Psychiatric Annals

CME Article 

Digital Health and Artificial Intelligence for PTSD: Improving Treatment Delivery Through Personalization

Matteo Malgaroli, PhD; Thomas Derrick Hull, Mphil; Katharina Schultebraucks, PhD


Posttraumatic stress disorder (PTSD) is a condition characterized by symptom heterogeneity and persistent avoidance patterns. These characteristics hinder the identification of candidates for early intervention and identification of treatment recommendations appropriate for an individual person's needs. Moreover, treatment barriers (particularly in areas with enduring disaster or conflict) further limit the delivery of behavioral interventions to ameliorate PTSD symptoms. To overcome these limitations, digital health approaches offer innovative ways for treatment delivery that extend beyond imitating face-to-face modalities. We discuss how digital health technologies have been increasingly used to bridge clinical needs. Another advantage of digital delivery is the automatic gathering of clinical data. We discuss how these clinical information pipelines can be analyzed using artificial intelligence to enhance prediction of PTSD, monitoring, and ad hoc treatment interventions. [Psychiatr Ann. 2021;51(1):21–26.]


Posttraumatic stress disorder (PTSD) is a condition characterized by symptom heterogeneity and persistent avoidance patterns. These characteristics hinder the identification of candidates for early intervention and identification of treatment recommendations appropriate for an individual person's needs. Moreover, treatment barriers (particularly in areas with enduring disaster or conflict) further limit the delivery of behavioral interventions to ameliorate PTSD symptoms. To overcome these limitations, digital health approaches offer innovative ways for treatment delivery that extend beyond imitating face-to-face modalities. We discuss how digital health technologies have been increasingly used to bridge clinical needs. Another advantage of digital delivery is the automatic gathering of clinical data. We discuss how these clinical information pipelines can be analyzed using artificial intelligence to enhance prediction of PTSD, monitoring, and ad hoc treatment interventions. [Psychiatr Ann. 2021;51(1):21–26.]

Exposure to potentially traumatic events (PTEs), including combat, accidents, or natural disasters, is intrinsic to the human condition. Over the course of a lifetime, 70% to 90% of people will experience one or more PTEs.1 Although multiple prospective studies have shown that psychological resilience is the norm after a PTE, a significant minority of survivors will suffer from poor adjustment and functioning difficulties, including post-traumatic stress disorder (PTSD).2 According to the Diagnostic and Statistical Manual of Mental Disorders, fifth edition (DSM-5),3 hallmark features of PTSD include intrusions and/or flashbacks about the PTE, internal/external avoidance of PTE reminders, nightmares and other re-experiencing, and cognitive and hyperarousal symptoms. Although structured behavioral interventions for PTSD are available, clinical trials comparing their effectiveness are often characterized by high nonresponse and dropout rates, with marginal differences among conventional forms of treatment.4 A factor strongly contributing to such an outcome is the heterogeneity of PTSD; in an effort to encompass all different traumatic stress presentations, the DSM-5 diagnostic algorithm results in more than 636,120 possible diagnostic combinations meeting PTSD criteria.5 People presenting different symptoms may, therefore, be offered the same treatment, with varying results. All PTSD symptoms also uniquely interact and mutually reinforce one another in a way that is not fully captured by the PTSD diagnostic label.6 In addition, PTSD has unclear boundaries with commonly associated maladaptive responses including complicated grief,7 which may require additional targeted interventions.8 Outcomes to the same PTE can also be heterogenous9 and dependent on comorbid medical outcomes;10 therefore, differences in the trauma event could call for different therapeutic approaches. In sum, the small differences observed between conventional and PTE-focused therapy outcomes could be attributed to poor specificity of treatment given the wide variety of possible symptom clusters.

Barriers to treatment access constitute additional factors contributing to poor PTSD outcomes. For many survivors, the immediate aftermath of a PTE is the only point of contact with the health care system,11 and currently no PTSD treatment recommendation algorithm for PTSD is available in routine clinical care. Such an absence leaves a large portion of people unserved who would otherwise benefit from early interventions. After this initial window, hallmark features of PTSD such as avoidance, fear, and guilt may further prevent symptomatic survivors from seeking treatment, ultimately leading to increased functional impairment. Geographical barriers are also serious impediments to receiving PTSD treatment. Areas with elevated PTSD rates are often afflicted by combat, civil war, or natural disaster, with each carrying a degree of risk that makes it arduous for clinicians to intervene where they are needed the most. In addition, almost one-quarter of US veterans live in rural areas,12 with limited availability for combat-exposure targeted treatment. Given these factors, a clear need emerges for new forms of improved treatment delivery, with the goal of creating a setting where diffused and personalized PTSD interventions are possible.

Psychiatry has been increasingly looking at technological solutions to solve the challenges of diagnostic and treatment heterogeneity. Digital health interventions in particular have been proposed to improve access to mental health care, particularly given the availability of technological media for delivery (such as computers, smartphones, and other mobile devices). The initial momentum of digital health has been dramatically increased during the coronavirus disease 2019 (COVID-19) pandemic, given the safety and travel concerns of in-person care.13 Technological delivery holds promise to overcome treatment barriers and enhance our current ability to detect PTSD symptoms and personalize treatment. Therefore, the purpose of this article is to examine some of the past, current, and possible future trends of digital health approaches to PTSD. Given the vast extent of the subject, we decided to focus on two aspects: (1) digital delivery of treatment, and (2) treatment personalization (also known as personalized medicine).

Digital Delivery

Digital delivery of mental health treatment is a branch of telemedicine focusing on at-distance interventions, clinical services, and symptoms monitoring. It is also referred to as telehealth, telemental health, teletherapy, and/or telepsychiatry. The digital aspect of delivery has grown as modern hardware allows for multiple forms of interaction: for example, a smartphone can be used for telephonic care, messaging, video calls, sending recordings, and completing self-report measures. New technologies are also constantly evolving and changing in the way they interface with our lives (eg, smartwatch). As such, digital delivery conducted by clinicians is divided into two main groups that depend on the temporal characteristics of patient-clinician interactions (rather than on the interaction medium): synchronous for real-time interactions (eg, live videoconferencing, telephone), and asynchronous for interactions that allow a delay in the response (eg, messages, audio/video recordings). Mobile technologies also increase access to self-help software (eg, mindfulness apps), some of which are assisted by artificial intelligence (ie, a chatbot). Importantly, even these modalities have become increasingly interrelated, given the possibility for the latest digital health platforms to provide multiple modalities: for example, a self-help app offering an asynchronous coach, or therapy platforms allowing both synchronous and asynchronous treatment with the same clinician. A summary of the interplay between these modalities for PTSD is shown in Figure 1. Examples of multimodality applications include Talkspace (, Betterhelp (, Ableto (, Ginger (, and 7 Cups (, although only some of them reported data related to PTSD.

Proposed digital health pipeline for prediction and treatment of posttraumatic stress disorder symptoms. EMR, electronic medical record.

Figure 1.

Proposed digital health pipeline for prediction and treatment of posttraumatic stress disorder symptoms. EMR, electronic medical record.

Although “at-distance” treatment has been recently increasing in popularity, it has been available for more than 60 years. The first applications of synchronous telepsychiatry were between two Nebraska hospitals in the 1950s, which was picked up and further developed at Massachusetts General Hospital and similar hospital groups.14 After a period of telephone and fax as the primary telehealth technologies, the diffusion of the internet in the 1990s allowed for broader implementations, particularly within the Veterans Affairs (VA) health system. Crucial characteristics of the VA, which justified and propelled the development of a telehealth program, were the ability to operate federally beyond state license limitations, combined with the need to reach a largely rural population.12 Given their early investment in telehealth, the VA was the first to produce large-scale studies to evaluate its efficacy, with results assessing almost 100,000 patients showing positive outcomes.15 Since then, dozens of randomized clinical trials have demonstrated the noninferiority and equivalence of synchronous delivery. The substantial evidence from these studies allowed for a relatively rapid transition for mental health care into full digital delivery during the COVID-19 pandemic.13 Such a feat was possible due to (1) the already available implementation of tele-health channels in routine care (although originally as adjunct technology), and (2) the promulgation of emergency laws to ease regulations around digital delivery. The American Psychiatric Association designed a toolkit for telepsychiatry containing up-to-date clinical, research, and legal information.16

When discussing the large body of research on the effectiveness of digital delivery, it is crucial to note that most studies focused on synchronous modalities, and videoconferencing in particular. In other words, they analyzed whether a consultation-liaison or psychotherapy could be administered remotely, without major changes to the format of traditional services. Asynchronous delivery, on the other hand, has required a greater modification of the clinical framework and changes in format, resulting in it mostly being relegated as an adjunct to synchronous treatment with good results.17 In regard to PTSD, studies have shown the efficacy of conducting written exposure and cognitive restructuring asynchronously.18 These approaches were primarily self-guided, placing the burden of producing written content on the patient and making it difficult to return feedback or provide human support. Currently, investigators are examining whether asynchronous delivery could be beneficial as a primary vehicle of patient-therapy interactions, particularly using multimedia messaging. In asynchronous messaging therapy,19 patients write (or send audio/video messages) to their therapists at their convenience, and therapists reply to patient messages during pre-established times each day. This approach alters the treatment frame from a once-a-week 50- minute block into many shorter blocks over the course of the week. There are several advantages to using messaging for treatment: (1) it increases ease of access by overcoming geographic and scheduling barriers; (2) patients can write to their therapist at any moment, enabling them to capture their experience as it has just happened rather than after a day or several in retrospect; (3) the greater frequency of therapeutic contact over the course of a week helps to keep the change process at the front of the patient's mind, and (4) it decreases wait times to complete treatment substantially, often down to just a couple hours. Preliminary studies on this medium for a large cohort of patients with anxiety and depression showed symptom remission consistent with treatment as usual.20 Similarly, patients diagnosed with PTSD that were administered asynchronous messaging therapy showed comparable symptoms amelioration.21 These results need to be further confirmed through randomized controlled trials but hold promise of more widespread access to the treatment of PTSD. Another important step forward for the integration of asynchronous therapy will be to design interventions that take advantage of the unique characteristics of messaging. Ideally, these interventions should start from already manualized/established PTSD interventions, such as cognitive processing therapy, and adapt them for messaging. This translation would also benefit from the data gathered from the messages log or transcript to build a base to inform treatment.

No overview of digital health applications for mental health would be complete without mentioning the ever-increasing abundance of self-help mental health apps. This digital field is also wide in its scope, with apps focusing on behavior change (eg, sleep quality), relaxation interventions (eg, mindfulness), emotion regulation and interpersonal skills exercises, and even chatbots that teach cognitive skills (such as WoeBot). Such is the interest in these apps that the VA designed one specifically for post-traumatic stress, called PTSD Coach ( Despite their number, only a few of these apps have robust research backing their effectiveness.22 This crucial fact, often underestimated by the general public, can also hinder therapists trying to make recommendations to patients. As the field has evolved, a consensus has slowly been reached on the need to study these applications in a manner that merges the public and private interests of mental health software.

Individualized Treatment

The analysis of PTSD using artificial intelligence is a surging field, thanks to its promise of identifying heterogeneity and integrating multimodal information.23 Due to their broad reach, digital health approaches offer the opportunity to improve PTSD treatment through the application of computational methods on their large datasets. The information pipeline for digital health can be harnessed in two ways: (1) identifying who is in need of treatment (either in the aftermath of trauma, or due to relapse) and (2) personalizing and titrating treatment to individual needs (ie, what works for whom).

Recent advances in algorithmic prediction showed that it is possible to identify PTSD outcomes prospectively, using in-depth biological multi-omics24 as well as routinely collected electronic medical record (EMR) data.11 Early identification of PTSD risk holds promise to take advantage of the “golden hour” after trauma, with its increased opportunities for fear extinction and memory reconsolidation.25 However, these studies used hospital data for prediction, which was collected only shortly after the traumatic event and then again later if the patient returned. As such, these data can only identify deterioration when the person self-refers to treatment. Data gathered more consistently after hospital discharge without requiring interaction from the user would increase coverage and enhance predictive models. One such possibility is the use of digital phenotyping, defined as the collection of information on personal characteristics by means of smartphone data proxies.26 Numerous features indicative of potential illness can be extracted from a smartphone (eg, typing speed), its incorporated devices (eg, geolocation, accelerometer, microphone), or the audio/video/text content loaded through the device. These features reflect behavioral, cognitive, and emotional characteristics and their variation through time. One recent study identified digital biomarkers that accurately classify provisional PTSD and depression diagnosis.27Figure 2 shows potential sources of specific data and their potential use as psychometric sources. Importantly, this approach can identify individual baselines for later comparison, as opposed to comparing individual data to group means. Salient information for PTSD would be related to sleep quality (for nightmares and intrusions), geolocation (for avoidance), and possibly the accelerometer (for agitation). With the right model, this information could be used to alert people (and their health care provider) of worsening symptoms and to offer “justin-time” treatment specifically tailored to the symptoms detected and their acuity. For example, a patient with PTSD experiencing worsening sleep could be offered sleep hygiene information to help improve falling and staying asleep or receive a check-in by their health care provider to determine if treatment is needed for nightmares or hypomanic deprivations (in rarer cases). Despite this potential, there are nontrivial difficulties in analyzing these digital phenotyping data, such as mitigating differences between devices, generational differences in use of digital technologies, and differences in smartphone use. Moreover, there are still no well-established mappings between these digital phenotypes and psychiatric traits. To do so, a steeper integration would be needed between digital phenotyping devices and medical health data. Currently, the integration between PTSD digital health data and EMR only exists as proof of concept.28 If such a system was implemented, multiple health care systems and patients could integrate different sources of information to improve the identification of digital phenotypes predictive of trauma reactions. This would allow treatment to move seamlessly between different tiers of intensity, depending on clinical needs and acuity of symptoms (Figure 1).

Digital phenotyping for psychiatry. Different components indicate candidate sources of information on affect, behavior, and cognition.

Figure 2.

Digital phenotyping for psychiatry. Different components indicate candidate sources of information on affect, behavior, and cognition.

Artificial intelligence methods can also be used to provide modeling of treatment components with different possible levels of analysis. The first level would consist of individualized intervention selection, predicting who is most likely to respond to trauma-focused therapies compared to those who will not or who are at greater risk of dropout. To compute this initial treatment recommendation, an analysis would integrate multimodal information about symptoms, digital phenotypes, EMR trauma event descriptions, and other behavioral markers for people known to respond well to certain types of therapies. The treatment regimen could then be actively switched based on in vivo treatment response, such as reducing or increasing the intensity of the therapeutic engagement, depending on individualized needs. Adjusting the treatment as described here is already the norm in clinical practice but having an algorithm to help generate information when the patient is outside of the session or hospital setting could help make treatment adjustments more timely. In addition, the implementation of such a decision-making process can organize reactions from individual cases into a shared knowledge base of treatment efficacy keyed to individual differences in the patient. Such generalization of knowledge is hard to obtain outside of a research setting, given differences in how the experience of individual practitioners is recorded.

The second level of analysis is the content of the treatment itself, made possible by delivering the treatment through a digital interface. Therapy process data can be used as a source to inform natural language processing models of psychotherapy, for both patients and clinicians. Algorithms to detect risky behavior and relapse would be a helpful first step, and one such implementation has been deployed in a clinical practice setting.29 More in-depth models could then identify trauma-specific narratives and assess how a patient reacts to interventions targeting them (eg, cognitive reframing). The analysis of these narratives can also be geared toward the mechanisms of change hypothesized by different evidence-based treatments for PTSD. For example, the identification of “stuck points,” or the “whys” of what happened may help further tailor interventions from cognitive processing therapy. Clinicians can also benefit from algorithms to automatically analyze their interaction with patients as well as provide a form of supervision. In this case, the model would analyze the factors shared among different types of psychotherapy, such as the warmth of the therapist or how safe the patient feels with the therapist. Computational models that study therapeutic content have already been developed,30 and they are able to segment an unstructured session into labeled sub-components (eg, psychoeducation). These models can help push forward our understanding of specific elements in psychotherapy to achieve more granularity in model parameters that attempt to predict which patients could benefit from particular approaches to psychotherapy. At the broadest level, these models could help in distinguishing patient preferences between directive versus supportive therapy styles. Then, once in treatment, models can parse specific interventions and the impact they have had on the patient, such as whether they agree with and follow-up on what was suggested or whether they shift topics or disagree and resist aspects of the treatment. Digital health platforms offer the possibility of understanding the psychotherapy process in a way that was simply time and cost-prohibitive in the past. As noted, early successes in this area suggest great promise in using current modeling techniques for enhancing treatment.


This article offered an overview of current and future applications of digital health delivery for PTSD, specifically (1) how digital treatment delivery can be accomplished through several different modalities, often interchangeably; (2) how individual characteristics can be gathered and used to guide how to offer, select, and titrate treatment; and (3) how these individual traits can be collected automatically and passively from the digital device used for the intervention or from EMRs. Work to date suggests that there are significant inefficiencies in simply deploying evidence-based packages to everyone without regard for their particular needs and presentation. This issue is further complicated by not having an early PTSD detection system broadly implemented in routine care. Digital health can fill these gaps by offering improved detection of probable PTSD, treatment delivery to overcome access barriers, and individualized treatment that effectively tailors intervention to the patient and patient-therapist dyad. It is worth emphasizing that people struggling with PTSD tend to be more difficult to reach relative to other disorders. Digital health can help address mental health needs in locations that are hard to access, such as refugee camps, areas close to combat zones, and the rural areas to which many US veterans return. Despite these strengths, it is important to point out that we are still in the early stages of applying computational methods for the detection and treatment of PTSD. Preliminary results, however, are encouraging. A digital health strategy that has the unique characteristics of disseminating treatment, aggregating treatment data for rapid analysis, and offering feedback in vivo could increase therapy success and outcome. The added flexibility of digital media can target multiple patient presentations to individualize treatment, ultimately improving the quality of life for people affected by PTSD.


  1. Shalev AY. Post-traumatic stress disorder. Disorder takes away human dignity and character. BMJ. 2001;322(7297):1301. doi:10.1136/bmj.322.7297.1301 [CrossRef] PMID:11403051
  2. Galatzer-Levy IR, Huang SH, Bonanno GA. Trajectories of resilience and dysfunction following potential trauma: a review and statistical evaluation. Clin Psychol Rev. 2018;63:41–55. doi:10.1016/j.cpr.2018.05.008 [CrossRef] PMID:29902711
  3. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders. 5th ed. American Psychiatric Publishing; 2013.
  4. Steenkamp MM, Litz BT, Marmar CR. First-line psychotherapies for military-related PTSD. JAMA. 2020;323(7):656–657. doi:10.1001/jama.2019.20825 [CrossRef] PMID:31999301
  5. Galatzer-Levy IR, Bryant RA. 636,120 Ways to have posttraumatic stress disorder. Perspect Psychol Sci. 2013;8(6):651–662. doi:10.1177/1745691613504115 [CrossRef] PMID:26173229
  6. McNally RJ, Robinaugh DJ, Wu GWY, Wang L, Deserno MK, Borsboom D. Mental disorders as causal systems: a network approach to posttraumatic stress disorder. Clin Psychol Sci. 2015;3(6):836–849. doi:10.1177/2167702614553230 [CrossRef]
  7. Malgaroli M, Maccallum F, Bonanno GA. Symptoms of persistent complex bereavement disorder, depression, and PTSD in a conjugally bereaved sample: a network analysis. Psychol Med. 2018;48(14):2439–2448. doi:10.1017/S0033291718001769 [CrossRef] PMID:30017007
  8. Simon NM, Hoeppner SS, Lubin RE, et al. Understanding the impact of complicated grief on combat related posttraumatic stress disorder, guilt, suicide, and functional impairment in a clinical trial of post-9/11 service members and veterans. Depress Anxiety. 2020;37(1):63–72. doi:10.1002/da.22911 [CrossRef] PMID:31916660
  9. Malgaroli M, Galatzer-Levy IR, Bonanno GA. Heterogeneity in trajectories of depression in response to divorce is associated with differential risk for mortality. Clin Psychol Sci. 2017;5(5):843–850. doi:10.1177/2167702617705951 [CrossRef] PMID:29034135
  10. Schultebraucks K, Wen T, Kronish IM, Willey J, Chang BP. Post-traumatic stress disorder following acute stroke. Curr Emerg Hosp Med Rep. 2020;8(1):1–8. doi:10.1007/s40138-020-00206-5 [CrossRef]
  11. Schultebraucks K, Shalev AY, Michopoulos V, et al. A validated predictive algorithm of post-traumatic stress course following emergency department admission after a traumatic stressor. Nat Med. 2020;26:1084–1088. doi:10.1038/s41591-020-0951-z [CrossRef] PMID:32632194
  12. Holder KA. Veterans in rural America: 2011–2015. Accessed December 3, 2020.
  13. Shore JH, Schneck CD, Mishkind MC. Telepsychiatry and the coronavirus disease 2019 pandemic-current and future outcomes of the rapid virtualization of psychiatric care. JAMA Psychiatry. 2020;77(12):1211. doi:10.1001/jamapsychiatry.2020.1643 [CrossRef] PMID:32391861
  14. Baer LD, Elford R, Cukor P. Telepsychiatry at forty: what have we learned?Harvard Rev Psychiatry. 1997;5(1):7–17. doi:10.3109/10673229709034720 [CrossRef]
  15. Godleski L, Darkins A, Peters J. Outcomes of 98,609 U.S. Department of Veterans Affairs patients enrolled in telemental health services, 2006–2010. Psychiatr Serv. 2012;63(4):383–385. doi:10.1176/ [CrossRef] PMID:22476305
  16. American Psychiatric Association. Telepsychiatry toolkit. Accessed December 9, 2020.
  17. Aguilera A, Bruehlman-Senecal E, Demasi O, Avila P. Automated text messaging as an adjunct to cognitive behavioral therapy for depression: a clinical trial. J Med Internet Res. 2017;19(5):e148. doi:10.2196/jmir.6914 [CrossRef] PMID:28483742
  18. Knaevelsrud C, Maercker A. Internet-based treatment for PTSD reduces distress and facilitates the development of a strong therapeutic alliance: a randomized controlled clinical trial. BMC Psychiatry. 2007;7(1):13. doi:10.1186/1471-244X-7-13 [CrossRef] PMID:17442125
  19. Hull TD, Mahan K. A study of asynchronous mobile-enabled SMS text psychotherapy. Telemed J E Health. 2017;23(3):240–247. doi:10.1089/tmj.2016.0114 [CrossRef] PMID:27797646
  20. Hull TD, Malgaroli M, Connolly PS, Feuerstein S, Simon NM. Two-way messaging therapy for depression and anxiety: longitudinal response trajectories. BMC Psychiatry. 2020;20(1):297. doi:10.1186/s12888-020-02721-x [CrossRef] PMID:32532225
  21. Malgaroli M, Hull TD, Wiltsey Stirman S, Resick P. Message delivery for the treatment of posttraumatic stress disorder: longitudinal observational study of symptom trajectories. J Med Internet Res. 2020;22(4):e15587. doi:10.2196/15587 [CrossRef] PMID:32347814
  22. Torous J, Andersson G, Bertagnoli A, et al. Towards a consensus around standards for smartphone apps and digital mental health. World Psychiatry. 2019;18(1):97–98. doi:10.1002/wps.20592 [CrossRef] PMID:30600619
  23. Schultebraucks K, Galatzer-Levy IR. Machine learning for prediction of posttraumatic stress and resilience following trauma: an overview of basic concepts and recent advances. J Trauma Stress. 2019;32(2):215–225. doi:10.1002/jts.22384 [CrossRef] PMID:30892723
  24. Schultebraucks K, Qian M, Abu-Amara D, et al. Pre-deployment risk factors for PTSD in active-duty personnel deployed to Afghanistan: a machine-learning approach for analyzing multivariate predictors [published online ahead of print June 2, 2020]. Mol Psychiatry. doi:10.1038/s41380-020-0789-2 [CrossRef]
  25. Vermetten E, Zhohar J, Krugers HJ. Pharmacotherapy in the aftermath of trauma; opportunities in the ‘golden hours’. Curr Psychiatry Rep. 2014;16(7):455. doi:10.1007/s11920-014-0455-y [CrossRef] PMID:24890991
  26. Onnela J-P, Rauch SL. Harnessing smart-phone-based digital phenotyping to enhance behavioral and mental health. Neuropsychopharmacology. 2016;41(7):1691–1696. doi:10.1038/npp.2016.7 [CrossRef] PMID:26818126
  27. Schultebraucks K, Yadav V, Shalev AY, Bonanno GA, Galatzer-Levy IR. Deep learning-based classification of posttraumatic stress disorder and depression following trauma utilizing visual and auditory markers of arousal and mood [published online ahead of print August 3, 2020]. Psychol Med. doi:10.1017/S0033291720002718 [CrossRef] PMID:32744201
  28. Open mHealth. Case study: post-traumatic stress (PTSD). Accessed December 3, 2020.
  29. Bantilan N, Malgaroli M, Ray B, Hull TD. Just in time crisis response: suicide alert system for telemedicine psychotherapy settings [published online ahead of print June 19, 2020]. Psychother Res. doi:10.1080/10503307.2020.1781952 [CrossRef] PMID:32558625
  30. Lee F-T, Hull D, Levine J, Ray B, McKeown K. Identifying therapist conversational actions across diverse psychotherapeutic approaches. In: Proceedings of the Sixth Workshop on Computational Linguistics and Clinical Psychology. Association for Computational Linguistics; 2019:12–23. doi:10.18653/v1/W19-3002 [CrossRef]

Matteo Malgaroli, PhD, is a Fellow, Department of Psychiatry, NYU Grossman School of Medicine. Thomas Derrick Hull, MPhil, is the Research Director, Talk-space, Inc.; and a Doctoral Candidate, Department of Clinical Psychology, Teachers College, Columbia University. Katharina Schultebraucks, PhD, is the Director of Computational Medicine and Artificial Intelligence and an Assistant Professor in Behavioral and Cognitive Sciences, Department of Emergency Medicine, Columbia University Medical Center.

Address correspondence to Matteo Malgaroli, PhD, Department of Psychiatry, NYU Langone Medical Center, One Park Avenue, 8th Floor, New York, NY 10016; email:

Disclosure: Matteo Malgaroli is a consultant for Groop Internet Platform. The remaining authors have no relevant financial relationships to disclose.


Sign up to receive

Journal E-contents