The Smartphone Health Innovation Program was a prospective, observational pilot study that evaluated the feasibility and acceptability of repeated smartphone-based EMA among people who use opioids. The study was conducted in New Haven, Connecticut, at the New Haven Syringe Services Program (NHSSP). The NHSSP is a community harm-reduction program that provides sterile syringe exchange, hygiene supplies, overdose-prevention education, naloxone distribution, and referrals to medical and substance-use treatment services for people who use opioids. Participants were followed for 30 days and completed an in-person baseline visit, a 30-day EMA monitoring period, a structured support check-in approximately 7 days after enrollment, and an end-of-study visit with a qualitative interview.

Participants were recruited using convenience sampling through flyers posted at the NHSSP and surrounding community locations, word-of-mouth referrals, and referrals from local service organizations in the New Haven area (including treatment and harm-reduction programs, such as the APT Foundation, Inc). Potential participants either approached research staff while visiting the NHSSP for services (eg, syringe exchange visits) or contacted the study team by phone. Eligibility screening was conducted in person at the NHSSP or by telephone. Participants were eligible if they met the following criteria: (1) were aged 18 years or older, (2) reported use of at least one illicit opioid (eg, heroin and fentanyl) within the past 30 days, (3) owned a personal smartphone compatible with the study app (Android or iOS), and (4) resided in the Greater New Haven area.

Eligible participants attended an in-person enrollment visit at the NHSSP. Trained research assistants confirmed eligibility and obtained written informed consent in a private office. Participants then completed baseline questionnaires and onboarding procedures during the same visit. The enrollment and onboarding visit lasted approximately 30 minutes. Baseline questionnaires were administered electronically on a tablet and were primarily self-administered. Research staff read questions aloud and assisted participants when literacy or comprehension barriers were present. Participants were enrolled between November 2024 and February 2025. They completed 30 days of EMA monitoring following enrollment, and end-of-study assessments were conducted between January 2025 and March 2025.

Quantitative data were analyzed using Python version 3.13.4 (Python Software Foundation). Participant characteristics and feasibility outcomes were summarized using descriptive statistics. Continuous variables were summarized using means and SDs, and categorical variables were summarized using counts and percentages. EMA adherence was calculated at the participant level, separately for morning and evening prompts. To assess changes in engagement over time, EMA prompt completion was aggregated into 4 weekly intervals and summarized across participants. EMA behavioral responses (drug use, craving, planned use, overdose events, and naloxone access) were summarized as counts of endorsed events across completed observations. Data management and statistical summaries were conducted using the pandas and NumPy libraries, and figures were generated using Matplotlib.

Qualitative interviews were audio-recorded and transcribed verbatim. Data were analyzed using inductive thematic analysis, following the steps described by Braun and Clarke [35]. Interview transcripts were reviewed multiple times to become familiar with the data. Two coders independently conducted open coding of all transcripts and met regularly to review and refine the coding framework. Any discrepancies were resolved through discussion with the senior coder (KG), and codes were gradually grouped into broader thematic categories. Consistent with recommendations for enhancing rigor in qualitative research, coding decisions and thematic interpretations were reviewed collaboratively throughout the analytic process [36]. Qualitative data were managed and analyzed using Dedoose version 9.0.54 (SocioCultural Research Consultants, LLC).

The study was approved by the institutional review boards (IRB protocol #2000036555) at Yale University and the University of Connecticut. All participants provided written informed consent prior to participation. Participants received US $25 for completing the baseline visit and US $25 for completing the end-of-study visit and interview. EMA completion was incentivized on a per-prompt basis. Participants received US $1 for each completed EMA prompt (maximum of 60 prompts). The maximum total compensation per participant was US $110 over the 30-day study period.

Overall, the acceptability and usability of the EMA protocol were high (Figure 2). All participants reported that the EMA prompts were easy to understand, felt private and confidential, and that the survey frequency was appropriate. All participants also indicated willingness to participate in future EMA studies and found occasional reminders from research staff helpful. Most participants (6/7, 86%) reported receiving push notifications on time and indicated that daily EMA prompts did not feel burdensome. EMA timing was considered convenient by 71% (5/7) of participants, and 71% (5/7) also reported that the survey questions were not distressing.

Participants also reported a few barriers to EMA completion. EMA timing conflicts with daily tasks and phone-related issues (eg, phone not working or lost) were each reported by 43% (3/7) of participants. Missed EMA notifications were reported by 29% (2/7) of participants, and lack of phone availability at the time of the EMA was reported by 14% (1/7) of participants. None of the participants reported low motivation, concerns about privacy, or prompt length as barriers to completing the EMA.

Seven participants completed the one-on-one semistructured exit interview. Thematic analysis yielded five themes that reflected the participants’ experiences with the EMA study: (1) ease of use and integration into daily routines, (2) self-awareness and reflection related to substance use, (3) perceived accountability through self-monitoring, (4) emotional discomfort related to substance use reflection, and (5) situational and technical barriers to EMA prompt completion.

Five participants described the EMA prompts as easy to complete and manageable within their daily routines. Participants mentioned that the prompts were brief and delivered at predictable times, which supported routine completion without being perceived as intrusive:

Four participants reported that repeated reporting increased their awareness of their substance use patterns and daily behaviors. Some participants noted that completing the EMA prompts encouraged them to reflect on their cravings, intentions to use substances, and substance use experiences throughout the day:

Two participants described how regular EMA participation created a sense of accountability, structure, or engagement in their daily routines, even without any intervention or direct interaction with the study staff:

Two participants described experiencing emotional discomfort when reflecting on their substance use through repeated EMA reporting. These experiences included feelings of guilt, sadness, or emotional distress associated with reporting substance use behaviors:

Two participants described situational or technical barriers that occasionally interfered with EMA prompt completion, including oversleeping, low phone battery, and missed alerts:

In this study, we explored the feasibility and acceptability of a 30-day smartphone-based EMA protocol among people who use opioids. Recruitment through the SSP was successful and represented an important strength of this study. Overall, 70% of the participants completed the study and follow-up interviews. EMA completion rates were high, with an overall compliance rate of 85%, and similar response rates for morning and evening prompts. Participants regularly reported opioid use, cravings, overdose experiences, and carrying naloxone through EMA, suggesting that twice-daily monitoring of opioid-related behaviors is feasible in community settings.

The EMA compliance rate observed in this study was higher than the pooled compliance rate reported in a meta-analysis of 126 effect sizes from 19,431 participants among EMA studies of people who use substances (75.1%, 95% CI 72.4%-77.7%) [23]. Prior EMA studies among people who use opioids have largely focused on momentary predictors of drug use, such as craving, mood, and stress, rather than on overdose-related experiences or harm-reduction behaviors. Compliance remained generally stable across morning and evening prompts and throughout the 30-day study period, with a modest decline in the third week—a pattern reported in other EMA studies and often attributed to participant fatigue [23-25]. In our study, compliance increased again during the final week, suggesting that participants were able to integrate the EMAs into their daily routines. Predictable EMA prompt timing, brief questionnaires, and regular reminders from the research staff may have contributed to these high response rates. Prior EMA studies often included larger samples and longer study durations, which may partly explain the differences in compliance [17,18,37]. In this context, the high compliance observed in our study may have been influenced by the small sample size, short study duration, and participants’ relatively high levels of education and digital literacy [23,38,39].

Importantly, participants reported their drug use behavior and experiences related to overdose and carrying naloxone through daily EMA prompts. Although only a small number of overdose events were reported during the study period, participants reported carrying naloxone on every occasion when an overdose was reported. However, the study did not assess whether naloxone was administered during these overdose events or whether it was used for reversal. These findings suggest that EMA may be a useful approach for capturing timely information on overdose experiences and harm-reduction responses. Prior EMA-based research among people who use opioids has focused on momentary predictors of drug use, such as craving, mood, pain, and contextual factors, but did not assess overdose experiences or whether participants were carrying naloxone [17,18,40-42]. While preliminary, our findings suggest that EMA may also be useful for capturing overdose-related experiences and harm-reduction behaviors in daily life. Such information may help improve the understanding of the circumstances surrounding overdose events and may inform future real-time prevention and intervention strategies.

Despite disclosing minimal barriers (EMA prompt timing, unavailability of phone, and missed prompt notifications), participants reported that overall participation was acceptable, minimally burdensome, and did not interfere with their daily activities. In similar EMA studies [43,44], privacy and confidentiality were reported as major barriers; however, none of the participants in our study raised these concerns. Participants further reported an overall positive experience, with all indicating that they would participate in a similar study in the future. In real-world settings, it would be helpful to consider the barriers they faced, such as prompt timing, missed push notifications, and the emotional discomfort associated with some questions. Fixed notifications, individualized prompt timing, and participant-friendly question wording may help reduce participant burden and improve engagement when EMA protocols are developed using a user-centered approach. In addition, providing participants with personalized feedback based on EMA data may be more impactful [45], which could motivate them to remain engaged for a longer period. Such EMA data may also support the development of mobile health–based just-in-time adaptive interventions, where timely information on drug craving, planned use, or contextual risk factors could be used to predict periods of elevated overdose risk and deliver timely harm-reduction support, such as reminders to carry and use naloxone [46,47]. However, further research is needed to determine how such approaches can be effectively implemented in real-world settings.

Several limitations should be considered when interpreting the findings of this study. First, the sample size was very small, and participants were recruited through a single SSP in New Haven, which limits the stability of the findings and restricts generalizability to the wider opioid-using population. Individuals engaged with harm reduction services may differ from those who are not connected to such programs, including having greater access to harm-reduction resources, higher overdose awareness, and more frequent engagement with health services, potentially limiting the generalizability of the findings to broader populations of people who use opioids. Second, the study sample included individuals with relatively high digital health literacy and smartphone access, which may not reflect the experiences of people who use opioids with lower or more unstable access to digital technologies due to social and structural vulnerabilities, such as housing instability and poverty. Third, EMA items, along with the poststudy acceptability questionnaire, were adapted and were not formally validated, which may limit comparability with other studies. Fourth, all EMA responses and poststudy assessment measures were self-reported and may have been influenced by social desirability or participant fatigue, potentially leading to overreporting of positive experiences and underreporting of barriers. Fifth, although participants reported carrying naloxone, the study did not assess whether naloxone was administered during overdose events. Sixth, the 30-day EMA protocol does not capture long-term adherence or engagement beyond the short monitoring period. Finally, participants received compensation for completing EMA prompts, which may have influenced response rates and engagement observed in this pilot study and may not reflect participation in real-world settings, where similar incentives may not be available.

Our findings suggest that a smartphone-based EMA protocol is feasible and acceptable among people who use opioids. Participants demonstrated high compliance with twice-daily EMA prompts and reported opioid use, overdose experiences, and naloxone possession. These findings highlight the potential of EMA to capture day-to-day patterns of substance use and overdose-related behavior. Future research should examine how EMA data can be leveraged to inform adaptive interventions, such as just-in-time adaptive interventions, to reduce the risk of overdose. Larger and longer-term studies are needed to further evaluate EMA for monitoring opioid use, supporting harm-reduction strategies, and understanding broader aspects of daily functioning and well-being among people who use opioids.