Posttraumatic stress is endured by an estimated 28% of people with HIV, globally [1]. Nearly two-thirds (64%) report symptoms consistent with diagnosable posttraumatic stress disorder (PTSD) [2]. The trauma burden can reach as high as 97.1%, with an average of 4.2 out of 10 recognized adverse childhood experiences (ACEs) on the ACEs Questionnaire, among women with HIV in clinical samples. Lifetime trauma has been linked, in a dose-response manner, to greater likelihoods of self-reported depressive and anxiety symptoms, as well as alcohol and illicit drug use [3]. Indeed, the origins of traumatic stress among many people with HIV in the United States are often attributable to multiple sources, including compounded early life adversities, economic hardship, and, in particular, firsthand violence exposures [4-6]. Brown et al [5], in South Carolina, reported 3 overarching components of traumatic life experiences among people with HIV: firsthand extreme violence and secondhand homicide and suicide exposures, physical and sexual assault, and accidents and natural disasters. Among people with HIV, these experiences can encompass childhood sexual abuse [4], racial trauma, incarceration [5], homelessness, rape, intimate partner and intrafamilial violence, and comorbid life-threatening illness [6].

Brezig et al [7], pointing to the disproportionate prevalence of violent (vs nonviolent) trauma among people with HIV, developed a syndemic theory of trauma alongside the acquisition, forward transmission, and later-life outcomes associated with HIV survivorship. Trauma, per the American Psychological Association, can manifest in heightened fear, vulnerability, rumination, disorientation, and dissociative tendencies [2]. Among some people with HIV, these symptoms are linked to anger, isolation, distrust [6,7], comorbid anxiety, depression, suicidal ideation [8,9], and hazardous drug and alcohol use [2,7].

Through a behavioral medicine lens, posttraumatic stress represents a crucial, robustly demonstrated, within-person determinant of antiretroviral therapy (ART) adherence [2-4]. Among women with HIV, acute trauma is linked to 4 times the odds of lapsed ART [7]. Within the Brown et al [5] sample, people with HIV who experienced past extreme violence–related or death-related trauma were 63% less likely (adjusted odds ratio 0.37, 95% CI 0.15-0.95) to demonstrate adherence than non–trauma-exposed participants; people with HIV exposed to any trauma were 58% less likely (adjusted odds ratio 0.42, 95% CI 0.21-0.86). Among Black gay and bisexual men in Baltimore, Maryland, and Jackson, Mississippi, 24% reported past-year physical, sexual, or psychological violence exposure, and recent and lifetime violence exposure was linked to ART nonadherence (adjusted prevalence risk ratio 1.59, 95% CI 1.25-2.01) [10]. Qualitatively similar effects, entangling lifetime trauma, mental health challenges, and substance use or misuse as drivers of poor ART adherence are consistently demonstrated across independent samples of people with HIV and US geographies [11-13], including New Orleans [14].

Narrowly individualistic perspectives on trauma among people with HIV may overlook important neighborhood-based sources of stress [15,16] and stressors that can change across time and space even within a single day [17,18]. In New Orleans, neighborhood-level alcohol outlet density has been linked to alcohol misuse among men with HIV and an increased likelihood of endorsing depressive symptoms among women with HIV [19]. GPS and daily diary data (from a very small feasibility sample; n=11) demonstrate that people with HIV spend an average of just 19% of their awake time within their own residential Census tracts, exposed to neighborhood crime rates ranging from 1 to 28 per 1000 population and alcohol outlet density rates of 0 to 22 per 1000 as they traversed, on average, 23 Census tracts [20]. Emerging evidence, typically captured via geographic ecological momentary assessment, shows that these dynamic place-based factors can affect mood and rumination among people with HIV in New Orleans [9,21].

Effective coping can interrupt traumatic stress and ART adherence lapses [22-25]. Specifically, adaptive coping, grounded in the coping framework of Lazarus and Folkman [26], distinguishes between “changeable” and “unchangeable” stressors. The former are approached via problem-focused strategies, such as structured problem-solving and effective communication skills, along with additional cognitive behavioral therapy–oriented strategies, while the latter are addressed via emotion-focused strategies, including cognitive restructuring, mindfulness meditation, and confronting irrational beliefs such as “all-or-nothing” thinking [23]. The Living in the Face of Trauma (LIFT) intervention by Sikkema et al [25], a Centers for Disease Control and Prevention “best evidence” intervention for improving coping with early life trauma stressors and HIV, as well as reducing trauma symptom burden, exemplifies this approach. Across a series of pilot studies [25] and fully powered randomized controlled trials [27], as well as an adaptation for women with HIV in Cape Town, South Africa [28], LIFT has demonstrated efficacy in reducing intrusive symptom burden, forward transmission risk behaviors, and avoidance and arousal PTSD symptoms, while improving ART adherence.

A 15-session, linear, in-person group intervention, LIFT, by definition, does not respond to momentary ecological stressors that people with HIV may encounter throughout the day. Just-in-time adaptive interventions (JITAIs), delivered via smartphone, and often reliant on ecological momentary assessment (EMA) inputs [29,30], can prompt users with brief psychoeducational and coping skills refreshers whenever necessary [24,31]. Similarly, geofencing interventions deliver intervention content triggered by spatial inputs, wherever necessary [32]. JITAIs have demonstrated preliminary efficacy in reducing ruminative episodes via cognitive behavioral therapy skills prompts and alcohol misuse among homeless adults in pilot studies [33,34], as well as robust efficacy profiles among adults transitioning from inpatient alcohol use disorder treatment [35], alongside a wide array of mental health endpoints [36,37]. Reported acceptability, feasibility, and engagement, including among sexually minoritized men living with HIV [38], are generally high [34,39-41], with many opportunities for incorporating prospective end user insights [42].

To our knowledge, no prior studies have examined the acceptability, feasibility, and broader user experience of a geographic ecological momentary assessment (GEMA)–leveraging JITAI that incorporates trauma-informed principles and personalizable geofencing to meet the needs of people with HIV. We report the findings of the pilot of NOLA (New Orleans, Louisiana) Gem, a JITAI adaptation of LIFT, trialed among a sample of people with HIV enrolled across Greater New Orleans. The NOLA Gem app was adapted from the Living in the Face of Trauma: An Intervention for Coping with HIV and Trauma manual, in consultation with Sikkema [27,28]. The NOLA Gem study covered Aims I to III. The findings of Aim I are reported in detail in Gasik et al [21]. The results of Aim II, the NOLA Gem design synthesizing the insights of Aims I and II (including a detailed user flow diagram and user story maps), and the pilot trial protocol are described in Kabir et al. Here, we report the results of Aim III.

The NOLA Gem app was designed with a control and a treatment version, which were masked to the user. The assignment was dependent on how the user was created in the admin panel by research staff. Both versions had GPS tracking; latitudinal and longitudinal data points were recorded at 1-minute intervals if the device moved more than 20 m for the duration of the 21-day pilot trial [43]. GEMA alone, combining GPS tracking and daily diary delivery via text message at 9:00 AM (due by 12:00 PM) and 6:30 PM (due by 9:30 PM) intervals, but lacking any interventional component, constituted the control condition.

The treatment arm of the NOLA Gem app leveraged the GEMA inputs as tailoring variables, calling on fixed decision rules determined in Aim II, to recommend from among 7 “just-in-time” adaptive coping skills (“Set a SMART Goal,” “Common Irrational Thoughts,” etc) mapped to predetermined proximal outcomes [39,44]. Each skill was adapted from a LIFT worksheet, culturally tailored to invoke drivers of resilience identified by New Orleanian people with HIV in Aim II. Endorsing more than “a little stress” on daily diary 37f (“Since your last diary, how much stress did you feel today from any of the following? Had an argument with friend, family member, or romantic partner”) elicited deep-linked recommendations for the “Adaptive Coping” and “Common Irrational Thoughts” in-app brief skills modules, to give just one example. In addition, NOLA Gem users were provided 7 self-directed sessions (estimated 4‐13 minutes in duration), grounded in the traditional, linear delivery of LIFT. These sessions offered psychoeducation on maladaptive or adaptive coping strategies, goal setting, and mindfulness meditation, among other topics. As no generalizable across-case geospatial stressors were evident in Aim I [21], the NOLA Gem treatment arm included a user-determined, essentially personalizable, geofencing feature. Users were asked at baseline to identify 3 locations that they found to be particularly stressful in their lives. Latitude and longitudinal coordinates for these locations were entered into the admin panel with a 150-m geofence; when users entered this geofence, they received a text message asking if they would like to practice their skills, with a link to the skills section of the app. The NOLA Gem JITAI features, built atop and including all aspects of GEMA, constituted the treatment condition for the duration of the 21-day trial period.

Illustrative screenshots of NOLA Gem are shown in Figure 1.

In both arms, the study app was Health Insurance Portability and Accountability Act–compliant and transmitted all data to a Health Insurance Portability and Accountability Act–compliant, Crowdbotics-hosted administrator dashboard.

Participants for Aims I and II were convenience-sampled from the longitudinal New Orleans Alcohol Use in HIV cohort, which, since 2017, has followed 365 people with HIV and hazardous alcohol use histories in outpatient care across Greater New Orleans [45]. For Aim III, this study enrolled participants from this cohort while diversifying recruitment to include a range (n=12) of local HIV outpatient clinics and community organizations serving people with HIV, which were provided with study recruitment flyers.

Participants were eligible if they (1) were 18 years or older; (2) maintained a current ART prescription, verified in person or via a photograph uploaded to a secure Qualtrics link; (3) owned a smartphone; and (4) indicated experiencing at least 1 incident of violence on the screener.

Study enrollment spanned from October 2023 to August 2024. Initially, eligible, consented participants were randomly assigned 1:1 to treatment (NOLA Gem + GEMA) or control (GEMA) arms. Once total enrollment reached 16, the remaining participants were assigned to the treatment arm. This step was taken to address compounding enrollment challenges posed by the ongoing COVID-19 pandemic, specifically the Delta and Omicron variants (during the former wave, Louisiana recorded the highest number of new cases relative to the total population in the United States [46]), and Hurricane Ida, which caused widespread evacuations and a prolonged power outage across Orleans Parish [47]. By favoring treatment-arm enrollment, we aimed to achieve a sufficiently sized pilot sample to gather preliminary mixed methods evidence of NOLA Gem’s acceptability, feasibility, and usability, despite these unforeseen obstacles, while maximizing the potential human benefits of research involvement for people with HIV in New Orleans.

The study arm was randomly assigned following eligibility verification, prior to the baseline appointment, to streamline onboarding. At the baseline appointment, participants were read an informed consent form aloud, and verbal consent was recorded in Research Electronic Data Capture (REDCap). The NOLA Gem app was then downloaded from the App Store (iOS) or Google Play (Android) onto participants’ phones using the institutional Wi-Fi. The research assistant enabled notifications and location tracking on participants’ devices, as well as location services for the NOLA Gem app. Onboarding to both control and treatment conditions was supported by illustrated guides. The baseline assessment was research assistant–administered, and responses were recorded in REDCap.

All procedures were reviewed and approved by the Socio-Behavioral Institutional Review Board of the Biomedical and Social Behavioral Institutional Review Boards at Tulane University (reference number 2022‐1861). All participants provided verbal informed consent. All recorded data were private, confidential, and anonymous. Participants were compensated with a US $30 Amazon gift card for completing their baseline appointment, US $20 via ClinCard for their immediate postassessment and 30-day follow-up visit (each), US $2 via ClinCard for each daily diary completed, and a US $10 Walmart gift card for the 90-day follow-up. Starting in March 2024, after switching to treatment-arm only enrollment, participants were also compensated with US $5 via ClinCard per session completed, with a US $15 bonus if all 7 were completed.

The overall pilot sample was generally in their mid- to late-50s, with slight over-representation by men (n=17, >55%), a plurality identifying as minoritized sexualities (n=10, <45.45% heterosexual/straight), and roughly 3 quarters Black (n=22, 73.33%). A very large majority (n=30, 93.33%) reported <US $50,000 past-year household income. Digital confidence scores were high, in the “strongly agree”–“agree” range on average.

The pilot sample’s demographic details, overall and by study arm, are shown in Table 1.

Acceptability, similarly, was generally very high across measures in the NOLA Gem + GEMA (treatment) group (n=22). Overwhelmingly, 16 (72.73%) participants reported the easy use of the app, with perceived helpfulness more equivocal, but a very small minority (n=2, 0%‐9.09%) perceiving the app as totally ineffectual. Bivariate logit regressions were null (male, P=.40; Black, P=.75; Latine, P=.26; lifetime alcohol use, P=.70; past 30-day drug use, P=.29; positive PTSD screen, P=.07; >median posttraumatic growth, P=.15; >median digital confidence, P=.22), suggesting no detectable social patterning to NOLA Gem acceptability among the pilot sample.

Nearly a third (n=7, 31.82%) reported GPS-related privacy concerns, though no statistically significant associations with privacy-related concerns were evidenced (male, P=.23; minoritized sexuality, P=.70; >median HIV stigma, P=.70; >median adverse childhood experiences, P=.30; as GEMA was universal, these models used the complete sample, ie, n=30). Consistent with the qualitative findings, nearly a quarter (n=7, 23.33%) found the daily diaries difficult to incorporate into their daily lives. Within the treatment group, 11 activated a custom geofence; though enthusiasm for this feature was mixed, no participants found it “not at all” helpful.

The app usage and satisfaction measures are summarized in Table 2.

Additional measures of multifactorial appeal are plotted in Figure 2.

Measures of NOLA Gem’s perceived helpfulness are plotted in Figures 3 and 4.

Because the control-treatment comparison was essentially additive, building features atop the already-demanding GEMA deployment, we tested for between-group differences in acceptability measures. Although, descriptively, GEMA measures tended to be higher (eg, treatment: mean 2.59, SD 0.73 vs control: mean 2.75, SD 0.46 for “able to open the app without confusion”), between-group t tests were uniformly null (P=.13‐.87).

NPS scores averaged (mean 7.70, SD 3.04). Promoters comprised 63.64% (n=14), and detractors comprised 22.73% (n=5) of the GEMA + NOLA Gem treatment sample, yielding an NPS metric of 40.91, approaching the generally accepted ≥50 “excellent” score [56].

At baseline, 32 participants were enrolled, with 30 (93.75%; this study’s analytic sample) completing the 21-day trial period and completing ≥1 daily diary. Sociodemographic attributes of noncompleters are withheld to maintain participant confidentiality.

NOLA Gem paradata, collected from the treatment arm, showed that in a majority of the 607 completed daily diaries, across all participants (n=22), no skill was recommended: 463 (76.28%). Only 3 (13.64%) of the users completed 0 skills, and in 1 instance, technical errors may have prevented data recording.

These usage distributions are shown in Table 3.

We trialed NOLA Gem, a trauma-informed JITAI leveraging GEMA inputs to provide people with HIV in New Orleans adaptive coping skills refreshers, against a GEMA-exclusive control, over a 21-day pilot period. In general, GEMA + NOLA Gem users (n=22) endorsed high acceptability, particularly in domains of perceived helpfulness, ease of comprehension, and ease of use. Of the 32 participants enrolled, 30 (93.75%) completed at least 1 daily diary, highlighting the app’s feasibility. Among GEMA + NOLA Gem (treatment-arm) participants, for every skill, completions exceeded recommendations, indicating a nontrivial rate of “pull,” or user-determined, app interactions, and suggesting valuable “Big E” (behavioral) engagement among a subset of dedicated users [39,44]. Most skills saw zero inflation among their “minutes spent” distributions, which were heavily skewed; this pattern might indicate user self-selection into accessible LIFT skills. Conversely, these users might miss out on new and potentially more beneficial coping skillsets. Qualitative user experience concerns converged on the frictions and interaction costs associated with the twice-daily EMA inputs: their sometimes-inconvenient timing and often-burdensome duration. While the relatively low skills recommendation rate among daily diary completers suggests those items may not be parsimoniously tapping frequently encountered stressors, the lower recommendation rates also prevent habituation—users “tuning out” a JITAI due to excess “push” notifications of no clear benefit [44]. Nevertheless, many participants declined to offer recommendations for next-generation changes, describing the app as “perfect” and acknowledging the value of the “just-in-time” adaptive coping skills delivery intuitively.

The protocolization, or defining the ruleset that determines how intraindividual dynamic inputs trigger just-in-time interventions [44], for each stressor-triggered adaptive coping skills prompt was formally mapped by the study team via feature hypothesis charting in 2022. Work on within-person JITAI optimization has advanced substantially since then, incorporating methods as diverse as reinforcement learning [71], N-of-1 model-predictive controller personalization algorithms [72], LLM-enabled free-text user preference updates in real time [73], and, most commonly, microrandomized controlled trials to isolate the most promising, in Nahum-Shani terms, “states of vulnerability/opportunity” and “states of receptivity” for momentary intervention delivery [74]. NOLA Gem, in contrast, was protocolized based on the findings of Aims I and II [21] and HIV biobehavioral domain expertise. These rulesets were static [55], even as they responded daily to EMA inputs. It is evident, based on our qualitative findings, that the sprawling battery of inflexibly scheduled daily diaries [55] was the primary driver of user burden and, as such, the foremost threat to acceptability. Despite some participants’ insistence that “nothing” in NOLA Gem should change, a clear opportunity exists to optimize the JITAI mechanisms of action, re-protocolizing and pruning surplus daily diary measures in an evidence-based manner.

Though, by most standards, our findings support the advancement to a fully powered randomized controlled trial, in light of the abovementioned emergent methods for JITAI optimization, an additional microrandomized controlled trial cycle using contemporary techniques may further enhance acceptability. Launching intraindividual microrandomized trials, the most robust stressor → state of receptivity → coping skill → proximal outcome → distal outcome dynamics can be modeled and protocolized, at a within-person and within-user base (eg, across an outpatient HIV clinic network) level [75]. Indeed, the required in-app feature set already exists, in the form of single-item “SkillCheck” quizzes tap each delivered skill’s success in improving a hypothesized JITAI-consistent proximal outcome.

Such improvements would benefit the prospective user base of people with HIV. However, the efficiencies built in by such a “pruning” cycle would benefit future clinical workflow integration—a potential prerequisite to a successful Phase III trial—and, ultimately, scalability and sustainment. The nonadoption, abandonment, scale-up, spread, and sustainability of technological innovation framework recognizes 7 domains. These include the condition, or health issue itself; the technology under consideration; its inherent value proposition, supply- or demand-side; adopters, including staff, patients, caregivers, and providers; organizational climate; political and regulatory considerations; and temporal aspects of implementation and organizational resilience (or lack thereof) [76]. Interventions that incorporate rampant complexity across the sum of these domains are shown, across empirical case studies and hermeneutic literature reviews [77], to rarely, if ever, achieve scale and sustainability. HIV prevention and treatment, despite past policy successes, remain subject to a volatile political climate [78]. People with HIV engaged in outpatient care endure poverty, violence, and medical and psychiatric comorbidities disproportionately [1,7,11,79]. Complexity, across the HIV continuum of care, is all but assured. Even at the pilot stage, it is incumbent upon technologists and HIV digital therapeutics investigators to prioritize simplicity, wherever possible, in intervention design and delivery. The NOLA Gem pilot study, and, in particular, the valuable insights contributed by NOLA Gem pilot users, lay the groundwork for concurrent optimization via simplification to advance next-generation HIV mobile health.

This study is subject to limitations as well. First, the pilot sample was not powered to detect between-group differences, either in GEMA versus GEMA + NOLA Gem acceptability (a proxy for differential interaction costs [61]), or predictors of disparate acceptability across subsets of outpatients with HIV. Null findings on these questions should be understood as highly preliminary. In particular, we are unable to extend the important work of Hubach et al [60], revealing EMA-related confidentiality concerns among a sample of rural men who have sex with men. A confluence of external factors—namely the COVID-19 variants still endemic in Louisiana, and Hurricane Ida [46,47]—delayed the trial launch substantially, hampering our ability to obtain our preregistered enrollment target of 60 participants, and to achieve balanced 1:1 random assignment. Though oversampling treatment-arm participants was ultimately to the benefit of these analyses, focused on the experiences of those NOLA Gem–assigned participants, our capacity to demonstrate equivocal acceptability across GEMA and JITAI paradigms is, similarly, undermined. We are also unable to gauge meaningful engagement metrics via paradata due to the small sample size, beyond a dichotomized “high” versus “low” engager subsetting [80]. Although these findings suggest future directions for NOLA Gem optimization, and support the need for a fully powered randomized controlled trial, attention toward protocolizing user-centered in-app adaptiveness to spur engagement is needed [45,75].

Trauma, particularly trauma rooted in firsthand violence exposures, disproportionately affects the health and wellbeing of people with HIV [1]. Interventions tailored to the needs of people with HIV must not only address environmental stressors across spatial and temporal dimensions [17-20], but they must also embrace principles of trauma-informed care [2,7]. To this end, we developed NOLA Gem, a smartphone-delivered JITAI adaptation of the effective LIFT group intervention [39]. This pilot study demonstrated the acceptability, feasibility, and usability of NOLA Gem, supporting future optimization, and a fully statistically powered evaluation of the app’s efficacy among outpatients with HIV.