Associations between recollected rates of Category D fruit machine use as a child and adult gambling in a young adult sample

There have been recent restrictions to the range of gambling products made available to UK children, with the industry association Bacta introducing a minimum age of 18 for Category D fruit machines, which can otherwise be legally used by children. Any potential government action on Category D fruit machines may want to consider limitations in the current evidence base. The present study addressed this issue, by conceptually replicating previous results associating recollected usage of Category D fruit machine usage as a child and adult gambling. Compared to previous studies, the present study used a younger adult sample, and also tested for potential moderation by subjective confidence. Three out of four main tested associations were significant and in the hypothesised direction, and a lack of significance on interaction terms yielded no evidence of potential moderation by subjective confidence. The present study therefore added to the limited evidence base on potential risks of legal Category D fruit machine usage as a child.

www.nature.com/scientificreports/Given these valid concerns about the evidence base for associations between childhood use of Category D fruit machines and adult gambling, I aimed to conceptually-replicate previous findings with some design improvements taken from previous research on other gambling products that are legally-available to children.First, the measured associations of H1-H4 below were done in a young adult sample aged 19-23 22 .Second, participants provided a measure of subjective confidence of their recollected childhood use of Category D fruit machines 23 , and interaction models were run to see if differential rates of subjective confidence across the sample could explain any observed positive associations (H5).
The following hypotheses were therefore preregistered: H1.That any level of recollected engagement with Category D fruit machines, versus not recollecting using them, will be associated with being an adult gambler.
H2.That for adults recollecting using Category D fruit machines, higher frequencies of recalled machine use will be associated with being an adult gambler.
H3.That any level of recollected engagement with Category D fruit machines, versus not recollecting using them, will be associated with higher Problem Gambling Severity Index (PGSI) scores among adult gamblers.

H4.
For adult gamblers who recollect using Category D fruit machines, that higher frequencies of use will be associated with higher PGSI scores.

H5.
That in models similar to those in H1-H4, but adding a main effect and an interaction for confidence, that the resulting interaction effects will be nonsignificant.

Method
Data, materials, and the preregistration are available from: https:// osf.io/ ams2c/.The study received ethical approval from the University of Bristol's School of Psychological Science Research Ethics Committee (#14,073), and all participants provided informed consent.The methods were carried out in accordance with relevant guidelines and regulations.An initial 998 responses were collected via Prolific.co, of which 37 (3.7%) self-reported as providing careless responses 24 , so the final sample size used was 961 participants.This initial sample size was selected heuristically based on the availability of research funds.In order to reduce the overall level of recollection biases, only participants aged 19-23 were recruited for the study (M = 21.5, SD = 1.3).Overall, 487 participants (50.7%) were female (one participant preferred not to say).Demographic information was collected automatically via Prolific.co.Participants were paid £0.40 each, and took an average of 93.3 s to complete the survey (£15.43/hourpro-rata).
The survey involved two blocks presented in random order.In one block, participants were asked if they had gambled in the past 12 months, and participants who responded 'yes' then proceeded to complete the PGSI 25 .This approach, of only giving the PGSI to participants reporting gambling recently, is the standard approach used in gambling prevalence surveys 26 .Overall, 606 participants (63.1%) reported gambling in the past 12 months, of whom 234 (38.6% of gamblers) were categorized as no-risk, 183 (30.2%) as low-risk, 131 (21.6%) as moderaterisk, and 58 (9.6%) were in the highest-risk category.
In the second block, participants were provided with a description of Category D fruit machines taken from a previous study 21 , and a relevant image.Participants were asked, "How often do you recall using category D fruit machines while being under the age of 18?", and responded on a five-point scale of: never, seldom, occasionally, frequently, very frequently 22 .Immediately after, participants provided their level of subjective confidence ("How confident are you in your recollection of this?") on a scale from 0 ("not at all confident") to 100 ("very confident"), as first done by another recent study 23 .
H1 and H2 were analyzed via logistic regression, due to the binary nature of the outcome variable (adult gambling, yes/no).H3 and H4 were analyzed via negative binomial regression, as this model can better account for the skewed distribution of PGSI scores than linear regression can 27 .For H5, models were run for all of the significant associations found across H1-H4, by adding main effects for subjective confidence and interactions between subjective confidence and the independent variable for Category D fruit machine use, and interpreted by the significance of the interaction term.A p-value of 0.05 was preregistered.

Results
Descriptive results on frequency of recollected use are shown in Table 1.Overall, 48.7% of the sample recollected using Category D fruit machines at least once as a child; with rates of use being descriptively higher among gamblers (54.0%) than non-gamblers (39.7%).Subjective confidence ratings were also high on average (M = 83.4,SD = 22.6).
Results of analyses for H1-H4 are shown in Table 2.As can be seen, support was found for H1 (Z = 4.25, p < 0.001), H2 (Z = 2.16, p = 0.031), and H4, (Z = 3.68, p < 0.001) but not H3 (Z = 1.40, p = 0.161).Overall, this suggested that merely recollecting using Category D fruit machines was not associated with higher PGSI scores among gamblers (H3).However, usage of Category D fruit machines was associated with being an adult gambler (H1 and H2), and higher rates of recollected Category D fruit machine use were associated with higher PGSI scores among gamblers (H4).
Results of analyses for H5 are shown in Table 3, where models add main effects for subjective confidence and interactions between subjective confidence and the main effects for Category D fruit machines used before.An interaction model was not run for H3, due to the lack of statistical significance seen earlier.None of the interaction effects were statistically significant (p's ≥ 0.121), and the estimated coefficients on the interactions were all extremely close to one.This means that H5 was supported; the significant associations found for H1, H2, and H4 did not appear driven by differential levels of subjective confidence.

Discussion
Different approaches have recently been taken with the UK's range of gambling products that are legally available to children.The legal age of National Lottery products was recently increased to 18 by government regulation, while a minimum age of 18 to use Category D fruit machines was implemented via industry self-regulation 12 .Industry self-regulation can be less effective than government regulation [13][14][15][16] , but any potential government action should also consider the robustness of any underlying evidence base.The present results largely supported previous research showing associations between rates of recollected Category D fruit machine usage as a child and adult gambling [19][20][21] , with three out of four associations significant and in the hypothesized direction.Furthermore, rates of subjective confidence were on average high (M = 83.4out of 100), and zero out of three interaction models were significant, adding further support to these observed associations.These findings are still subject to various limitations.The measure of subjective confidence only tapped into potential conscious memory distortions, and cannot rule out potential biases that participants were unaware of.Participants were recruited from a crowdsourcing platform, and so this sample was non-representative 28 .The findings are also correlational, meaning that the present results do not show causal effects between Category D fruit machine use and adult gambling.The present study only measured one gambling product, and other products such as coin pusher and crane grab machines are used legally by a high proportion of UK children 29 , which means that the issue of legal availability of gambling products to children is broader than just Category D fruit machines.
In conclusion, the present research largely conceptually-replicated and extended previous findings showing that legal usage of Category D fruit machines as a child is a risk factor for experiencing gambling-related harm as an adult.As a result, UK policymakers may want to consider whether recent industry self-regulation should be reinforced via additional government regulation, where a legal age of 18 for Category D fruit machines would bring them into line with the legal age of use of other gambling products.

Table 1 .
Descriptive results on frequency of recollected use.

Table 2 .
Main effect models showing associations between engagement and frequency of engagement with adult gambling outcomes.Logistic regression output are odds ratios, negative binomial regression output are incidence rate ratios.Output in the left column refer to H1 (top) and H2 (bottom), while output in the right column refer to H3 (top) and H4 (bottom).

Table 3 .
Factorial models showing associations between engagement and frequency of engagement with adult gambling outcomes.Logistic regression output are odds ratios, negative binomial regression output are incidence rate ratios.Confidence and frequency measures were mean-centered.Output in the left column refer to H1 (top) and H2 (bottom).A model was not run for H3 (right column, top) due to the insignificance of the main effect.H4 is shown in right column, bottom.