Striving Towards National Lower-Risk Gambling Guidelines: An Empirical Investigation Among a Sample of Swedish Gamblers

Abstract

Several countries, including Canada and Australia, have developed public health-based lower-risk gambling limits to differentiate lower-risk from higher-risk gambling. This study aimed to identify a preliminary set of lower-risk gambling limits (gambling frequency, duration, expenditure, expenditure as a proportion of personal net income, and diversity), and investigate if gambling types are linked to additional harms, in a Swedish context. The study involved secondary analyses of two online survey studies using the Gambling Disorder Identification Test (GDIT). Receiver operating curve analyses were conducted in relation to both + 1 and + 2 gambling-related harms in a sample of 705 past-year gamblers. Potential lower-risk limits ranges identified were: gambling frequency of “2–3 times a week” to “4 or more times a week” (8–16 times monthly); gambling duration of 6 to 15 h per month; gambling expenditure of 2,000 SEK (approximately $USD190) per month; gambling expenditure as a proportion of personal net income of 5%; and gambling diversity of only one problematic gambling type. Gambling on slots and sports betting were associated with gambling-related harms. The lower-risk limits in the current study were higher than in previous studies, which may be explained by the large proportion of support- or treatment-seeking gamblers with high rates of problem gambling and problematic online gambling in the study sample. An international consensus-based framework on gambling consumption is warranted, with lower-risk limits validated in future empirical studies using larger datasets collected from the Swedish general population.

Supplementary Information

The online version contains supplementary material available at 10.1007/s10899-024-10372-w.

Gambling is a major public health concern globally (Ukhova et al., 2024). The prevention of Gambling Disorder (GD) (American Psychiatric Association, 2013) is informed by public health perspectives, which aim to identify its determinants and subsequent harms (Korn & Shaffer, 1999). There is evidence that gambling-related harms, such as financial loss, relationship breakdown, and psychological distress (Langham et al., 2016), can occur below the GD diagnostic thresholds (Browne & Rockloff, 2017; Delfabbro & King, 2019), thereby providing the opportunity for public health initiatives to impact on a larger segment of the population (Currie et al., 2006, 2009). One such initiative to reduce alcohol-related harm has been the adoption of lower-risk drinking guidelines (Room & Rehm, 2012; Wechsler et al., 1995), which are promoted to the general public and clinical populations to help individuals adapt safe behaviors in response to accurate information (Room & Rehm, 2012).

The Identification of Lower-Risk Gambling Limits

An emerging literature for gambling lower-risk limits in population-representative studies conducted in Canada (Currie et al., 2006, 2008, 2009, 2012, 2017), Germany (Brosowski et al., 2015), Australia (Dowling et al., 2021; Dowling et al., 2021), and Norway (Langeland et al., 2022) have identified relatively consistent limits across multiple gambling consumption indices: gambling frequency (0.6–5 times per month or 2–4 activities per year); gambling duration (400–454 min per year); gambling expenditure (USD$93–$720 per year); and gambling expenditure as a proportion of income (1–3% of gross household income/0.8–10% of gross personal income). More recently, lower-risk limits were identified via linked analyses with 11 high-quality population-level datasets across eight Western countries: gambling no more than 1% of household income, gambling no more than 4 days per month, and avoid regularly gambling at more than 2 types of games (i.e., gambling diversity, number of gambling types) (see https://gamblingguidelines.ca/) (Hodgins et al., 2023). Other studies extending this research to specific groups, including post-treatment-seeking gamblers (Weinstock et al., 2007), university students (Weinstock et al., 2008), and community/psychiatric outpatients (Quilty et al., 2014) have identified similar lower-risk limits. In contrast, studies using behavioral data from gambling website customers have identified limits that are higher than those derived from prevalence study data (Jonsson et al., 2022; Louderback et al., 2021).

These studies, which suggest that lower-risk gambling limits generally yield moderate levels of classification accuracy, have most often assigned equal importance to both sensitivity (accurate identification of people experiencing harm) and specificity (accurate identification of people not experiencing harm). There are some exceptions, however, whereby some studies have conducted analyses that maximize specificity, with a view to increasing the credibility of the limits by reducing the proportion of false positives (Currie et al., 2017; Dowling et al., 2021a; Hodgins et al., 2023), while others have conducted analyses that maximize sensitivity, arguing that the consequences of setting thresholds too high are more serious than the consequences of setting thresholds too low, consistent with public health’s ”precautionary principle” (Dowling et al., 2021). The extant research has also defined gambling-related harm using diagnostic criteria or subsets of items from measures of problem gambling severity, with more recent research employing validated measures of gambling-related harm (Dowling et al., 2021). Limits are generally robust to these definitional variations in gambling-related harm, although superior parameters are generally produced when gambling-related harm is defined as two or more harms (Currie et al., 2006, 2008, 2009; Dowling et al., 2021a).

Harmful Gambling Types

While most of the existing evidence base has identified lower-risk limits applicable across all gambling activities, research has suggested that there are considerable variations in risk associated with specific types of gambling. Various studies indicate that certain gambling activities pose higher risk and associated harm (Binde et al., 2017; Browne et al., 2023; Currie et al., 2006, 2009; Delfabbro et al., 2020; Dowling et al., 2005; Markham et al., 2016; Russell et al., 2023; Wall et al., 2021). Specifically, these studies have found that electronic gaming machines (EGMs or slots) contribute more to the development of gambling problems and harm than other gambling activities, mainly because of their structural characteristics, such as rapid playing speeds and payout intervals, use of multi-credit or -lines, note acceptors, and credited wins. They also reveal that casino table games, sports betting, and racing are high-risk activities relative to other forms of gambling.

Study Aims

To date, no studies have specifically attempted to derive lower-risk gambling limits in a Swedish context. The aim of this study was therefore to identify a preliminary set of lower-risk limits for gambling consumption indices in Sweden. A secondary aim was to investigate which problematic gambling types were associated with harm beyond the effect of the gambling consumption indices.

Methods

Participants and Study Design

This study secondarily analyzed data from two previous online survey studies conducted in Sweden that included the Gambling Disorder Identification Test (GDIT) (Molander et al., 2021). The first study involved gamblers recruited from treatment- and helpline support-seeking contexts, self-help groups, and social media (Molander et al., 2021). Inclusion criteria included: ≥18 years old and past-year gambling participation (excluding the self-help group participants). The second study involved support-seeking gamblers recruited via advertisements on the Swedish gambling helpline (Wall et al., 2025). Inclusion criteria included: a desire for support for their own gambling and visiting the helpline’s website. Both studies received approval from the Regional Ethics Board of Stockholm, Sweden (ref. no. 2017/1479-31 and 2022-03651-01).

The combined sample comprised 705 past-year gamblers from: treatment services (n = 78, 11.1%), helpline support services (n = 291, 41.3%), self-help groups (n = 47, 6.7%), and social media (n = 289, 41.0%). As indicated in Table 1, GDIT scores were classified as 45.5% for recreational/at-risk gambling, 6.8% for problem gambling, and 47.7% for GD (6.7% mild, 6.7% moderate, 34.3% severe). The mean age of the sample was 33.8 years (SD = 12.6), with 72.6% men. Approximately one-third were university-educated (32.8%) and half were in cohabiting households (52.1%). Two-thirds were employed (66.0%), and the median monthly income was approximately 20,000 Swedish crowns (SEK; USD$1,800). One-quarter of the sample had gambling debts, with a reported median debt of 300,000 SEK.

Table 1.

Participant characteristics

	Gambling severitya					Total N = 705 (100%)
	Recreational /at-risk n = 321 (45.5%)	Problem gambling n = 48 (6.8%)	Gambling Disorder
			Mild n = 47 (6.7%)	Moderate n = 47 (6.7%)	Severe n = 242 (34.3%)
Age, M (Sd)	31.4 (12.9)	32.0 (10.6)	35.4 (15.3)	36.2 (11.0)	36.6 (11.7)	33.8 (12.6)
Gender, %
Men	79.8%	77.1%	78.7%	72.3%	61.2%	72.6%
Women	19.0%	22.9%	19.1%	27.7%	37.6%	26.2%
Can/will not disclose	1.2%	0.0%	2.1%	0.0%	1.2%	1.1%
Highest level of education, %
University	42.4%	22.9%	29.8%	25.5%	24.0%	32.8%
High school	45.2%	60.4%	57.4%	63.8%	59.5%	53.2%
Junior high school	9.7%	8.3%	10.6%	4.3%	12.8%	10.4%
Civil status, %
Cohabiting	56.4%	52.1%	40.4%	46.8%	49.6%	52.1%
Children	35.8%	41.7%	42.6%	48.9%	55%	44.1%
Source of income, %
Employed	61.4%	62.5%	66.0%	59.6%	74.0%	66.0%
Studies	28.0%	20.8%	21.3%	14.9%	6.6%	18.9%
Otherb	10.6%	16.7%	12.8%	25.5%	19.4%	15.2%
Net income, past month
SEK, median (IQR), range	18,500 (17,500), 0- 85,000 SEK	20,000 (17,250), 45- 52,000 SEK	23,330 (18,750), 0- 75,000 SEK	21,500 (15,275), 0- 84,500 SEK	21,000 (11,380), 0- 100,000 SEK	20,000 (15,700), 0- 100,000 SEK
Having gambling debts, %	9.0%	27.1%	53.2%	57.4%	80.2%	40.9%
SEK, median (IQR), range	250,000 (595,000), 20 - 8,500, 000 SEK	725,000 (859,000), 300–2,000,000 SEK	225,000 (583,750), 1 - 1500,000 SEK	350,000 (537,500), 100- 1,400,000 SEK	300,000 (640,000), 0- 5,000,000 SEK	300,000 (640,000), 0- 8,500,000 SEK

Note.^aGambling severity assessed using the GDIT

^bThis category comprises unemployment insurance, income support, sickness compensation, sickness benefit, pension, and other sources of income

IQR = Interquartile range

SEK = Swedish crowns

Measures

Participants responded to an online survey, comprising demographic characteristics (age, gender, highest level of education, civil status, source of income, net income, gambling debts). Gambling consumption indices (frequency, duration, expenditure, expenditure as a proportion of income, diversity [i.e., number of problematic gambling types]), gambling harm, and specific problematic gambling types were measured by the GDIT (see Table 2). The GDIT (Molander et al., 2023) consists of 14 items with frequency- and time-based responses across three subscales: Gambling Behavior (items 1–3, scored 0–6), Gambling Symptoms (items 4–10, scored 0–4), and Negative Consequences (items 11–14, scored 0, 2 or 4). The maximum GDIT total score is 62. The GDIT also includes items in an appendix measuring past-month expenditures and gambling types with negative consequences (broken down via internet/online and physical venue). The GDIT is in the public domain, available at www.gditscale.com.

Table 2.

Gambling consumption indices and harm measurement

Gambling consumption indices
Gambling frequency	Gambling frequency was assessed using the first item (GDITitem1) of the Gambling Behavior subscale: “How often do you gamble?“, with response options of “Never” (0), “Monthly or less” (1), “2–4 times a month” (2), “2–3 times a week” (3), “4 or more times a week” (4), “Daily” (5), “Several times a day” (6).
Gambling duration	Gambling duration was assessed using the gambling frequency item (GDITitem1) and the second item (GDITitem2) of the Gambling Behavior subscale: “How much time do you spend gambling on a typical day?“, with responses of “No time” (0), “Less than an hour” (1), “1–2 hours” (2), “3–4 hours” (3), “5–6 hours” (4), “7–9 hours” (5), and “10–24 hours” (6). Gambling duration was derived by first recoding GDITitem1 and GDITitem2 into frequencies per month, i.e., 0, 0.5, 1.5, 3.5, 5, 8 and 17 h, and 0, 1, ,3, 10, 16, 30 and 30 days, respectively. These recoded variables were then multiplied with each other.
Gambling expenditure/gambling expenditure as a proportion of income	Gambling expenditure was assessed using one item in the GDIT appendix “How much money did you wager last month on gambling?” (GDITwager item). The expenditure as a proportion of income was estimated by dividing the appendix items (GDITwager item) by “What was your income after tax last month (including salary and grants)?” (GDITincome item), multiplied by 100.
Gambling diversity	Gambling diversity was assessed in a checkbox grid in the GDIT appendix, specifying engagement in problematic gambling activities (Slots, Poker, Casino table games, Sports betting, Horse betting, Bingo, Lotteries, and Instant lotteries/Scratch cards), assessed online or at venues. These items were summarized into binary variables for access mode (any online or venue gambling, regardless of type), or specific activities (regardless of access mode).
Gambling harm	Gambling harm was measured using the Negative Consequences subscale, encompassing past-year gambling-related worsened mental health (GDITitem12), experiences of financial problems (GDITitem11), or relationship problems (GDITitem13), or work or school difficulties (GDITitem14). The past-year negative consequences/harms endorsed in GDITitems11−14 (excluding responses “Yes, but not in the past year”) were summarized into a numeric variable (1–4 harms). Analyses were then conducted using cut-offs of + 1 harms (to prioritize sensitivity) as well as + 2 harms (in line with most previous research on lower-risk gambling limits.
Problematic gambling types	The checklist in the GDIT appendix used to calculate gambling diversity was employed to measure problematic gambling types. These were dichotomized into a binary variable for each specific gambling type (collapsed online and venue gambling). For descriptive purposes, they were also dichotomized into a binary variable for online or venue gambling (collapsed specific gambling types).

Statistical Analysis

An analysis plan was published (https://aspredicted.org/vf8mz.pdf), outlining the study’s statistical methods a priori. The final study sample deviated slightly from the plan due to the removal of data from four participants. Monthly incomes over 100,000 SEK (n = 14), one reported monthly wager of 150,000,000 SEK, and one reported gambling debt of 40,000,000 SEK were deemed outliers by visual inspection and were replaced with missing values. Regression analyses examining whether problematic gambling activities were associated with harm beyond the effect of the gambling consumption indices, were also subsequently added as a complementary post-hoc analysis.

Descriptive data were presented in mean, standard deviations, and percentages. Numeric variables, including SEK, were presented using median, interquartile range, and range values. Lower-risk gambling limits were identified using Receiver Operating Curve (ROC) analyses across the gambling consumption indices (frequency, duration, expenditure, expenditure as a proportion of income and diversity) in relation to + 1 and + 2 harms. Sensitivity and 1-specificity were plotted for each level of gambling consumption, and the area under the curve (AUC) was estimated for the resulting ROC graph. AUC is a general index of a test’s classification performance, interpreted according to the following guidelines: small (0.50–0.70), moderate (0.70–0.90), and high (> 0.90) classification accuracy (Swets et al., 2000). In the current study, AUC estimates ranged from 0.67 to 0.77, and all values except two demonstrated moderate classification accuracy (i.e., above 0.70). Lower-risk gambling consumption limit ranges were determined based on the highest Youden Index values across + 1 and + 2 harms. The Youden Index provides equal weighting to sensitivity and specificity, thereby minimizing both false positives and false negatives (Ruopp et al., 2008; Youden, 1950).

To examine whether problematic gambling activities were associated with harm beyond the effect of the gambling consumption indices, a stepwise logistic regression approach was employed. First, bivariate associations were estimated between specific gambling types (collapsed online and venue gambling) and + 1/+2 harms, using separate regression models. Second, all gambling types were included in a combined model (Model 1). Finally, the gambling consumption indices were included in a total model (Model 2), where the risk for harms was adjusted for by specific problematic gambling activities. For this full model, participants with a past-month percentage of income wagered larger than 100 were excluded, leaving n = 602 (see Table 3). The magnitude of odds ratios (ORs) (and their reciprocal estimates) was interpreted as small (OR = 1.68), medium (OR = 3.47), or large (OR = 6.71) (Fey et al., 2023).

Table 3.

Association between gambling participation, at-risk indicators, and harms/negative consequences

Predictors	Bivariate associations OR (95% CI), n = 705	Model 1 OR (95% CI), n = 705	Model 2, full model OR (95% CI), n = 602
+ 1 harms
Gambling consumption indices
Frequency			1.03 (0.94–1.13)
Duration (hours per month)	-	-	1.33 (1.21–1.47)***
Expenditure			1.24 (1.10–1.39)***
Expenditure as a proportion of income	-	-	1.02 (0.92–1.13)
Diversity (number of gambling types)	-	-	1.01 (0.91–1.13)
Problematic gambling types
Slots	7.80 (5.59–10.99)***	8.21 (5.72–11.93)***	1.28 (1.18–1.38)***
Poker	1.65 (1.07–2.54)*	1.07 (0.62–1.85)	0.96 (0.87–1.07)
Casino table games	3.33 (2.31–4.85)***	1.90 (1.21–3.00)**	1.07 (0.97–1.18)
Sports betting	1.65 (1.14–2.42)**	2.31 (1.43–3.72)***	1.11 (1.01–1.21)*
Horse betting	1.48 (0.91–2.41)	1.20 (0.64–2.21)	0.99 (0.88–1.12)
Bingo	1.48 (0.71–3.12)	0.84 (0.32–2.26)	0.96 (0.81–1.13)
Lotteries	1.00 (0.56–1.78)	0.67 (0.26–1.59)	1.02 (0.89–1.19)
Scratch cards	1.30 (0.74–2.27)	0.66 (0.29–1.48)	0.97 (0.83–1.12)
	Model fit	R2 = 0.32	R2 = 0.56
+ 2 harms
Gambling consumption indices
Frequency			1.11 (1.01–1.21)*
Duration (hours per month)			1.18 (1.07–1.3)**
Expenditure			1.17 (1.04–1.31)**
Expenditure as a proportion of income			0.93 (0.83–1.03)
Diversity (number of gambling types)			1.01 (0.9–1.13)
Problematic gambling types
Slots	6.08 (4.25–8.79)***	6.62 (4.48–9.95)***	1.25 (1.16–1.35)***
Poker	1.37 (0.86–2.14)	0.84 (0.47–1.44)	0.96 (0.87–1.06)
Casino table games	2.56 (1.78–3.70)***	1.41 (0.90–2.20)	1.05 (0.95–1.16)
Sports betting	1.91 (1.29–2.81)**	2.61 (1.59–4.30)***	1.16 (1.06–1.27)***
Horse betting	1.65 (1.00–2.71)*	1.15 (0.60–2.16)	0.99 (0.88–1.11)
Bingo	1.65 (0.76–3.46)	0.82 (0.30–2.12)	0.96 (0.81–1.13)
Lotteries	1.4 (0.75–2.52)	0.89 (0.34–2.21)	1.04 (0.9–1.2)
Scratch cards	1.74 (0.98–3.07)	0.99 (0.43–2.22)	1.01 (0.87–1.17)
	Model fit	R2 = 0.24	R2 = 0.40

*p < .05, **p < .01, *** p < .001

OR = Odds ratios, CI = Confidence intervals, R² = Nagelkerkes quasi R² (Nagelkerke, 1991)

Model 1 included Slots, Poker Casino table games, Sports betting, Horse betting, Bingo, Lotteries and Scratch cards (regardless of access mode). The full Model 2 included past month expenditure as a proportion of income (dichotomized > 5%), past month duration (dichotomized > 6 h), gambling diversity (dichotomized > 1 gambling types), and Slots, Poker Casino table games, Sports betting, Horse betting, Bingo, Lotteries and Scratch cards (regardless of access mode)

Results

Descriptive Statistics

Table 4 shows that the GDIT total score for the sample was 21.1, which is classified in the mild GD category. Participants reported a mean gambling frequency of “2–3 times a week” and a mean past-month gambling duration of 47.5 h (SD = 98.1, range = 0-510). The median past-month gambling expenditure was 600 SEK (approximately USD$60), and the median expenditure as a proportion of income was 5.3%. For gambling diversity, the mean number of gambling types was 1.3. Online gambling was more frequently reported compared to gambling at venues (68.2% versus 26.7%).

Table 4.

Gambling characteristics (N = 705)

	Gambling severitya					Total sample
	Recreational /at-risk	Problem gambling	Gambling Disorder
			Mild	Moderate	Severe
GDIT total score, M (Sd)	6.4 (3.6)	16.7 (1.4)	21.6 (1.5)	26.9 (1.5)	40.8 (7.9)	21.1 (16.3)
Gambling behavior
Gambling frequency (GDITitem1), mean (frequency)	1.5 (2–4 times a month)	1.9 (2–4 times a month)	2.8 (2–3 times a week)	2.9 (2–3 times a week)	4.2 (4 or more times a week)	2.6 (2–3 times a week)
Gambling duration, hours past month, mean (sd)	3.2 (7.3)	10.2 (14.3)	43.5 (91.7)	35.1 (50.8)	116.9 (134.7)	47.5 (98.1)
Gambling expenditures, past month
SEK wagered,  median (IQR), range	162 (500), 0–150,000 SEK	500 (2,000), 0-534,125 SEK	3,000 (8,900), 0–1,500,000 SEK	5,000 (11,500), 0-200,000 SEK	10,000 (17,625), 0–900,000 SEK	600 (7,970), 0–1,500,000 SEK
Percent of monthly income wagered,  median (IQR)	1.1 (4.0)%	5.4 (12.0)%	15.6 (39.6)%	32.4 (59.7)%	50.0 (83.3)%	5.6 (50.0)%
Gambling diversity
Number of problematic gambling types,  mean (sd), range	0.69 (1.17), 0–8 types	1.60 (1.67), 0–8 types	1.45 (1.14), 0–6 types	1.77 (1.56), 0–8 types	1.91 (1.37), 0–8 types	1.29 (1.42), 0–8 types
Problematic gambling types
Online gambling(%)	40.2%	81.2%	83.0%	87.2%	96.3%	68.2%
Venue gambling (%)	21.8%	20.8%	40.4%	25.5%	31.8%	26.7%
Slots (%)	14.6%	56.2%	53.2%	66.0%	74.4%	44.0%
Poker (%)	8.1%	22.9%	19.1%	17.0%	17.8%	13.8%
Casino games (%)	9.3%	20.8%	34.0%	36.2%	36.4%	22.8%
Sports betting (%)	14.0%	16.7%	19.1%	21.3%	26.0%	19.1%
Horse betting (%)	6.9%	16.7%	10.6%	12.8%	13.6%	10.5%
Bingo (%)	3.4%	4.2%	2.1%	4.3%	5.8%	4.3%
Lotteries (%)	6.9%	14.6%	2.1%	6.4%	7.0%	7.1%
Scratch cards (%)	5.6%	8.3%	4.3%	12.8%	9.9%	7.7%

Note.^aGambling severity assessed using the GDIT

Identification of Lower-Risk Gambling Limits

Lower-risk limits were derived using ROC analyses in relation to both + 1 and + 2 gambling-related harms. A gambling frequency of “2–3 times a week” to “4 or more times a week” (8–16 times monthly) demonstrated sensitivities of 0.54 and 0.71, specificities of 0.75 and 0.87, and Youden’s indices of 0.40 and 0.45 (see supplementary Table 1). Regarding duration, gambling for 6 to 15 h per month demonstrated sensitivities of 0.69 and 0.80, specificities of 0.67 and 0.82 and Youden’s indices of 0.51 and 0.74 (see supplementary Table 2). In terms of expenditures, a past-month wagering of 2,000 SEK showed sensitivities of 0.69 and 0.70, specificities of 0.79 and 0.70, and Youden’s indices of 0.48 and 0.40 (see supplementary Table 3). A ratio of 5% of expenditure as a proportion of income corresponded to sensitivities of 0.74 and 0.71, specificities of 0.69 and 0.58, and Youden’s indices of 0.43 and 0.29 (see supplementary Table 4). For diversity, gambling on one problematic gambling type showed sensitivities of 0.92 and 0.95, specificities of 0.49 and 0.42, and Youden’s indices of 0.41 and 0.37 (see supplementary Table 5).

Harmful Gambling Types

The gambling consumption indices were included in logistic regression analysis, where the risk for harms was adjusted for participation in specific gambling types (Model 2; see Table 3). The results indicated that two types of gambling—slots and sports betting—were associated with harms. Moreover, the analysis showed that gambling expenditure as a proportion of income and duration during the past month were significant predictors of gambling-related harms (p < .001). However, gambling frequency and diversity were not associated with harms (See Table 3).

Discussion

Study Findings

This study aimed to identify a preliminary set of lower-risk limits for gambling consumption for Sweden. The results identified limits relating gambling between “2–3 times a week” to “4 or more times a week” (8–16 times per month), between 6 and 15 h per month, 2,000 SEK (approximately USD$190) per month, and 5% of personal net income. These limits are generally higher than those identified in previous research internationally. This is likely explained by the large proportion of treatment- or support-seeking gamblers with high rates of gambling problems in the study sample, given that problem gambling severity and gambling-related harm are overlapping constructs (Browne & Rockloff, 2017; Delfabbro & King, 2019). Relatedly, there was a high proportion of problematic online gamblers in the study sample, whereby problematic online gambling was reported by nearly seven out of 10 participants. Previous studies that have employed behavioral data from online gamblers have identified higher limits than those identified in the general population, possibly because online gambling has a higher risk profile (Jonsson et al., 2022). Although it therefore remains unclear as to how applicable the limits identified in the current study are to the general Swedish population, they offer insights into lower-risk gambling levels in a health-care or support context, which holds merit in relation to clinical applications.

In contrast, the results suggest that engaging in one single problematic gambling type could be related to harm, which is a lower threshold than those identified in previous studies. This finding can be partly explained by the reporting of only problematic gambling types, as recommended in a consensus agreement among gambling researchers (Walker et al., 2006). Further, when investigating engagement in specific problematic gambling types, slots and sports betting were associated with harm. The current study therefore adds to a growing empirical base suggesting that gambling participation in specific types of gambling, such as slots and sports betting, is harmful (Binde et al., 2017; Browne et al., 2023; Currie et al., 2006, 2009; Delfabbro et al., 2020; Dowling et al., 2005; Markham et al., 2016; Russell et al., 2023; Wall et al., 2021), which may have implications for the development of lower-risk limits for specific gambling types (Brosowski et al., 2015; Dowling et al., 2021a; Dowling et al., 2021; Quilty et al., 2014).

Study Implications

This study was the first step in a larger pursuit to establish national lower-risk gambling guidelines in Sweden. Such guidelines are important to establish for several reasons (Currie et al., 2006, 2008, 2009, 2012, 2017; Quilty et al., 2014; Weinstock et al., 2007, 2008). First, from a public health perspective, lower-risk gambling limits can serve as benchmarks for non-harmful gambling within the general population. Such limits can be used to generate public discussion regarding gambling norms, enable consumers to compare their gambling with these norms, and allow consumers to make informed choices. They can also assist gamblers to reduce their own gambling behavior by increasing their awareness about levels of risk, highlighting potential harms should they exceed the limits, and enhancing motivation to seek help. Second, the establishment of lower-risk gambling guidelines is important from the perspective of the ethical provision of gambling products and industry standards of duty of care. According to the Swedish Gambling Act (Gambling Act 2018 (2018:1138), n.d.), licensed gambling companies are obliged to protect their customers from harms, and subsequently intervene when there is reason to do so. The absence of a uniform set of lower-risk gambling guidelines at a societal level might, from this point of view, result in varying levels of risk employed by different operators and a duty of care that is inconsistently applied. Third, lower-risk limits can be used to monitor the prevalence of gambling-related harm and investigate the impact of secondary intervention efforts. Finally, lower-risk gambling limits are important to establish from a clinical perspective, as such limits can be used as a cost-effective method of screening for people at high-risk of gambling harm in clinical settings and treatment outcome studies and also be applied to gamblers who select non-abstinence treatment goals.

Study Strengths and Limitations

A strength of the current study is the use of the GDIT instrument to measure gambling-related harm, rather than using subsets of items from problem gambling severity measures. As the first to identify lower-risk gambling limits using this measure, this study adds to the emerging research employing validated measures of gambling-related harm (Dowling et al., 2021). Moreover, by incorporating responses based on frequencies or time (Molander et al., 2024), the GDIT enables a comprehensive and systematic assessment of consumption indices relevant for lower-risk gambling limits. In terms of limitations, the study sample was not drawn from the general population and was over-represented in treatment- and support-seeking gamblers, with high rates of gambling problems and problematic online gambling behavior. The limits identified in this study should therefore be considered preliminary, pending validation by future studies with data collected from large, population-representative samples.

Conclusion

Lower-risk gambling limits differentiating lower-risk from high-risk gambling behavior, which are identified empirically, can be converted to lower-risk gambling guidelines, which are promoted to the general public. From an authority perspective, public health considerations are important for the establishment of guidelines at a societal level, such as promoting credible thresholds which can be easily remembered. To date, there exists no universal definition of lower-risk gambling and different at-risk indicators, timeframes and currencies have been employed across studies. Consensus-based methods are therefore arguably the next research step in the development of establishing lower-risk gambling guidelines in Sweden. In such studies, Swedish and international gambling researchers, as well as various representatives of Swedish gambling-relevant authorities, can reach consensus on a set of lower-risk gambling principles. Similar research identifying expert and public perspectives regarding the number of limits, the gambling consumption indices, the timeframes employed, and the harms that should be considered, as well as issues relating to the measurement of these constructs and the promotion of such limits have been conducted in other jurisdictions (Currie et al., 2008; Dowling et al., 2021a). This framework can then be validated in future empirical studies, using larger datasets collected from the Swedish general population.

Electronic Supplementary Material

Below is the link to the electronic supplementary material.

Author Contributions

Nicki Dowling: conceptualization, methodology, writing- original draft, review & editing. Peter Wennberg: conceptualization, writing-review & editing. Håkan Wall: conceptualization, writing-review & editing. Olof Molander: conceptualization, formal analysis, data curation, methodology, writing- original draft, review & editing. All authors contributed to and approved the final manuscript. Study title: Striving towards national lower-risk gambling guidelines: An empirical investigation among a sample of Swedish gamblers.

Funding

Open access funding provided by Karolinska Institute.

The authors have no conflicts of interest to declare, and reports receiving funding from the following sources: In the last 3 years, author ND reports having received research and consultancy funding from multiple sources (unrelated to the study), including via hypothecated taxes from gambling revenue. ND has received research funding from the Victorian Responsible Gambling Foundation, New South Wales Office of Responsible Gambling, Tasmanian Department of Treasury and Finance, Gambling Research Australia, Independent Research Council of Svenska Spel, Health Research Council of New Zealand, and New Zealand Ministry of Health. She has been the recipient of a Deakin University Faculty of Health Mid-Career Fellowship. She has not knowingly received research or consultancy funding from the gambling, tobacco, or alcohol industries or any industry‐sponsored organization. Author PW reports no funding. Author HW was financed by the Swedish Research Council for Health, Working Life, and Welfare (Forte, grant number 2023 − 00898). Author OM reports having received funding from the Independent Research Council of Svenska Spel (unrelated to the current study).

Data Availability

Due to the nature of this research, participants of this study did not agree for their data to be shared publicly, so supporting data is not available.

Declarations

Ethical Approval

Approvals from the Regional Ethics Board of Stockholm, Sweden (ref. no. 2017/1479-31 and 2022-03651-01). An analysis plan was published (https://aspredicted.org/vf8mz.pdf and https://osf.io/zsq3v/) prior to study analyses.

Competing Interests

The author declares no competing interests.