Why is harm important?
In most jurisdictions around the world, it is recognised that gambling (while a recreational activity for the majority of participants) can also cause considerable harm for some individuals, their families and the broader community. Although there is ongoing debate concerning the overall balance of harms and benefits associated with gambling, regulators recognise a mandate to implement and enforce policies and practices that prevent, minimise or reduce the harms associated with gambling. Such a view is also consistent with broader adoption over the last 20 years of public health approaches to gambling that emphasise the need to identify broader population-level approaches to prevent gambling harm, but also services that target those already negatively affected by gambling. However, despite this, it has only been recently that concerted attempts have been made to measure harm in a systematic way. In this brief article, I will summarise some of the recent developments in harm research and the important contributions they have made. I will also outline some of the conceptual and measurement challenges which this new area raises. Drawing upon points which I have previously raised at IAGR conferences, I will summarise some of my current thinking about the state-of-play in this area and the main insights which I have gained from this recent work.
What is gambling-related harm?
The types of harm associated with gambling are generally well-recognised. In 1999, the Productivity Commission in Australia made the first dedicated attempt to measure gambling-harm comprehensively. According to the Commission, harm can be divided into several categories. Harm can personal (depression, anxiety, suicidality); financial (debts, bankruptcy, inability to meet daily living expenses); inter-personal and familial (family conflict); vocational (loss of productivity or employment); and legal (e.g., committing offences to finance gambling). Most of these categories will be familiar to regulators because questions concerning problems of this nature are usually included in nearly all population prevalence studies around the world. Some items appear in standardised instruments such as the South Oaks Gambling Screen or Problem Gambling Severity Index (PGSI). Others are presented as standalone questions. In either case, what these questions yield is an estimate of the proportion of the population that experiences various forms of harm. As I pointed out at IAGR, however, a limitation of much of this work is that it has typically only measured a narrow range of harm. The standardised measures usually only contain a small number of items relating to harm. Most other questions (e.g., have you lost a job due to gambling) usually attract such a low level of endorsement (e.g., 0.2%) that the data is not useful to inform changes in population-level harm over time. Items of this nature also are typically only reported by the most severe problem gamblers, so that one does not capture the experiences of other gamblers whose behaviour might still be amenable to modification (a principal focus of public health as well as responsible gambling regulation).
Contemporary harm research
It was with this situation in mind that the Victorian Responsible Gambling Foundation funded an important study into gambling harm in 2016 (Browne et al., 2016). This work, which has been subsequently replicated in New Zealand and Sweden and has led to the development a 10-item measure called the Short Harms Scale (Browne et al., 2017) which has been administered in a number of population surveys around Australia. Using a sample of over 2000 gamblers, the Browne et al. (2016) work developed a list of 72 gambling harm items that spanned all of the dimensions identified by the Productivity Commission (1999). Gamblers were asked to indicate (yes/ no) whether they had experienced a particular harm because of their gambling in the previous 12 months. Items were then analysed to examine: (a) the prevalence of different forms of harm across the PGSI groups and (b) which forms of harm best differentiated between different levels of gambling risk. A particularly important feature of this work was that it developed a continuum of items ranging from mild to severe. For example, financial harm was captured using items ranging from “My savings are reduced because of gambling” to those relating to bankruptcy. Health impacts from harm could range from feeling distress or shame to being suicidal or being admitted to the emergency room.
An important strength of this work was that it generated measures that had a higher prevalence / base-rate and which appeared to capture some of the harms associated with lower risk gambling (ie., people who were not yet problem gamblers). This work therefore has offered the potential to generate harm scores that could be compared across time to examine the efficacy of harm-minimisation strategies. It has also widened policy debates away from the severe end of the gambling continuum to allow greater analysis of the impacts of gambling on at-risk populations who might be a focus for preventative measures (e.g., low and moderate risk gamblers as identified by the PGSI).
On the whole the results seemed valid and logical. Problem gamblers, as would be expected, were much more likely to endorse the items and were generally the only gamblers who endorsed the severe harm items. Low risk gamblers, on the other hand, rarely endorsed any of the severe harm items, but did endorse some of the less severe items such as whether they had used their savings for gambling, had less money for other leisure or activities because of gambling. PGSI scores were generally positively corrected with harm scores: higher PGSI indicated more harm.
Calculating the burden of harm
None of the above work is considered contentious and has made an important contribution to the field. However, as I have pointed at IAGR, a challenging part of this work is that it also involved the calculation of burden or harm estimates that yielded some quite surprising results. In the original 2016 study, what the researchers did was to ask each participant to rate their own bundle of self-reported harm and to indicate how many years of life out 10 they would give up to be free of the harms identified. A person might say ‘1’ or ‘2’ or some other figure. In more recent studies, e.g., in the most recent Tasmanian Social and Economic Impact Study, the researchers administered the Short Harm scale and worked out everyone’s mean score. Problem gamblers generally scored a few points out of 10, whereas lower risk and recreational gamblers had very low scores (.05 in Tasmania). In all of their studies, the researchers observe that the prevalence of low-risk gamblers is much higher than for problem gamblers . Thus, as they showed, if one multiplies very small quantum of harm reported by low risk gamblers by the total number of these gamblers in the population ( a large number) then the total burden of harm for these people will come out higher than for problem gamblers. It is a bit like identifying 50 in a room with very little harm (1 out 10) and 3 severe cases each with 8 out 10 units of harm. One can show that 50 x 1 = 50 is clearly larger than 3 x 8 = 24. Using this type of calculation, the researchers were able to show that 85% or all harm was attributable to low and moderate risk gambling- a quite counter-intuitive finding. They also presented analyses which show that the burden or harm or disease associated with low risk gambling was higher than a number of major conditions, e.g., arthritis and diabetes.
Concerns about burden of harm estimates
As I have indicated at IAGR, while the methods used to reach these conclusions are quite innovative and often accepted in public health contexts, there are challenges. The first issue is where one draws a line between harm and opportunity cost. A number of items in this harm research and in the Short Harm Scale appear to be closer to opportunity cost than true harm (e.g., I spend less time / money doing other things than gambling). As I pointed out: would it not be possible to apply similar arguments to the harms associated with regular sports attendance on weekends? I raised a question as to what would happen if the low risk harm or opportunity cost items were excluded given that it appeared to be these items that contributed to much of the harm reported by low risk gamblers. Some insight into what happens is provided by the results of the recent NSW prevalence study. This study only included only moderate and severe harm items and not the Short Gambling Harm Scale (SGHS). In contrast to the Tasmanian study that used the SGHS, the vast majority of the harm was found to be borne by the moderate risk and problem gamblers. In other words, where the burden of harm falls is quite sensitive to the selection of harm items.
A second issue is whether it is valid to multiply small micro-elements of harm by large population figures and compare it with totals based on severe harms multiplied by small numbers of people. The harms are not qualitatively the same. Can one really say that 10 x 1 units of minimum harm (which may be reducing other leisure activities) is equivalent 1 x 10 units of the most severe harm (an individual who is bankrupt)? I argued, perhaps somewhat facetiously, that some of the burden of harm work seemed rather like comparing 50 sneezes to 1 case of the flu. However, my concerns about this practice is borne out in the recent Tasmanian prevalence study which found that the highest burden of harm (in terms of lost years of quality of life) was reported by recreational gamblers (who scored 0 on the PGSI). Despite the fact that recreational gamblers only scored .05 out 10 on the Short Harm Scale, there were so many of them that the total product of harm was found to be higher than moderate risk and problem gambling put together. To my mind, this presented something of a reducio ad absurdum of a methodology that, while logical on the surface, makes the assumption that one can compare different types of harm in the same analysis.
The third problem is that it is highly questionable whether methods that ask people to rate their gambling harm against the symptom descriptions of other disorders (which respondents had probably never personally experienced) can yield valid results. How does one rate whether low risk gambling is worse than arthritis or major depressive disorder if one has never experienced either condition?
My current working hypothesis is that the main burden of real harm associated with gambling is still borne by moderate risk and problem gamblers and is probably not as severe as many other disorders (including alcoholism). However, I believe that the recent harm work has made important contributions to broaden the discussion about how we conceptualise gambling harm. Such work will be further enhanced as the measures of harm are refined and more carefully calibrated and is likely the focus of ongoing work by the designers of the Short Harm Scale. It is also the focus of some current work which we are undertaking at the University of Adelaide.
What have we learned?
As I pointed out at IAGR in Jamaica last year, harm does appear to lie on a continuum and that is possible to capture how the harms or impacts of gambling change as one moves up or down the level of risk. For low risk gamblers, there is probably little real harm, but a gradual prioritisation of gambling over activities and commitments. Gambling is becoming more important for the person and there may be a gradual narrowing of interests. For moderate risk gamblers, there is a sense of growing ‘pressure’. The serious impacts of gambling have not usually emerged, but the person is starting to be short of money, late on bills, seeing mood changes, and a loss of productivity. Problem gamblers (although still only 10-15%) are generally the only group to report the serious harms at any meaningful level: bankruptcy, loss of utilities, or the loss of assets. Thus, when one plots the endorsement of gambling harms against the level of gambling risk (e.g., as based on the PGSI), one generally observes a gradually or linear increase in endorsement of the softer harms; a slightly more J-shaped curve for moderate risk harm items and a steeper J-curve for severe items (i.e., a J- curve means that the type of harm is seldom endorsed by lower risk groups but endorsed by higher risk groups (See Figure 1 below). To my mind, this differentiation of the type of harm item (rather than just a focus on the relatively magnitudes of the burden of harm) might provide regulators with potentially more useful ways to capture the impacts of changes in policies, practices and gambling behaviours. In addition to looking for changes in severe harm in problem gamblers, there is scope now to look for evidence of changes in the various emerging ‘pressures’ reported by moderate-risk gamblers as well as the changing experiences of low-risk gamblers (who are probably not harmed in the true sense), but who remain a group of fundamental interest in discussions around the prevention of harm.
Author: Professor Paul Delfabbro
School of Psychology
University of Adelaide