Evaluating effectiveness and engagement strategies for asynchronous, messaging, trauma-focused therapy for posttraumatic stress disorder: Study design and methodology for a hybrid effectiveness-implementation randomized controlled trial

Abstract

Posttraumatic stress disorder (PTSD) can be a chronic and debilitating condition with significant individual and societal costs. Despite the availability of effective, trauma-focused treatments like Cognitive Processing Therapy (CPT), access and sustained engagement remain limited due to logistical, financial, and stigma-related barriers. The current study utilized a hybrid effectiveness-implementation randomized control trial (N = 300) to evaluate the clinical effectiveness and feasibility of CPT-Text, an asynchronous, therapist-delivered messaging version of CPT, compared to culturally informed treatment-as-usual (CI-TAU) delivered via secure text on the Talkspace platform. Participants were also randomized to receive either standard engagement reminders (RAU) or a novel engagement incentive (RAU + I), which encourages altruism by offering participants the opportunity to “pay forward” credits for free therapy to another PTSD-affected individual through their continued participation. Primary outcomes include PTSD symptom severity (PCL-5) and engagement (e.g., treatment completion and messaging frequency), with secondary outcomes assessing depression, functioning, satisfaction, substance use. We hypothesize that CPT-Text will outperform CI-TAU in symptom reduction, and that RAU + I will enhance engagement and outcomes. Exploratory analyses will examine individual-level moderators and motivation as a mechanism of change. A novel component of this trial includes development of a large language model–based fidelity assessment tool for CPT-Text, allowing scalable evaluation of treatment delivery. Findings will inform scalable, effective, and accessible PTSD interventions that meet the needs of individuals underserved by traditional mental health care, and provide insight into how human-delivered, technology-mediated therapy can be both clinically robust and broadly accessible. Evaluating Asynchronous Messaging Therapy for PTSD.

1. Introduction

Posttraumatic stress disorder (PTSD) is a prevalent, chronic condition linked to significant distress, functional impairment, and broader public health and economic costs [1,2]. The COVID-19 pandemic has contributed to new cases and worsened symptoms among those with existing mental health issues [3–5], increasing ER visits, suicide attempts, and overdoses [6]. Fortunately, trauma-focused, cognitive-behavioral therapies like Cognitive Processing Therapy (CPT) are well-established, first-line treatments across diverse populations and settings [7,8].

Despite strong evidence for their effectiveness, access to evidence-based treatments (EBTs) for PTSD remains low [9–11]. Barriers include limited availability of trained therapists, transportation, time, cost, stigma, and competing responsibilities [12,13]. In-person sessions during work hours are especially difficult to maintain. Many avoid care due to stigma, long waitlists, or inflexible scheduling. Although trauma-focused PTSD treatments are effective via video conferencing [14–16], access remains limited especially for those without a computer, high-speed internet, or daytime availability.

Online and mobile app interventions offer alternatives to traditional PTSD therapy and can help overcome access barriers. Meta-analyses support their effectiveness, especially for trauma-focused, therapist-guided interventions [14]. However, engagement remains low with usage drops to 3–4 % for standalone apps and about 9 % with peer support [17]. There is a clear need for more engaging, efficient, and acceptable care options.

Asynchronous messaging therapy may help overcome treatment barriers. With smartphones and texting widely used, messaging mental health services are growing [18]. While synchronous messaging has shown effectiveness comparable to in-person therapy [19,20], asynchronous messaging—where responses don’t happen in real time—offers greater flexibility and discretion [21]. As an adjunct to in-person therapy, it improves retention [22]. Therapist-delivered messaging may enhance access and engagement compared to unsupported digital tools [23,24].

Asynchronous messaging therapy can be effective, especially when grounded in evidence-based interventions. In large studies, 66 % of the clients treated for depression and anxiety via asynchronous messaging reported significant improvement [21], and nearly half of those treated for PTSD showed clinically significant improvement [25,26]. In the latter study, the therapy modalities used to address PTSD were at the discretion of the therapist, and individuals could engage with their therapists 5 days per week through video messaging or texting, with texting being far more common. We piloted a structured version of CPT delivered via asynchronous messaging (CPT-Text) [25], comparing 28 CPT-Text participants to a matched comparison sample of 23 receiving treatment-as usual (TAU). CPT-Text led to faster and greater improvements in PTSD and depression, with large between-group effect size. Message volume and video use were similar or lower than TAU, supporting feasibility. While TAU lacks a defined endpoint, CPT-Text’s structured, modular format (63 % completion rate) may enhance engagement by supporting goal completion.

There remains a pressing need to understand how to increase engagement in digital mental health (DMH) to ensure adequate doses of effective interventions. Enhancing motivation for using DMH may be critical for increasing engagement [27]. Individuals need to interact with the platform enough to receive a sufficient “dose” to experience sufficient clinical gains [28,29]. The Behavioral Intervention Technology Model [30] highlights motivation (behavioral intention) as a key mechanism influencing behavior. This is especially relevant for PTSD and depression, which are characterized, and maintained by, high avoidance [28]. Motivation may impact engagement and adherence in DMH interventions, thereby improving clinical outcomes. In our study, participants are randomized to either CPT-Text or a control, and to one of two engagement strategies: reminder as usual (RAU) by the therapist or RAU plus an incentive (RAU + I). In the RAU + I group, participants are told that sustained engagement earns therapy credits for others with PTSD who need financial help. This altruistic incentive condition aligns with research on consumer motivation [31,32] and altruism in PTSD [33]. We will evaluate how these strategies affect engagement and treatment response, and test motivation as a potential mechanism.

1.1. Research objectives and hypotheses

The primary objective of the study is to examine the effectiveness of trauma-focused CPT when delivered by messaging (CPT-Text), compared to a control condition–messaging-based Treatment As Usual Condition guided by culturally informed trauma treatment practices (culturally-informed TAU)–using a randomized, Hybrid Type 1 implementation-effectiveness trial methodology. Messaging-based therapies are being delivered on the Talkspace HIPAA-compliant, secure texting platform. We hypothesize that clients who receive CPT-Text will experience greater symptom change relative to those who receive CI-TAU (Hypothesis 1). We also have an exploratory objective to examine client characteristics as moderators of (a) engagement and (b) treatment outcomes to inform efforts to triage clients into forms of care that are most likely to benefit them.

A secondary objective is to examine the novel method of engagement, RAU + I, compared to RAU. We will compare their impact on treatment response and engagement and examine motivation as a potential mechanism. We hypothesize that clients in the RAU + I condition will experience greater symptom change and engagement than those in RAU (Hypothesis 2). We also aim to explore interactions between treatment and engagement conditions.

A final, exploratory aim of the study is to build and evaluate a computational, language-based model of CPT-Text fidelity to advance efforts to scale fidelity assessment.

2. Materials and methods

2.1. Participants

Participants for this study include 360 individuals over the age of 18 residing in the United States and enrolled as a Talkspace user (www.talkspace.com). All participants must have a Criterion A event as measured by the Life Events Checklist for DSM-5 (LEC-5) [34], significant symptoms of PTSD as evidenced by a score of 33 or above on the PTSD Checklist (PCL-5). Participants must own a personal device for messaging and reside in a state with current study therapist capacity. Based on the focus of the funding mechanism, during the study recruitment phase, potential participants were asked whether their PTSD symptoms began or increased during the COVID-19 pandemic (per self-report), although this is not a basis for exclusion. Participant contact information, residence, and IP addresses provided by participants are verified following best practices in screening out fraudulent responses in internet-based research [35,36].

Exclusion criteria are minimal in order to increase the generalizability of the results: (1) sub-threshold symptoms of PTSD or do not meet criteria for PTSD; (2) acute risk for suicidal thoughts and/or behaviors as measured by the Patient Health Questionnaire-9 (PHQ-9) [37] item 9 and, if required for further screening, the Columbia Suicide Severity Rating Scale Lifetime-Recent Screen (C-SSRS) [38]; (3) positive screening on the electronic screen for psychosis or substance use (Psychosis Screening Questionnaire; [39]) indicating needing prioritization of treatment and/or a higher level of care; (4) non-US residence; (5) invalid phone number; (6) invalid or non-residential mailing address; (7) multiple screening attempts with the same IP address under different aliases; (8) other evidence of identity misrepresentation, such as inconsistent or suspicious contact information that cannot be verified.

2.2. Study design and procedures

This hybrid effectiveness-implementation randomized controlled trial was designed to (1) investigate clinical effectiveness of messaging-based therapies for individuals with PTSD and (2) examine a novel engagement strategy. In this 2 × 2 factorial design, eligible participants are randomly assigned to one of two treatment conditions (CPT-Text or CI-TAU) and one of two engagement conditions (RAU or RAU + I). Participants complete online symptom and functioning assessments at baseline and Week 5, 9, 13, 17, and 25. Those who drop out of treatment continue to receive assessments for inclusion in the intent-to-treat (ITT) analyses. See Fig. 1 for a diagram of participant flow. Participants receive Talkspace’s standard of care, including unlimited messaging with therapist responses occurring during therapist working hours, regular symptom assessments, and engagement reminders (e.g., messages encouraging patients to log on and communicate with their therapist). As per Talkspace standard of care, participants are matched with a therapist licensed in their state and can request to change therapists during the active treatment phase. Participants receive free treatment for the duration of the study.

Fig. 1.

Evaluating asynchronous messaging therapy for PTSD.

2.2.1. Ethical oversight

The study procedures were reviewed and approved by the Institutional Review Boards (IRB) of The University of Texas Health San Antonio and Stanford University. Clinical trial registration was completed at ClinicalTrials.gov (NCT05037175).

2.2.2. Recruitment and screening

The study was conducted in partnership with Talkspace. Participants were recruited via and online platforms including Mental Health America, ResearchMatch, Prolific, Nativve, CliniContact, and ClinicalTrials.gov. Our recruitment targeted diverse trauma-exposed groups including first responders, medical professionals, essential workers during COVID-19, those with limited mental health access, and vulnerable populations. Individuals with any form of Criterion A trauma (e.g., assault, combat, natural disaster, etc.) are eligible. Participants completed an electronic REDCap screening survey which study staff review to assess Criterion A, PTSD symptoms and eligibility. Before randomization, participants must create a Talkspace account. The research team follows up via phone, text, or video to verify identity, device access, and screen for potential fraud.

2.2.3. Randomization

In our 2 × 2 factorial randomized trial, eligible participants are randomly assigned to one of the four conditions (CI-TAU with RAU, CI-TAU with RAU + I, CPT-Text with RAU, CPT-Text with RAU + I) with equal probability of assignment. Randomization occurs after the baseline assessment using a pre-generated listed created with a random number generator.

2.3. Measures

The primary outcome measure is PTSD symptom severity as measured by the Posttraumatic Stress Disorder Checklist for DSM-5 (PCL-5). The PCL-5 is a 20-item self-report measure that evaluates the degree to which individuals have been bothered by PTSD symptoms tied to their most currently distressing traumatic event [34]. The monthly version was administered at baseline and the weekly version thereafter (Table 1).

Table 1.

Schedule of assessment measures.

	Baseline	Week 1	Week 5	Week 9	Week 13	Week 17	Week 25
Demographics	x
DAST-10	x				x		x
AUDIT-10	x
CUDIT	x
Psychosis Screen	x
C-SSRS	x
Behavioral Intention	x		x	x
B-IPF	x		x	x	x	x	x
PCL-5			x	x	x	x	x
PHQ-9	x		x	x	x	x	x
PTCI	x		x	x	x
PREPARE	x
OBTSI	x
CSS	x
SRA-9	x
WAI-SF		x	x	x	x
CHS		x	x	x	x
CMO		x	x	x	x
CSQ			x		x
Treatment History					x	x	x
OBSTI-SF				x		x
AUDIT-C			x	x	x	x	x
CUDIT-SF			x	x	x	x	x

Note. DAST = Drug Abuse Screening Test; AUDIT = Alcohol Use Disorders Identification Test; CUDIT = Cannabis Use Disorder Identification Test; C-SSRS = Columbia Suicide Severity Rating Scale Lifetime Screen; B-IPF = Brief Inventory of Psychosocial Functioning; PCL-5 = PTSD Checklist for DSM-5; PHQ-9 = Patient Health Questionaire-9; PTCI = Posttraumatic Cognitions Inventory; PREPARE = Protocol for Responding to and Assessing Patients’ Assets, Risks, and Experiences; OBTSI = Oppression-Based Traumatic Stress Inventory; CSS = Coronavirus Stressor Survey; SRA-9 = Self-Reported Altruism Scale-9; WAI-SF = Working Alliance Inventory-Short Form; CHS = Cultural Humility Scale; CMO = Cultural Missed Opportunities Scale; CSQ = Client Satisfaction Questionnaire.

Secondary outcomes include depressive symptoms (PHQ-9; [37,40]), psychosocial functioning (Brief Inventory of Psychosocial Functioning [B-IPF]; [41]), alcohol use (Alcohol Use Disorders Identification Test-10 and -C [AUDIT-10, AUDIT-C]; [42]), cannabis use (Cannabis Use Disorder Identification Test and Cannabis Use Disorder Identification Test Short Form [CUDIT, CUDIT-SF]; [43]), and drug use (Drug Abuse Screening Test-10 [DAST-10]; [44]). The Self-Reported Altruism Scale-9 (SRA-9; [45]) was administered at baseline as a potential moderator of engagement for the incentive condition. Client satisfaction with the intervention (Client Satisfaction Questionnaire [CSQ]; [46]) will be measured and compared between the two intervention conditions and free responses will be examined and summarized to examine user perceptions of each messaging therapy condition.

Based on a review of the research literature, we identified other mediators and moderators that might be associated with treatment engagement and outcome include motivation and intention (Behavioral Intention [BI]; [47,48]), posttraumatic cognitions (Posttraumatic Cognitions Inventory [PTCI]; [49]), COVID-19–related stressors (Coronavirus Stressor Survey [CSS]; [50]), social needs (Protocol for Responding to and Assessing Patients’ Assets, Risks, and Experiences [PRAPARE]; [51]), therapeutic alliance (Working Alliance Inventory-Short Form [WAI-SF]; [52]), experiences of discrimination and/or identity-based trauma (Oppression-Based Traumatic Stress Inventory [OBTSI]; [53]), and patient perceptions of therapist cultural humility and missed opportunities to address cultural issues (Cultural Humility Scale [CHS]; [54] and Cultural Missed Opportunities Scale [CMO]; [55]).

2.3.1. Assessing and defining engagement

Our primary engagement outcomes are (1) completion of the intervention (defined as 13 weeks of CPT-Text or CITT, or completion of all CPT Modules which ever comes first, or early achievement of a PCL-5 score < 20, which indicates good end-state functioning [56]) and (2) the number of days the client messages divided by total intervention days (days until treatment completion or 13 weeks, whichever comes first). Previous research has established 8 sessions as an adequate dose of EBT for PTSD [57]; thus, we will also report the proportion of participants who completed 8 or more modules of CPT-Text. We will also examine the total word counts at both the therapist and client level, comparing between conditions. It is possible that the videos, worksheets, and standardization could make CPT-Text more efficient in terms of therapist and client time.

2.4. Treatment conditions

The treatment phase is 13 weeks for both treatment conditions, which allows for a week of onboarding, assessment, and goal setting followed by 12 weeks for active intervention.

2.4.1. CPT-text

CPT-Text is an asynchronous messaging therapy version of CPT, a gold-standard, ~12-session, trauma-focused, cognitive therapy that teaches clients to examine unhelpful beliefs about the trauma as well as beliefs about themselves and the world that were altered as a result of trauma. In CPT-Text, session material is delivered as modules, delivered over days of daily interaction between therapist and participant, with approximately one week of messaging for each traditional CPT session. Each module includes psychoeducation and introduction of a new skill or topic, which builds on the previous information and skills. Participants receive psychoeducation video links via Talkspace and an electronic workbook with handouts, written explanations of the concepts and activities, and CPT worksheets completed electronically and viewable to the therapist. Therapists encourage the participant to complete each module using the provided materials and assist clients in reflecting on their beliefs through Socratic questions and feedback on their worksheets [25,26]. Participants work at their own pace, with therapists sending the next skill or module after the participant practices the previous one.

2.4.2. Culturally informed trauma treatment as usual (CI-TAU)

Culturally Informed Trauma Treatment as Usual (CI-TAU) is a control condition in which participants receive treatment as usual by therapists, who also received training in culturally responsive care. Thus, therapists may select the intervention. Based on previous research and Talkspace’s analysis of treatment as usual, it is likely CI-TAU will include elements of supportive and client-centered interventions, problem solving, and elements from EBTs such as cognitive-behavioral therapy [26].

2.4.3. Therapist training

The Talkspace platform engages a vetted, trained network of over 5000 master’s-level and higher, experienced, licensed, and insured mental health therapists, who provide care for relationship issues, workplace stress, and diagnosed conditions such as depression, anxiety, PTSD, substance use disorders, and most other behavioral health conditions. Talkspace therapists were invited to receive training to serve as therapists for the research study. Therapists with previous training in CPT or an interest in learning CPT were trained in the delivery of CPT-Text and also received training in culturally informed CPT. Therapists without previous training in CPT were provided with five hours of didactic training in culturally responsive care with ongoing consultation and served as therapists in the CI-TAU condition. In total, we trained 40 Talkspace therapists from around the U.S. Therapists were compensated for client care for participation in the study through Talkspace. We monitor treatment and provide consultation opportunities in groups and ad hoc to therapists, but attendance care decisions are ultimately up to the therapists and their participants. Essentially, the study observed routine care among therapists who have been trained in CPT-Text or CI-TAU.

2.5. Engagement conditions

To examine a novel engagement strategy intended to address low levels of ongoing engagement with DMH technologies, client participants are randomized to one of two different strategies: (1) reminders as usual (RAU) on the Talkspace platform or (2) RAU + Retention Incentives (RAU + I).

2.5.1. Reminder as usual

As per Talkspace guidelines, therapists are available to client participants during the business week. Therapists send a personalized message to the client participant who have not engaged or messaged during the business week to encourage them to re-engage.

2.5.2. Retention incentive

Participants are told at baseline that if they message with their therapist regularly each month, a discount will be applied to another study participant or Talkspace user with PTSD. If participants do not re-engage after 24 h from the RAU (3 days without engagement), they receive an automated text-message reminding them to message their therapist regularly (on average, every other business day) in order to facilitate this discount in the next month. This incentive was designed to tap into people’s sense of altruism, which may be more effective than reminders as usual for people with PTSD.

2.6. Data analytic strategy

2.6.1. Analysis of study aims

To assess Aim 1, which is to test the hypothesis that participants in CPT-Text will experience greater symptom relief than those in CI-TAU, we will assess the Intent-to-Treat (ITT) effect of the intervention (CPT-Text vs. CI-TAU) on the severity of PTSD symptoms based on mixed effects growth modeling. The change (slope) in PTSD symptoms (PCL-5) will be modeled as the dependent variable, treating the intervention assignment status (CPT-Text vs. CI-TAU) as the predictor. All individuals with outcomes measured at one or more assessment points will be included in our analysis. We will also explore various baseline patient characteristics, particularly trauma type, PHQ-9 score, age, and education level, as potential moderators of treatment effect on PTSD. For this investigation, we will employ the MacArthur framework for moderator analysis [58] embedded in mixed effects modeling, following the eligibility and analytical criteria for determining moderators.

For Aim 2, we will assess the ITT effect of engagement boosting strategies (RAU vs. RAU + RAU + I) on PTSD. Additionally, we will examine the interaction effect, which will show whether the intervention effect (CPT-Text vs. CI-TAU) can be significantly enhanced by adding a novel engagement strategy. We will employ the MacArthur approach for mediation [58] as well as modern causal mediation approaches that utilize potential outcomes and propensity scores (e.g., [59–62]). The engagement measures are (1) completion of the intervention (defined as 13 weeks of CI-TAU or CPT-Text, or early achievement of PCL-5 score < 20, which indicates good end-state functioning [63]) and (2) the proportion of total intervention time (until treatment completion or 12 weeks, whichever comes first,) for which the client messaged their therapist at least once every other business day. The second strategy will be used to examine a measure of engagement that is observed only under the CPT-Text condition, i.e., completion of CPT-Text intervention modules. In this investigation, we will be able to assess the intervention effect for those who would show sufficient engagement in offered treatments (e.g., 8 or more modules out of 12 total). Specifically, we will employ the method called compiler average causal effect (CACE) estimation to identify the ITT effect for the subpopulation of interest – those who would show enough engagement when offered CPT treatment modules (e.g., [[64–66]). We will conduct various sensitivity analyses, including the use of bounds and alternative identifying assumptions (e.g., [67–70]) to support the validity of the CACE estimate as causal effect estimate. The role of engagement in Aim 2 will be further explored by incorporating the baseline motivation and engagement boosting strategy (with and without incentives) as potential moderators, and by incorporating the change in motivation (BI) and altruism as mediators.

2.7. Fidelity assessment

For our exploratory aim of building and validating a computational model of treatment fidelity for CPT-Text, we will adapt the Therapist Adherence and Competence (TAC; [71]) protocol to label individual therapist messages from messaging transcripts with core CPT skills. This adaptation allows for message-level fidelity measurement, enabling fine-grained temporal analyses (i.e., evaluating the effect of the order in which CPT skills are delivered). Items from the TAC will be revised, added to, and clustered into higher-level categories, then refined through iterative transcript rating and consensus discussions. Once interrater reliability is achieved, a subset of transcripts (10 %) will be rated by human coders to assess fidelity and train a large language model to perform fidelity rating by classifying therapist messages into CPT skill category. We will ensure balanced representation across therapists and engagement levels, and selectively annotate additional messages to ensure sufficient examples of each label. A separate test set will be used to evaluate model performance, with F1 scores ≥0.40 indicating adequate accuracy. Finally, we will report CPT-Text fidelity using both the 10 % human-rated fidelity dataset and for 100 % of the transcripts rated using the computational model of treatment fidelity

2.7.1. Assessment of CI-TAU fidelity

Due to the importance of treatment receipt, a construct of fidelity that has been less studied, we will examine the Cultural Humility Scale [[54] and Cultural Missed Opportunity Scale [55] scores to determine whether participants perceived their therapists as providing culturally responsive and appropriate treatment. To determine the most frequent activities and therapist behaviors in the CI-TAU conditions, secondary analyses and papers will focus on automated coding of common therapeutic elements present in all transcripts, using frameworks of therapist behaviors drawn from multiple theoretical orientations and taxonomies of therapist activities (e.g., [72,73].

2.8. Power analysis

We estimated power focusing on our primary aim to test the intention to treat (ITT) effect of treatment strategy (CPT-Text or CI-TAU) on PTSD symptoms measured by the PCL-5. Given this one clear primary hypothesis, we consistently used the nominal significance level (alpha = 0.05, two-tailed) without adjusting for multiple testing. We estimated power taking into account expected attrition over time (about 40 % by end of treatment or 13 weeks). Our preliminary study showed a sizable intervention effect over TAU (Cohen’s d = 0.84) on PTSD. Given the preliminary nature of this information that is not based on an RCT, we used a more conservative effect size of d = 0.5 for our power estimation for the primary ITT effect of intervention on PTSD. For the secondary ITT effect of engagement strategy (i.e., with vs. without incentive), we assumed a more modest effect size of d = 0.3. For the interaction effect (intervention by incentive), we assumed d = 0.6 (assuming a small intervention effect (d = 0.2) without incentive and a large effect (d = 0.8) with incentive). We assumed intra-class correlation of ICC = 0.6 across repeated measures based on the established test-retest reliability of 0.8–0.9 of the PCL-5 [74]. Under this scenario, with initial N = 360, assuming about 40 % attrition by week 13 [25], the estimated power to detect the main effect of intervention (primary, CPT-Text vs. CI-TAU) on PTSD is 0.99 (alpha = 0.05, two-tailed). The estimated power to detect the main effect of adding incentives is 0.84 (alpha = 0.05, two-tailed). The estimated power to detect the intervention by incentive interaction effect is 0.89 (alpha =0.05, two-tailed). We assume that about 60 % of intervention participants will show sufficient engagement. Given the assumed ITT effect of d = 0.5, the projected intervention effect for this sufficiently engaging group is d = 0.83 (CACE = ITT/engagement rate = 0.5/0.6 = 0.83. See [66]). Under this scenario, the estimated power to detect the intervention effect for those with sufficient engagement (CACE) is 0.99 (alpha = 0.05, two-tailed). For our exploratory analyses of various potential moderators and mediators, we will put more emphasis on gathering information on hypothesis generation and clinical significance (effect size) rather than on making inference based on statistical significance (i.e., p-value).

3. Discussion

In this study, we are investigating the effectiveness of a version of CPT that was adapted for asynchronous messaging, as delivered by professional therapists. We compared it to messaging therapy as usual, provided by trauma specialists who had been trained in culturally responsive care CI-TAU. Additionally, we are comparing different engagement strategies to better understand approaches that may support adequate engagement in asynchronous, technology-based treatment. Results will inform efforts to provide accessible forms of interaction with licensed therapists, for individuals who are not able to access, or not interested in traditional forms of trauma therapy.

A recent sequential randomized, multiple assignment randomized trial comparing messaging-based therapy to videoconferencing therapy for depression and anxiety indicated that individuals engaged over more weeks in messaging based therapy, and that differences in treatment outcomes were not found between the two treatment formats [75]. Our study will expand efforts to understand the impact of messaging-based therapy by comparing a guideline-consistent PTSD treatment, adapted for messaging format, to an active control. With high rates of dropout among trauma-exposed individuals and individuals who use technology to support their mental health, our findings can inform efforts to identify treatment formats for PTSD that are engaging, effective and accessible.

Consistent with a hybrid-I implementation and effectiveness study design, this study primarily focuses on effectiveness and will investigate treatment fidelity and client satisfaction with CPT-text as compared to treatment as usual. Comparison of client satisfaction measure will allow assessment of acceptability of a structured protocol as compared to a less structured treatment as usual. Additionally, planned exploratory work to develop and validate a computational model of fidelity to CPT-Text will allow for scalable fidelity assessment and a fine-grained exploration of the intensity and sequencing of key CPT elements. This represents a novel contribution to the field, as fidelity assessment is traditionally labor-intensive, costly, and difficult to implement reliably at scale. By leveraging large language models to label therapist messages with unique elements of CPT, our approach could streamline fidelity monitoring across full trials or clinical programs. Beyond scalability, computational fidelity assessment may open new avenues for understanding evidence-based treatments by enabling fine-grained analyses of how specific skills, delivered in particular sequences, influence outcomes across different patient subgroups. This could ultimately help yield insights into what aspects of evidence-based treatments works for whom and under what circumstances, and aid in the refinement of existing – and development of new – evidence-based treatments for mental disorders.

We note that this study was designed and started before the broad public release of generative AI-based chatbots, and with the exception of automated onboarding messages and reminder messages from the messaging platform, all interactions on the platform occur between the participants and a human therapist.

Data availability

Data will be made available on request.