Most gambling harm is attributable to electronic gaming machines (EGMs) because of their inherently risky design features, such as high speed of betting, and relatively high participation rate (Browne et al., 2023). Importantly, EGM-related harm is not confined only to players with a clinically diagnosable gambling disorder, but also extends to those experiencing less severe gambling problems (Browne et al., 2016, 2017; Canale et al., 2016; Salonen et al., 2018). Few people with a severe disorder seek professional gambling treatment before a crisis point, and low-moderate risk gamblers rarely use formal help even when experiencing gambling harm (Bijker et al., 2022). However, gamblers are generally more amenable to using self-regulatory strategies (SRS) to control their gambling and believe they are able to implement them (Bagot et al., 2021; Hing et al., 2012; Lubman et al., 2015). It is therefore important that SRSs promoted to gamblers are evidence-based as part of a broad suite of public health measures aimed at safer gambling consumption, products, environments, and policies.

Self-regulatory gambling strategies have been conceptualised as two categories that seek broadly different goals (Rodda et al., 2019): (1) behaviour change strategies to reduce, regain control over, and resolve an existing gambling problem and (2) protective behavioural strategies to prevent gambling harm from occurring by limiting and maintaining control over gambling. In practice, these two types of SRSs show considerable overlap, as evident below.

Behaviour Change Strategies

Research has explored strategies used to regain control over gambling and address a gambling disorder. In a study of 43 ‘resolved problem gamblers’ (Hodgins & El-Guebaly, 2000), commonly used strategies included stimulus control, treatment seeking, cognitive strategies and social support. In a focus group study (Thomas et al., 2010), higher-risk gamblers reported they required intensive strategies like abstinence, replacing gambling with healthier activities and help-seeking, whereas lower-risk gamblers reported that setting limits, maintaining awareness and keeping gambling social were sufficient to maintain control. A survey of 238 ‘social gamblers’ and 68 ‘problem gamblers’ identified five types of strategies: cognitive approaches, direct action, social experience, avoidance and limit setting (Moore et al., 2012). Participants with a gambling problem who were actively trying to reduce their gambling were the most likely to use the strategies. However, while setting limits is a common strategy, higher-risk gamblers report relatively lower adherence to this strategy (Abbott et al., 2014; Thomas et al., 2010).

Rodda et al. (2017); Rodda, Hing, et al. (2018b) analysed online counselling sessions and forum posts related to problem gambling. They identified six primary change strategies: cash control, social support, avoiding or limiting gambling, engaging in alternative activities, changing thoughts and beliefs and self-assessment and self-monitoring. They then administered an inventory of 99 behaviour change strategies to 489 people who had experienced a gambling problem (Rodda, Bagot, et al., 2018a). Those meeting criteria for past-year problem gambling reported all strategies as more helpful compared to lower-risk gamblers, except for planning and financial control strategies. The problem gambling group rated cognitive strategies as most helpful, such as reminding themselves of the consequences of gambling. Conversely, non-problem gamblers reported that setting financial limits was most helpful (Knaebe et al., 2019).

Protective Behavioural Strategies

Several studies have focused on protective behavioural strategies to limit gambling, stay in control and prevent harmful consequences. Lostutter et al. (2014) administered the Gambling Protective Behavior Scale to 1922 US college student gamblers. Harm reduction strategies, such as resisting chasing losses and keeping track of spending, were associated with lower gambling quantity and problem gambling severity. Avoidance strategies, such as not carrying bank cards and refraining from gambling when feeling down, were associated with lower gambling frequency but not quantity or problem severity. These results suggest that higher-risk gamblers seeking to abstain from gambling may be more likely to use avoidance strategies, whereas lower-risk gamblers prefer harm reduction strategies. Amongst 860 regular gamblers in Victoria Australia (Hing et al., 2017), lower-risk gamblers were more likely to report using harm reduction strategies, including setting limits, gambling for pleasure rather than to win money and balancing gambling with other activities.

Wood and Griffiths (2015) examined ‘positive play’, based on players who do not report problematic gambling behaviours. Their Positive Play Scale (Wood et al., 2017) identified four factors—honesty and control, pre-commitment, personal responsibility, and gambling literacy—which correlate negatively with disordered gambling (Delfabbro et al., 2020; Tabri et al., 2020; Tong et al., 2020). However, the scale mostly assesses responsible gambling beliefs, rather than a set of actionable strategies that gamblers can adopt. Rodda et al. (2019) investigated the strategies that 184 gamblers used to adhere to their limits on EGMs. Lower-risk gamblers used certain strategies more frequently, such as avoiding chasing losses and viewing gambling as entertainment. Higher-risk gamblers more frequently asked family or friends to look after cards or cash in the venue. Finally, Hing et al. (2019) surveyed 1174 gamblers in Canada to assess 43 potential SRSs. Certain strategies, such as stopping gambling if it is not enjoyable and setting a dedicated gambling budget, predicted lower gambling harm. Conversely, strategies including researching gambling strategies and using gambling to make money were linked to increased harm.

Gaps in Current Evidence

Prior research provides useful insights into the SRSs that gamblers use. However, the evidence base supporting their efficacy is weak due to several design limitations. First, all studies have been cross-sectional, identifying correlations but not causation between SRS-use and subsequent gambler-risk status. Second, findings are obscured by conceptual overlap between behaviour change strategies and protective behavioural strategies (Rodda et al., 2019). This may explain why conflicting results exist for whether lower-risk or higher-risk gamblers are more likely to use SRSs. Some SRSs tend to be used only by people with a gambling problem who want to reduce their gambling, such as avoidance, cognitive and help-seeking strategies. Thus, whether lower-risk gamblers use more or fewer SRSs than higher-risk gamblers depends on the set of strategies measured. Third, including SRSs used only by people wanting to address a gambling problem further obscures results because their use correlates with higher-risk rather than lower-risk gambling. Less frequent gamblers may not use some strategies simply because they have no need to do so. This indicates the importance of focusing research on players who are vulnerable to experiencing gambling harm. Otherwise, including gamblers not needing to use SRSs is likely to cloud results. Finally, the purpose of protective behavioural strategies is to minimise harm from gambling, but only two studies have included a measure of gambling harm (Delfabbro et al., 2020; Hing et al., 2019). The current study attempted to overcome these limitations.

The Current Study

The two studies presented here were conducted in the Australian state of New South Wales (NSW) where EGMs are legally available only in land-based clubs, hotels and casinos (and not online). Study 1 aimed to identify a set of SRSs that best predict less harmful gambling amongst EGM players who are most vulnerable to EGM-related harm. Its specific objectives were to (a) examine the use of SRSs, (b) identify the most efficacious SRSs and (c) examine the use of the most efficacious SRSs and by personal characteristics (age, gender, PGSI and EGM frequency). These findings informed study 2 which aimed to test the efficacy of the most efficacious SRSs from study 1 as a brief intervention in a randomised controlled trial. The objectives for study 2 were to examine the effects on three outcome variables (expenditure on EGMs, time spent playing EGMs, and harm from EGM play) of (a) the assignment of any treatment condition vs the control condition, (b) the assignment of the individual SRSs vs the control condition and (c) the frequency of SRS utilisation.

Study 1: Survey of EGM Gamblers

Methods

Study 1 was approved by Central Queensland University Human Research Ethics Committee (#22741).

Recruitment

Online panel aggregator, Qualtrics, recruited participants via online convenience panels in November and December 2020. Potential participants were emailed a link to an information sheet and online survey. The population of interest was regular EGM gamblers, that is, people who are most likely to experience harm from EGM use, who lived in the state of the funding body. Inclusion criteria were informed consent, living in NSW, aged 18 years or older and gambling on EGMs at least monthly. Of the 2053 respondents who fully completed the survey, 21 were removed for failing quality checks that assessed attention and straight-lining through questions. This left 2032 participants for analysis.

Measures

Where relevant, measures referred to the most common terminology for EGMs in NSW, the ‘pokies’.

Self-Regulatory Strategies (SRSs)

Participants were asked their agreement or disagreement to using each of 45 SRSs in relation to their pokies gambling (Table 2). The SRSs were distilled from a comprehensive list of SRSs promoted to gamblers and assessed in prior research (Hing et al., 2019; Hing, Russell, & Hronis, 2016b) and further refined based on recent research (Rodda et al., 2019; Rodda, Hing, et al., 2018b).

The Problem Gambling Severity Index (PGSI; Ferris & Wynne, 2001) was administered in relation to the last 12 months. We used the validated scoring of ‘never’ = 0, ‘sometimes’ = 1, ‘most of the time’ = 2 and ‘almost always’ = 3, and the validated categories of ‘non-problem gambler’ = 0, ‘low-risk gambler’ = 1–2, ‘moderate-risk gambler’ = 3–7 and ‘problem gambler’ = 8–27. Cronbach’s alpha was .95.

GHS

The Short Gambling Harms Screen (SGHS; Browne et al., 2018) and the Unimpeachable Gambling Harms Scale (UGHS; Murray-Boyle et al., 2021) were combined for analysis to form the Gambling Harms Scale (GHS). While the 10-item SGHS is a reliable and validated measure of gambling-related harm, it includes some items that arguably describe minor harms. The 10-item UGHS was therefore included to add further probes that are incontrovertibly harmful and serious (e.g. ‘late payment on bills’). Respondents were asked if, over the last 12 months, they had experienced each of the 20 harms as a result of their pokies gambling (no/yes), with higher scores indicating more harms. Cronbach’s alpha was .94.

Risk Factors for Gambling Harm

Several known risk factors for experiencing significant gambling harm (Browne et al., 2019) were measured: frequency of pokies gambling, frequency of playing pokies alone, whether any adults in their household had a gambling problem when the respondent was growing up, the Gambling Urge Scale (Raylu & Oei, 2004; Cronbach’s alpha was .96.), Gambling Fallacies Measure (2003; Cronbach’s alpha was .70.), a single item to rate the importance of religion or spirituality in their life (Likert: 1 not at all to 5 extremely important) and the Barratt Impulsiveness Scale—Brief (Steinberg et al., 2013; Cronbach’s alpha was .73).

Demographics

Participants reported their age and other demographics (Table 1).

Table 1 Study 1: Participant characteristics
Full size table

Participant Characteristics

The sample comprised 59.7% males and 40.3% females (Table 1). Age ranged from 18 to 87 years (m = 41.1 years). Table 1 details other demographic characteristics.

The mean number of harms reported from the GHS was 5.60 (SD = 6.01) from a possible range of 1–20. The mean score on the Gambling Fallacies Measure was 4.82 (SD = 2.46) from a possible range of 0–10, with higher scores indicating greater resistance to gambling fallacies. The mean Gambling Urge Scale score was 19.40 (SD = 10.61) from a possible range of 6–42, with higher scores indicating higher urges. Around half the participants (47.5%) rated religion or spirituality as moderately to extremely important in their lives. The Barratt Impulsiveness Scale mean score was 17.29 (SD = 4.08) from a possible range of 8–32, with higher scores indicating higher impulsivity.

Analysis

Individuals experiencing gambling-related harm may employ SRSs to improve upon their outcomes. This self-selection in using SRSs complicates a simple analysis to identify which SRSs may be effective in reducing harm. Some of the worst affected gamblers are likely to employ some good strategies that are nevertheless not 100% effective. To address this confounding issue, the analyses employed a propensity matching approach to create two matched groups of persons that are either harmed or not harmed by gambling. Critically, after selection and propensity weighting, both groups have an equal chance (or propensity) for being harmed by gambling based on the known risk factors measured (Browne et al., 2019). SRSs that are more frequently used by the unharmed group can thus more confidently be attributed to the use of such strategies, since the propensity matching has controlled (by degrees) for the issue that some people ‘at risk’ are more likely to use a variety of SRSs out of need. This approach provides results that are more accurate and usable for people, such as clinical practitioners, who are looking for what SRSs to recommend.

As a first step, participants were matched one-to-one across both groups according to the predicted probabilities for their risk for being harmed by gambling. People who could not be matched were discarded (n = 148, 7.3%), as is common in propensity matching (Leite, 2016). After this step, however, the unharmed group still had a lower overall propensity for being harmed by gambling relative to the harmed group. Discarding unmatched cases cannot eliminate all risk discrepancies between the two groups.

In the second step, cases were weighted inversely with respect to their propensity for risk of gambling harm, in the case of harmed gamblers 1/(p) and for unharmed gamblers 1/(1-p). For example, an unharmed gambler whose behaviour and traits led us to expect them to be at relatively high risk of gambling harm was up-weighted. Similarly, a harmed gambler whose behaviour and traits indicated a relatively low-risk was down-weighted. This weighting acts to make the two groups equivalent in terms of known risk factors, removing the effect of these confounding variables, and makes them more directly comparable when evaluating the effects of SRSs. In the third step, after matching and weighting, we evaluated SRSs by a simple comparison of their prevalence among (weighted) harmed and unharmed gamblers. In short, SRSs that are used more frequently by the unharmed group, inclusive of weighting, are inferred to be effective at preventing gambling-related harm.

The fourth step assessed whether the use of the most efficacious SRSs differed by gambler characteristics. Non-parametric tests examined the relationships between SRS scores (total number of endorsed SRSs) and the predictors. The relationship between gender and SRS scores was examined using a Mann-Whitney U test. Spearman’s correlation examined age and SRS score. Kruskal-Wallis tests examined the relationship between SRS scores and EGM gambling frequency.

Results

Use of SRSs

Table 2 shows the use of the 45 SRSs. The most used SRSs were ‘I usually play low denomination pokies’ (73.1%), ‘When I have a large win on the pokies, it is time for me to quit’ (72.6%), and ‘I keep a household budget’ (70.9%).

Table 2 Study 1: Proportion of the sample who endorsed each SRS
Full size table

Identification of the most efficacious SRSs

We first constructed a propensity model of the likelihood of participants experiencing harm. Because the PSM framework requires two defined groups, we implemented a 0–2 versus 3+ categorisation based on the Gambling Harm Scale (GHS). Browne et al. (2020) found that scores 1–2 showed a small but significant decrement to health utility, whereas scores 3+ showed both a significant difference and a clinically meaningful effect size. Therefore, 0–2 was used to indicate lesser harm vs 3+ indicating greater harm.

Table 3 summarises the risk factors based on a logistic regression. All effects were significantly associated with the probability of being significantly harmed by EGM play.

Table 3 Study 1: Risk factors for experiencing significant gambling harms
Full size table

Case matching based on the predicted probabilities was then applied to the 2032 cases, across the not-harmed (technically less harmed) and more harmed groups, leading to 148 unmatched cases, and 942 matched cases in each group. Weighting was then applied, as detailed earlier. Lastly, a weighted average frequency in the use of each SRS was calculated in each group, and the difference between use by harmed and unharmed gamblers was calculated. This difference (larger means a stronger association with avoiding harm) was then used to rank and evaluate the SRSs. Table 4 summarises all 17 SRSs that were associated with decreased harm, after matching and weighting. The differential P scores describe the difference in the probability that an SRS would be employed by an unharmed gambler, as opposed to a harmed gambler. This heuristic can be used to capture the association between SRS use and the avoidance of harm.

Table 4 Study 1: Propensity model of experiencing significant harms
Full size table

Use of the Most Efficacious SRSs and By Personal Characteristics

This analysis examined use of the most efficacious SRSs and by age, gender, PGSI and EGM frequency. Participants ranged from using all 17 SRSs to none, with a mean score of 10.65 (SD = 4.06). There was no significant difference between the number of SRSs used between males (m = 10.60, SD = 4.12) and females (m = 10.71, SD = 3.96). There was a significant but negligible-strength positive relationship between age and number of SRSs used (rS = .073, p = .001), with older participants tending to use more SRSs. There were no significant differences in the mean SRS scores across EGM gambling frequency. The number of identified SRSs used was associated with a significantly lower likelihood of being in the moderate-risk or problem gambling categories of the PGSI, rS = − .18, p < 0.01.

Study 2: Randomised Controlled Trial

Methods

Study 2 was approved by Central Queensland University Human Research Ethics Committee (#22959).

Recruitment

Online panel aggregator, Qualtrics, recruited participants via online convenience panels to complete three surveys, each 1 month apart, in June to August 2021. As with study 1, the focus was on regular EGM players who lived in NSW. Inclusion criteria were aged 18 years or older, playing EGMs in the last 4 weeks, living in NSW and having an interest in better controlling how much they spend on EGMs. Of the 1238 participants who completed the wave 1 survey, 103 were excluded based on data quality checks as well as 47 who did not opt-in to receive SMS messages as part of the experimental design. Data quality checks detected four duplicate responses at wave 2 and seven in wave 3 and these duplicates were removed. Sample sizes after exclusions totalled 1088 (wave 1), 756 (wave 2) and 725 (wave 3).

Sample Characteristics

The sample was reasonably balanced by gender, had a mean age of 32.7 years (range 18–83), and most respondents were married/de facto, were university-educated and worked full-time (Table 5).

Table 5 Study 2: Demographic characteristics at Wave 1
Full size table

Procedure

For parsimony, the 17 most efficacious SRSs identified in study 1 were reduced to 13 SRSs for the RCT by combining similar items. Table 6 lists the 13 SRSs, their codes and number of participants and data points per SRS. In wave 1, approximately two-thirds of respondents were randomly allocated to one of the 13 SRSs and asked: ‘For the NEXT 4 WEEKS, please try to consistently use this practice when you play the pokies’. About one-third of participants were allocated to the control group. Randomisation to the 14 groups was stratified by gender and age (18–34 and 35+ years) and reported weekly hours of EGM play (< 16 and 16+). At the end of each survey, the test group was reminded to use their allocated SRS during the next 4 weeks, while the control group was simply reminded to ‘gamble responsibly’. Respondents were also sent an SMS with the same message between waves.

Table 6 Study 2: SRSs, codes, and number of participants and observations per SRS
Full size table

Measures

After screening questions based on the inclusion criteria, the following measures were administered.

Demographics (wave 1): age, gender and the characteristics in Table 5.

EGM playing behaviour in the last 4 weeks (waves 1–3): number of hours spent playing EGMs (open-ended text box) and EGM expenditure (defined as losses; open-ended text box).

Short Gambling Harms Screen (waves 1–3): The 10-item SGHS (Browne et al., 2018) was modified to ask about harms experienced within the last 4 weeks as a result of the respondent’s EGM play (no/yes). Cronbach’s alpha was .81 (Wave 1), .87 (Wave 2) and .95 (Wave 3).

Use of assigned SRS (waves 1–3, test group only): How often they used their assigned SRS during the last 4 weeks (never, sometimes, most of the time, always)

Analysis

The analysis involved three main steps. A nested experimental design where multiple observations were nested within participants was employed, with the primary level being a comparison of exposure to each of the tested SRS messages (N = 733, codes 1–13) with a control message, ‘gamble responsibly’ (N = 355, code 0). This first analysis evaluated the effects of the assignment of any treatment condition vs the control condition.

Second, comparisons between individual SRSs were conducted to see which were potentially most highly associated with better gambling outcomes. This evaluated the effects of assignment of the individual SRSs vs the control condition.

Third, data were also collected on the frequency with which participants used their allocated SRS. This allowed a secondary repeated measures observational analysis to evaluate the effects of frequency of SRS utilisation on the outcome variables. Our assumption was that actual use of the SRS, rather than simply being assigned to use the SRS, should be associated with better gambling outcomes.

Three key outcomes were employed:

  1. 1.

    EGM Spend: spend on EGMs during the prior period. Transformed using the formula log(x + 1) to stabilise error variance

  2. 2.

    EGM Time: number of hours spent playing EGMs during the prior period. Transformed using the formula log(x + 1) to stabilise error variance

  3. 3.

    SGHS: scores on the SGHS, untransformed

Time and spend on gambling are directly implicated in gambling harm and gambling problems (Neal et al., 2005). The SGHS is a direct measure of harmful outcomes that SRSs are intended to prevent (Browne et al., 2018).

The repeated measures design was handled using robust linear mixed effects (RLME) modelling, using the robustlmm package in the R statistical programming environment. Since each participant received the same SRS for the duration of the experiment, the data structure can be understood as hierarchical, with multiple observations nested within participants. That is, the design was repeated measures on the same outcomes for each participant. We considered models in which SRS was treated either as a random factor within the treatment condition (i.e. the SRS was considered representative of a large number of SRSs that might have been included in the study, but the set was not comprehensive) or as a fixed effect with 13 levels (i.e. the set of SRSs tested was deemed to be a complete set of possible strategies that could be used). For random effects included in the models below, variances, rather than standard deviations, are reported.

Results

Effects of the Assignment of Any Treatment Condition vs the Control Condition

Table 7 summarises the analyses for the broadscale treatment effect: whether allocation to the treatment conditions (any of codes 1–13) was associated with a differential change in gambling outcomes over time in comparison to the control condition (code 0). There was no improvement over time in any outcome for people assigned to an SRS message condition. This conclusion was manifest in no significant interactions between the variables time (i.e. T2 vs. T1 and T3 vs. T1, respectively) and test (i.e., SRS messages vs. control).

Table 7 Study 2: Summary of RLME models testing for an interaction between experimental condition (test versus control) and time, with a random effect for subject nested within SRS
Full size table

Moreover, model comparisons between the base model (a), including only main effects, and the interaction model (b) that included an additional interaction effect were not significant in each case. This indicates that, in aggregate, allocation to one of the SRS conditions did not result in a detectable change in gambling outcomes during the study period relative to the control condition.

Effects of Assignment of the Individual SRSs vs the Control Condition

The above analysis was repeated using a 14-level factor ‘SRS’ (codes 0–13) in place of the 2-level factor ‘group’ (Table 8). These analyses included a random effect for participants only. An analysis of deviance test providing an omnibus comparison of the interaction model with the main-effects only model found only a marginally significant difference for EGM spend, χ2(26) = 39.03, p = .048. There were no significant time × SRS interactions for the dependent variables of EGM Time or SGHS.

Table 8 Study 2: Models evaluating the effects of assignment of the individual SRSs vs the control condition
Full size table

Given the significant omnibus test for EGM spend, we considered interpretation of the fixed effects for EGM spend. Inspection of the beta coefficients showed significant decreases in EGM spend for assignment to the following SRSs: (1) T3xSRS2 (B = − .939, p = .009) ‘When you play the pokies, always set aside a fixed amount to spend’; (2) T2xSRS4 (B = − .822, p = .025), T3xSRS4 (B = − .799, p = .034) ‘Make sure your leisure time is busy with other hobbies, social activities and/or sports’ and (3) T3xSRS6 (B = − 1.170, p = .002) ‘Don’t go and play the pokies just to avoid being bored’. Unlike the omnibus chi-square statistic quoted above, these p-values associated with individual beta coefficients do not take into account the multiple comparisons being made within the single regression model.

Effects of Frequency of SRS Utilisation

The above analyses are predicated entirely on assignment of participants, at random, to experimental conditions. However, not all participants adhered to the requested protocol of implementing their assigned SRS during the RCT. Of the 1715 observations in the test condition, 390 reported never (1) using the SRS during that period, 695 sometimes (2), 381 most of the time (3) and 249 always (4). The dataset was therefore analysed as repeated measures relating paired observations of frequency of SRS utilisation and each gambling outcome, rather than an experimental manipulation alone. In this scheme, we compared the simple effect of frequency of SRS use, with the joint effect of which SRS was allocated and the frequency with which that SRS was employed. Importantly, people’s use of SRS was still related to their assigned experimental condition, but the present analysis allowed that people might differentially use the SRSs to which they were assigned. This allowance can be considered as ‘treatment adherence’.

These analyses (Table 9) therefore examined the joint effects of treatment assignment and treatment adherence. Since treatment adherence (i.e. whether people used the SRS) is not an experimental effect, these results are correlational in nature. Results that are highly significant (p < .01) hold even after considering that multiple tests were performed. This analysis is valuable since it stands to reason that SRSs ‘work’ because people employ them rather than just ‘think’ about them.

Table 9 Study 2: Summary of models of the effect of SRS allocation and frequency of use on gambling outcomes
Full size table

Comparing (a)/(b) models in Table 9, there was a significant improvement in fit for EGM spend, χ2 (25) = 50.211, p = .002, and the SGHS, χ2 (25) = 52.741, p = .001. There was no significant improvement for EGM time, χ2 (25) = 32.194, p = .1525. Detailed evaluation of significant beta coefficients for these two outcomes can be made with respect to Table 9. SRS4 ‘Make sure your leisure time is busy with other hobbies, social activities and/or sports’ stands out as having a significant effect on EGM spend and the SGHS, at the .01 threshold for both main effects of SRS and frequency × SRS interactions. Thus, the frequency with which people used this SRS was related to better outcomes on spend and harms experienced. SRS3, ‘Make sure you take regular breaks every 30 minutes when you are playing the pokies’ also showed consistent main effects for EGM spend, as well as an interaction for EGM spend. Thus, frequently adhering to regular breaks was associated with lower spending. Lastly, SRS8 ‘Keep a household budget’ had a significant main effect and frequency interaction effect for the SGHS. People who more frequently kept a budget had lower gambling-related harm.

Discussion

Study 1 identified a group of SRSs that best predicted less harmful gambling amongst 2032 frequent EGM players. The individual use of 17 SRSs and the total number of these SRSs used were both associated with decreased EGM-related harm. On average, participants used 10.7 of the 17 effective SRSs. There was no significant difference in their use by gender or EGM gambling frequency, and only an extremely weak association with age.

Study 2 tested the efficacy of 13 protective SRSs identified from study 1 when delivered as a brief intervention to EGM players wanting to reduce harmful play. In the first wave, the 1088 respondents were randomly allocated to either one of the 13 SRS test conditions or the control condition (‘gamble responsibly’). Outcome measures comprised EGM expenditure, time spent playing EGMs and EGM-related harm. Assignment to any SRS treatment condition (in aggregate) did not result in a detectable change in gambling outcomes relative to the control condition. However, significant decreases in EGM spend were observed for assignment to the following SRSs: ‘When you play the pokies, always set aside a fixed amount to spend’ (SRS2), ‘Make sure your leisure time is busy with other hobbies, social activities and/or sports’ (SPG4), and ‘Don’t go and play the pokies just to avoid being bored’ (SRS6). When evaluating the effects of the frequency of utilising the assigned SRS, three SRSs had significant effects on one or more gambling outcomes. Increased frequency of using SRS4, ‘Make sure your leisure time is busy with other hobbies, social activities and/or sports’, resulted in a significant reduction in EGM spend and EGM-related harm. Increased frequency of using SRS3, ‘Make sure you take regular breaks every 30 minutes when you are playing the pokies’, resulted in a significant reduction in EGM spend. Increased frequency of using SRS8, ‘Keep a household budget’, resulted in a significant reduction in EGM-related harm. Of additional interest is that the SRSs that resulted in reduced EGM spend and EGM-related harm in study 2 were also strongly negatively associated with gambling harm in study 1.

Overall, the results suggest that not all SRSs that might be promoted to gamblers are likely to lead to beneficial gambling outcomes, but that some SRSs have greater efficacy. This emphasises the importance of research that identifies the optimal set of SRSs that should be promoted to gamblers. This study has provided arguably the strongest evidence to date of the efficacy of certain SRSs for EGM players. However, the findings are subject to limitations. The samples may not be representative of the NSW population of frequent EGM players. The findings are also based on self-report data, which may be subject to recall and social desirability biases. In line with good scientific practice, replication studies are needed to confirm the findings in different samples and jurisdictions and to assess gambling outcomes from SRS-use over the medium and longer term. Nonetheless, the study advances current knowledge about the potential effectiveness of SRSs, since its design overcame several limitations of earlier research.

While further research would be beneficial, using the study’s findings to refine the SRSs that are currently promoted to EGM players would improve on current advice. This is because the currently promoted SRSs have very little evidence supporting their efficacy and generally have been selected on an ad hoc basis. Instead, the five SRSs (2, 3, 4, 6 and 8) that resulted in reduced EGM spend and/or EGM-related harm could be helpfully communicated on help service websites, in brochures and signage in gambling venues, on gambling websites and apps and in public education materials. The five efficacious SRSs could also be used as a ‘call to action’ in responsible gambling messages, since widely used messages have been criticised for being superficial and lacking helpful advice (Hing, Nuske, et al., 2016a; Newall et al., 2022; Sproston et al., 2015). The five SRSs could also provide the basis for a consumer self-assessment tool with automated feedback, to encourage consumers to assess and self-regulate their gambling by using the promoted SRSs. The SRSs might assist treatment providers by identifying actionable strategies to help their clients make behavioural changes to reduce financial impacts and harm from their gambling.

Conclusion

Most harm from gambling is attributable to EGMs (Browne et al., 2023). Consumer protection from this harm is largely based on an informed choice model (Blaszczynski et al., 2004) that relies on players self-regulating their gambling. Strategies to assist this self-regulation are currently widely promoted (Hing, Russell, & Hronis, 2016b) but have very little evidence to support their efficacy. This study has advanced the evidence base to support the use of five SRSs that are empirically associated with reduced EGM spend and/or EGM-related harm.

However, we caution that adherence to these SRSs does not guarantee that a person’s gambling will be harm-free. We also acknowledge that these strategies may be perceived as placing increased responsibility on people to self-regulate their gambling. We emphasise that our focus on SRSs in this research is not intended to downplay the role of industry or governments in reducing gambling harm by providing safer gambling products, environments and policies. Instead, it provides people who gamble with harm minimisation advice. In short, effective SRSs are just one ingredient in a broader public health approach needed to reduce gambling harm.