Wearable fitness technologies (eg, smartwatches and smart rings) are increasingly popular among young adults but may be missing crucial behavioral health data to guide behavior change. Over half (52%) of young adult consumers in the United States own commercial wearables and use them for fitness, stress or weight management, sleep, and other wellness goals [1]. Their appeal lies in their ability to reliably measure physiological signals (eg, heart rate, blood oxygen, and skin temperature) in a noninvasive, accessible way via biometric sensors [2]. However, physiological data are provided to users without information about concurrent behaviors that impact physiological states and patterns [3], including alcohol use. This leaves users without guidance for how to change risky behaviors to support their wellness goals.

Alcohol use disorder (AUD) onset and rates of heavy drinking peak during young adulthood, but young adults are often more concerned about their general wellness than alcohol use behaviors [4-7]. However, risky alcohol use can impede wellness goals, like sleep improvement and cardiovascular recovery [8,9]. Alcohol’s harmful effects across the body are well-established and include effects on cardiovascular health, sleep, and immune function [10]. Alcohol use can contribute to cardiac arrhythmias [8] and poor sleep quality in young adults, and vice versa [9,11,12]. Inadequate sleep may lead to increased, problematic alcohol use in young adults [12-15] and to relapse for people with AUDs [16,17]. Furthermore, behavioral interventions for insomnia may reduce alcohol use in adults who drink heavily [15], especially digital insomnia programs [18] and broader digital sleep interventions [5]. Therefore, wearable fitness technologies may support sleep and other wellness goals by targeting related risky behaviors, like alcohol use.

Personalized feedback from wearable technologies may help young adults make connections between their alcohol use and their wellness goals, like improved sleep [19]. In reviews and meta-analyses of clinical trials [20,21], digital personalized feedback interventions result in small but meaningful reductions in young adults’ drinking. Feedback interventions often involve normalized feedback comparing young adults’ perceptions of peer drinking and actual peer drinking levels, which highlights that young adults’ peers often drink less than they expect [20]. This feedback then facilitates comparison of their own self-reported drinking to actual norms. Personalized feedback for alcohol reduction is also increasingly integrating multiple personal data streams, enabling comparison between drinking and other facets of young adults’ experiences [21]. Combining physiological data (eg, alcohol’s effects on heart rate and sleep) and self-reported behavioral data (eg, number of drinks consumed) can provide highly personalized feedback and insights that can increase user engagement in interventions [3], which is critical to intervention effectiveness [22].

Feedback in digital health interventions tends to be delivered (1) when a behavior is occurring or (2) after a behavior occurs [23], which facilitates reflection-in-action or reflection-on-action, respectively, to promote insight [24]. Retrospective delivery of feedback allows users to think about their behavior in the larger context of related events, feelings, and motivations and to consider relations between their experiences and physiological or behavioral data [25]. Feedback with a record of experiences over time can facilitate “descriptive reflection” (revisiting behaviors) and “explanatory reflection” (explaining behaviors). Furthermore, feedback showing correlations or causal patterns (eg, associations between alcohol use and poor sleep) can encourage “dialogic reflection” [25]. Reflecting on causal patterns between alcohol use and hindered wellness goals may boost motivation to change alcohol consumption among young adults who are unconcerned about their alcohol use [5].

Popular commercial devices are uniquely positioned to promote reflection and insight among young adults who might not otherwise seek help with risky behaviors, like alcohol use [25]. Given the potential for wearable technologies to promote wellness and decrease risky behaviors simultaneously [5], it is essential to study young adults’ experiences with integrated physiological and behavioral feedback on alcohol use. User engagement is critical to the success of digital health tools [22], and intervention designers must understand how to optimize feedback to encourage long-term engagement from young adults.

The current study is the first randomized controlled trial (RCT) of a wearable, personalized feedback intervention for young adults with risky drinking that combines: (1) physiological metrics of sleep, heart rate variability (HRV), and resting heart rate via wearable photoplethysmography biosensors in the Oura Ring (Oura Health Oy) and (2) behavioral daily diary self-monitoring of sleep and alcohol use.

Our primary evaluation aim was to describe young adults’ perceptions of the acceptability, feasibility, and perceived effectiveness of the Oura Ring wearable, the Oura Ring mobile app, smartphone diary self-monitoring, and personalized feedback and tailored advice reports, with a focus on participants’ experiences in the feedback group. In addition, we also had some exploratory aims. First, we aimed to compare user experiences between the feedback group (full access to the Oura Ring mobile app and feedback reports every 2 weeks) and the assessment group. Second, we aimed to compare user experiences of different intervention components and, finally, explore young adults’ health coaching preferences for personalized feedback.

For this pilot, parallel-group RCT, our goal was to evaluate users’ experiences with a wearable, personalized feedback intervention leveraging physiological data (cardiovascular and sleep) and behavioral data for alcohol reduction. Participants were randomly assigned 1:1 to either the feedback (n=30) or assessment (n=30) group. In 2022, all participants wore the commercial wearable Oura Ring, Gen2 (Oura Health Oy) daily for 6 weeks, completed daily smartphone diaries about their sleep, alcohol or substance use, and health behaviors, and completed follow-ups at weeks 6 and 10. Study staff members gave SMS text reminders to participants to sync Oura Ring data 3‐4 times per week and to complete smartphone diaries (if not yet completed).

The feedback group had full access to the Oura Ring mobile app via smartphone. The app included daily personalized, biometric feedback on sleep (ie, sleep stages, wakefulness, timing, efficiency, latency, and duration), physical activity (ie, calories burned and steps), cardiovascular recovery (HRV and resting heart rate), respiratory rate, and body temperature. Furthermore, the app provided composite health scores in the areas of “sleep,” “activity,” and “readiness” based on proprietary algorithms and in-app sleep tips and activity prompts. The assessment group only had partial access to the Oura Ring mobile app, including general wellness tips, but they did not have access to personalized, biometric feedback in the app (eg, daily sleep and cardiovascular feedback). Based on our previous work [4-7], we judged this to be the best control as it provides the experience of wearing the ring, having knowledge of being monitored, and using the app.

The feedback group also received personalized feedback and tailored advice reports every 2 weeks, derived separately and delivered by our research team, on integrated physiological Oura Ring data and behavioral smartphone diary data. Personalized feedback included trends of alcohol and other substance use based on self-report (eg, heavy drinking or substance use occasions, drinks per week, and peak blood alcohol level), alongside cardiovascular recovery and sleep data over each 2-week period (refer to the studies by Fucito et al [4,7] for more details on similar feedback reports in previous research). Reports included data visualization to reveal patterns among sleep, cardiovascular, and alcohol and substance use data streams. Within reports, participants were also given brief, evidence-based advice tailored to their data, such as sleep hygiene, controlled drinking, stress management, and exercise strategies. The assessment group did not receive feedback reports every 2 weeks from the research team, but received 1 delayed feedback report postintervention at week 10. The assessment group participants were unblinded given their knowledge that features of the Oura Ring app were blocked and that they were not receiving feedback reports throughout the intervention period. Furthermore, study team members were unblinded when administering participant appointments and interviews.

Participants were recruited through online advertisements on open-access social media sites (eg, Snapchat [Snap Inc], Instagram [Meta], Facebook [Meta], and Reddit) and offline via community flyers in universities, gyms, and other public spaces. Although advertisements did not explicitly seek out young adults who wanted to reduce their drinking, they did target young adults with heavier drinking levels as central to the study. Interested volunteers were directed to complete an online screener. The study’s affiliation with Yale School of Medicine was apparent in recruitment and screening materials. To be eligible, participants needed to (1) be 18‐25 years old, (2) be fluent in English, (3) report ≥4 heavy drinking occasions (≥5 drinks for men and ≥4 for women) in the past 28 days, (4) be at risk of harmful drinking (Alcohol Use Disorder Identification Test- Consumption [26] ≥7 for men or ≥5 for women), and (5) own a smartphone. Potential participants were excluded if they had (1) current (active) alcohol or sleep treatment; (2) clinically severe AUD withdrawal or substance use disorder other than cannabis in the last 12 months as assessed by diagnostic interview; (3) severe symptoms of a mental health disorder (MHD), for example, current psychosis or suicidality; (4) history of sleep disorders; (5) job with a night or rotating shift that prevented a consistent sleep schedule; or (6) travel >2 time zones during study participation or a month before.

Of the 81 participants who were screened online for eligibility, 21 did not meet the inclusion criteria due to insufficient drinking levels (n=15), severe MHDs (n=2), and planned travel (n=1; Multimedia Appendix 1). Furthermore, 3 participants were no longer interested. Those 60 participants who met online screening eligibility were invited to attend an intake to confirm eligibility face-to-face. All 60 eligible participants who were enrolled and randomized into groups ultimately completed the treatment, and 59 completed follow-up. Given that this was a pilot trial, neither a power analysis nor other sample size calculations were undertaken. We judged that 60 participants, including 30 in the feedback group, would be sufficient to assess intervention feasibility.

To assess user experiences, participants completed self-assessed, web-based exit surveys and face-to-face exit interviews designed for this study (Multimedia Appendix 2). Acceptability was defined as survey ratings of intervention satisfaction and likeability and interview sentiment and descriptions of preference. Feasibility was determined via survey ratings of intervention comfort, schedule workability, life interference, and interview descriptions of understandability. Perceived effectiveness encompassed survey ratings of intervention helpfulness alongside interview descriptions of helpfulness, behavior influence, and behavior change. The timing and some content of assessments varied by group. To assure quality, exit interviews were given face-to-face, and exit surveys were self-assessed by participants under the supervision of a study team member.

All participants completed exit surveys at week 6. Participants were emailed a web link to the exit survey, which they could complete via smartphone or computer. Exit surveys included original Likert scale and Likert-type questions written by the study team regarding the acceptability, feasibility, and perceived effectiveness of the overall program, wearing the Oura Ring, and completing the smartphone diaries. These user experience questions were based on validated measures [27,28] and have been used in a variety of previous online user experience studies [5,29,30]. In the week 6 exit survey, the feedback group participants responded to additional questions about the acceptability, feasibility, and perceived effectiveness of the Oura Ring mobile app and the personalized feedback and tailored advice reports received every 2 weeks. Questions about the perceived effectiveness of the intervention referred to the helpfulness of information and tips for reducing alcohol and improving sleep or cardiovascular health. At week 10, the assessment group participants were not asked questions about the Oura Ring mobile app due to their limited access, but they were asked questions about the delayed feedback report that they received. After trial initiation, additional survey questions were completed by a smaller subset of participants, including their willingness to purchase the Oura Ring (n=51) and feedback reports (n=53), the likelihood of recommending the Oura Ring to others (n=51), the likability and understandability of alcohol or substance information in feedback reports (n=32), and the helpfulness of behavioral tips (such as sleep and alcohol tips) in feedback reports (n=28‐31).

Both groups also completed exit interviews with a study team member (MA). The feedback group participants completed exit interviews during week 6 follow-ups. Interview questions focused on changes in overall health, sleep, and alcohol or substance use during the study, comparisons to peers, helpfulness of received intervention components (Oura Ring, smartphone diaries, and personalized feedback reports), interests and preferences for health coaching, and suggestions for future research. The feedback group participants also answered questions about the Oura Ring mobile app and provided comparisons among components. Thematic saturation occurred before interviewing all feedback group participants, but all participants were asked to take part in interviews to ensure all user experiences were captured. Exit interviews were not initially planned for the assessment group but were added after study initiation to their week 10 follow-up to gain a richer understanding of their experiences and reactions to the delayed feedback report. However, this protocol addition resulted in the first assessment group participants (n=4) not being offered interviews. Assessment group participants were not asked questions about the Oura Ring mobile app because they did not have full access, and they were not asked to compare intervention components. Consistent with iterative qualitative research methods [31], some interview questions for both groups (eg, exploratory health coaching questions) were adaptively added during the evaluation process in response to users’ experiences.

We used an innovative convergent mixed methods approach [5] incorporating artificial intelligence (AI)–driven natural language processing (NLP) to evaluate this wearable, personalized feedback intervention for young adults with risky drinking. Exit surveys and exit interviews were analyzed in parallel to assess the convergence of findings. For our primary aim, we examined the descriptive results of exit surveys to evaluate the acceptability, feasibility, and perceived effectiveness of the overall program and its intervention components. Then, for our exploratory aims, we assessed (1) predictive results of exit surveys, specifically whether the acceptability, feasibility, or perceived effectiveness of the program varied based on study group, and (2) descriptive results of exit surveys and interviews comparing intervention components and health coach preferences.

Concurrently with exit survey analyses, we also analyzed the content of exit interviews using AI-driven NLP and researcher-coded qualitative analysis. We used NLP first to characterize exit interviews, which included (1) topic modeling analysis with Latent Dirichlet Allocation (LDA; [32]) and (2) sentiment analysis with the Finn Årup Nielsen (AFINN) lexicon [33]. With a given number of topics, LDA determines the most likely topics within each document and the most likely terms within each topic [32]. The number of topics (k) used in our LDA was determined through a preliminary analysis using 3 methods [34]. Overall topic prevalence, or the topic’s proportion of a given document on average, is characterized by Ɵ_k. Following the LDA, 2 study team members (FJG and OKE) named each topic based on recursive reading of interviews most likely to include each topic. These coders compared the names to ensure trustworthiness. For the sentiment analysis, we used the AFINN lexicon, which assigns values ranging from −3 to 3 based on their negative to positive valence, with 0 indicating neutrality in words. Documents are each given an index score based on the net value of included terms from the AFINN lexicon [33].

Based on the scope of the NLP results, study team members then conducted a rapid qualitative analysis on specific exit interview questions to target remaining areas of research interest. In total, 7 study team members (FJG, OKE, SK, MF, LL, and Holly Boyle and Sophia Sniffin) participated in the deductive qualitative coding process informed by the rigorous and accelerated data reduction (RADaR) technique for rapid qualitative analysis [35]. This technique involves a series of spreadsheets and data tables in which qualitative passages are successively reduced to derive themes [35]. A randomly selected subset (10/50, 20%) of interviews was independently coded by multiple coders to assess interrater reliability. Prereconciliation meeting Cohen kappa values between coding pairs ranged from 0.72 to 0.82, and postreconciliation meeting kappa values ranged from 0.90 to 0.97. The coding framework was revised based on reconciliation discussions, and remaining interviews were divided among coders who engaged in ongoing consultation and discussion to reduce coder drift and maintain trustworthiness. The researcher-coded qualitative findings from exit interviews were compared with quantitative NLP results from exit interviews using joint display methods, specifically an integrated results matrix [36]. For interview thematic results, we distinguish between primary aim results, which focus on intervention acceptability, feasibility, and perceived effectiveness, and exploratory aim results, which concentrated on comparisons between trial groups or intervention components and young adults’ health coaching preferences.

This research was approved by the Yale University institutional review board (2000030417). All eligible participants discussed an informed consent form in detail with a study team member before agreeing to take part in the study, and informed consent was obtained from every participant. During the informed consent process, the intervention’s focus on alcohol use and related physiological metrics was made explicit. Participants’ identifying information was kept private and confidential and stored only on a secure university server. All data used for the user experience evaluation were deidentified before analysis. Participants could earn up to US $279 if they completed all study activities (ie, smartphone diaries, study visits, and wearing and returning the Oura Ring).

A total of 60 participants took part in the RCT and completed the exit survey (Figure 1). Almost all feedback group participants (n=29) and most assessment group participants (n=21) completed the exit interview. Half of the participants (30/60, 50%) were men, and 81.6% (49/60) were White, with a mean age of 22.02 (SD 1.98) years (Table 1). The majority (40/60, 66.6%) were students, and most (41/60, 68.3%) were employed. Over three-fourths (47/60, 78.3%) met criteria for an AUD, 21.7% (13/60) for another substance use disorder, and 15% (9/60) for an MHD. No demographic variable differed significantly between the assessment and feedback group participants.

On a 100-point scale, participants in both groups reported high acceptability (mean 84.17, SD 17.81) and perceived effectiveness (eg, promoting hope [mean 70.42, SD 25] and meeting program goals [mean 75.83, SD 21.57]) of the overall program (Table 2). Almost all participants (58/60, 96.7%) said that they would recommend the program to a family member. Although participants in both groups found the overall program to be highly feasible (eg, schedule duration [mean 77.92, SD 25.25] and schedule workability [mean 91.67, SD 15.03]), assessment group participants reported some higher aspects of overall program feasibility compared with feedback group participants, including schedule workability (mean difference=8.30, P=.03) and visit comfortability (mean difference=10.00, P=.007) of visits.

Related to their experiences of wearing the Oura Ring, participants in both groups reported high acceptability (eg, likeability [mean 76.25, SD 22.75] and willingness to continue wearing [mean 90.83, SD 17.20]) and feasibility (eg, ring comfortability [mean 68.33, SD 29.06] and no itchiness [mean 93.52, SD 13.77]). Most participants (44/51, 86.3%) said they would recommend the Oura Ring to a family member, and only 33.3% (20/60) noted marks on their skin from wearing it. Feedback group participants with full access to the Oura Ring mobile app also reported high acceptability (27/30, 90% liked the app) and moderate effectiveness (eg, activity prompt helpfulness [mean 52.88, SD 30.27] and readiness tip helpfulness [mean 67.00, SD 29.51]) of in-app recommendations and prompts.

Regarding the smartphone diaries, participants in both groups also reported moderate acceptability (mean 59.17, SD 25.20) and moderately high feasibility (eg, willingness to continue diaries in an app [mean 64.17, SD 28.51] and easiness [mean 82.50, SD 20.74]). Although diaries were highly rated in general, a descriptive comparison of ratings indicates that smartphone diaries may have been less acceptable and feasible than other intervention components (Table 2). Furthermore, members of both groups rated information in the feedback report as highly acceptable (eg, layout acceptability [mean 75.00, SD 24.33] and sleep or cardio information likeability [mean 92.67, SD 13.25]), feasible (eg, overall understandability [mean 70.26, SD 20.12] and alcohol or substance use information understandability [mean 84.38, SD 18.78]), and effective (eg, substance use tip helpfulness [mean 69.64, SD 24.87] and sleep or cardio information helpfulness [mean 88.36, SD 17.66]). A descriptive comparison of perceived effectiveness ratings (Table 2) indicates that the information in feedback reports may have been more effective than that in the Oura Ring app rated by feedback participants.

Feedback group participants had high self-reported adherence to intervention components, including using the app every day on average (mean 83.33, SD 21.71) on a 0‐100 scale from weekly to multiple daily use. The most frequently used aspects of the Oura Ring mobile app were sleep data (28/30, 93.3%), activity data (21/30, 70%), and cardiovascular recovery data (15/30, 50%). Less frequent in-app activities included tagging workouts (12/30, 40%), clicking on personal trend data (7/30, 23.3%), and interacting with story or meditation content (1/30, 3.3%). Most feedback group participants (24/30, 80%) also reported having read all 3 personalized feedback reports, and all feedback group participants (30/30, 100%) read at least 1 report. There was no significant difference between the baseline drinking levels of feedback group participants who read all 3 reports and those who did not (P=.35). Among all participants, they self-reported that they most frequently used sleep tips from the feedback reports (43/60, 71.7%), followed by alcohol or substance use tips (25/60, 41.7%), physical activity tips (19/60, 31.7%), stress management tips (18/60, 30%), and diet tips (14/60, 23.3%).

Mixed methods evaluation results converged about users’ perceptions of the wearable physiological and behavioral feedback intervention (Table 3). Participants described the overall program as having high acceptability, feasibility, and perceived effectiveness in exit surveys and interviews. Wearing the Oura Ring was described as highly acceptable and feasible in the survey and as moderately to highly effective across both the survey and interview. These ratings included the Oura Ring mobile app for feedback group participants. Smartphone daily diaries tracking behavioral data were described as moderately to highly acceptable and feasible in the survey and as highly effective in the interview; therefore, the evaluation of this component had less cross-method convergence. The feedback reports were described as highly or moderately highly feasible and effective in both the exit survey and interviews and highly acceptable in the survey. Both methods used to analyze interviews (NLP and qualitative analysis) also showed that participants reported learning insights about their sleep deficits from the feedback reports.

Comparison of the feedback and assessment groups’ experiences revealed different findings depending on the method. Although the groups did not significantly differ on most exit survey ratings of the program, assessment group participants rated some aspects of program feasibility (comfortability and workability) more highly than feedback group participants. Also, per NLP with exit interviews, feedback group participants may have had higher perceived program effectiveness (reported behavior change).

Our exploratory analysis comparing program components revealed preferences for different aspects of program components. For example, some feedback group participants described the Oura Ring as most effective (influential on behavior) when asked in the exit interview. However, on their exit surveys, feedback group participants tended to rate the personalized information in the feedback reports as more effective (helpful) than the tips and prompts in the Oura Ring app. Also, although participants described the smartphone diaries as similarly effective (helpful) as other components (Oura Ring and feedback reports), they rated the acceptability and feasibility of the smartphone diaries as lower. These findings are consistent with previous research on lower levels of engagement in self-report data [37]. Actively completing the smartphone diaries (behavioral data) may have been more challenging than passively wearing the Oura Ring (physiological data). However, participants found the integration of both physiological and behavioral data streams in personalized feedback reports to be especially helpful.

Mixed methods analysis of exit interviews also revealed differences in participants’ preferences for health coaching about their personalized feedback. NLP and qualitative analysis of exit interviews indicated that assessment participants preferred to read their feedback reports independently without a health coach. The qualitative analysis revealed additional preferences, including feedback group participants’ interest in clinician health coaching and assessment group participants’ preference for peer coaching. The only discrepancy between evaluation methods was the topic highlighted within the NLP, which noted that feedback group participants are also interested in peer health coaching.

This evaluation paralleled previous research on young adults’ greater concern about improved wellness, such as improved sleep [6], rather than alcohol use. Study participants were motivated to join the study due to curiosity and interest in their wellness and personalized feedback, as opposed to a desire to reduce their drinking. Curiosity about highly personalized feedback also played a role in maintaining engagement in the intervention after its initiation. This aligns with previous findings that personalized feedback can enhance engagement [3]. Consistent with previous research, interventions focused on wellness goals may be more accessible and appealing to young adults than those primarily targeting alcohol use [5]. Despite their explicit focus on fitness and wellness, wearable devices have the potential to contribute to reducing risky behaviors.

Commercial wearable devices, like the Oura Ring, could incorporate more active monitoring of self-reported behaviors, such as alcohol use, to provide highly personalized feedback and foster motivational change in young adults. A key focus of our study was integrating behavioral self-monitoring and feedback, as this is not available in Oura. Although Oura users can make implicit associations by examining and tagging their physiological data, there is no explicit integrated feedback. Whereas it is especially challenging in general to encourage behavioral health app users to maintain engagement [38], our findings of high acceptability and feasibility support the integration of behavioral self-report data with passively collected physiological data. In particular, the combination of alcohol-related behavioral data and physiological data related to sleep and cardiovascular recovery could highlight connections between these data streams [23,25]. Our findings indicated that participants found the experience of active self-monitoring through smartphone diaries to be acceptable, feasible, and perceived as effective. Furthermore, they reported gaining insights from the integration of these data with their passively collected physiological data from the Oura Ring. Some noted they appreciated learning through integrated information and receiving tailored coaching. Insights into contextual factors that influence physiological data, such as sleep deficits, may promote dialogic reflection and enhance motivation to change risky behaviors [19,25].

Personalized feedback can be optimized to better promote insight and enhance change readiness. In this study, feedback group participants received daily health data and recent trends on the Oura Ring mobile app, along with more retrospective, integrated feedback in written reports every 2 weeks. Participants found both the feedback reports and the Oura Ring mobile app effective. On one hand, they reported preferences for the personalized insights in the integrated feedback reports; whereas, on the other hand, they liked the easy functionality of the Oura Ring and app.

Young adults may be interested in an option that combines the benefits of these intervention components (feedback reports and the Oura ring or app) via highly personalized, integrated in-app feedback. Mobile apps for wearable devices could offer active behavioral monitoring that is flexible according to the amount of time young adults are willing and able to answer self-report questions. Then, apps could offer integrated data reports at different time scales (eg, daily reports and longer trends) to leverage reflection-in-action and reflection-on-action [25]. Enhanced feedback options could also include opportunities for health coaching via educated peers or clinicians. Participants in both study groups showed some interest in peer coaching, and feedback group participants were more interested in clinician health coaching. Depending on the complexity of some data relationships (eg, HRV after a heavy drinking episode), it may be important to consult a coach to interpret and gain insights from personalized feedback reports.

Our results should be considered in the broader tradition of personalized feedback interventions for alcohol reduction. Normative feedback interventions compare young adults’ own drinking and perceptions of peer levels with actual peer levels, and these interventions may have small but meaningful impacts on alcohol reduction [21]. These interventions are theorized to address young adults’ social pressure to drink as a mechanism of change; however, highly personalized feedback on physiological and alcohol data may address overall wellness motivations to change. Given that young adults are generally unconcerned about their drinking [6], our findings reveal that the integration of personalized feedback on physiological metrics may increase the appeal of personalized alcohol feedback. Accessible, personalized feedback that promotes reflection [25] and engagement [3] may enhance young adults’ awareness of the connections between their behaviors and aspects of their wellness, like sleep and cardiovascular health. Further, integrated physiological and behavioral feedback has implications for other issues impacted by lifestyle behavior change, such as cardiovascular disease prevention.

Although our mixed methods user evaluation approach leveraged AI-driven approaches to enable breadth and depth [5], there were limitations in our methodology. Our sample size was relatively small for an RCT, especially for quantitative evaluation analyses. Furthermore, our sample consisted mostly of students from a single geographic location, which may not be representative of other young adults. Additionally, some survey and interview questions were introduced iteratively, limiting them to a subset of participants (eg, health coaching and component comparisons). The prevalence of these themes may have differed if they had been presented to the entire sample from the outset. Finally, as a phase 1 study primarily focused on feasibility, the duration of the intervention was only 6 weeks; however, a longer duration (>8 wk) would have been ideal to fully test the effect of an intervention intended to promote behavior change.

Our results support the inclusion of self-report behavioral data in commercial wearable devices. Participants found the intervention acceptable, feasible, and effective, including the completion of smartphone diary self-monitoring. Many found that personalized feedback reports integrating their physiological and behavioral data were helpful and promoted insights about their sleep and other wellness goals. Wearable devices may lack important functionality by not capturing the behaviors that contribute to wellness goals, such as improved sleep, cardiovascular recovery, or fitness. Additionally, targeting risky and prevalent behaviors, such as alcohol use, through wearable devices could be a more appealing intervention for young adults who are less concerned about heavy drinking than about improving overall wellness.