and 2129 (1485 female)
responses were obtained from participants reporting on harms arising from the gambling of
a significant other (hereafter affected others). The age distributions in these two participant
groups were very similar, with mean ages of 46.0 and 45.8, and 50 % of participants aged
between 33 and 58 in both groups. Participants in the affected others group reported their
relationships to the person whose gambling had affected them, and the prevalences of those
relationships are presented in Table 2.
PGSI Functioning Across Measurement Groups
A relatively strong assumption can be made that both gamblers and affected others in the
current sample have similar demographic backgrounds, being either online panel partici-
pants or people close to them. We would therefore expect the observed PGSI sample means
Table 2 Affected others’ repor-
ted relationships to gamblers Relationships to gamblers N %
Child 58 2.7
Close friend 375 17.6
Parent 408 19.2
Sibling 142 6.7
Spouse 809 38.0
Other close family member 264 12.4
Close co-worker/colleague 73 3.4
J Gambl Stud
123
and variances for the second-hand reporting done by the affected others to be similar or
close to the first-hand reporting done by the gamblers themselves. Similarly, if the retro-
spective version of the PGSI is functioning equivalently, we would expect not to observe
large differences between current and retrospective reporting. There were four groups
reporting on PGSI scores: current versus retrospective reporting and self-reporting versus
reporting by an affected other, allowing us to make a number of comparisons regarding
equivalent functioning of the PGSI with respect to groups.
Thirty-nine percent (1206) of gamblers reported on gambling problems and harms
experienced currently (i.e. in the last 12 months), while the remainder reported retro-
spectively on a period (median =9 years, inter-quartile range 4–15 years) earlier in their
lives. The current group had a significantly higher mean PGSI score (11.2) than the
retrospective group (9.5); (t(2540.59) =7.76, p\0.01). However, group mean differ-
ences accounted for only 1.9 % of variance in PGSI scores.
Twenty-six percent (561) of participants in the affected-others group reported on cur-
rently experienced harms, and the rest affected others reported retrospectively on a period
(median =10 years, inter-quartile range: 4–15 years) earlier in their lives. No significant
difference was observed between the means of current (11.6) and retrospective (12.1) PGSI
reporting done by affected others (t(949.35) =1.68, p=.09). The variances of PGSI
scores of current and retrospective reports by affected others also did not differ signifi-
cantly (F(560, 1568) =1.09, p=.18). Table 3shows CFI and RMSEA fit indices for a
sequence of three CFA models, testing for measurement invariance on the PGSI for four
contrasts between the four groups of participants. PGSI items were treated as ordinal
indicators. As detailed in Table 3, each row of the table corresponds to a model with a
progressively stronger assumption regarding the item-level measurement invariance of the
PGSI across groups. The best fits were observed for the weakly invariant measurement
model (Model 2), suggesting that item loadings, but not item (ordinal) response thresholds
were invariant across participant groups.
The mean PGSI score reported first-hand by gamblers (10.1) was 1.8 points lower than
that reported second-hand by affected others (11.9), though the effect of first-hand versus
second-hand reporting only accounted for 2.2 % of variance in PGSI scores. The variance
of retrospective PGSI scores reported by affected others compared to gamblers did not
differ significantly (F(1569, 1869) =1.05, p=.23). These results suggest that both the
Table 3 Model comparisons testing for equivalent item functioning for PGSI reporting across groups
Group contrast Subset (df) Current versus retrospective Gamblers versus affected others
Gamblers Affected others Current Retrospective
CFI RMSEA CFI RMSEA CFI RMSEA CFI RMSEA
Model 1 (54) .939 .148 .831 .180 .939 .148 .924 .150
Model 2 (62) .942 .134 .853 .157 .942 .134 .930 .134
Model 3 (79) .849 .192 .821 .153 .849 .192 .675 .257
Model 1: Configural invariance. The same factor structure is imposed on all groups
Model 2: Weak invariance. The factor loadings are constrained to be equal across groups
Model 3: Strong invariance. The factor loadings and intercepts are constrained to be equal across groups
J Gambl Stud
123
retrospective and second person PGSI reporting versions used provided a reasonably valid
indication of the true level of problems experienced.
IRT Parameters, PGSI Categories, and Correlations
Tables 4,5,6,7,8and 9summarize the prevalences of specific harms, for each of the six
domains considered. IRT severity and discrimination parameters are presented, and the
specific harms are ordered with respect to IRT based severity, within each domain. Hence,
the IRT severity parameter places the specific harm on a continuum of ‘harmfulness’ that is
indicated by the whole group of items in each domain from least to most severe. Moreover,
the IRT discrimination parameter indicates how well the item discriminates between low
and high levels of harmfulness. Data for gamblers (Tables 4,5,6) and affected others
(Tables 7,8,9) were analyzed separately, and are therefore presented in different tables. In
addition, the tables also show the percentages and 95 % confidence intervals of each harm
probe for the four PGSI categories (i.e., no identifiable problems, low risk, moderate risk,
problem gamblers). Further, the tables present the point-biserial correlation coefficient
with respect to the PGSI and the general domain Likert harm item. Finally, the item-total
correlation for that harm domain (excluding the current item) is also given, providing a
classical assessment of the reliability of that indicator in reflecting a presumed underlying
dimension of harm within each domain. This set of statistics provide a picture of the
functioning of each item, with respect to the domain and the problem gambling status. To
illustrate, we consider one example, the item ‘Red. Ess. Exp.’ (less spending on essential
expenses such as medications, healthcare and food), when administered to gamblers
(Table 4). This item was the most effective probe for discriminating higher versus lower
levels of financial harm (discrimination parameter = 3.28). It was sensitive for capturing
medium to low levels of financial harm (severity parameter = 0.85). 1.7 % of non-problem
gamblers responded positively to this item, as compared to 30.8 % of problem gamblers
(as measured by the PGSI). The highest prevalence increase between PGSI categories for
this item was between moderate risk and problem gamblers. Consistent with the IRT
results, this item had the highest item-total correlation (.56), the second highest correlation
with the general financial harm Likert item (.40), and the second highest correlation with
the PGSI (.38), among the financial harm items.
As far as the severity parameters were concerned, items such as ‘bankruptcy’ (4.10),
‘excluded study’ (2.27), ‘emergency treatment’ (3.63), ‘escape’ (1.10), ‘actual ending’
(2.49), and ‘children unsupervised’ (2.36) indicated the most severe harm for gamblers,
within the financial, work/study, health, emotional/psychological, relationship, and other
domains respectively. In comparison, ‘bankruptcy’ (3.09), ‘lost job’ (2.78), ‘suicide’
(2.83), ‘failure’ (1.50), ‘neglected responsibilities’ (1.72), and ‘shame culture’ (2.94) were
indicative of the most severe harm for affected others within each of the six corresponding
domains. Hence, ‘bankruptcy’ was the most severe financial harm for both gamblers and
affected others.
According to the discrimination parameters estimated for gamblers, items including
‘reduced essential expenses’ (3.28), ‘absent’ (2.39), ‘service’ (2.28), ‘worthlessness’
(2.89), ‘increased conflict’ (2.39), and ‘outcast’ (2.19) were the most effective in dis-
criminating between low and high levels of harmfulness for their corresponding harm
domains. In comparison, ‘reduced beneficial expenses’ (3.51), ‘absent’ (2.80), ‘service’
(2.92), ‘worthlessness’ (2.57), ‘increased conflict’ (2.42), and ‘took money’ (2.84)
appeared the best discriminator for affected others within the corresponding domains.
Hence, ‘absent’ was the most reliable discriminator for work/study harms on both gamblers
J Gambl Stud
123
Table 4 Financial and work/study harms (gamblers)
Item IRT parameters PGSI categories Correlations
Severity Dscrm Non
problem
Low risk Moderate
risk
Problem PGSI General Total
Financial harms
Bankrup 4.10 0.80 1.7
(0.3, 6.6)
0.0
(0.0, 3.0)
0.7
(0.3, 1.7)
7.0
(5.9, 8.2)
.21 .23 .18
Loss
Utilities
3.53 0.66 1.7
(0.3, 6.6)
0.6
(0.0, 4.0)
2.4
(1.5, 3.8)
15.1
(13.6, 16.8)
.24 .17 .20
Emerg.
Acc.
3.26 1.23 0.0
(0.0, 3.9)
0.6
(0.0, 4.0)
0.4
(0.1,1.1)
5.2
(4.3,6.3)
.19 .17 .24
Add.
Employ.
3.2 0.88 0.0
(0.0, 3.9)
0.0
(0.0, 3.0)
2.5
(1.6, 3.9)
10.8
(9.4, 12.2)
.23 .17 .24
Loss
Assets
3.06 0.79 0.8
(0.0, 5.3)
0.6
(0.0, 4.0)
2.3
(1.4, 3.6)
14.8
(13.3, 16.5)
.27 .26 .25
Welfare 2.27 1.01 2.5
(0.7, 7.8)
0.0
(0.0, 3.0)
3.0 (2.0, 4.5) 18.1
(16.4, 19.8)
.30 .29 .30
Sold Items 1.28 1.42 3.4
(1.1, 9.0)
1.3
(0.2, 5.0)
3.3
(2.2, 4.8)
30.2
(28.2, 32.3)
.41 .37 .44
Inc. CC.
Debt
1.04 0.83 3.4
(1.1, 9.0)
3.8 (1.6,
8.5)
17.0
(14.5,19.8)
42.2
(40.1,44.5)
.36 .36 .32
Red. Ben.
Exp.
0.85 2.81 1.7
(0.3, 6.6)
1.9
(0.5, 5.9)
10.6
(8.6, 13.0)
31.1
(29.1, 33.2)
.36 .37 .55
Red. Ess.
Exp.
0.85 3.28 1.7
(0.3, 6.6)
1.9 (0.5,
5.9)
8.7
(6.9,10.9)
30.8
(28.8, 32.9)
.38 .40 .56
Late Bills 0.71 1.98 3.4
(1.1, 9.0)
1.9
(0.5, 5.9)
14.6
(12.3, 17.2)
39.4
(37.2,41.6)
.36 .43 .52
Red. Rec.
Exp.
0.02 1.72 8.5
(4.4,
15.4)
19.7
(14.0,27.0)
47.3
(43.8,50.7)
55.1
(52.9, 57.3)
.23 .27 .42
Red. Sav. -0.41 0.87 7.6
(3.8,
14.4)
21.0
(15.1,
28.4)
51.5
(48.0, 55.0)
65.9
(63.8, 68.0)
.27 .33 .29
Red.
Spend.
-0.69 1.32 16.1
(10.2,
24.3)
30.6
(23.6,
38.5)
69.2
(66.0,72.3)
71.3
(69.2, 73.3)
.21 .27 .33
Work/study harms
Exc. Study 2.27 1.32 0.0
(0.0, 3.9)
0.6
(0.0, 4.0)
2.9
(1.9, 4.3)
12.2
(10.8, 13.8)
.18 .33 .28
Lost Job 2.17 1.54 2.5
(0.7, 7.8)
0.0
(0.0, 3.0)
1.7
(1.0, 2.9)
11.1
(9.8, 12.6)
.24 .44 .33
Conflict 2.09 1.99 3.4
(1.1, 9.0)
0.0
(0.0, 3.0)
1.4
(0.8, 2.6)
8.4
(7.2, 9.7)
.20 .34 .40
Hin. Job.
Seek
2.04 1.38 0.0
(0.0, 3.9)
1.3
(0.2, 5.0)
3.4
(2.3, 4.9)
14.5
(13.0, 16.2)
.25 .35 .36
Resources 1.76 2.25 2.5
(0.7, 7.8)
2.5
(0.8, 6.8)
1.8
(1.1, 3.0)
11.7
(10.3, 13.2)
.24 .35 .47
Lack Prog. 1.63 1.92 1.7
(0.3, 6.6)
0.6
(0.0, 4.0)
4.3
(3.1, 6.0)
15.8
(14.3, 17.5)
.23 .40 .46
Time
a
1.36 1.88 2.5
(0.7, 7.8)
1.3
(0.2, 5.0)
7.0
(5.4, 9.0)
21.7
(19.9, 23.6)
.28 .41 .45
J Gambl Stud
123
and affected others. This was also the case for ‘service’ for health harms, ‘worthlessness’
for emotional/psychological harms, and ‘increased conflict’ for relationship harms.
Among problem gamblers identified by the PGSI, ‘reduced spending’ was reported
as the most prevalent financial harm (71.3 %; 95 % confidence interval: 69.2%, 73.3%; for
simplicity reason please refer to Tables 4,5,6,7,8and 9for the rest 95 % confidence
intervals). Moreover, ‘reduced performance’ (30.4 %), ‘reduced sleep worry’ (48.3 %),
‘shame’ (60.8 %), ‘reduced time’ (51.5 %), and ‘pay money’ (18.4 %) also displayed the
highest prevalence for those problem gamblers, within the work/study, health, emotional/
psychological, relationship, and other domains respectively. On other hand, ‘reduced
spending’ (40.6 %), ‘reduced performance’ (27.5 %), ‘reduced sleep worry’ (42.7 %),
‘distress’ (66.2 %), ‘increased tension’ (52.0 %), and ‘violence’ (16.3 %) demonstrated the
highest prevalence among problem gamblers’ affected others, within their corresponding
domains. Hence, ‘reduced spending’ was the financial harm most frequently reported by
both problem gamblers and their affected others. Likewise, for both groups, ‘reduced
performance’ was the most frequently reported work/study harm, and ‘reduced sleep
worry’ was the most frequently reported health harm.
For gamblers, harms such as ‘sold items’ (.41), ‘absent’ (.30)/‘reduced performance’
(.30), ‘reduced sleep worry’ (.42), ‘extreme distress’ (.50), ‘increased tension’ (.39), and
‘pay money’ (.31) had the highest correlations with their reported PGSI, within corre-
sponding domains. Across these domains, on other hand, ‘sold items’ (.25), ‘reduced
performance’ (.20), ‘stress problems’ (.23)/‘reduced sleep worry’ (.23), ‘extreme distress’
(.29), ‘neglected responsibilities’ (.17), and ‘took money’ (.18) were harms possessing the
highest correlations with PGSI reported by affected others. Hence, ‘sold items’, ‘reduced
performance’, ‘reduced sleep worry’, and ‘extreme distress’ were, respectively, the most
reliable financial, work/study, health, and emotional/psychological consequences of
increasing gambling problems, among both gamblers and affected others.
For gamblers, harms including ‘late bills’ (.43), ‘reduced performance’ (.50), ‘depres-
sion’ (.48), ‘extreme distress’ (.51), ‘increased conflict’ (.49), and ‘pay money’ (.40)
displayed the strongest correlations with their corresponding general domain Likert harm
items. Similarly, ‘reduced spending’ (.51), ‘reduced performance’ (.53), ‘depression’ (.51)/
‘reduced sleep worry’ (.51), ‘extreme distress’ (.54), ‘increased conflict’ (.46)/‘increased
tension’ (.46), and ‘violence’ (.54), as reported by affected others, also had the strongest
correlations with their general domain items. Hence, ‘reduced performance’, ‘depression’,
‘extreme distress’, and ‘increased conflict’ were, respectively, the most reliable predictors
of general work/study, health, emotional/psychological, and relationship harms, for both
gamblers and affected others.
Table 4 continued
Item IRT parameters PGSI categories Correlations
Severity Dscrm Non
problem
Low risk Moderate
risk
Problem PGSI General Total
Absent 1.27 2.39 0.8
(0.0, 5.3)
1.9
(0.5, 5.9)
4.6
(3.3, 6.3)
21.7
(19.9, 23.6)
.30 .46 .52
Late 1.21 2.07 1.7
(0.3, 6.6)
1.9
(0.5, 5.9)
7.2
(5.6, 9.3)
24.4
(22.5, 26.4)
.29 .44 .49
Red. Perf. 1.09 1.56 2.5
(0.7, 7.8)
3.2
(1.2, 7.7)
12.5
(10.4, 15.0)
30.4
(28.4, 32.5)
.30 .50 .42
J Gambl Stud
123
Table 5 Health and emotional/psychological harms (Gamblers)
Item IRT parameters PGSI categories Correlations
Severity Dscrm Non
problem
Low risk Moderate
risk
Problem PGSI General Total
Health harms
Emerg.
Treat.
3.63 0.72 0.0
(0.0, 3.9)
0.6
(0.0, 4.0)
1.6
(0.9, 2.7)
12.2
(10.8, 13.8)
.24 .20 .15
Overeating 2.88 0.76 5.1
(2.1, 11.2)
0.6
(0.0, 4.0)
7.6
(5.9, 9.7)
15.4
(13.9, 17.1)
.16 .19 .19
Suicide 2.84 1.25 1.7
(0.3, 6.6)
1.9
(0.5, 5.9)
0.7
(0.3, 1.7)
7.5
(6.4, 8.7)
.19 .28 .25
Self-Harm 2.48 1.97 0.8
(0.0, 5.3)
1.3
(0.2, 5.0)
0.6
(0.2, 1.5)
4.9
(4.0, 5.9)
.20 .27 .33
Living
Cond.
2.37 1.06 1.7
(0.3, 6.6)
0.6
(0.0, 4.0)
3.6
(2.5, 5.2)
15.1
(13.6, 16.8)
.23 .28 .27
Service 2.11 2.28 0.8
(0.0, 5.3)
0.0
(0.0, 3.0)
0.6
(0.2, 1.5)
7.2
(6.1, 8.5)
.26 .29 .40
Medical
needs
1.69 1.98 0.8
(0.0, 5.3)
0.0
(0.0, 3.0)
1.3
(0.7, 2.4)
15.2
(13.7, 16.9)
.34 .34 .45
Hygiene 1.60 1.79 1.7
(0.3, 6.6)
0.6
(0.0, 4.0)
3.9
(2.7, 5.5)
17.5
(15.9, 19.3)
.29 .33 .45
Alcohol 1.32 0.85 5.1
(2.1, 11.2)
12.7
(8.1, 19.2)
22.2
(19.4, 25.2)
32.2
(30.2, 34.3)
.18 .25 .30
Malnutrition 1.30 1.29 0.8
(0.0, 5.3)
3.8
(1.6, 8.5)
11.2
(9.2, 13.6)
28.4
(26.5, 30.5)
.29 .31 .39
Tobacco 1.18 0.92 2.5
(0.7, 7.8)
6.4
(3.3, 11.7)
19.8
(17.2, 22.7)
35.0
(32.9, 37.2)
.25 .28 .32
Physical
activity
0.90 1.44 4.2
(1.6, 10.1)
5.1
(2.4, 10.1)
17.0
(14.5, 19.8)
36.0
(33.8, 38.1)
.30 .35 .43
Stress
Problems
0.89 1.87 5.1
(2.1, 11.2)
3.2
(1.2, 7.7)
12.3
(10.2, 14.8)
34.1
(32.0, 36.3)
.36 .42 .49
Red. Sleep
Gamb.
0.79 1.43 5.1
(2.1, 11.2)
3.2
(1.2, 7.7)
15.3
(13.0, 18.0)
40.6
(38.4, 42.8)
.36 .38 .43
Depression 0.75 1.77 3.4
(1.1, 9.0)
3.2
(1.2, 7.7)
12.5
(10.4, 15.0)
40.2
(38.0, 42.4)
.37 .48 .48
Red. Sleep
Worry
0.54 1.56 2.5
(0.7, 7.8)
1.3
(0.2, 5.0)
16.6
(14.2, 19.4)
48.3
(46.1, 50.6)
.42 .45 .44
Emotional/psychological harms
Escape 1.10 1.52 4.2
(1.6, 10.1)
3.2
(1.2, 7.7)
7.2
(5.6, 9.3)
32.2
(30.2, 34.3)
.40 .48 .46
Worthless. 0.86 2.89 3.4
(1.1, 9.0)
1.3
(0.2, 5.0)
5.2
(3.8, 7.0)
32.3
(30.2, 34.4)
.44 .47 .62
Vulnerable 0.77 1.99 4.2
(1.6, 10.1)
4.5
(2.0, 9.3)
11.1
(9.1, 13.5)
37.4
(35.3, 39.6)
.40 .43 .55
Ext.
Distress
0.64 1.59 4.2
(1.6, 10.1)
3.2
(1.2, 7.7)
9.0
(7.2, 11.3)
46.3
(44.1, 48.6)
.50 .51 .48
Failure 0.47 1.87 3.4
(1.1, 9.0)
3.2
(1.2, 7.7)
14.1
(11.9, 16.7)
49.0
(46.8, 51.3)
.43 .46 .53
Hopeless. 0.43 1.73 3.4
(1.1, 9.0)
1.3
(0.2, 5.0)
17.5
(15.0, 20.3)
50.2
(48.0, 52.4)
.42 .45 .51
Distress 0.27 1.90 6.8
(3.2, 13.3)
3.8
(1.6, 8.5)
24.5
(21.6, 27.6)
53.2
(51.0, 55.4)
.38 .45 .55
Anger 0.09 1.70 2.5
(0.7, 7.8)
3.8
(1.6, 8.5)
34.1
(30.9, 37.5)
58.2
(56.0, 60.4)
.36 .41 .52
J Gambl Stud
123
Based on the item-total correlations calculated for gamblers, harms like ‘reduced
essential expenses’ (.56), ‘absent’ (.52), ‘stress problems’ (.49), ‘worthlessness’ (.62),
‘increased conflict’(.55)/‘neglected responsibilities’ (.55), and ‘took money’ (.44)
demonstrated strongest associations with the rest items in their corresponding domains.
Similarly, ‘reduced beneficial expenses’ (.61), ‘absent’ (.51), ‘depression’ (.52), ‘vulner-
able’ (.54), ‘increased conflict’ (.54), and ‘took money’ (.49), as reported by affected
others, also possessed the strongest item-total correlations in their respective domains.
Hence, ‘absent’, ‘increased conflict’, and ‘took money’ were, respectively, the most reli-
able indicators in reflecting the underlying dimensions of work/study, relationship, and
other harms, for both gamblers and affected others.
Number of Harms with Respect to PGSI
The number of specific harms overall, and broken down by domains, were regressed
against the PGSI. As results, the linear regression slopes estimated for each harm domain
and across all domains are listed in Table 10. Non-linear loess smoothed curves are dis-
played in Fig. 2. A common pattern arising from these regressions was greater slopes being
found among gamblers who reported their own PGSI , as compared to affected others who
reported others’ PGSI (in approximately 2:1 ratios). For example, each one-point increase
in PGSI was associated with 0.26 (vs. 0.14) more financial harms, 0.14 (vs. 0.07) more
work/study harms, 0.28 (vs. 0.13) more health harms, 0.24 (vs. 0.11) more relationship
harms, and 1.34 (vs. 0.64) more harms across all domains for gamblers (vs. affected
others). Hence, the amount of experienced harms was more strongly related to PGSI for
gamblers than affected others. This is an unsurprising finding, given the potential link
between gamblers’ PGSI and harms upon affected others is more distal than harms upon
gamblers themselves. Furthermore, the strength of this link also depends on the nature of
the relationships between gamblers and affected others.
Discussion
The results of the survey point to a high level of correspondence in gambling harms
experienced by gamblers and affected others. In addition, harms in all domains tended to
accumulate more quickly to gamblers than to affected others as gambling problems
increased. The results also provide a great deal of insights into the specific harms from
each domain checklist: in terms of their prevalences, as indicators of harm severity levels,
Table 5 continued
Item IRT parameters PGSI categories Correlations
Severity Dscrm Non
problem
Low risk Moderate
risk
Problem PGSI General Total
Shame 0.04 1.79 5.1
(2.1, 11.2)
3.8
(1.6, 8.5)
32.8
(29.6, 36.1)
60.8
(58.6, 62.9)
.38 .41 .51
Regret 0.01 1.17 3.4
(1.1, 9.0)
17.2
(11.8,
24.2)
46.9
(43.5, 50.4)
55.8
(53.6, 58.0)
.22 .33 .42
J Gambl Stud
123
Table 6 Relationship and other harms (gamblers)
Item IRT parameters PGSI categories Correlations
Severity Dscrm Non
problem
Low
risk
Moderate
risk
Problem PGSI General Total
Relationship harms
Actual ending 2.49 0.84 0.8
(0.0, 5.3)
1.9
(0.5, 5.9)
3.3
(2.2, 4.8)
19.5
(17.8, 21.3)
.27 .34 .22
Belittled 1.69 1.80 1.7
(0.3, 6.6)
1.3
(0.2, 5.0)
4.1
(2.9, 5.7)
15.6
(14.0, 17.3)
.28 .33 .42
Threat ending 1.42 1.52 1.7
(0.3, 6.6)
2.5
(0.8, 6.8)
4.3
(3.1, 6.0)
24.8
(23.0, 26.8)
.32 .46 .40
Isolation 1.08 1.31 4.2
(1.6, 10.1)
4.5
(2.0, 9.3)
10.5
(8.5, 12.8)
34.4
(32.3, 36.6)
.34 .42 .41
Red.
enjoyment
0.99 1.55 2.5
(0.7, 7.8)
3.8
(1.6, 8.5)
11.7
(9.6, 14.1)
33.9
(31.8, 36.0)
.35 .37 .47
Increased
conflict
0.85 2.39 4.2
(1.6, 10.1)
3.2
(1.2, 7.7)
9.9
(8.0, 12.2)
33.4
(31.3, 35.6)
.35 .49 .55
Reduced
events
0.80 1.27 3.4
(1.1, 9.0)
6.4
(3.3,
11.7)
19.5
(16.9,
22.4)
40.1
(37.9, 42.3)
.31 .34 .43
Increased
tension
0.67 2.15 1.7
(0.3, 6.6)
3.2
(1.2, 7.7)
14.0
(11.7,
16.6)
40.4
(38.2, 42.6)
.39 .47 .53
Neglected
Resp.
0.57 2.19 4.2
(1.6, 10.1)
3.2
(1.2, 7.7)
16.4
(14.0,
19.1)
43.3
(41.1, 45.5)
.36 .48 .55
Reduced
Time
0.33 1.62 6.8
(3.2, 13.3)
7.6
(4.2,
13.3)
27.0
(24.1,
30.2)
51.5
(49.2, 53.7)
.34 .46 .47
Other harms
Children
Unsup.
2.36 1.99 0.8
(0.0, 5.3)
0.6
(0.0, 4.0)
1.0
(0.4, 2.0)
5.6
(4.7, 6.8)
.18 .24 .35
Arrested
driving
2.35 1.86 0.0
(0.0, 3.9)
0.6
(0.0, 4.0)
0.8
(0.4, 1.8)
6.5
(5.5, 7.7)
.16 .23 .32
Shame
culture
a
2.31 1.77 3.4
(1.1, 9.0)
0.6
(0.0, 4.0)
1.7
(1.0, 2.9)
7.1
(6.0, 8.3)
.17 .28 .35
Violence 2.20 1.62 0.8
(0.0, 5.3)
0.6
(0.0, 4.0)
1.3
(0.7, 2.4)
9.9
(8.7, 11.4)
.20 .36 .35
Theft
Government
2.06 1.52 2.5
(0.7, 7.8)
0.6
(0.0, 4.0)
3.0
(2.0, 4.5)
12.3
(10.9, 13.9)
.24 .39 .35
Children
Neglected
2.03 1.46 1.7
(0.3, 6.6)
1.9
(0.5, 5.9)
4.3
(3.1, 6.0)
13.0
(11.5, 14.6)
.18 .28 .34
Red. Connec.
Cult.
2.03 1.71 1.7
(0.3, 6.6)
1.9
(0.5, 5.9)
2.9
(1.9, 4.3)
10.8
(9.4, 12.2)
.17 .26 .36
Outcast
a
1.91 2.19 0.8
(0.0, 5.3)
0.6
(0.0, 4.0)
1.1
(0.5, 2.1)
10.0
(8.8, 11.5)
.20 .29 .41
Red. Contrib.
Cult.
1.86 2.01 0.8
(0.0, 5.3)
0.6
(0.0, 4.0)
3.0
(2.0, 4.5)
11.3
(10.0, 12.8)
.19 .28 .41
Crime
a
1.72 2.06 0.8
(0.0, 5.3)
0.6
(0.0, 4.0)
2.1
(1.2, 3.3)
13.9
(12.4, 15.5)
.26 .36 .41
Pay money 1.59 1.82 1.7
(0.3, 6.6)
0.6
(0.0, 4.0)
2.7
(1.7, 4.1)
18.4
(16.7, 20.2)
.31 .40 .41
J Gambl Stud
123
and as consequences of gambling problems. We shall discuss each of the harm domains
separately below.
Financial Harms
Among all examined financial harms, bankruptcy was found to be the most severe one for
both gamblers and affected others. However, it was also a relatively unreliable indicator of
harm for both groups—most likely due to its low prevalence, and the fact that bankruptcy
is affected by a number of other factors, such as one’s capacity to borrow money and incur
debt. Reduced spending (including on recreational activities) was among the least severe
financial harms, supporting both the intuitive idea and the findings from Langham et al.
(2016); that the first impact of even moderate gambling spending is to reduce funds
available for other activities. Reduced spending on essential and beneficial items were the
most reliable indicators of financial harms, for both groups. This accords with what one
might propose as the definition of financial gambling harms, which is to divert money away
from expenditure necessary to provide for basic needs and additional wants. The preva-
lence of this ‘everyday’ inability to spend money on essential and beneficial items also
varied strongly with respect to evolving problem gambling status. Interestingly, the
prevalence of both indicators was quite low among the non-problem gamblers (1.7 % and
1.7 % respectively), but much higher among the affected others by non-problem gamblers
(18.8 % and 9.4 %). This prevalence rose dramatically to around one-third of problem
gamblers (30.8 % and 31.1 %), and to a lesser degree, to around one quarter of affected
others by problem gamblers (23.0 % and 23.8 %). Another harm of particular interest was
sold items, which had the strongest correlations to the PGSI, among both gamblers and
affected others. Hence, selling items to fund gambling should be the most reliable financial
consequence of increasing gambling problems for both groups.
Work/Study Harms
Within this domain, being excluded from study, losing one’s job, and conflict at work were
the most severe harms for both gamblers and affected others. On the contrary, being absent,
being late, or reduced performance for work or study were the least severe, and also very
reliable, indicators of work/study harms for both groups. Thus, as gambling problems
begin to occur, a reliable early warning sign would appear to be the action of skipping
work in order to gamble—or to deal with the consequences of a close one’s gambling.
Interestingly, absenteeism could be related to a number of co-morbid harms (e.g., financial,
health, emotional/psychological harms), however as an isolated behavior, it offers a ready,
early indicator for risk of harms among both gamblers and affected others. Performance
reduction was also the most reliable work/study consequence of increasing gambling
Table 6 continued
Item IRT parameters PGSI categories Correlations
Severity Dscrm Non
problem
Low
risk
Moderate
risk
Problem PGSI General Total
Took money 1.57 2.04 1.7
(0.3, 6.6)
0.6
(0.0, 4.0)
2.9
(1.9, 4.3)
17.1
(15.5, 18.9)
.28 .39 .44
J Gambl Stud
123
Table 7 Financial and work/study harms (affected others)
Item IRT parameters PGSI categories Correlations
Severity Dscrm Non
problem
Low risk Moderate
risk
Problem PGSI General Total
Financial harms
Bankrup. 3.09 1.37 0.0
(0.0, 13.3)
0.0
(0.0, 10.0)
0.9
(0.2, 2.8)
3.7
(2.8, 4.8)
.08 .23 .23
Add.
Employ.
2.84 1.29 3.1
(0.2, 18.0)
0.0
(0.0, 10.0)
2.1
(0.9, 4.5)
5.5
(4.4, 6.8)
.09 .19 .24
Emerg.
Acc.
2.73 1.58 3.1
(0.2, 18.0)
0.0
(0.0, 10.0)
1.2
(0.4, 3.3)
4.1
(3.2, 5.2)
.10 .22 .26
Loss
Assets
2.08 1.41 6.2
(1.1, 22.2)
0.0
(0.0, 10.0)
3.3
(1.8, 6.0)
11.1
(9.6, 12.8)
.16 .32 .35
Loss
Utilities
2.05 1.73 3.1
(0.2, 18.0)
2.3
(0.1, 13.5)
1.8
(0.7, 4.1)
8.8
(7.5, 10.4)
.14 .27 .37
Welfare 1.75 1.80 6.2
(1.1, 22.2)
0.0
(0.0, 10.0)
3.6
(2.0, 6.4)
12.1
(10.5, 13.9)
.19 .34 .41
Inc. CC.
Debt
1.52 1.39 12.5
(4.1, 29.9)
2.3
(0.1, 13.5)
10.8
(7.8, 14.8)
18.5
(16.6, 20.6)
.15 .39 .40
Sold
Items
1.48 1.84 12.5
(4.1, 29.9)
0.0
(0.0, 10.0)
1.2
(0.4, 3.3)
17.1
(15.3, 19.1)
.25 .37 .48
Red. Ess.
Exp.
0.96 3.11 18.8
(7.9, 37.0)
0.0
(0.0, 10.0)
8.4
(5.8, 12.1)
23.0
(21.0, 25.3)
.16 .44 .59
Red. Ben.
Exp.
0.89 3.51 9.4
(2.5, 26.2)
4.5
(0.8, 16.7)
12.3
(9.1, 16.5)
23.8
(21.7, 26.1)
.13 .43 .61
Red. Sav. 0.85 1.30 9.4
(2.5, 26.2)
6.8
(1.8, 19.7)
16.3
(12.6,
20.8)
34.2
(31.8, 36.6)
.22 .47 .43
Late Bills 0.80 2.60 3.1
(0.2, 18.0)
2.3
(0.1, 13.5)
16.3
(12.6,
20.8)
28.4
(26.2, 30.8)
.18 .48 .58
Red. Rec.
Exp.
0.46 2.04 12.5
(4.1, 29.9)
20.5
(10.3, 35.8)
36.4
(31.3,
41.9)
37.5
(35.0, 40.0)
.06 .41 .52
Red.
Spend.
0.46 1.82 9.4
(2.5, 26.2)
11.4
(4.3, 25.4)
26.2
(21.6,
31.3)
40.6
(38.1, 43.1)
.18 .51 .52
Work/study harms
Lost Job 2.78 1.73 6.2
(1.1, 22.2)
0.0
(0.0, 10.0)
0.9
(0.2, 2.8)
3.2
(2.4, 4.2)
.08 .34 .25
Conflict 2.59 1.50 0.0
(0.0, 13.3)
0.0
(0.0, 10.0)
1.2
(0.4, 3.3)
5.8
(4.7, 7.1)
.13 .27 .29
Exc.
Study
2.53 1.60 9.4
(2.5, 26.2)
0.0
(0.0, 10.0)
2.7
(1.3, 5.3)
5.0
(4.0, 6.3)
.09 .33 .28
Hin. Job.
Seek
2.36 1.51 3.1
(0.2, 18.0)
0.0
(0.0, 10.0)
5.4
(3.3, 8.6)
6.7
(5.5, 8.1)
.08 .29 .33
Resources 2.28 1.73 0.0
(0.0, 13.3)
0.0
(0.0, 10.0)
1.5
(0.6, 3.7)
6.7
(5.5, 8.1)
.12 .28 .34
Time 2.05 1.71 6.2
(1.1, 22.2)
0.0
(0.0, 10.0)
2.7
(1.3, 5.3)
9.0
(7.6, 10.6)
.15 .30 .36
Lack
Prog.
1.63 2.31 6.2
(1.1, 22.2)
6.8
(1.8, 19.7)
5.1
(3.1, 8.2)
11.1
(9.6, 12.8)
.11 .43 .48
J Gambl Stud
123
problems, as well as the most reliable predictor of general work/study harm, for both
participant groups. Additionally, performance reduction was reported by the highest per-
centages of low-risk, moderate-risk, and problem gamblers (as identified by PGSI) , as well
as their affected others.
Health Harms
Attempted suicide, requiring emergency treatment, overeating, and self-harm were the
most severe indicators of health-related harms among both gamblers and affected others.
However, none of these were as reliable as increased use of health services in discrimi-
nating between low and high levels of harmfulness. Early indicators of health-related
harms included reduced sleep due to worry, stress, and depression. These were among the
most reliable indicators of health-related harms, which more generally appeared to be those
health impacts associated with emotional distress. Further, reduced sleep due to worry not
only had the highest correlations with PGSI within the health domain, but was also the
health harm most frequently reported by both problem gamblers and their affected others.
Hence, loss of sleep is a health-related impact that could trouble both gamblers and people
around them early on, then keep occurring as gambling problems become exacerbated.
Emotional/Psychological Harms
Within this domain, feelings of failure, worthlessness, escaping, extreme distress and
vulnerability were the most extreme harms for both gamblers and affected others. Feelings
of regret and shame were reliable early indicators of harms for gamblers themselves, whilst
feelings of anger and hopelessness were among the negative emotions that tended to be
first felt by those affected. These differing emotions for the two groups make sense, given
the different roles, responsibilities, and perceived controls among the gamblers and the
affected others. A sobering and somewhat surprising result was that feelings of worth-
lessness were among the most reliable indicators of emotional/psychological harms not
only for gamblers, but also for affected others. This suggests that affected others tend to
share and internalize the threat to self-regard that uncontrollable gambling instigates.
These results are consistent with emerging awareness in terms of the psychological well-
being among problem gamblers. For example, recent research indicates that pathological
gambling could be associated with higher scores of anxious and depressive symptoms of
psychological health (Jauregui et al. 2016).
Table 7 continued
Item IRT parameters PGSI categories Correlations
Severity Dscrm Non
problem
Low risk Moderate
risk
Problem PGSI General Total
Late 1.53 2.60 0.0
(0.0, 13.3)
2.3
(0.1, 13.5)
6.6
(4.3, 10.0)
11.7
(10.2, 13.5)
.16 .37 .50
Absent 1.51 2.80 3.1
(0.2, 18.0)
0.0
(0.0, 10.0)
3.6
(2.0, 6.4)
12.2
(10.6, 14.0)
.19 .44 .51
Red. Perf. 0.97 1.71 3.1
(0.2, 18.0)
9.1
(3.0, 22.6)
16.9
(13.1,
21.4)
27.5
(25.2, 29.8)
.20 .53 .42
J Gambl Stud
123
Table 8 Health and emotional/psychological harms (affected others)
Item IRT parameters PGSI categories Correlations
Severity Dscrm Non
problem
Low risk Moderate
risk
Problem PGSI General Total
Health harms
Suicide 2.83 1.84 3.1
(0.2, 18.0)
0.0
(0.0, 10.0)
0.0
(0.0, 1.4)
2.7
(2.0, 3.7)
.09 .17 .24
Self-harm 2.66 2.03 3.1
(0.2, 18.0)
0.0
(0.0, 10.0)
1.8
(0.7, 4.1)
2.4
(1.7, 3.4)
.07 .20 .27
Emerg.
Treat.
2.63 1.62 3.1
(0.2, 18.0)
0.0
(0.0, 10.0)
1.8
(0.7, 4.1)
4.4
(3.5, 5.6)
.10 .21 .27
Overeating 2.44 1.05 6.2
(1.1, 22.2)
0.0
(0.0, 10.0)
6.3
(4.1, 9.7)
11.6
(10.0, 13.3)
.10 .23 .25
Alcohol 2.22 1.09 6.2
(1.1, 22.2)
9.1
(3.0, 22.6)
9.3
(6.5, 13.1)
12.6
(11.0, 14.4)
.07 .20 .29
Living
Cond.
2.16 1.65 9.4
(2.5, 26.2)
2.3
(0.1, 13.5)
2.7
(1.3, 5.3)
8.0
(6.7, 9.5)
.12 .28 .34
Hygiene 2.13 2.33 3.1
(0.2, 18.0)
4.5
(0.8, 16.7)
1.8
(0.7, 4.1)
5.0
(4.0, 6.3)
.12 .25 .40
Service 2.01 2.92 0.0
(0.0, 13.3)
0.0
(0.0, 10.0)
1.2
(0.4, 3.3)
4.9
(3.9, 6.1)
.12 .29 .42
Medical
needs
1.87 2.72 3.1
(0.2, 18.0)
0.0
(0.0, 10.0)
1.2
(0.4, 3.3)
6.8
(5.6, 8.2)
.15 .30 .45
Physical
activity
1.73 1.63 6.2
(1.1, 22.2)
4.5
(0.8, 16.7)
3.6
(2.0, 6.4)
13.9
(12.2, 15.8)
.15 .31 .41
Tobacco 1.73 1.28 9.4
(2.5, 26.2)
4.5
(0.8, 16.7)
8.7
(6.0, 12.4)
16.9
(15.0, 18.9)
.12 .29 .35
Malnutrition 1.47 1.77 9.4
(2.5, 26.2)
4.5
(0.8, 16.7)
9.6
(6.8, 13.5)
16.3
(14.5, 18.2)
.12 .37 .42
Red. Sleep
Gamb.
1.39 1.56 9.4
(2.5, 26.2)
0.0
(0.0, 10.0)
9.6
(6.8, 13.5)
19.9
(17.9, 22.0)
.19 .35 .42
Depression 0.93 2.22 6.2
(1.1, 22.2)
4.5
(0.8, 16.7)
16.9
(13.1, 21.4)
25.9
(23.7, 28.2)
.17 .51 .52
Stress
problems
0.92 1.98 12.5
(4.1, 29.9)
2.3
(0.1, 13.5)
11.7
(8.6, 15.8)
28.5
(26.3, 30.9)
.23 .49 .49
Red. Sleep
Worry
0.46 1.48 9.4
(2.5, 26.2)
15.9
(7.2, 30.7)
24.7
(20.2, 29.8)
42.7
(40.2, 45.3)
.23 .51 .42
Emotional/psychological harms
Failure 1.50 1.92 9.4
(2.5, 26.2)
0.0
(0.0, 10.0)
5.7
(3.6, 8.9)
15.4
(13.6, 17.3)
.17 .33 .42
Worthless. 1.38 2.57 12.5
(4.1, 29.9)
0.0
(0.0, 10.0)
7.5
(5.0, 11.1)
14.2
(12.5, 16.1)
.13 .36 .48
Escape 0.91 1.67 15.6
(5.9, 33.5)
0.0
(0.0, 10.0)
15.4
(11.7, 19.8)
29.4
(27.1, 31.7)
.17 .51 .46
Vulnerable 0.84 2.45 9.4
(2.5, 26.2)
6.8
(1.8, 19.7)
15.1
(11.5, 19.5)
27.1
(24.9, 29.4)
.15 .45 .54
Ext.
Distress
0.57 2.05 12.5
(4.1, 29.9)
2.3
(0.1, 13.5)
16.9
(13.1, 21.4)
37.5
(35.0, 40.0)
.29 .54 .53
Shame 0.55 1.34 12.5
(4.1, 29.9)
15.9
(7.2, 30.7)
28.0
(23.3, 33.2)
39.3
(36.8, 41.8)
.16 .37 .45
Anger 0.21 1.12 12.5
(4.1, 29.9)
13.6
(5.7, 28.0)
32.5
(27.6, 37.9)
49.4
(46.9, 52.0)
.19 .38 .40
J Gambl Stud
123
Relationship Harms
Among both gamblers and affected others, experiencing greater relationship conflict was
the most reliable discriminator and predictor of relationship harms. Conflict within rela-
tionships is generally regarded as a reliable indicator of underlying disagreements or
relationship problems, and it is reasonable that it may serve as a reliable signal of gambling
problems. Nevertheless, there were illuminating differences between the two groups for
some indicators. For example, neglect of responsibilities was one of the earliest and least
severe indicators of relationship harms for gamblers themselves. However, neglect of
responsibilities appeared to be the most severe relationship harm for affected others. This
has an intuitively appealing interpretation in terms of gambling problems causing a
‘cascade’ of responsibility neglect through the social networks around gamblers. Initially,
gamblers are able to compensate for their time and money investments in gambling by
relying on those around them to absorb the duties. However, as pressures on those around
them increase with more severe gambling problems, they in turn will become more likely
to neglect their responsibilities—a second order relationship effect.
Other Harms
Unlike other domains, the domain for other harms did not have an underlying construct
attached, due to the diversity of harms allocated under this domain. Therefore IRT results
for this domain should be interpreted with great caution. Focusing on the relation-
ships between harm indicators and the PGSI; feeling compelled to commit a crime, not
intending to pay back money, and taking money without asking first were most reliably
associated with the PGSI among gamblers. This was also the case for affected others,
among whom having experiences with violence and neglecting the needs of children were
also more strongly associated with the PGSI. In general, associations within this domain,
both in terms of reliability and with respect to the PGSI, were much lower than other
domains. This reflects the fact that harms within this domain were diverse, very specific,
and with quite low prevalences.
Implications
The outcome of this survey has important implications for detecting, treating, and
addressing gambling related harms and problems. For both gamblers and people who are
close to them, if they are only experiencing those specific harms with low IRT severity
scores, that should provide them a valuable early signal. That is, early detection based on
Table 8 continued
Item IRT parameters PGSI categories Correlations
Severity Dscrm Non
problem
Low risk Moderate
risk
Problem PGSI General Total
Hopeless. 0.19 1.23 28.1
(14.4,
47.0)
15.9
(7.2, 30.7)
31.6
(26.7, 37.0)
49.4
(46.9, 52.0)
.22 .38 .43
Distress -0.50 1.20 21.9
(9.9, 40.4)
27.3
(15.5,
43.0)
48.2
(42.7, 53.7)
66.2
(63.7, 68.5)
.20 .40 .39
J Gambl Stud
123
Table 9 Relationship and other harms (affected others)
Item IRT parameters PGSI categories Correlations
Severity Dscrm Non
problem
Low risk Moderate
risk
Problem PGSI General Total
Relationship harms
Neglected
Resp.
1.72 1.40 9.4
(2.5, 26.2)
4.5
(0.8, 16.7)
6.6
(4.3, 10.0)
15.8
(14.0,
17.8)
.17 .24 .39
Actual ending 1.27 1.23 18.8
(7.9, 37.0)
0.0
(0.0, 10.0)
15.1
(11.5, 19.5)
25.1
(23.0,
27.4)
.15 .40 .36
Isolation 1.25 1.41 12.5
(4.1, 29.9)
9.1
(3.0, 22.6)
16.0
(12.3, 20.5)
22.8
(20.8,
25.1)
.13 .32 .42
Red.
Enjoyment
1.25 1.20 12.5
(4.1, 29.9)
13.6
(5.7, 28.0)
15.7
(12.0, 20.1)
25.7
(23.5,
28.0)
.14 .25 .40
Threat ending 1.12 1.25 12.5
(4.1, 29.9)
4.5
(0.8, 16.7)
20.2
(16.1, 25.0)
27.0
(24.8,
29.3)
.14 .37 .37
Belittled 1.12 2.14 9.4
(2.5, 26.2)
2.3
(0.1, 13.5)
16.3
(12.6, 20.8)
20.5
(18.5,
22.6)
.10 .36 .52
Reduced
events
1.05 1.91 6.2
(1.1, 22.2)
15.9
(7.2, 30.7)
12.7
(9.4, 16.8)
24.1
(22.0,
26.3)
.15 .30 .52
Reduced time 0.96 1.28 21.9
(9.9, 40.4)
18.2
(8.7, 33.2)
21.1
(16.9, 26.0)
30.0
(27.7,
32.3)
.13 .31 .44
Increased
conflict
0.21 2.42 15.6
(5.9, 33.5)
13.6
(5.7, 28.0)
32.8
(27.9, 38.2)
46.7
(44.2,
49.3)
.16 .46 .54
Increased
tension
0.04 2.27 12.5
(4.1, 29.9)
9.1
(3.0, 22.6)
42.2
(36.8, 47.7)
52.0
(49.5,
54.5)
.13 .46 .51
Other harms
Shame
culture
2.94 1.25 0.0
(0.0, 13.3)
0.0
(0.0, 10.0)
1.2
(0.4, 3.3)
5.6
(4.5, 6.9)
.12 .26 .27
Arrested
driving
2.86 1.96 3.1
(0.2, 18.0)
0.0
(0.0, 10.0)
0.6
(0.1, 2.4)
2.1
(1.5, 3.0)
.07 .14 .24
Outcast 2.77 1.84 6.2
(1.1, 22.2)
0.0
(0.0, 10.0)
2.7
(1.3, 5.3)
2.4
(1.7, 3.3)
.06 .16 .28
Red. Contrib.
Cult.
2.58 1.72 0.0
(0.0, 13.3)
0.0
(0.0, 10.0)
2.4
(1.1, 4.9)
4.2
(3.3, 5.4)
.07 .20 .33
Red. Connec.
Cult.
2.51 1.54 6.2
(1.1, 22.2)
2.3
(0.1, 13.5)
4.2
(2.4, 7.1)
5.3
(4.2, 6.5)
.05 .26 .34
Crime 2.46 2.26 0.0
(0.0, 13.3)
0.0
(0.0, 10.0)
0.6
(0.1, 2.4)
3.3
(2.5, 4.4)
.15 .22 .34
Theft
government
2.30 2.04 0.0
(0.0, 13.3)
0.0
(0.0, 10.0)
1.2
(0.4, 3.3)
5.1
(4.0, 6.3)
.11 .24 .35
Children
Unsup.
2.20 2.29 0.0
(0.0, 13.3)
2.3
(0.1, 13.5)
1.2
(0.4, 3.3)
4.9
(3.9, 6.2)
.08 .20 .39
Children
neglected
1.73 1.82 3.1
(0.2, 18.0)
2.3
(0.1, 13.5)
5.4
(3.3, 8.6)
12.2
(10.7,
14.0)
.11 .30 .41
J Gambl Stud
123
these minor harms could drive them toward acting and preventing potentially more
deleterious consequences from occurring. Meanwhile, as the study findings indicate that
harms could build up about twice as fast among gamblers compared to affected others, a
creative detection approach would be to ‘estimate’ the degree of problems a gambler is
facing by measuring the amount of harms being experienced by her/his close family or
friends. This approach might be particularly valuable when gamblers try to lie or deny
about the real problems (Velleman et al. 2015).
Other findings may also lead to useful insights for organizations that provide treatment
or support for gamblers or affected others. These providers can prioritize their resource
allocation toward those high-prevalence harms under different PGSI categories, and design
corresponding strategies that can reduce or minimize their impacts. They may also pay
special attention to those specific harms that could affect motivations or barriers to
treatment/help-seeking. For example, not only was shame an early harm indicator, it was
also most frequently reported emotional harm by problem gamblers. Since shame has been
identified as a key barrier against help-seeking (Evans and Delfabbro 2005; Gainsbury
et al. 2014), it would be imperative for the treatment/support providers to develop targeted
public awareness campaigns to address this issue.
Conclusions and Limitations
Overall the present findings have pointed to some large commonalities in gambling-related
harms occuring to gamblers and affected others. The most notable difference between the
two groups appears to be in quantity, rather than quality of experienced harms. In
Table 9 continued
Item IRT parameters PGSI categories Correlations
Severity Dscrm Non
problem
Low risk Moderate
risk
Problem PGSI General Total
Violence 1.68 1.44 3.1
(0.2, 18.0)
2.3
(0.1, 13.5)
6.6
(4.3, 10.0)
16.3
(14.5,
18.3)
.13 .54 .38
Pay money 1.55 1.97 12.5
(4.1, 29.9)
2.3
(0.1, 13.5)
6.0
(3.8, 9.3)
14.1
(12.4,
16.0)
.14 .29 .42
Took money 1.53 2.84 3.1
(0.2, 18.0)
0.0
(0.0, 10.0)
4.8
(2.9, 7.9)
11.3
(9.8, 13.1)
.18 .31 .49
Table 10 Linear regression slopes
Slopes PGSI (self-report) PGSI (of other)
Financial harm domain 0.26 0.14
Work/study harm domain 0.14 0.07
Health harm domain 0.28 0.13
Emotional/psychological harm domain 0.31 0.14
Relationship harm domain 0.24 0.11
Other harm domain 0.12 0.06
All harm domains 1.34 0.64
J Gambl Stud
123
particular, regressions of harms on PGSI among affected others have only generated slopes
approximately half the size of those among gamblers. In other words, gamblers can
seemingly ’export’ about half of the harms they have experienced to people in their close
social networks.
The results also provided evidence supporting the conceptual differences between
gambling harms and PGSI-screened gambling problems. Particularly worth mentioning are
those specific harms within each domain that possessed the highest correlation with PGSI,
only one of which turned out the most severe harm within the corresponding domain (i.e.,
neglected responsibilities within the relationship domain, and only for the affected others).
The checklists of harms developed for different domains, could also serve as practical and
useful tools for future research into gambling related harms, on either gamblers or affected
others.
There are still a few limitations to the present study. Given the complexity of the current
findings and the small sample size of certain subgroups, we did not break down the
heterogeneity of affected others in our analysis. However, harms experienced by affected
others could vary considerably depending on the type of relationship they have with the
gamblers. Hence, an important direction for future research is to examine the effects of
relationship category/proximity on the type/extent of harms occurring to affected others, as
well as their willingness to stay close to/supportive of the person who gambles.
Another limitation lies in the lack of cross-cultural comparisons in our study design.
During recent past, a growing gambling literature has pointed to the important roles
culture/ethnicity can play in shaping gambling and relevant phenomena (e.g., Chamberlain
et al. 2016; Dhillon et al. 2011; Medeiros et al. 2015; Orford et al. 2005; Svetieva and
Walker 2008). Investigating harm experiences and implementing the developed harm
checklists cross-culturally, therefore, would lead to deeper insights on ways of evaluating
and reducing harms for gamblers and affected others in different parts of the world.
Admittedly, affected others surveyed in this study assessed PGSI second-hand, and the
potential issues in this approach need to be acknowledged. Argubly, participants evaluating
the problems of another person would be less likely to minimize problems, leading to a
potentially greater mean score. This was reflected in the slightly higher mean PGSI score
reported by affected others. However, this difference accounted for only 2.2 % of the
Fig. 2 Fitted regression lines
J Gambl Stud
123
variance. In other respects, the PGSI appeared to function equivalently between participant
groups. Therefore, despite this limitation, we consider the responses of affected others to
provide a reasonably sound representation of harms that accrue with increasing PGSI.
The findings of this study provide critical evidence that similar harms can occur to both
gamblers and people close to them. It also presents detailed profiles of evolving harms as
problem gambling severity increases, and identifies the type of harms that most effectively
discriminate between different levels of gambling problems. These findings are of rele-
vance to treatment and support providers in identifying and addressing harm-minimization
needs of both gamblers and affected others, and helping both break away from the ‘bad’ of
gambling.
Acknowledgment This research was funded by a grant from the Victorian Responsible Gambling Foun-
dation. A previous version of this paper has been submitted as part of a research report to the Victorian
Responsible Gambling Foundation.
Compliance with Ethical Standards
Conflict of interest En Li has received research grants from the Victorian Responsible Gambling Foun-
dation and Gambling Research Australia. Matthew Browne has received grants from the Victorian
Responsible Gambling Foundation, the New Zealand Ministry of Health and Gambling Research Australia.
Erika Langham has received grants from the Victorian Responsible Gambling Foundation, the New Zealand
Ministry of Health and Gambling Research Australia. Matthew Rockloff has received grants from the
Queensland Treasury, the Victorian Treasury, the Victorian Responsible Gambling Foundation, the New
Zealand Ministry of Health and Gambling Research Australia. Vijay Rawat declares no conflicts of interest.
Ethical Approval All procedures performed in this study were in accordance with the ethical standards of
the institutional and/or national research committee and with the 1964 Helsinki declaration and its later
amendments or comparable ethical standards.
References
Blaszczynski, A. (2009). Problem gambling: We should measure harm rather than ‘cases’. Addiction,
104(7), 1072–1074.
Browne, M., Langham, E., Rockloff, M. J., Li, E., Donaldson, P., & Goodwin, B. (2015). EGM jackpots and
player behaviour: An in-venue shadowing study. Journal of Gambling Studies, 31(4), 1695–1714.
Chamberlain, S. R., Leppink, E., Redden, S. A., Odlaug, B. L., & Grant, J. E. (2016). Racial-ethnic related
clinical and neurocognitive differences in adults with gambling disorder. Psychiatry Research, 242,
82–87.
Currie, S. R., Miller, N., Hodgins, D. C., & Wang, J. (2009). Defining a threshold of harm from gambling for
population health surveillance research. International Gambling Studies, 9(1), 19–38.
Darbyshire, P., Oster, C., & Carrig, H. (2001). The experience of pervasive loss: Children and young people
living in a family where parental gambling is a problem. Journal of Gambling Studies, 17(1), 23–45.
Dhillon, J., Horch, J. D., & Hodgins, D. C. (2011). Cultural influences on stigmatization of problem
gambling: East Asian and Caucasian Canadians. Journal of Gambling Studies, 27(4), 633–647.
Dickson-Swift, V. A., James, E. L., & Kippen, S. (2005). The experience of living with a problem gambler:
Spouses and partners speak out. Journal of Gambling Issues,. doi:10.4309/jgi.2005.13.6.
Dixon, M. J., Harrigan, K. A., Santesso, D. L., Graydon, C., Fugelsang, J. A., & Collins, K. (2014). The
impact of sound in modern multiline video slot machine play. Journal of Gambling Studies, 30(4),
913–929.
Dowling, N. A., Suomi, A., Jackson, A. C., & Lavis, T. (2015). Problem gambling family impacts:
Development of the Problem Gambling Family Impact Scale. Journal of Gambling Studies,. doi:10.
1007/s10899-015-9582-6.
Evans, L., & Delfabbro, P. H. (2005). Motivators for change and barriers to help-seeking in Australian
problem gamblers. Journal of Gambling Studies, 21(2), 133–155.
Ferland, F., Fournier, P. M., Ladouceur, R., Brochu, P., Bouchard, M., & Pa
ˆquet, L. (2008). Consequences
of pathological gambling on the gambler and his spouse. Journal of Gambling Issues, 22, 219–229.
J Gambl Stud
123
Ferris, J., & Wynne, H. (2001). The Canadian Problem Gambling Index: Final report. Submitted for the
Canadian Centre on Substance Abuse.
Gainsbury, S., Hing, N., & Suhonen, N. (2014). Professional help-seeking for gambling problems:
Awareness, barriers and motivators for treatment. Journal of Gambling Studies, 30(2), 503–519.
Hodgins, D. C., Stea, J. N., & Grant, J. E. (2011). Gambling disorders. The Lancet, 378(9806), 1874–1884.
Holdsworth, L., Nuske, E., Tiyce, M., & Hing, N. (2013). Impacts of gambling problems on partners:
Partners’ interpretations. Asian Journal of Gambling Issues and Public Health, 3(1), 1–14.
Jacobs, D. F., Marston, A. R., Singer, R. D., Widaman, K., Little, T., & Veizades, J. (1989). Children of
problem gamblers. Journal of Gambling Behavior, 5(4), 261–268.
Jauregui, P., Urbiola, I., & Estevez, A. (2016). Metacognition in pathological gambling and its relationship
with anxious and depressive symptomatology. Journal of Gambling Studies, 32(2), 675–688.
Kalischuk, R. G., Nowatzki, N., Cardwell, K., Klein, K., & Solowoniuk, J. (2006). Problem gambling and its
impact on families: A literature review. International Gambling Studies, 6(1), 31–60.
Krishnan, M., & Orford, J. (2002). Gambling and the family: From the stress-coping-support perspective.
International Gambling Studies, 2(1), 61–83.
Langham, E., Thorne, H., Browne, M., Donaldson, P., Rose, J., & Rockloff, M. (2016). Understanding
gambling related harm: A proposed definition, conceptual framework, and taxonomy of harms. BMC
Public Health,. doi:10.1186/s12889-016-2747-0.
Li, E., Rockloff, M. J., Browne, M., & Donaldson, P. (2016). Jackpot structural features: Rollover effect and
goal-gradient effect in EGM gambling. Journal of Gambling Studies, 32(2), 707–720.
Lorenz, V. C., & Shuttlesworth, D. E. (1983). The impact of pathological gambling on the spouse of the
gambler. Journal of Community Psychology, 11(1), 67–76.
Lorenz, V. C., & Yaffee, R. A. (1988). Pathological gambling: Psychosomatic, emotional and marital
difficulties as reported by the spouse. Journal of Gambling Behavior, 4(1), 13–26.
Medeiros, G. C., Leppink, E. W., Yaemi, A., Mariani, M., Tavares, H., & Grant, J. E. (2015). Electronic
gaming machines and gambling disorder: A cross-cultural comparison between treatment-seeking
subjects from Brazil and the United States. Psychiatry Research, 230, 430–435.
Orford, J., Templeton, L., Velleman, R., & Copello, A. (2005). Family members of relatives with alcohol,
drug and gambling problems: A set of standardized questionnaires for assessing stress, coping and
strain. Addiction, 100(11), 1611–1624.
Polokill. (2013). Favorite quotes from Breaking Bad. Retrieved from http://www.imdb.com/list/
ls056274521/.
Rizopoulos, D. (2006). ltm: An R package for latent variable modeling and item response theory analyses.
Journal of Statistical Software, 17(5), 1–25.
Rockloff, M. J., Browne, M., Li, E., & O’Shea, T. (2014). It’s a sure bet you’re going to die: Existential
terror promotes gambling urges in problem players. Gambling Research, 26(1), 33–43.
Rodgers, B., Caldwell, T., & Butterworth, P. (2009). Measuring gambling participation. Addiction, 104(7),
1065–1069.
Rosseel, Y. (2012). lavaan: An R package for structural equation modeling. Journal of Statistical Software,
48(2), 1–36.
Salonen, A. H., Castre
´n, S., Alho, H., & Lahti, T. (2014). Concerned significant others of people with
gambling problems in Finland: A cross-sectional population study. BMC Public Health,. doi:10.1186/
1471-2458-14-398.
Svetieva, E., & Walker, M. (2008). Inconsistency between concept and measurement: The Canadian
Problem Gambling Index (CPGI). Journal of Gambling Issues, 22, 157–173.
Vandenberg, R. J., & Lance, C. E. (2000). A review and synthesis of the measurement invariance literature:
Suggestions, practices, and recommendations for organizational research. Organizational Research
Methods, 3(1), 4–70.
Velleman, R., Cousins, J., & Orford, J. (2015). Effects of gambling on the family. In H. Bowden-Jones & S.
George (Eds.), A clinician’s guide to working with problem gamblers (pp. 90–103). London, New
York: Routledge/Taylor & Francis Group.
Wohl, M. J., Gainsbury, S., Stewart, M. J., & Sztainert, T. (2013). Facilitating responsible gambling: The
relative effectiveness of education-based animation and monetary limit setting pop-up messages
among electronic gaming machine players. Journal of Gambling Studies, 29(4), 703–717.
J Gambl Stud
123
A preview of this full-text is provided by Springer Nature.
Content available from Journal of Gambling Studies
This content is subject to copyright. Terms and conditions apply.
