Drinking location moderates the association between social group size and alcohol consumption among young adults: An event-level study

Abstract

Introduction.

Previous studies have demonstrated relationships between social and environmental characteristics of the drinking context and alcohol use. However, the use of event-level data to investigate individual and joint relationships between such characteristics and alcohol use remains a gap in the literature. This study aimed to examine associations between drinking context (location and social group size) and alcohol consumption, and estimate the relationship between the interaction of context and alcohol consumption.

Methods.

Using an Internet-based cellphone-optimised assessment technique, 183 Swiss young adults (mean: 23 years; range: 17–37 years) completed hourly assessments from 8 pm to midnight Thursday through Saturday for five consecutive weeks. Participants contributed 3454 hourly questionnaires. The number of drinks, the number of friends present and location (off-premise–home, outdoors; on-premise–bars, restaurants) were assessed based on the previous hour. Multilevel mixed-effects models were used to assess the relationships of interest.

Results.

Being off-premise compared to on-premise was associated with fewer hourly drinks consumed (b = −0.44, P < 0.001). Greater numbers of friends present were associated with more drinks consumed (b = 0.02, P < 0.001). The association between number of friends and number of drinks consumed was significantly stronger for off-premise compared to on-premise locations (b = 0.03, P < 0.001).

Discussion and Conclusions.

Compared to off-premise locations, on-premise locations are associated with more hourly drinks consumed. However, the positive relationship between social group size and drinks consumed is significantly stronger for off-premise locations compared to on-premise locations. Findings have implications for tailored interventions focused on reducing alcohol consumption by young adults.

Introduction

Alcohol use is impacted by social and physical environments beyond an individual’s personal attributes [1,2]. Event-level research using ecological momentary assessment (EMA) can be used to understand real-time context of the drinking environment [3,4], and such studies have set forth multiple contextual characteristics that are associated with heavy alcohol consumption [5]. These characteristics include timing (Saturdays and later in the evening) and aspects of the social group (the number of friends present, mixed-gender groups and the number of drinking companions) [6–10]. The type of location where an individual drinks may also affect drinking outcomes, but evidence in this respect often comes from studies relying on retrospective recall [11].

Multiple studies have demonstrated that drinking in on-premise establishments (bars, nightclubs, restaurants) is associated with greater alcohol consumption and risk for alcohol-related problems [12–14], although such work has typically relied on retrospective data. One event-level study has also demonstrated that spending more time in bars may be more likely to have accelerated drinking patterns due to the constant environmental cues of alcohol availability and others’ drinking [7]. Location-specific rationale for this relationship includes on-premise marketing to attract specific clientele [15] and advertising drink specials, such as discounts and promotions, that increase the risk of drinking to the point of intoxication [16].

Drinking in off-premise locations (e.g. homes and outdoor settings) can also have distinct effects on drinking behaviour [13,14]. First, pre-drinking (also called pre-gaming or pre-loading) in the home prior to drinking elsewhere later in the evening has been linked to greater quantity of consumption and intoxication [17,18]. Next, drinking in the home can occur in the context of house parties, which may contribute to increased consumption through access to greater quantities of alcohol and participation in drinking games with other party-goers [12]. Finally, off-premise locations may increase drinking volume due to these settings typically being unsupervised and without rules surrounding drinking [19], thus, presenting fewer barriers to overconsumption.

Social group context is also associated with individuals’ drinking outcomes. Compared to drinking alone, drinking in a group, even with as few as two other people, may increase quantities of alcohol consumed [20]. Previous EMA work has also shown the positive relationship between increasing the number of friends present and alcohol consumed [7]. Such effects may arise through beliefs of what is socially acceptable and imitation of others in the group [21], descriptive drinking norms of social groups [22,23], and trust that friends will be sources of safety during engagement in heavy drinking [24].

While previous studies have suggested social groups and location (both on-premise and off-premise) have the capacity to influence alcohol consumption, the use of event-level data to investigate the interaction between locations and social group size remains a gap in the literature [11]. On-premise drinking locations are highly regulated environments, and drinkers may self-select into these locations with the intent to consume alcohol [15] regardless of who they are with, reflecting a potentially weaker influence of the social group on drinking behaviour, in comparison to other locations. It is also possible that the influence of environmental cues, such as alcohol-focused marketing and live music, on alcohol consumption may supersede that of the social group in an on-premise setting. In contrast, the social group could potentially have a greater impact on drinking behaviour in off-premise locations, considering that these locations are less supervised [19], individuals can bring their own drinks, and a greater social group could indicate a party [12]. Moreover, off-premise settings like private homes typically lack elements of a sensory environment that are conducive to influencing alcohol consumption by individuals. As such, engaging in drinking in an off-premise setting may result in consumption being more impacted by the social group. Thus, investigating potential interactions between locations and social group size and their joint impact on drinking behaviour is an important research question to address. Insights gained from this analysis could make a contribution to preventing and reducing heavy drinking by, for example, informing which factors to target in tailored and place-based interventions.

This study uses EMA data from a sample of young adults of legal drinking age to address the following aims: (i) examine the relationship between location type (off-premise versus on-premise) and alcohol consumption; (ii) estimate the association between social group size and consumption across locations; and (iii) investigate how the association between social group size and alcohol consumption may differ based on the location of drinking. For the first aim, we hypothesise that attending on-premise locations (bars, nightclubs and restaurants) will be associated with greater alcohol consumption compared to off-premise locations (home, outdoors). In relation to the second aim, we hypothesise greater numbers of friends in the social group will be positively associated with alcohol consumption. Finally, for the third aim, we hypothesise that the positive association between social group size and alcohol consumption will be significantly greater in off-premise compared to on-premise locations.

Methods

Design

Data were collected as part of an EMA study investigating the drinking behaviour of young adults on weekend evenings [25]. The study consisted of a baseline internet questionnaire and a series of questionnaires completed on participants’ personal cellphones. Participants were college-age students from three higher education institutions from French-speaking Switzerland. Full details about recruitment procedures are described elsewhere [6,7,9,10,26]. Briefly, the 276 enrolled participants were prompted via SMS text messages to complete five cellphone-browser-based assessments from 8 pm until midnight on Thursday, Friday and Saturday for five consecutive weeks. A prompt for an additional assessment of the prior night’s post-midnight activity was sent to participants each following morning, but these assessments were not included in the present study since they did not include questions about the number of friends present. Each evening assessment asked about behaviour in the previous hour, except for the first assessment that asked about the prior 3 h (5–8 pm). Participants were excluded if they did not complete any cellphone assessments (n = 24), did not have any alcohol use during the study period (n = 16) or completed less than three assessments per evening (n = 53). Missing hourly assessments for included participants were imputed using Markov Chains [6] and accounted for 12% of all assessments. The overall sample consisted of 183 participants with data for 7205 hourly evening assessments (hereafter called ‘hourly observations’ or ‘observations’).

Participants

The present work used both participant- and observation-level data. From the overall sample, we excluded 3064 hourly observations (42.5%) where the respondent was alone, resulting in a subsample of 4141 hourly observations with no missing data on the constructs of interest. Additional details about how location was operationalised and related exclusions (n = 687 observations; 9.5%) are noted below. The final analytic sample included 183 participants with 3454 observations (11% [n = 397] imputed). Previous work using data from this sample has compared differences between respondents with complete data and those excluded because of missing observations [18]. The groups were found to be similar in relation to gender and average alcohol consumption, and excluded participants were slightly younger than those who were included (22.1 versus 23.1 years; P = 0.02).

Measures

Alcohol consumption

Alcohol consumption was measured as the total number of drinks consumed during a given hour. In each assessment, participants reported how many of each of the following they consumed in the past hour: beer, wine or champagne, aperitifs (e.g. port) or liqueurs, (straight) spirits, self-mixed drinks or cocktails, and alcopops (pre-mixed drinks); and response categories ranged from ‘0’ to ‘5 or more’ (the latter was coded as 5.5). The amounts for each drink type were added together to generate the total number of drinks consumed. In line with previous studies [6,10,27], two-thirds of reported consumption during the first assessment period was used to approximate consumption before 8 pm.

Social group size

Social group size reflects the total number of male and female friends present during a given hour. Response options for each of the two gender-specific questions ranged from ‘0’ to ‘more than 20’ (coded as 23.5) and were summed to obtain the total number of friends [10,26]. Any observations with a social group size of zero were excluded because our analyses also examined gender composition of the present social group (see below).

Location

Location corresponds to the participants’ primary location, where they spent most of the time during a given hour. For each assessment period, participants were asked how much time they spent, in 15-min increments from 0 to 60 (and 30-min increments from 0 to 3 h for the first evening assessment, 5–8 pm), at each of the following location categories: ‘at a home (my place or someone else’s)’; ‘travelling (bus, car, on foot)’; ‘bar, restaurant, pub, or nightclub’; ‘outdoors, public park, or natural setting’; ‘cultural or sporting venue (cinema, theatre, stadium, etc.)’; and ‘at work, in class’. The time increments for locations during the first assessment period were adjusted to be proportional to an hour. For example, being at home for 2 h was adjusted to 40 min.

We determined primary location according to where a participant spent the majority of his or her time (40 min or more) during an assessment period. By including only observations where at least two-thirds of the hour were spent in one location with one group of people, we could examine the isolated effect of location type and social group size on alcohol consumption, assuming that the majority of drinking reported in that hour likely occurred at that location.

Primary locations were categorised as ‘on-premise’ or ‘off-premise’ (=0 and =1, respectively). Primary locations that were ‘bar, restaurant, pub, or nightclub’ or ‘cultural or sporting venue (cinema, theatre, stadium, etc.)’ were categorised as ‘on-premise’. For example, if a participant reported spending 20 min in travel and 40 min at a bar, that observation was assigned to the ‘on-premise’ category (n = 948). Primary locations that were ‘at a home (my place, or someone else’s)’ or ‘outdoors, public park, or natural setting’ were categorised as ‘off-premise’. For example, if a participant reported spending 40 min at home and 20 min in travel, that observation was assigned to the ‘off-premise’ category (n = 2506). We excluded observations where participants did not spend at least 40 min at any of the reported locations (n = 450); for example, an observation with 30 min at home, 15 min in travel, 15 min in a bar would be excluded. Primary locations that were either ‘travelling (bus, car, on foot)’ or ‘at work, in class’ were also excluded from the analyses (n = 237). Compared to observations in the current analyses, excluded observations had significantly lower alcohol consumption (0.9 vs. 0.7 drinks, t = 3.7, P < 0.001).

Covariates

Covariates for the current analyses were based on previous work exhibiting associations between individual and contextual characteristics and alcohol use. The following covariates were included: participant gender (participant level; female = 0, male = 1); weekday (evening level: Thursday = 0, Friday = 1, Saturday = 2) [6,9]; and social group gender composition (hourly observation level: mixed-gender = 0, all-male = 1, all-female = 2) [10].

Statistical methods

The analytic sample included 183 participants and 3454 hourly observations that included at least one friend present and a primary location categorised as either on- or off-premise. For descriptive statistics, means and medians of each variable were calculated for the total sample and for on- and off-premise locations separately. Based on the skewed distributions of the count data for alcohol consumption and social group size, the nonparametric equality-of-medians test was used to compare the medians of the number of drinks and the number of friends across location type. Spearman’s rho was used to examine correlations between the number of drinks and the number of friends within each location type.

We used multilevel mixed-effects negative binomial regression to account for the nested data (hourly observations nested within evenings nested within participants). We first estimated the main effects model according to the focal independent variables, which were location and social group size (both hourly observation level) and covariates. The covariates were participant gender (participant level), weekday (evening level) and gender composition of the social group (hourly observation level). We then used an interaction model to examine whether an individual’s location moderated the relationship between their social group size and alcohol consumption, accounting for the same covariates noted above. Specifically, we examined whether the association between the number of friends present and the number of drinks consumed in a given hour (observation level) differed depending on whether the individual reported being at an on-premise or off-premise location. Reported effect sizes were unstandardised regression coefficients (b), standard errors (SE) and incidence rate ratios (IRR). Analyses were conducted using Stata 14.2 [28].

Results

Descriptive analyses

Participants’ average age was 23.1 years (SD = 3.1), and 53% (n = 97) were female. At the observation level, approximately three-quarters of hourly observations were categorised as off-premise locations (Table 1). Across all observations, the median number of drinks was zero and the median number of friends present was three. When comparing across locations, the median number of drinks and the median number of friends present were both lower for off-premise compared to on-premise (χ² = 202.9 and 207.6, respectively; P values <0.001). All drink types except alcopops were more commonly consumed in on-premise locations compared to off-premise (Table 2). Of those observations that included at least one drink of a given type, there were no differences in the quantity of different drink types across location except a marginal difference in spirits and mixed drinks (less than two drinks in on-premise compared to two drinks in off-premise, P < 0.10).

Table 1.

Descriptive estimates of the number of drinks and the number of friends by location (n = 3454 hourly observations)

	Number of observations % (n)	Number of observations with 1+ drinks % (n)*	Number of drinks: M (Median) ± SD (IQR)**	Number of friends: M (Median) ± SD (IQR)**	Correlation coefficient (ρ)***
Overall	100 (3454)	44.5 (1504)	0.9 (0) ± 1.4 (0–1)	6.0 (3) ± 9.0 (1–6)	0.38
On-premise (bars, venues)	27.4 (948)	63.1 (598)	1.3 (1) ± 1.4 (0–2)	9.4 (4) ± 11.6 (2–13)	0.19
Off-premise (home, outdoors)	72.6 (2506)	36.2 (906)	0.7 (0) ± 1.3 (0–1)	4.7 (2) ± 7.4 (1–4)	0.40

^*Location-specific proportions of observations with any alcohol consumption different across location type at P < 0.001 (χ² test).

^**Medians significantly different across location type at P < 0.001 (nonparametric equality-of-medians test).

^***All correlations between the number of drinks and the number of friends significant at P < 0.001. IQR, interquartile range; ρ, Spearman’s rho assessing correlations between the number of drinks and the number of friends.

Table 2.

Descriptive estimates of drink type by location (n = 3454 hourly observations)

	On-premise (n = 948)		Off-premise (n = 2506)
	1+ drinks % (n)	Number of drinksa M (SD)	1+ drinks % (n)	Number of drinksa M (SD)
Total	63.1 (598)	2.0 (1.33)	36.2 (906)**	2.1 (1.43)
Beer	30.6 (290)	1.7 (0.98)	13.7 (342)**	1.7 (1.02)
Wine	22.8 (216)	1.9 (1.09)	17.6 (440)**	1.9 (1.05)
Apertifs and liqueurs	2.9 (27)	1.4 (0.75)	1.5 (37)*	1.7 (1.04)
Spirits and mixed drinks	15.8 (150)	1.8 (1.26)	6.9 (173)**	2.0 (1.30)***
Alcopops	0.7 (7)	1.4 (0.53)	1.0 (25)	1.4 (0.58)

Significant differences across location type:

^*P < 0.01.

^**P < 0.001.

^***P < 0.10.

^aIncludes only observations with 1+ drinks.

Multilevel negative binomial regression analyses

Results showed a significant relationship between location and alcohol consumption, where being in off-premise locations was associated with fewer drinks consumed per hour compared to being in on-premise locations (Table 3, Main effects model). A high number of friends present, Saturday evening and male gender were also associated with more drinks consumed. Compared to social groups who were mixed-gender, both all-male and all-female social groups were associated with fewer drinks consumed.

Table 3.

Multilevel negative binomial regression models predicting hourly number of drinks consumed (three-level models)

	Main effects model		Interaction model
	b (SE)	IRR (95% CI)	b (SE)	IRR (95% CI)
Hourly observation level
Number of friends	0.02 (0.003)	1.02 (1.01, 1.03)*	0.01 (0.00)	1.00 (1.00, 1.01)
Location type
On-premise	Ref	–	Ref	–
Off-premise	−0.44 (0.06)	0.64 (0.58, 0.72)*	−0.75 (0.07)	0.47 (0.41, 0.54)*
Number of friends ×
On-premise			Ref	–
Off-premise			0.03 (0.01)	1.03 (1.02, 1.04)*
Social group gender
Mixed	Ref	–	Ref	–
All male	−0.45 (0.07)	0.64 (0.55, 0.74)*	−0.42 (0.07)	0.66 (0.57, 0.76)*
All female	−0.63 (0.08)	0.53 (0.46, 0.62)*	−0.59 (0.08)	0.56 (0.48, 0.65)*
Evening level
Weekend day
Thursday	Ref	–	Ref	–
Friday	0.11 (0.09)	1.12 (0.93, 1.34)	0.11 (0.09)	1.12 (0.93, 1.34)
Saturday	0.35 (0.09)	1.42 (1.19, 1.68)*	0.36 (0.09)	1.43 (1.21, 1.70)*
Individual level
Gender
Female	Ref	–	Ref	–
Male	0.46 (0.12)	1.59 (1.26, 2.00)*	0.46 (0.12)	1.59 (1.27, 2.00)*

^*P < 0.001. CI, confidence interval; IRR, incidence rate ratio (e.g. 1.55 means 55% greater number of drinks compared to reference category); SE, standard error.

A second model was used to examine whether the association between social group size and alcohol consumption differs according to location (Table 3, Interaction model). We found a significant interaction between location type and alcohol consumption. The association between increasing numbers of friends and hourly alcohol consumption was significantly stronger for off-premise compared to on-premise locations. As displayed in Figure 1, there was a positive association between the number of friends and alcohol consumption for both types of locations, but the positive association was significantly steeper in off-premise compared to on-premise locations.

Figure 1.

Predictive margins of the number of drinks per hour according to the number of friends present, by location type.

Discussion

Using event-level data, this study examined whether the number of friends present has a differential relationship with young adults’ hourly alcohol consumption in on-premise versus off-premise locations. Corroborating our first hypothesis, being in on-premise locations (pubs and nightclubs) was associated with greater hourly consumption compared to being in off-premise locations (home and outdoor environments). The results are, in part, explained by the fact that drinking is more likely to occur in on-premise locations compared to home and outdoors locations. Individuals may feel compelled to consume more alcohol in on-premise locations due to norms of drinking being acceptable or encouraged [22] and by other patrons’ main intention to drink [15]. The on-premise physical environment might also encourage alcohol consumption through drink specials [16], sensory stimulation by lights and sounds [29], and music with alcohol references in the lyrics [30].

Concerning the second hypothesis, the main effect of social group size was positively associated with hourly alcohol consumption, mirroring results from previous EMA work [9]. Moreover, the moderation results were in line with our third hypothesis and showed a significant interaction of location and social group size on alcohol consumption. As hypothesised, for off-premise locations, having additional friends present was associated with a greater number of hourly drinks consumed, whereas in on-premise locations, the number of drinks consumed increased only slightly in the presence of additional friends. These results persisted even when accounting for other previously-established characteristics associated with alcohol consumption, including day of the week and gender composition of the social group [6,7,10].

The stronger association between social group size and alcohol consumption in off-premise locations may be explained by a number of factors, one of which is the context of large social gatherings. Although alcohol use is less frequent at home or outdoors, when drinking does occur, the quantity of alcohol consumed appears to be more strongly associated with social group size, compared to bars and restaurants. This finding reflects previous studies that have demonstrated that off-premise locations can have an impact on drinking behaviour, particularly through parties and social gatherings [12–14]. Mirroring some of the conclusions from Thrul et al. [31], the significant interaction between social group size and off-premise location might be indicative of events like house parties, pre-drinking (or ‘pre-gaming’) before going to an on-premise location [18,32], or gatherings of friends in a public outdoor space.

Another explanation for the greater association between social group size and alcohol consumed in off-premise locations is the absence of barriers that prevent excessive alcohol use. Although the presence of intoxicated individuals increases a person’s drinking regardless of location [12], off-premise locations are both unregulated and unsupervised [19] and may provide a setting in which social groups exert unchecked influence on excessive use. Each attendee may bring his or her own alcohol into the setting, either for themselves or for the group. The presence of more attendees may thus expose individuals to larger quantities of available alcohol [12], which may contribute to increased consumption. In addition, off-premise locations typically do not have the supervisory policies that are typical of nightlife venues, such as not serving patrons who appear overly-intoxicated [33] and having ‘bouncers’ or other security present [34]. As such, off-premise locations do not have structural features that may prevent individuals from drinking at risky levels.

The key findings from this work have implications for future research. In particular, research might focus on understanding the moderating relationship of the off-premise environment in a more nuanced way using EMA data. That is, are there differential effects of being a host versus an attendee; drinking in a supervised environment (e.g. presence of parents, on-campus housing proximally close to administrative oversight) versus being unsupervised; and whether or not the gathering is considered a party or some other celebratory occasion? With the latter question, work could also seek to clarify what is meant by ‘party’, given that it could have variable and subjective meanings, ranging from a small gathering to a larger collection of friends and acquaintances at a house party. In addition, the present work also categorised all outdoor environments into the off-premise category, so future research might seek to qualitatively understand the context and effects of different outdoor spaces, including public parks and camp sites.

Our findings also speak to areas where improvements might be made when intervening on risky drinking behaviours. Our work adds context to the previously-demonstrated associations between drinking in off-premise locations and subsequent alcohol-related consequences through quantity of consumption by showing that social group size may be important to consider on the path to negative consequences of alcohol use. Taken together, these results may inform in-person and mobile interventions to reduce risky drinking among young adults [35,36], especially with emphasising how people and place together might affect a person’s drinking outcomes and subsequent harms. This study also speaks to the opportunity to use location-specific brief interventions focused on celebratory gatherings or party environments [37]. An example might be teaching hosts the importance and social responsibility of monitoring attendees’ intake and strategies for doing so, such as denying access to alcohol in cases of over-intoxication. Such interventions may be bolstered by legislation focusing on socially-responsible hosting [38,39] as well as incentives to compel attendees’ to drink less for a chance to win a cash prize [40].

Concerning limitations of this work, this study included a sample of young adults from one European country, so the findings may not be generalisable to populations of other age groups and countries with different alcohol cultures or legislations. We also could not account for friends’ alcohol consumption as well as any special occasions in relation to the location or group size. We were also not able to track whether the friend group changed within the hour. For the measurement of the number of drinks consumed, participants were provided a visual with examples of standard drink sizes with their baseline questionnaire. However, they were not reminded of this information with every hourly assessment, so we do not know how participants may have quantified larger (pints versus cans) or stronger (double shot vs single shot) drinks. Future research should seek to confirm the present results while accounting for the real amount of alcohol contained in each drink. The study also could not examine the effects of other situations where locations and earlier time of day may interact (like brunching, day-drinking, celebrating after a sports club practice or game). Similarly, we were unable to examine the relationships of interest for any time after midnight, which may potentially involve even heavier alcohol consumption. Future research should investigate details surrounding locations and expand the time periods covered by assessments.

A strength of this work is that it used an EMA design to collect data on location, group size and alcohol consumption over the course of weekend evenings, in the natural environment, and in near real-time. As such, these data have a higher level of ecological validity and limited recall bias compared to previous cross-sectional studies [3,4].

Conclusions

This work uses event-level data to examine the effects of locational and social context on alcohol consumption in a sample of young adults and highlights the effects of the intersection of physical and social characteristics in the drinking environment [11]. Drinking in off-premise social environments may have low regulatory oversight or lack of alcohol use monitoring, and therefore put individuals at risk for excessive alcohol consumption. Findings call attention to the need for messaging and interventions to prevent potential harms of excessive drinking in unmonitored environments.