Previous frameworks stem from findings on inter-individual differences in neurophysiological, genetic, personality, social, and environmental factors and their interaction in contribution to adolescent alcohol use [20-23]. The Incentive Sensitization Theory (IST) [24] proposes alcohol as a potentially rewarding stimulus that is liked at first (ie, its use is positively reinforced), for instance, due to its positive impact on mood and evolves into a so-called wanting in the course of addiction development, where its use is negatively reinforced for instance through the amelioration of negative mood states or craving [24]. Greater incentive salience is assigned to alcohol-related cues (eg, the favorite bar or the favorite alcoholic drink), which results in greater attention to these cues compared with other rewarding cues or conditions [25]. Importantly, these cues can then trigger the urge to drink alcohol [26]. Especially adolescents show hyperreactive responses to alcohol-related rewarding cues [7,27]. This model represents a theoretical approach to explaining how alcohol use transitions from early use to habitual use over time. However, specifically in the very early periods of alcohol use, that is, the onset and first experiences, information on these processes is rather scarce. Recently, through advanced methodologies, it is becoming possible to receive much more fine-grained information at the intraindividual level and in daily life to explain when and why adolescents consume alcohol.

Methods to map daily life experiences, trajectories, and movements in daily life are increasingly available. These encompass ecological momentary assessments (EMAs) and GPS tracking (geospatial measures), enabling the concurrent investigation of interindividual mechanisms identified as risk factors for alcohol use (eg, IST) alongside behavioral patterns. These patterns include movements within the daily environment and exposure to specific environmental triggers (eg, nightclubs, youth gatherings, or social contexts such as being in close proximity to peers). This combination is an important addition to traditional diary methods as a resource for gaining information on individuals’ health status over time. Due to the potential to capture momentary states multiple times throughout the day, EMA methodology allows to obtain a high-resolution representation of self-reports, including emotions and thoughts as well as daily behavior and information on contexts, including social situations. Its intuitive usability further ensures greater attractiveness and, consequently, increased compliance and ecological validity compared with retrospective assessments as well as paper and pencil applications [28,29]. This also becomes significant in the realm of health research, as it provides more sensitive data regarding critical daily triggers for symptoms. Consequently, it aids in deciding on the most adaptive interventions. Previous studies using EMA have demonstrated that in adolescent and young adult alcohol users, contextual factors [30], such as the presence of peers or being in specific locations like bars or restaurants [31], along with mood-related factors, including heightened negative affect [32-34], are highly influential in initiating alcohol use. In addition, smartphone technology allows us to assess geospatial movement patterns as proxies for exploratory behaviors and environmental context factors in which specific behaviors occur (ie, alcohol-associated trigger points) and combine this with information gathered from using EMA. This provides the possibility to investigate and identify triggers with not only respect to individual feelings or thoughts but also specific contexts, including information on where (eg, private or public environments) or when (eg, time of the day, day in the week) adolescents consume alcohol [35]. These triggers can also encompass specific situational cues that are associated with alcohol use [36-38]. This information can be gained through exploratory movement patterns, which are estimated, for example, through roaming entropy, an index of “the variability in an individuals’ physical location over the course of a day” [39]. Even in the early period of alcohol use, these data could be beneficial to explore how potential trigger points develop over time. In terms of alcohol use, geospatial measures have successfully been used in app-based alcohol use disorder treatment in adults by sending alerts if patients are near alcohol-related trigger points that are risky places for lapse, relapse, or craving [40-43]. Furthermore, studies have demonstrated that neighborhood disadvantages are associated with increased substance use among adolescents [44], and adolescents tend to consume alcohol in proximity to locations where it is available for purchase [45]. However, environmental exploratory movement patterns, such as roaming entropy, have not been considered in these studies. Thus, together with EMA data, geospatial measures might allow additional insight into adolescent drinking behavior and driving factors and can add to our understanding of transitions from nonrisky to risky alcohol use, and thus inform the development of early prevention approaches, for example, in the form of ecological momentary interventions (EMI). Therefore, this study aimed to delineate the influence of daily variability in affect, craving, rumination, and presence of others (social context), together with exposure to potential environmental alcohol triggers and the characterization of environmental movement patterns on early adolescent daily alcohol use.

Participants were included in the study if they were either 14 or 16 years old at the time of enrollment, aiming to capture early drinking patterns typically initiated around the age of 14 years, and the period when alcohol consumption becomes more frequent and prevalent, around age 16 years [4]. This approach facilitated a better understanding and comparison of these developmental stages. In addition, participants had to be fluent German speakers, right-handed, safe to perform magnetic resonance imaging tasks (eg, no metal in the body or claustrophobia), and without any present history of psychiatric or somatic disorders [46]. Out of 72 recruited participants, 70 participants (twenty-nine 14-year-olds, thirty-six 16-year-olds, 49% female) underwent the entire assessment battery. This included a series of questionnaires (eg, sociodemographic information of participants and their parents, general alcohol consumption, stress, mindfulness, experience with meditation techniques, emotion regulation, personality, quality of life, and rumination), behavioral tasks, neuropsychological tasks in an magnetic resonance imaging scanner, and EMA. Participants received a reimbursement of €50 (approximately US $58) for their participation in the entire study. For the present purpose, only the EMA was of interest, and 65 participants underwent the entire EMA assessment.

Within our subproject of the IMAC-Mind consortium [64], participants were recruited through advertisements in schools in Mannheim, Germany, through social media, and with the help of the local registration office of Mannheim. Both participants and their primary caretaker gave written informed consent before study participation. The ethics committee of Medical Faculty Mannheim, Heidelberg University, approved the study (2007-024-N-MA).

There was no significant difference in alcohol use between participants doing the EMA under COVID-19 restrictions compared with participants doing the EMA before COVID-19 occurred (t₅₀=–0.372, P=.71), and no significant differences in alcohol use due to sex (t₅₀=0.199, P=.85). However, there was a significant effect for age (t₅₀=–2.910, P=.005), with 16-year-olds (mean 1.61, SD 1.66) reporting significantly more alcohol drinking days than 14-year-olds (mean 0.548, SD 0.72). We noted a significant positive correlation (r=0.49, 95% CI 0.26 to 0.68; P<.001) between reported alcohol days within the EMA and AUDIT sum score. Additionally, there was a significant sex difference in rumination (t₅₀=–2.513, P=.02), with males having lower scores (mean 1.56, SD 0.513) than females (mean 2.08, SD 0.907) and a significant difference between participants doing the EMA under COVID-19 restrictions compared with participants doing EMA before COVID-19 occurred in negative affect (t₅₀=–2.091, P=.04) with participants doing EMA before COVID-19 having lower scores (mean 0.89, SD 0.67) than participants doing EMA under COVID-19 restrictions (mean 1.26, SD 0.55) and social context (t₅₀=2.027, P=.048) with participants doing EMA before COVID-19 having higher scores (mean 0.549, SD 0.215) than participants doing EMA under COVID-19 restrictions (mean 0.438, SD 0.162). All other variables were comparable between males and females and between 14- and 16-year-olds (all P>.05; more details in Table 1 and Table 2, and Figure S2 in Multimedia Appendix 1). Furthermore, no significant differences were observed in any of the variables between participants who did not meet the inclusion criteria of compliance (>50%) and those who did meet these criteria (all P>.05).

An overview of the model statistics can be found in Table 3. The model comprises a total of 587 observation days, which is the sum of the days on which participants completed at least 4 prompts per day. There is high variability between participants, with a range of 4 days included in the model for the participant with the fewest days and 14 days for the participant with the most days (mean observation days per participant: 11.29, SD 3.15). Comparing the deviance of the conditional or full model (with fixed and random effects) to the marginal or null model (containing only fixed effects), we see an overall significant difference. The pseudo R² of the generalized linear mixed models for both conditional and marginal models show weak to moderate fit. While some predictors remained stable, there was variability across models with different amounts of prompts included (Table S2 in Multimedia Appendix 1 shows detailed results of models with 1, 2, 3, 5, or 6 prompts included). In the model analyzing 4 prompts per day (more details in Table 3), predictors of alcohol use were positive affect (b=0.685, P=.007), rumination (b=0.586, P=.02), weekend day (b=1.082, P<.001), and RE (b=8.126, P=.02). All other predictors did not reach significance in this model.

The aim of this study was to examine the influence of daily life experiences, such as affect, craving, rumination, and social context, together with the exposure to potential environmental alcohol triggers (TP) and environmental movement patterns (RE) in contributing to daily alcohol use in healthy adolescents in the very early period of alcohol experience. Our exploratory results show that daily variations in positive affect, rumination, and RE can predict drinking behavior in healthy 14- and 16-year-old adolescents. We identified core concepts that might represent a pattern associated with a higher risk for the development of risky alcohol use.

Interestingly, positive affect (not negative affect) was significantly positively associated with alcohol use on that day in our sample, with robust effects across varying levels of compliance rates (more details in Table S2 in Multimedia Appendix 1). From previous research concerning affective states on alcohol use (stress reactivity theory [5]), it would be expected that alcohol use would follow negative affect, representing a coping strategy to deal with unwanted and aversive emotions [65,66]. In contrast, our results indicate that during early alcohol use, with a low level of frequency, that is, seldom use, adolescents use alcohol when experiencing positive emotions during the day, suggesting it might be used to enhance positive emotions. This is in line with previous work on alcohol-drinking motives in adolescents [18,67]. In addition, adolescents often underestimate the risks of alcohol use and have positive attitudes toward alcohol [68]. However, it is important to note that our data does not allow for the derivation of directionality between positive affect and alcohol use. This limitation stems from the retrospective assessment of alcohol use in this study, conducted without further situational or contextual information. It is plausible that either affect may change as a result of alcohol use, or that situational features prompt both positive affect and alcohol use (eg, a celebratory event, more details in Table S3 in Multimedia Appendix 1), or that adolescents experiencing positive emotions are more likely socialize and thus consume alcohol with peers. Future studies could address this by adopting a more detailed assessment of alcohol use, including information on when, where, and in what context alcohol was consumed, or by using event-related sampling schemes to specifically collect data when alcohol is used.

Furthermore, another central factor in association with alcohol use was rumination, with higher rumination being associated with a higher probability of alcohol use on that day. This association was rather weak compared with the other predictors, and rumination ratings were generally low, but the results are still robust across differing compliance thresholds (more details in Table S2 in Multimedia Appendix 1). Thus, our data suggests that already in early, as of yet non-hazardous alcohol-using adolescents, rumination is a factor that contributes to momentary alcohol use. As we outlined above, we cannot infer directionality from our data. However, it can be speculated that even at early use, alcohol might be used to relieve negative thoughts, posing a risk factor for the development of risky drinking patterns. This is in accordance with a body of work investigating rumination as a predisposing factor of alcohol use in patients with alcohol use disorder [69] and also in adult [70] and adolescent [71,72] regular drinkers. Furthermore, our results on rumination and positive affect might be cautiously interpreted in regard to the IST [72], representing the rather prominent “liking” mechanism of early alcohol use (experiencing positive affect) [73,74] and a risk potential already at early stages for a “wanting” mechanisms (relieving rumination), which might later generalize from momentary to more stable patterns, as more habitual alcohol use develops.

Our results regarding environmental exploratory movement patterns, such as RE and potential alcohol trigger points (TPs), suggest that, unlike individuals with alcohol use disorder or heavy drinkers [31,42], adolescents in the earlier stages of alcohol use may not yet have established stable environmental cues or locations that consistently trigger alcohol use, neither on weekdays nor on weekends (more details in Table S4 in Multimedia Appendix 1). Furthermore, places potentially associated with alcohol use were visited similarly during both weekdays and weekends (more details in Figure S3 in Multimedia Appendix 1). However, it is important to consider that we used a “standardized” categorization of trigger points across the entire sample. Future studies could benefit from assessing more personalized potential trigger points for adolescents by directly asking them where they typically consume alcohol and associating these locations with EMA-based information such as affect or rumination.

Nonetheless, our results suggest that these TPs might indeed be developing, as indicated by higher RE. RE emerged as a consistently robust predictor of alcohol use across various models (more details in Table S2 in Multimedia Appendix 1). It could be speculated that this association is driven by the fact that RE may lead to increased exposure to potential alcohol contexts, such as alcohol advertisements, bars, drinking opportunities, or social contexts. However, there may be other factors contributing to this relationship. For instance, higher RE could be linked to the motivation to drink, such as prosocial drinking motives [67,75], or it could reflect a decrease in daily structure and reduced parental supervision [76,77] or a lack of scheduled activities [78-80], such as regular participation in sports clubs. In addition, it is possible that RE varies substantially within individuals from day to day due to planned behaviors related to alcohol consumption on drinking days compared to nondrinking days [81]. Exploring the relationship between RE and these factors influencing alcohol use warrants further investigation in future studies, potentially using event-related sampling schemes to align movement patterns with alcohol use patterns. In addition, in order to learn more about the development of alcohol use disorders, it would be beneficial to examine whether a similar association with alcohol use remains in adolescents with alcohol use disorders.

Finally, our results indicate that alcohol use also seems to be dependent on adolescents’ age and day of the week. These findings corroborate previous work showing alcohol use increases with age [82] and that adolescents mainly use alcohol on weekends [83,84] when there are usually no obligations to be fulfilled, but there is instead leisure time to spend with peers. Importantly, our results showing adolescents drink more on weekend days might reflect a particular behavioral pattern of adolescence that can change over time (less drinking, but also during the week) or become more pronounced (development towards alcohol addiction).

Our results have implications for prevention approaches, which should sensitize adolescents to alcohol use and its consequences and risks at an early stage before alcohol use manifests in hazardous or risky patterns. At this stage, adolescents might be less aware of the risks concerning alcohol, such as a gradual beginning that might cascade into problematic alcohol use [85]. A relevant concept of prevention might be mindfulness, which has already been successfully used in alcohol addiction treatment [86,87]. However, our results on trait attention regulation, as a relevant part of mindfulness [88], do not show a relevant influence on daily alcohol use (more details in Table S5 in Multimedia Appendix 1). It can be assumed that both mindfulness and attention regulation might be situation-dependent (eg, someone is stressed or in a hurry vs relaxed) and thus may also have a situation-dependent influence on alcohol use.

Several limitations of this study should be noted. First, we assessed alcohol use only once a day in the morning, inquiring about alcohol use on the previous day. Consequently, we cannot infer the directionality underlying our findings nor establish any causal relationship. In addition, we are unable to link specific alcohol use episodes to momentary ratings or motives (EMA), social contexts, event appraisal, fluctuations in self-esteem, environmental locations (geospatial information), or the amount of alcohol consumed, which would provide additional crucial information for understanding the occurrence of alcohol-related harm [89]. Future studies could benefit from using participant-initiated assessments (event-related sampling), such as when drinking episodes commence or when participants report a specific context or emotional state [31]. This would provide insights into how the appraisal of occurring events and daily fluctuations in self-esteem contribute to and interact with emotional states and environmental exploratory movement behaviors during certain alcohol consumption occasions. This, in turn, would offer a comprehensive understanding of additional factors that contribute to the initiation of alcohol use in adolescents on a daily basis and help to investigate potential comorbidities with alcohol use behaviors, such as depression, where self-esteem is discussed as a potential critical factor as well [54]. This could also help in further examining alcohol-related coping strategies and their interaction with individual factors.

Second, our study partially coincided with the COVID-19 pandemic in Germany [63], so the generalizability of our results is limited due to COVID-19 restrictions regarding environmental and social factors [90]. Nevertheless, we identify factors influencing alcohol use even under these extremely restricted circumstances. Thus, our results might be underestimated compared with times when access to alcohol and peer communities is easier.

Furthermore, we lacked information on the socioeconomic status of participants and their families. We attempted to estimate this based on the type of school attended by the adolescents (more details in Table S6 in Multimedia Appendix 1), but future studies may benefit from directly inquiring about socioeconomic status, such as family income, as this factor may also influence adolescents’ drinking behavior [91]. In addition, geospatial information like GPS shows a certain inaccuracy, especially in a dense urban environment like Mannheim, which leads to ambiguity as to what a participant is currently seeing or interacting with. This is exacerbated by the reliance on publicly available crowdsourced data sources, which correctness and relevance are not guaranteed, especially in view of temporal gaps between the EMA study and the data processing or temporal development within the assessment phase.

Our results provide a first indication of how emotional and other alcohol-related factors might be related to adolescent activity profiles, that is, RE, to predict very early adolescent alcohol use. However, some of the questions raised remain unanswered. Additional EMA studies are warranted to validate the associations identified herein, including within other, more diverse (adolescent) populations, and to assess whether completing an EMA could potentially influence subsequent behavior, such as through postassessment questionnaires [92]. Once these findings are corroborated, low-threshold smartphone or app-based EMIs or ecological momentary prevention approaches could be derived from them. These interventions may assist adolescents in the early stages of alcohol use and potentially prevent the progression to risky or even abusive alcohol consumption during this vulnerable phase by bolstering resilience factors, such as psychoeducation or mindfulness [93].

Geospatial data might contribute to our understanding of developing risky alcohol use in addition to the aforementioned trigger points; in fact, internal contexts (ie, emotional trigger states) might be identified by linking emotional experience or changes in emotional states with geolocation (construction of emomaps [94]). This might be an important step for future research and would enable a link between environmental causes and behavior, such as alcohol use [95]. Building such emotional maps could be achieved by combining GPS positioning with EMA assessment as in our study, thereby enabling the exploration of the interaction of different alcohol theories with the real-life experiences of participants.

Furthermore, we observed reduced effects in certain variables when applying a stricter threshold for EMA participation, possibly due to fewer observations (more details in Table S2 in Multimedia Appendix 1). Future approaches could potentially enhance participant compliance by offering incentives, a common challenge in EMA studies [48,96], such as reducing the number of prompts per day. Our model comparison indicates that, with a fixed time-based sampling method, an optimal balance for prompts per day exists, ensuring adequate momentary data collection for statistical analysis and comparison. Study designs could be further strengthened by using a longitudinal design and a more age-diverse sample so that the effects on alcohol development could be observed over the period of adolescence.

Through the use of an EMA and geospatial approach, we identified daily life positive affect, rumination, weekend days, and environmental exploratory movement behaviors, as indicated by roaming entropy, as prominent factors associated with very early adolescent drinking behavior during the initial stages of drinking pattern development. Specifically, exploring such movement patterns could serve as an important avenue for future research, offering insights into the development of trigger points from early alcohol use to potentially hazardous use or use disorders.

While environmental factors may not yet play a significant role in the early stages of alcohol use, they are recognized as critical mechanisms in later stages of alcohol use. This understanding can inform the development of future research and may be used in prevention approaches during the early stage of alcohol use.