Understanding the interplay of social context and physiology on psychological outcomes in trauma-exposed adolescents: a protocol and baseline data report

ABSTRACT

Background: Adolescence is a period of increased risk for trauma, with a subset of adolescents developing post-traumatic stress disorder (PTSD). Research has explored the interplay of biology (e.g. heart rate, skin conductance) and social context as critical influences in the early development of PTSD. Prior research is limited, however, by assessments at discrete points and reliance on self-report. Research is needed to model the dynamic interplay of social context and biology.

Method: We have developed an integrated ecological method for characterizing biosocial processes using: (1) wristband health trackers; (2) the Electronically Activated Recorder (EAR), which samples ambient sounds to capture social context; and (3) adolescents’ own use of online social networks (OSNs). Participants consist of 191 adolescents (13–18 years) recruited in-hospital after traumatic injury (e.g. physical assault, serious vehicular accident). For 2 weeks post-discharge, participants wore a wristband health tracker and the EAR (later coded for social environment, interaction, affect, content, and behavioural indicators of acute threat construct). At the 2 week follow-up, they returned the devices, permitting download of OSN activity, and participated in a clinical interview. At 6 weeks, 6 months, and 9 months post-trauma, participants provided an OSN download of the prior 2 weeks, completed laboratory procedures to characterize fear learning and physiological reactivity indicators, and underwent semi-structured interviews of stress-sensitive disorders and self-reported measures of social context. In vivo social context was captured through the EAR/OSN, and will be modelled in concert with heart rate and skin conductance to reflect the dynamic process of biomarkers and social processes over time.

Results: The research protocol and cohort characteristics are described herein.

Conclusion: The combination of ecological and laboratory-based methods used here provides rich, clinically relevant characterization of interindividual and intraindividual biopsychosocial predictors and correlates of post-trauma adjustment, with direct implications for theory and intervention.

HIGHLIGHTS

• Social support after a trauma for adolescents in the digital age includes both in-person and virtual social interactions.

• This innovative research protocol examines the post-trauma interplay of social and biological factors in adolescents in the days following trauma as influences on both the development of post-traumatic stress outcomes and fear learning processes.

• The methods and cohort described here provide new ways of thinking about measuring social context after trauma.

1. Background

Trauma-exposed adolescents are at risk for a host of deleterious outcomes, particularly stress-sensitive disorders such as acute stress disorder (ASD), post-traumatic stress disorder (PTSD), depression, anxiety, and substance use disorders (Olff et al., 2025). Child and adolescent prevention studies have failed to provide strong support for interventions that can be implemented during the critical hours to days post-trauma, and very few studies of the early post-trauma period have applied the level of resolution needed for informed and innovative intervention development (Forbes, Bisson, Monson, & Berliner, 2020; Phelps et al., 2022). Theory and available evidence point to salient influences on the development of PTSD: individual stress response and social context (Nugent, Amstadter, & Koenen, 2010; Olff et al., 2025). Research has highlighted a range of peritraumatic and early post-traumatic biological responses, with evidence for the predictive utility of early indicators of reactivity including factors, such as heart rate (HR) and skin conductance (SC); HR and SC reflect, in part, the degree of physiological arousal a person is experiencing, with some evidence that high levels of arousal observed acutely post-trauma may reflect a high vulnerability to more distressing and impairing PTSD symptoms in the future (Olff et al., 2025). An individual’s ability to recover from traumatic or other adverse circumstances is also highly variable based on experiential factors, including available social support [both in-person interactions and online social network (OSN) behaviours] (Olff et al., 2025). Culturally and contextually relevant social connections provide empathic relational support and facilitate meeting one’s needs when it is often most difficult to do so. With OSN being a fairly new, yet highly utilized domain for social communication, online interactions are likely to be an important part of an adolescent’s social landscape.

This present study is situated within a broader goal of interrupting the development of PTSD symptoms by identifying novel, potentially malleable targets for intervention during the acute post-trauma period. This investigation sought to characterize the ways in which social context and individual stress responses interact to attenuate or exacerbate the development of PTSD symptoms following a potentially traumatic event. Adolescents were recruited during their acute trauma-related emergency department (ED) visit, and followed prospectively with follow-up appointments at 2 weeks, 6 weeks, 6 months, and 9 months post-presentation. Throughout, focal constructs were assessed multi-modally: the protocol integrated naturalistic observation [i.e. Electronically Activated Recorder (EAR), scraped OSN activity data], laboratory-based stress paradigms for stress-indicative biomarkers (i.e. dot-probe test, Cyberball, acoustic startle response), clinical interviews, and self-reported psychometrically validated measures. Integrating these complementary data sources maximizes the strengths of each, triangulating information to yield more nuanced within-person characterization. Findings from this research will translate into novel and empirically informed intervention approaches.

1.1. Adolescence as a stress-sensitive period of development

Relative to adults and children, adolescents are at increased risk for experiencing trauma, with some evidence that rates of PTSD among adolescents are nearly twice as high as those among adults (Breslau, Wilcox, Storr, Lucia, & Anthony, 2004; Nooner et al., 2012). Approximately 80% of adolescents in the USA meet Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) criteria for exposure to a potentially traumatic event (Kilpatrick et al., 2003; Nooner et al., 2012). Among those exposed, studies find that between 3% and 57% of adolescents develop PTSD symptoms that meet criteria for diagnosis, with substantial variation by trauma type (e.g. 57% meet criteria following sexual abuse, 10% following a natural disaster), sex assigned at birth (adolescent females are twice as likely to meet criteria as male counterparts), social support available, and physical health conditions (Nooner et al., 2012). Adolescent PTSD is associated with a host of negative outcomes, including persistent hormonal, neurotransmitter, and neuroanatomical abnormalities, as well as deleterious physical health outcomes, otherwise avoidable academic difficulty, and suicidal thoughts and behaviours (Carrion et al., 2002; De Bellis et al., 2009; De Bellis, Baum, et al., 1999; De Bellis, Keshavan, et al., 1999; Giaconia et al., 1995; Kilpatrick et al., 2003; Lemieux & Coe, 1995; Mathews et al., 2009; Mazza, 2000; Mazza & Reynolds, 1999; Saltzman et al., 2006; Wu et al., 2009; Yehuda & LeDoux, 2007; Zatzick et al., 2008). The effects of trauma and PTSD persist into adulthood, predicting the severity and complexity of PTSD as well as depression, substance abuse, and anxiety (Cloitre et al., 2009; Colman et al., 2013; Li D'acry, & Meng 2016; Mandelli, Petrelli, & Serretti, 2015; McKay et al., 2021; McLaughlin, Conron, Koenen, & Gilman, 2010). Yet, many trauma-exposed individuals do not experience deleterious outcomes, pointing to heterogeneity in response to trauma that, if better understood, could be applied to intervention. Efforts to prevent PTSD in trauma-exposed children and adolescents have achieved limited success (Kassam-Adams, 2014; Maccani, Delahanty, Nugent, & Berkowitz, 2012; Nugent, Christopher, et al., 2010). One possible explanation for limited success is that frameworks used to guide interventions are insufficiently nuanced, failing to include subtle timing considerations or to incorporate the dynamic interplay of individual and social context. Given the chronic, debilitating course for adolescents who develop PTSD, research that provides more nuanced characterization of processes that can become treatment targets is needed.

1.2. Acute physiological arousal and PTSD

Theoretical models for the development of PTSD highlight the early post-trauma period, particularly physiological arousal during this period, as a critically important influence on PTSD development (Feldner, Monson, & Friedman, 2007; Kassam-Adams, 2014). In the acute (minutes to hours post-trauma) period, high physiological arousal may promote overconsolidation of trauma memories (Jones & Barlow, 1990; Keane et al., 1985; McCleery & Harvey, 2004; Morris & Rao, 2013; Pitman, 1989; Pitman et al., 1993; Pitman & Orr, 1989). In the early (days to weeks post-trauma) period, these overconsolidated memories may increase intrusive recollections and re-experiencing symptoms of PTSD (Pitman, 1989; Pitman & Orr, 1989; Pitman et al., 1993). Should repetitive re-experiencing (and reconsolidation) be paired with increased physiological arousal, associative pairings between re-experiencing and distressing physiological arousal may be reinforced (Jones & Barlow, 1990; Keane et al., 1985; Pitman et al., 1993). Current theories propose that distressed individuals avoid thoughts/reminders of the trauma, preventing natural habituation. Consistent with this theory, we have reported associations between acute (transport and ED) HR samples and paediatric PTSD (Kassam-Adams et al., 2005; Nugent, Christopher, & Delahanty, 2006). However, risk conferred by these early biomarkers is moderated by environmental context (Nugent, Ostrowski, Christopher, & Delahanty, 2007). Characterization of physiological activity beyond the ED visit – and the factors that may influence this activity  – during the initial weeks is critical. One investigation assessed adults within weeks of trauma, showing that SC response to a neutral stimulus was higher among individuals with ASD (a common precursor to PTSD) relative to individuals with PTSD as well as non-traumatized and traumatized controls (Felmingham, Rennie, Gordon, & Bryant, 2012). These findings may point to a developmental process whereby SC differences are pronounced within the first 2 weeks, prior to the window of diagnosable PTSD. However, this was neither measured continuously nor measured in concert with other processes that might become targets for intervention. The present research will characterize physiological activity, over the course of 2 weeks post-trauma in concert with key treatment modifiable contextual factors (e.g. social support, both in person and online) to better understand this critical period.

1.3. Social context, social media, and the changing social world of adolescents

Although evidence across multiple studies and meta-analyses supports moderate to large effect sizes of social context on PTSD, for example parent PTSD, social support from parents, teachers, and friends, family functioning, and social withdrawal, interpersonal processes have only recently been incorporated into theories of PTSD development (Alisic et al., 2011; Brosbe, Hoefling, & Faust, 2011; Cox, Kenardy, & Hendrikz, 2008; Nijdam & Wittmann, 2022; Trickey, Siddaway, Meiser-Stedman, Serpell, & Field, 2012). Social context may influence biological/learning processes by attenuating (or enhancing) physiological arousal and challenging (or reinforcing) avoidance. Several investigations with non-clinical populations have shown that physiological reactivity and social interactions are closely tied (Christenfeld et al., 1997; Ditzen & Heinrichs, 2014; Ewart & Kolodner, 1991; Gerin et al., 1992; Lepore et al., 1993), with evidence that positive interactions with social support in advance of, or during, a stressful experience can mitigate laboratory stressors (Che, Cash, Chung, Fitzgerald, & Fitzgibbon, 2018; Ewart & Kolodner, 1991; Kambara, Mitsuishi, & Harada, 2024; Lepore et al., 1993). Laboratory research also points to the possibility that stressful experiences may result in an increase in social approach behaviours (Taylor et al., 2000; Taylor et al., 2006). Although these laboratory studies inform models of stress, and two of the most important peritraumatic predictors of PTSD in adolescents are social support and early biological response (Alisic et al., 2014; Brosbe et al., 2011; Trickey et al., 2012), investigations rarely examine the interplay of social and biological processes. For instance, we found that overall parent distress, and particularly parent PTSD, predicted later child PTSD even after accounting for early child biomarkers (Nugent et al., 2007). On the other hand, supportive family interactions could be protective (Montpetit, Bergeman, Deboeck, Tiberio, & Boker, 2010; Xiong, Milios, McGrath, & Kaltenbach, 2022). Existing research related to social interactions is extremely limited by reliance on self-report. Yet, as we have described, the experience of trauma itself may impact ability to accurately report, recruit, or benefit from, social support (Nugent, Amstadter, & Koenen, 2010). Post-trauma social interactions may be determined partly by the young person’s own behaviours (Monson et al., 2012). In a pilot study to demonstrate feasibility, the EAR method was used to study parent–child interactions in 71 families after a child’s hospitalization owing to a physical trauma (Vasileva et al., 2022). Furthermore, audio (EAR) data from a small pilot study of 13–18-year-olds to support this research application reinforced the richness of ecological data collection, capturing relevant examples of social context such as bidirectional arguments about adherence to medical regimens and comments about money or disagreements among family members about who would be responsible for participant care. Objective, observational modelling of the bidirectional nature of interpersonal interactions is needed to inform theory and intervention.

OSNs, such as texting and Instagram, increasingly augment adolescents’ face-to-face interactions with peers (and sometimes parents) (Drentea & Moren-Cross, 2005; Doty & Dworkin, 2014; Jang & Dworkin, 2014). Over time, adolescent face-to-face and OSN interactions are overlapping, even reciprocally reinforcing, yet unique processes (Reich, Subrahmanyam, & Espinoza, 2012). In adults, aspects of OSN use have been associated with increased risk for rumination, negative feelings, and depressive symptoms, and the same has been suggested for adolescents (Feinstein et al., 2013; Lee, 2014). Some providers working with adolescents have suggested that OSN use may negatively affect the still-developing social skills of adolescents (Graham, 2013). More likely, it is not simply OSN use itself but specific aspects of OSN use by the adolescent and their network that determines whether the OSN’s influence is deleterious (Olff et al., 2025). For example, whereas one investigation found that most adolescent texting content is positive/neutral, participants with an increased frequency of exchanging texts with negative talk were more likely to show internalizing and anxious depression (Underwood, Ehrenreich, More, Solis, & Brinkley, 2015). Although adolescent studies have shown some potential deleterious effects of OSNs, they also point to the potential benefits of OSNs in attenuating depressed mood (Frison & Eggermont, 2016; Rice, Kurzban, & Ray, 2012). For example, instant messaging may be particularly helpful for introverted adolescents, permitting them to express distressed emotions and receive support and advice from their social network (Dolev-Cohen & Barak, 2013). Internet support groups for depression have been associated with decreased depression (Houston, Cooper, & Ford, 2002), and experimentally induced increases in status updating on participant Facebook accounts have been associated with reduced loneliness (Deters & Mehl, 2013). Researchers have only recently begun to describe the use of OSN following trauma, particularly following disasters (Ben-Ezra et al., 2013; Casacchia et al., 2012; García-Ramírez, Bogen, Rodríguez-Guzmán, Nugent, & Orchowski, 2021; Olff et al., 2025; Procopio & Procopio, 2007; Rodríguez-Guzmán, García-Ramírez, Bogen, Orchowski, & Nugent, 2021), and to propose ways in which OSNs may be especially important during times of transition and stress (Mikal, Rice, Abeyta, & DeVilbiss, 2013). During times of stress, messages exchanged can reflect caring, empathy, and connectedness. Friends may use messages to offer emotional validation and support, as well as instrumental support, such as offering food or a lift somewhere. Furthermore, OSNs naturalistically document pre-trauma social interactions, such that studies can evaluate social context prior to the trauma without retrospective or self-report bias. Very few studies have examined observed naturalistic adolescent OSN use across multiple platforms as related to psychological well-being as one component of a greater social context. This is a missed opportunity for both understanding benefits and drawbacks, and the ways in which it may and may not be helpful for post-trauma psychological adjustment.

1.4. Integrating physiology and social context data

One interpretation of available evidence is that social context interacts with and modifies neurobiological processes, and results in more adaptive stress responses than early biomarkers would have predicted (Cohen & Wills, 1985; Ditzen & Heinrichs, 2014; Lepore et al., 1993; Taylor, 2006; Von Dawans, Fischbacher, Kirschbaum, Fehr, & Heinrichs, 2012). If so, interventions may be helpful to the degree that they are aimed at social context, irrespective of an individual’s concomitant biology. It is also possible, however, that biologically mediated processes, such as fear bias and interpretation of social context, serve to impact self-reporting of social context and behaviour towards social others (Bar-Haim, Lamy, Pergamin, Bakermans-Kranenburg, & Van Ijzendoorn, 2007; Masten, Telzer, Fuligni, Lieberman, & Eisenberger, 2012). Furthermore, if theories of biological underpinnings of PTSD are accurate, it is possible that a developmental process unfolds over the weeks to months post-trauma to reshape the biases underpinning social interactions. The dot or visual probe task (Mogg, Bradley, & Hallowell, 1994), an experimental paradigm (and a National Institutes of Health Common Data Element for PTSD) measures visual attention bias (Mogg & Bradley, 1999). Most studies show attention biases towards threat in trauma-exposed individuals (Bar-Haim et al., 2007; Lau et al., 2012). Such biases could affect social relationships by increasing perceptions of (and/or attention to) negative or threatening cues from others; this could reduce perceptions of support and even increase maladaptive behaviours, such as conflict or withdrawal. Another, complementary, method for assessing fear bias is through eye-tracking, which captures gaze patterns and has shown an association between PTSD symptoms and the duration of looking at threatening faces (Lee & Lee, 2013). Although childhood trauma is associated with differences in eye-tracking in adulthood (Irwin, Green, & Marsh, 1999), no studies to date have examined PTSD as related to eye-tracking in young. Furthermore, to date, no studies of attention bias, using probe tasks or eye-tracking, permit prospective characterization of how these biases develop over time after trauma.

Consistent with the Research Domains Criteria (RDoC) (Irwin et al., 2010) threat construct, we propose that fear bias has considerable overlap with physiological reactivity to a variety of stimuli, and particularly social stimuli. Even when visual cues (such as facial expressions) are not available, as in OSNs, trauma may sensitize biological systems to ‘threatening’ social context. For example, studies using a social exclusion paradigm (Cyberball, a virtual ball-tossing game) have demonstrated effects on stress hormones and brain activity (Kawamoto et al., 2013; Masten et al., 2011; Silk et al., 2014; Riem, Bakermans-Kranenburg, Huffmeijer, & van IJzendoorn, 2013), and reactivity to the task is associated with, and predictive of, depression in adolescents (Silk et al., 2013; Masten et al., 2011). However, there is evidence that social context may protect against the distress of Cyberball social exclusion (Masten et al., 2012). Although the first few weeks post-trauma are likely to be critical for understanding the development of PTSD, studies have shown that the initial months post-trauma are critical for determining symptom resolution versus chronic symptoms (Amstadter et al., 2011; Nugent et al., 2006). We propose that biases are developed and reinforced during this time, and that understanding their development is critical for understanding chronicity. We will further supplement the strengths of this method over repeated assessments by also assessing OSN use in the 2 weeks prior to each laboratory assessment. Longitudinal laboratory assessments permit rich characterization of how physiological and social processes, and related biases, may maintain early symptoms.

1.5. Innovative yet theoretically rooted directions in trauma research

Given the importance of the early post-trauma period, particularly with respect to physiological influences on memory consolidation/reconsolidation, in vivo assessment of HR and SC during the weeks post-trauma is critical to the scientific study of the phenomenology of PTSD development. Furthermore, the opportunistic contextual sampling used here permits sampling of social environments triggered by an individual’s increase in HR and SC. The present methods for assessing social context (EAR, OSN) provide an innovative, objective, and complementary approach that, non-invasively and with minimized participant burden, permits a rich characterization of social processes that unfold in the aftermath of trauma (and even pre-trauma OSN use). In vivo observational methods are particularly important when measuring evaluative constructs (e.g. conflict) that, in self-report, are prone to socially desirable responding (e.g. impression management, self-deceptive enhancement). Moreover, the use of the EAR and OSN in combination is a major innovation, as the present research is the first investigation to simultaneously examine the functioning of these ‘different’ social worlds during a time of critical importance. While OSN use is a complex and ongoing interaction, its observed benefits for adolescents during the early post-trauma period may include the immediacy that OSNs permit, an ability to reach out to (and receive support from) numerous friends at once, and ways to reinforce or supplement in person social interactions. Furthermore, OSN interactions may be sampled in a cost-effective manner and may be assessed through pre-trauma interactions; the implications of understanding how pre-trauma OSN use may predict, or how post-trauma OSN use may shape the development of PTSD symptoms are paradigm shifting for intervention development. By simultaneously examining physiology and social context, the present research further informs our understanding of the ways that social interactions and biological processes may reciprocally impact one another. This understanding can then be applied to novel intervention developments, implemented within the acute and/or early post-trauma period (e.g. hybrid or online-mediated skills groups or community-building sessions to augment individual therapy, chat-based support during this acute period). Furthermore, the present investigation combined the strengths of assessments of in vivo naturalistic processes during the acute/early post-trauma period with the strengths of laboratory assessments of perceptual biases.

This translational investigation was developed to characterize the early interplay of biomarkers and social context influencing symptom development in 191 trauma-exposed adolescents. The project aimed to accomplish the following goals: (1) to examine how social interactions bi-directionally influence the development of PTSD symptoms; (2) to examine physiological markers of arousal in the first weeks post-trauma as they relate to social context and developing symptoms of PTSD; and (3) to complement naturalistic assessment with laboratory assessments that provide in-depth understanding of the role of socially relevant threat constructs in maintaining PTSD. The present translational investigation models the early interplay of HR and SC that influences traumatic stress symptom development.

1.5.1. Manuscript aims

The current manuscript has four purposes. First, it summarizes the working conceptual model of youth PTSD development and describes the investigators’ operationalization for the study aims. Secondly, it provides an overview of research procedures for confidentiality, privacy, and working with OSN data as an increasingly relevant domain to consider in assessing stress cycles for adolescents. Thirdly, it documents the aims, methods, planned analyses, and cohort characteristics of the study, with findings forthcoming. Fourthly, it highlights the feasibility of longitudinal research with trauma-exposed adolescents that includes ecological, laboratory paradigm, and self-reported components.

1.5.2. Study aims

First, the present research aims to examine how social interactions (EAR, OSN) may influence the development of symptoms of PTSD. We hypothesize that OSN use in the 2 weeks post-trauma will be associated with fewer symptoms at follow-ups. We also hypothesize that adolescents who spend more in-person time with others in positive interactions will evidence fewer symptoms over the course of follow-ups, relative to adolescents who are alone or whose interactions are conflictual. We will also test our hypothesis that pre-trauma social interactions (indicated by pre-trauma OSN use) will predict observed social (OSN and EAR) interactions following the trauma. Our second aim is to examine psychophysiological indicators in the first weeks post-trauma as related to social context and developing symptoms of PTSD. We hypothesize that social interactions (EAR, OSN) will be associated with attenuation of HR and SC. We also hypothesize that increased HR and SC during the 2 weeks post-trauma is associated with increased risk for a stress-sensitive disorder. Our third aim is to complement the strengths of naturalistic assessment with laboratory assessments (psychophysiology, attention bias, Cyberball) that provide in-depth understanding of the role of socially relevant fear bias in maintaining symptoms of stress-sensitive disorders. We expect that naturalistically characterized biological processes (high HR/SC) associated with stress-sensitive disorders will also be associated with fear bias (physiological reactivity and attention bias). Finally, we hypothesize that, over time, persistently higher fear bias will be associated with higher symptom levels that do not resolve.

2. Research protocol

The present study took place in the ED of a children’s hospital in the Northeastern USA. Participants were recruited during their ED visit for a potentially traumatic event. After assessing eligibility, research staff approached participants for the informed consent process, including caregiver consent and adolescent ascent procedures for minors or consent for self for 18-year-old participants. At the time of enrolment (i.e. baseline), participants completed brief assessments of demographics. At this time, participants and their caregivers were also oriented to the EAR and smartwatch, and instructed to wear it for 2 weeks post-discharge. At the 2 week follow-up, the young people provided a download of OSN data from 2 weeks pre-trauma to the 2 week follow-up, and participated in a semi-structured clinical interview to assess symptoms of stress-sensitive disorders. At 6 weeks, 6 months, and 9 months post-trauma, they completed laboratory procedures and semi-structured diagnostic interviews of stress-sensitive disorders, and self-reported on measures of social context.

2.1. Recruitment and enrolment

To be eligible for study inclusion, participants must have experienced a potentially traumatic event resulting in ED presentation, be 13–18 years of age, and have sufficient English language skills to complete study procedures. The present study operationalized ‘potentially traumatic event’ as an event sufficient to warrant activation of the trauma surgery team, hospitalization, and/or placement in a critical care room. In addition, participants who did not need high levels of medical activation but were involved in an adverse event known to be potentially traumatizing (e.g. a shooting) were also included in recruitment efforts. Exclusion criteria were any current psychotic symptoms, developmental delay or pervasive developmental disorder, and substance use that would interfere with participation. Regarding suicidality, participants who endorsed suicidal ideation and otherwise met criteria were considered eligible, unless they required more intensive support that would interfere with participation, based on a clinician’s assessment.

Research assistants utilized chart review and worked with hospital staff to screen patients for eligibility, as the criteria that determined eligibility for this study were often information collected and documented in the electronic medical record upon presentation to the ED. Research assistants approached patients, either in the ED or in their private hospital room following admission, to provide more information about the study, confirm eligibility, and lead consent and assent protocols. Written informed consent was obtained from a parent or legal guardian and the adolescents provided informed assent; both parent and adolescent consent/assent was required prior to participation. In total, 191 adolescents were enrolled.

2.2. Procedures

2.2.1. Schedule and retention

Following recruitment, the current study scheduled an orientation ‘baseline’ appointment within several days of hospital discharge. Follow-up appointments were scheduled for 2 weeks, 6 weeks, 6 months, and 9 months post-trauma (Figure 1). Initial data were assessed at 2 weeks post-discharge, rather than immediately following the index event, so that analyses would focus on the ‘end’ of the acute period, rather than the beginning, when elevations in physiological responses may be more ‘normative’ or expected to be homogeneously elevated for all patients.

Figure 1.

Strobe diagram. EAR = Electronically Activated Recorder.

The study employed multiple strategies to ensure retention and adherence. First, we collected multiple forms of contact information to maximize avenues for communication (e.g. home telephone, mobile phone, e-mail). Between the 6 week, 6 month, and 9 month follow-ups, participants received postcards to remind them of the study and to encourage them to contact the study staff with changes in contact information. Secondly, participants were compensated for their time and for returning the devices. If needed, transportation was provided via secure rideshare, at no cost to the participant, to complete follow-up assessments. Upon the onset of the coronavirus disease 2019 (COVID-19) pandemic, protocols were adapted in alignment with national and institutional guidelines for minimizing risk.

2.3. Baseline procedures

2.3.1. Chart review

Participants were asked to consent to the collection of data from their medical record. The study team extracted data related to the index medical visit (i.e. event description, injury severity, loss of consciousness, site or sites of injury, early physiological biomarkers such as HR, indicators of distress, medications administered, chronic conditions). To supplement the detailed assessment of past trauma included with the Clinician Administered PTSD Scale for Children and Adolescents (CAPS-CA), we recorded prior potentially traumatic events available in the chart.

2.3.2. EAR orientation and procedures

Baseline procedures include providing participants with the EAR, as described in depth below. Participants and caregivers were oriented to the device, trained in proper use and study expectations, and given answers to any questions that they had. Participants completed a brief knowledge check quiz to verify their understanding of these devices to complete the appointment.

In addition to in-person orientation to the EAR at the baseline appointment, participants received copies of instructions on how to wear the EAR and reminders of the voluntary nature of their participation. Participants were encouraged to wear the EAR as much possible during appropriate times (i.e. when not in school, during contact sports, or at work) and to only take it off overnight and when proper functioning is jeopardized. Research staff called to remind adolescents about the EAR. The EAR was set to record 30 s every 12 min between 9 a.m. and 11 p.m., or approximately 5% of the participants’ waking time. The study implemented established safeguards to protect privacy and safety and to ensure data confidentiality by developing extensive protocols for deidentifying (and verifying deidentification) data (i.e. removing all names of people, places, social media handles, etc.). Research assistants were trained on protocols, and asked to keep instructions physically present while completing data tasks to ensure fidelity to safety protocols. In addition, 15% of the data were selected at random for double verification by the research coordinators. Protocols were also guided by diversity, equity, and inclusion principles (Kaplan et al., 2020; Kaplan et al., 2024; Robbins, 2017). To ensure that data were captured in culturally accurate and properly contextualized ways, protocols were proactively directive regarding how to include slang phrases and non-English verbalizations in transcriptions and coding (Kaplan et al., 2020; Kaplan et al., 2024). Comment spaces were included in the processing platform to allow for notes and questions for discussion. In addition, team members attended weekly meetings that involved reflecting on and discussing their own identities and biases, and the ways in which they may be influencing transcription and coding (Kaplan et al., 2020; Kaplan et al., 2024). A wiki was used to document discussion for later reference to maintain consistency. Protocols with similar procedures and parameters have resulted in good tolerance of and adherence with the method (Manson & Robbins, 2017; Mehl & Holleran, 2007). Sound files were downloaded at the 2 week visit. Prior EAR studies have found good intercoder reliabilities for standard Social Environment Coding of Sound Inventory (SECSI) variables (including conflictual conversations) (Kaplan et al., 2020; Mehl, Gosling, & Pennebaker, 2006). Cohen’s kappa was calculated to assess interrater reliability across codes. On average, Cohen’s kappa was .80. By code, kappa values are as follows: Sleep = 0.87, Phone = 0.83, Watch Talk  = 0.80, Trauma Talk = 0.90, With another person = 0.85, Location = 0.78, and Action = 0.75.

Discrepancies were adjusted through discussion, and final decisions made by research coordinators with guidance from study leadership. All sound files were coded by research staff who were trained and supervised by study leadership.

2.3.3. Ambulatory physiological activity

Android devices were used, based on their open-source development platform, which provides significant flexibility to utilize state-of-the-art mobile/wearable technologies. The ‘wearable internet-of-things (IoT) framework’ enables us to automate the sensor sampling from mobile and wearable devices such as smartwatches (Lam et al., 2022; Onnela & Rauch, 2016). Using this framework, we have recently created an Android EAR (anEar) app that is tailored to meet the requirements of opportunistic contextual sampling of the wearable device’s sensors. The software has incorporated intelligent settings such as periodic recording, set pattern lock to prevent access, blackout modes to prevent unwanted recordings, delay between recordings, and duration of recordings. The resultant app is highly robust, continuing to record as programmed even after battery has been discharged (once the device is recharged), an important consideration in adolescent research.

2.4. Follow-up schedule and procedures

2.4.1. Assessment of electronic communication

At the 2 week follow-up, participants were asked to bring the mobile device that they use for texting and other forms of private messaging. Participants were asked to permit a download of their OSN messages and activity from the last month (i.e. 2 weeks pre-trauma and 2 weeks post-trauma). Messages comprise text-based communications through SMS/MMS, iMessage, Twitter (direct messages and posts), Instagram (direct messages, posts, comments), Snapchat (metadata only), WhatsApp (in earlier instances), and Facebook communication across Facebook Messenger and ‘wall comments’. If at least 20% of participants reported the use of a new app rising in popularity, our team developed download procedures for the app. TikTok, although widely used at present, did not reach this threshold during the recruitment period. The data download and extraction is automated and implemented for a specified date range, without the experimenters manually retrieving individual messages. Data from apps such as Facebook can be assessed by asking the adolescent to log in to their account on a study computer and permit download. We developed a script to remove all non-text content and then create a text file consisting of activity between two dates and replace names with numeric codes. The downloaded file is temporarily stored on the study computer while the script deidentifies the content and stores only the timeframe of interest in a text file. The original file is destroyed upon transferring the deidentified data to an encrypted disk. The resulting data comprise a series of time-stamped messages tagged by the name of the application and a numeric code representing each unique participant and their contacts.

Text files were used to quantify aspects of electronic communication, including the frequency of messages submitted by the participant, count of responses to participant-initiated contacts, count of unique individuals, and frequency of messages to the participant. Time-series analysis will be used to reveal patterns of behaviour in relation to the time of the trauma. Sentiment analysis was used to assess the emotional valence of trauma-focused conversations with social support. Conversations were also qualitatively human-coded to supplement analyses conducted by Linguistic Inquiry and Word Count (LIWC) (Chung & Pennebaker, 2012; Pennebaker, Francis, & Booth, 2001).

2.4.2. Physiological data

Biopac MP150 for Windows (Biopac Systems) was used to collect psychophysiological data. Experimental stimuli were presented using SuperLab 4.0 for Windows (Cedrus) and synchronized with psychophysiological data acquisition using a digital input/output (DIO) card (Measurements Computing).

2.4.3. Acoustic startle response

The acoustic startle response (eyeblink component) was measured via electromyography (EMG) of the right orbicularis oculi muscle. Data were sampled at 1 kHz, and the signal was filtered using low- and high-frequency cut-offs of 28 and 500 Hz, respectively, and a notch filter at 60 Hz. The peak amplitude of the EMG startle response was scored 20–120 ms after the startle probe onset. Two 5 mm Ag/AgCl pre-gelled disposable electrodes were positioned approximately 1 cm under the pupil and 1 cm below the lateral canthus. A ground electrode was placed on the mastoid bone. The startle probe (noise burst) was a 106 dB sound pressure level, 40 ms burst of white noise, with near-instantaneous rise time.

2.4.4. Fear-potentiated startle session

Fear acquisition has been used successfully with adults for years (Gamwell, Nylocks, Cross, Bradley, Norrholm, & Jovanovic 2015). To adapt this technique for adolescents, we reduced the intensity of the aversive unconditioned stimulus (US) to a 100 ms airblast with an intensity of 80 psi directed to the larynx. This intensity was used with high-risk children in prior research and produced robust startle potentiation while being well tolerated (Briscione, Jovanovic, & Norrholm, 2014). The conditioned stimuli (CSs) were coloured shapes presented for 6 s on a computer monitor. The startle probe was presented after 6 s and followed by the airblast 500 ms later. One CS was paired with the airblast (CS+), while the other was not (CS−). The CS+ was reinforced 100% of the time. The session consisted of three blocks, each with three CS+ trials, three CS− trials, and three noise-alone (no CS) trials. Intertrial intervals were randomized from 9 to –22 s. Fear-potentiated startle was calculated as percentage change from baseline startle to the noise-alone trials.

2.4.5. Assessment of bias in social attention

The 15 min Dot Probe task was presented on computer using E-prime (Psychology Software Tools). This is a common data element for PTSD (www.phenxtoolkit.org). Participants responded by right- or left-sided key-press to indicate the location of a target (a ‘*’ sign) that appeared in one of two boxes, located on the right or left side of a fixation cross. Prior to target presentation, face pictures appeared in each box. By comparing reaction times to probes that followed various picture cues, a measure of vigilance or attention bias was generated. Trials lasted for 2000 ms, with a 1000 ms intertrial interval that randomly varied (± 250 ms). The experiment was two blocks of 80 trials, separated by 12 s fixation. Each trial included a ‘cue’ slide, consisting of face pairs, followed by a ‘target’ slide. Face pairs were presented in one of three combinations: two paired ‘neutral’ faces; an ‘angry’ and a ‘neutral’ face; or a ‘happy’ face and a ‘neutral’ face. The face stimuli included male and female faces, and an equal number of Caucasian and African American faces. The field of view was counter-balanced. We used a Tobii TX300 eye-tracker to assess eye movements, fixation, and pupil dilation during tasks.

2.4.6. Interview measures

The DSM-5 version of the CAPS-CA, paired with the Life Events Checklist, was used to assess trauma exposure and PTSD diagnosis and symptom severity (Nader et al., 1996). ASD symptoms were assessed using the Acute Stress Disorder Interview (ASDI), which is widely used an has excellent psychometric properties in adults and adolescents (Bryant et al., 1998; Harvey & Bryant, 2002; Jovanovic et al., 2013; Orsillo, 2001). We assessed for the range of DSM-5 stress-sensitive psychiatric outcomes, as well as possible comorbidities, using the Schedule for Affective Disorders and Schizophrenia for School-Age Children  – Present and Lifetime Version (K-SADS-PL) (Kaufman et al., 1997), a widely used and well-validated assessment tool. In addition, the Child and Adolescent Services Assessment (CASA) (Ascher et al., 1996) is a semi-structured interview assessing service use for mental health problems across multiple sectors (e.g. juvenile justices, schools). Adequate reliability has been demonstrated for this measure (intraclass coefficient = .74−.76).

2.4.7. Self-report assessments

Measures were selected to reflect the stress buffering model, which posits a protective role for social support (Cohen & Wills, 1985; Cassel, 1976), and the relationship perspective (Thompson, Flood, & Goodvin, 2006; Pierce et al., 1991; Sarason & Sarason, 2009), which proposes social benefits to be related to interpersonal qualities and reciprocal interactions. Adolescents completed the Multidimensional Scale of Perceived Social Support (MSPSS) (Zimet et al., 1988), a 12-item assessment of perceived support from family, a special person, and friends, which has high internal consistency and demonstrated reliability (Canty-Mitchell & Zimet, 2000). Cronbach’s alpha values reflect strong internal consistency at 2 weeks (.94), 6 weeks (.96), 6 months (.96), and 9 months (.95). Participants’ sense of embeddedness and connectedness was assessed using the UCLA Loneliness Scale, which has shown good generalizability and construct validity among adolescents, with acceptable Cronbach’s alpha values at 2 weeks (.70), 6 weeks (.76), 6 months (.78), and 9 months (.78) (Knight et al., 1988). The RDoC acute threat construct (as well as relevant personality influences) was further assessed with the State–Trait Anxiety Inventory for Children (STAI) (Seligman et al., 2004; Spielberger et al., 1973), a widely used measure shown to be sensitive to environmental influences such as treatment (Seligman et al., 2004; Spielberger, et al., 1973). Cronbach’s alpha values showed good internal consistency at 2 weeks (.84), 6 weeks (.85), 6 months (.82), and 9 months (.82).

2.4.8. Response to social exclusion task (Cyberball)

At 9 months post-trauma, participants completed the Cyberball task, a brief (5 min), widely used measure of response to social exclusion (Williams & Jarvis, 2006). Cyberball was the final task of the follow-up assessment. Participants were told that they were participating in an online game with two (virtual confederate) ‘adolescents who have logged on from somewhere else in the world’. Consistent with standard instructions, participants were told that Cyberball is a task to help practise mental visual skills for a ‘later task’. After reading the instructions, participants viewed a screen in which they ‘entered’ a game where the confederate players were represented by avatars who interacted with one another and excluded the participant from virtual play. Participants’ response to the task was assessed using both psychophysiological assessment (described above) and constructs previously found to be influenced by Cyberball, including sense of belonging, control, meaningful existence, self-esteem, and PANAS-X negative affect.

2.5. Recruited cohort

Of a total of 1157 participants screened, 416 did not meet eligibility criteria, 508 declined to participate, and 233 were consented; in total, 191 participants were enrolled and completed at least one follow-up visit in the study (see Figure 1 for the STROBE diagram). Reasons that eligible families provided for declining to participate ranged from being ‘too busy’, to living too far away for follow-ups, to families feeling overwhelmed by medical or logistical aspects of the qualifying traumatic event. In addition, during a period of time in 2020, a subset of families reported concerns about participation in research, given COVID.

Demographically, the sample included almost equal representation of natal males (46%) and females (54%), and was predominantly cisgender, with about 3% of participants identifying as a non-binary gender or not being sure of their gender identity. Participants reported their race and ethnicity, with roughly half of the sample (54%) reporting their race to be White, 30% Hispanic, 28% Biracial, 12% Black/African-American, 3% American Indian/Alaskan Native, and 1% Asian, and 3% of the sample did not wish to respond. Half of the sample (50%) identified as heterosexual or straight, 11% as queer, and 6% as an unlisted identity, and 32% declined to answer (Table 1).

Table 1.

Sample characteristics.

Characteristic	n	%
Sex assigned at birth
Male	88	45.8
Female	103	53.6
Missing	1	0.5
Self-reported gender
Boy/man	87	45.3
Girl/woman	99	51.6
Other	3	1.6
Unsure	1	0.5
Missing	2	1
Sexual orientation
Heterosexual or straight	96	50
Gay or lesbian	9	4.7
Bisexual	11	5.7
Not sure	4	2.1
None of the above	11	5.7
Missing	61	31.8
Ethnicity
Hispanic	57	29.7
Not Hispanic	134	69.8
Missing	1	0.5
Race
American Indian/Alaskan Native	5	2.6
Asian	2	1
Black or African American	23	12
White	104	54.2
More than one race	53	27.6
Missing	5	2.6

As shown in Table 2, just over half of the sample presented in the ED owing to a motor vehicle collision. Other frequent index events included an accident resulting from participating in a sport or other physical activity (e.g. all-terrain vehicle collision, soccer injury), being struck by a motor vehicle while walking or cycling, being physically/sexually assaulted or shot by another individual or group (e.g. wounds resulting from physical altercation at school), or another unintentional injury resulting from daily activities (e.g. slipped on icy stairs).

Table 2.

Sample index trauma type frequencies.

Reason for emergency department visit	n	%
Motor vehicle collision	91	61
Accident during sport/physical activity	39	26
Pedestrian or cyclist struck by motor vehicle	25	17
Gunshot wound and/or other physical/sexual assault	25	17
Other unintentional injury	11	7

Eighty seven per cent of participants were retained at study follow-ups. Retention was unrelated to all demographic variables (all p < .05). Fourteen per cent of the sample met criteria for ASD 2 weeks post-discharge (Table 3). By the 6 week follow-up, 8% reported symptoms meeting PTSD criteria. Eleven per cent met criteria for a current mood disorder and 19% met criteria for a history of mood disorder. Anxiety disorders were identified in 21% of participants, and behavioural disorders (i.e. attention deficit hyperactivity disorder, oppositional defiant disorder, conduct disorder) in 12%. The only substance that participants endorsed using in ways aligned with criteria for a use disorder was cannabis, and 7% met criteria for cannabis use disorder.

Table 3.

Sample psychiatric diagnoses.

	n	%
Week 2	191
Acute stress disorder	26	14
Week 6	191
Mood disorders	60	31
Current major depressive disorder	21	11
Past major depressive disorder	37	19
Persistent depressive disorder	2	1
Anxiety disorders	40	21
Panic disorder	4	2
Agoraphobia	1	1
Evidence of social anxiety disorder	11	6
Phobic disorder	7	4
Evidence of generalized anxiety disorder	17	9
Behavioural disorders	23	13
Attention deficit hyperactivity disorder (ADHD)	22	12
Oppositional defiant disorder	1	1
Substance use disorders	14	7
Alcohol use disorder	0	0
Cannabis	14	7
Trauma and stressor-related disorders		8
Post-traumatic stress disorder (PTSD)	15	8

2.6. Analytic plan

2.6.1. Overview

Analyses will be conducted using R Version 2.2.1 (R Foundation for Statistical Computing, 2005) and Mplus Version 2.2.1 (Muthén & Muthén, 2004). As appropriate, alternate estimators to address expected and observed characteristics (i.e. Poisson distribution, zero-inflation, etc.) can be easily used in the selected software packages. Wherever possible, models will utilize simultaneous estimation to minimize multiple testing and match known covariances among predictors. Given the dangers of post-hoc hypothesis testing in high-dimensional datasets, the interpretation of results will be adjusted using the step-down procedure (Benjamini & Yekuteli, 2001).

2.6.2. Defining/preparing modelled variables

We focus primarily on PTSD symptoms but will assess and model the course of other disorders (depression, anxiety, substance use/abuse) that have been shown to be highly comorbid with PTSD and to increase following trauma. Raw biometric data extracted from the watch include HR and SC as well as time stamps and variables aimed at data cleaning (steps, skin temperature). We will use accelerometer (steps) data to identify periods of increased physical activity to be excluded from composite models. EAR and OSN data are time-stamped and coded using the SECSI system to permit quantification of EAR/OSN interactions. Our latent construct of fear bias is indicated by physiological reactivity during laboratory tasks that present threat cues (i.e. startle, HR/SC during social exclusion, HR/SC during angry faces, attention bias assessed through reaction time and eye-tracking). If analyses do not support the presence of a unitary fear bias construct, we will separately model physiological response (i.e. one’s responsiveness when a threat is perceived) and attention bias (i.e. the attentional resources directed towards tracking potentially threatening cues) assessments (Buck et al., 2013; Mogg et al., 1994). In addition to theory and prior work (Mogg & Bradley, 1999), conventions for latent variable modelling will guide analytic decision making: we will estimate and compare the fit of single-factor and multi-factor models of fear bias, and retain the model with superior fit. Analyses will also consider potentially important covariates (medications, demographics such as gender and race/ethnicity, injury severity, trauma type (e.g. interpersonal trauma versus other), and medical procedures). Covariates will be assessed for variance inflation.

2.6.3. Analytic plan for Aim 1

To test the influence of social context on PTSD development, we capitalize on carefully matched design and theory in which multi-modal data sources are temporally aligned and evaluated through latent growth-curve modelling (LGM), which is ideally suited to examination of: (a) both intra-individual and inter-individual change over time, (b) parallel growth processes, and (c) indirect effects (Selig & Preacher, 2009). To test the relationship between OSN use and later symptoms, we will conduct LGM, with the intercept modelled at the time of trauma and OSN assessments parameterized as time since trauma; ASD and PTSD will be regressed on intercept and slope. We expect to also observe an effect of slope, as we expect that resilient adolescents will show an initial burst of OSN activity post-trauma that will gradually decrease to pre-trauma levels over the course of 2 weeks. To test our hypothesis that adolescents who spend more in-person time with others in positive interactions will evidence fewer symptoms over the course of follow-ups, relative to adolescents who are alone or whose interactions are conflictual and withdrawn, we will examine coded EAR data in the first weeks post-trauma as a predictor of subsequent symptoms over time. To examine pre-trauma social interactions measured using OSN as a predictor of post-trauma interactions, we will model pre-trauma OSN use as a latent summary variable, regressed on intercept and slope of OSN measured post-trauma.

2.6.4. Analytic plan for Aim 2

To test our hypothesis that social interactions (EAR/OSN) will be associated with attenuation of SC and HR, we will adapt parallel LGM procedures to examine the coupling of these outcomes and temporal sequencing (Bauer, Preacher, & Gil, 2006; MacKinnon, Fairchild, & Fritz, 2007; Mallinckrodt, Abraham, Wei, & Russell, 2006). Next, to test the relationship between HR/SC in the first weeks post-trauma and symptoms at follow-up, we will again conduct LGM, with the intercept modelled at the time of trauma and growth parameterized as days post-trauma. We expect to observe significant effects of both intercept and slope, such that elevated acute heart rate and skin conductance will be associated with higher initial posttraumatic stress symptoms as well as continued symptoms over time.

2.6.5. Analytic plan for Aim 3

To assess the course of early HR/SC as associated with subsequent fear biases, we will use LGM to model the course of HR/SC over time, with the 2 week assessment indicated as the intercept and latent fear bias at 6 weeks regressed on the intercept and slope of early biomarkers. The intercept is indicated at 2 weeks, rather than at the trauma, to focus analyses on the course and ‘end’ of the acute period, rather than the beginning of the acute period, during which elevations may be more ‘normative’. The 6 week assessment is selected because it represents the first assessment within the window of diagnosable PTSD. To characterize the expected longitudinal relationship between fear bias, which we expect will maintain symptoms, and PTSD symptoms, we will model a parallel growth process of fear bias and PTSD symptoms over the course of the 6 week, 6 month, and 9 month time-points.

3. Discussion

The multi-method approach described in in the current protocol permits a richer characterization of the within-person and ecological factors influencing the development of PTSD than has been previously possible for a youth sample. This protocol outlines an operationalization of the predominant working framework for PTSD development among adolescents. In doing so, it provides an overview of procedures for the integration of sensitive and fairly novel forms of data (e.g. OSN, EAR). We demonstrate the feasibility of this study, evident in the retention, and the possibility of partnering with young and their families on complex longitudinal studies of the influence of social media on long-term mental health outcomes. The clinical significance of this study is bolstered by participants’ longitudinal engagement.

The present research investigation has numerous important implications for our understanding of the influences on PTSD development on adolescents exposed to trauma. Our measurement of multiple real-time markers allows identification of ‘digital phenotypes’ – specific patterns of OSN use and biometric data – that have the potential to characterize markers of risk as they unfold (Onnela & Rauch, 2016). Findings from this research are expected to further conceptual and phenomenological understanding of PTSD development in adolescents, and the study been designed to provide a rich characterization that may be translated into intervention.

When used ethically and accurately, commercially available wearables may provide an important opportunity for research on trauma response. By using a commercially available watch (as compared with commonly deployed research physiological cardiovascular measures such as a bio-harness), we increased feasibility, maximizing adolescent willingness to wear the tool, minimizing study burden, and decreasing the possibility that the wearable itself will become a traumatic reminder, because adolescents are more likely to ‘forget’ that they are wearing a watch than they would be with a bio-harness. In addition, findings developed using commercially available technology may be more rapidly translated to usable intervention. For example, for smartwatches, a number of relevant applications are already available, including apps for deep breathing and relaxation, as well as apps that can ‘push’ a notification to the wearer under specific conditions, such as if a substantial increase in HR is observed. The use of the same measurement tools in longitudinal cohort studies and in intervention development, as will eventually be used commercially, may accelerate the translation pathway.

OSN use, itself, may also be a target for identification of risk and secondary prevention. Society is becoming increasingly concerned about the negative ramifications of OSNs on young development. Our investigation is granularly assessing the benefits as well as harms of OSNs in the post-trauma period, while maintaining privacy and the highest ethical standards. By incorporating assessment of OSN context prior to the trauma, our research may identify pre-trauma ‘risk’ OSN profiles, much like past research has looked at biomarker risk profiles. It is conceivable that our investigation could lead to a clinical paradigm in which a trauma-exposed patient consents to an automated computerized assessment of PTSD risk based on OSN profiles, at the same time as they are undergoing medical care, and could be offered tailored OSN-based interventions that enable them to better utilize OSN tools to promote resilience (e.g. reminders that arise as needed regarding social engagement, prompts to engage in physiological relaxation to reduce reactive responding) (Kutok et al., 2021; Nugent et al., 2023). OSNs may also be able to be integrated into traditional outpatient psychotherapy in the post-traumatic period (e.g. reviewing interactions with clinician to practise reflection or skills, or for incorporation in trauma narrative or safety plans). Unlike social encounters in daily life that cannot be readily observed by a clinician during a weekly therapy appointment, OSNs may provide a mechanism through which patients can review with their clinician difficult interactions that may have occurred during the week, allowing the clinician to have a richer conversation with the patient about their experience of the interaction, and the reaction of the patient’s social network to the event. Of course, much like biomarkers that go beyond basic clinical care, any decisions about the use of social networking data must lie first and foremost with the patients and their families.

Similarly, EAR audio data could be a tool for psychotherapy in the post-traumatic period, providing objective ecological samples that can be reviewed during therapy to discuss ways in which individuals may better elicit and accept social support and reduce conflict (Alisic et al., 2015). Key audio files could be identified using sensor data, such as the smartwatch methods used here, with clinicians collaboratively reviewing audio samples occurring following elevated physiological activity with patients (Mangelsdorf, Conroy, Mehl, & Alisic, 2019; Mangelsdorf, Conroy, Mehl, Norton, & Alisic, 2020).

3.1. Limitations

We emphasize the importance of maintaining privacy, safety, equity, and patient/family consent in research and intervention efforts that involve collection and coding of social interaction data (such as research using OSN and EAR) (Goldberg et al., 2022; Kaplan et al., 2024). In our experience, the research should be presented as a collaborative venture in which participants may choose what they do and do not wish to share. Although ethical concerns are heightened for research on the post-trauma period, we believe that for all social interaction data collection, both the consent process and every follow-up should present participants with the ability to select interactions (audio samples occurring during a particular time, OSN samples occurring during particular dates or with specific conversation partners) that they are not willing to share with researchers. Furthermore, participants were asked not to wear the EAR device during times when it may have compromised their privacy (e.g. during school, work) or safety (e.g. during contact sports). This trade-off, though ethically necessary, limits data collection during potentially important social contexts for young. It is also important to consider equity and inclusion during both the study design and the coding phase, which inherently involves subjective decisions (Kaplan et al., 2024). For example, our provision of smartwatches to the participants reduces socioeconomic bias in enrolment (compared to a study based on a ‘bring your own device’ model of data collection) (Nagappan et al., 2024). Although we did not integrate a youth advisory board for this investigation, our experiences in this work have led us to begin to incorporate methods that involve working directly with young people in the design, implementation, and dissemination of research findings. In addition, with the rapidly evolving landscape of social media use, platforms and apps that are now widely used were not scraped during data collection, as they were not yet ubiquitous. We suspect this will be an ever-present challenge in social media research, and believe that the current findings retain internal validity nonetheless.

3.2. Public health, scientific, and translational significance

The public health significance of this research is substantial given the negative outcomes of trauma, the increasing attention to the ED as a point for identification of risk and referral, and the inclusion of social context in our digital phenotyping. The scientific significance of this work is its potential contribution to frameworks for understanding how longitudinal social (both in-person and OSN) context and HR/SC in adolescents in the weeks following trauma may impact the development of PTSD and related concerns. The incorporation of laboratory assessments of physiological reactivity and bias, particularly related to social processes, at the 2 week, 6 week, 6 month, and 9 month follow-ups provides rich information about a critical time period in PTSD progression. The concept of a shared neurocircuitry underlying seemingly disparate functions of physiological and social behaviour is consistent with the RDoC’s ‘effort to define basic dimensions of functioning … to be studied across multiple units of analysis, from genes … to behaviors, cutting across disorders as traditionally defined’, and is familiar to affective neuroscience (Afraimovich, Young, Muezzinoglu, & Rabinovich, 2011; Davidson, 2003), which applies designs such as the present investigation to integrate our understanding of social, biological, and psychological processes. The clinical significance of this research is enhanced by the focus on mechanisms that may later become targets for treatment (Bakermans-Kranenburg, Van IJzendoorn, Mesman, Alink, & Juffer, 2008; Beach et al., 2009; Beauchaine, Neuhaus, Brenner, & Gatzke-Kopp, 2008; Brody, Beach, Philibert, Chen, & McBride Murry, 2009; S. Feldstein Ewing, H. LaChance, A. Bryan, & K. E. Hutchison, 2009; S. W. Feldstein Ewing, H. A. LaChance, A. Bryan, & K. E. Hutchison, 2009; Nugent et al., 2023).

Funding Statement

This project was supported by National Institutes of Mental Health [grant number R01MH108641] awarded to NRN. KI's work was supported by Eunice Kennedy Shriver National Institute of Child Health and Human Development [grant number T32HD101392].

Disclosure statement

Dr Armey is a compensated member of Ilumivu’s Scientific Advisory Board. Dr Nugent is an unpaid member of Ilumivu’s Scientific Advisory Board with stock options. No potential conflict of interest was reported by the remaining authors.

Data availability statement

Data are available through the NDA repository Understanding the Interplay of Social Context and Physiology on Psychological Outcomes in Trauma-Exposed Adolescents #2409.