A Therapy System for Post-Traumatic Stress Disorder Using a Virtual Agent and Virtual Storytelling to Reconstruct Traumatic Memories
- 1.4k Downloads
Although post-traumatic stress disorder (PTSD) is well treatable, many people do not get the desired treatment due to barriers to care (such as stigma and cost). This paper presents a system that bridges this gap by enabling patients to follow therapy at home. A therapist is only involved remotely, to monitor progress and serve as a safety net. With this system, patients can recollect their memories in a digital diary and recreate them in a 3D WorldBuilder. Throughout the therapy, a virtual agent is present to inform and guide patients through the sessions, employing an ontology-based question module for recollecting traumatic memories to further elicit a detailed memory recollection. In a usability study with former PTSD patients (n = 4), these questions were found useful for memory recollection. Moreover, the usability of the whole system was rated positively. This system has the potential to be a valuable addition to the spectrum of PTSD treatments, offering a novel type of home therapy assisted by a virtual agent.
KeywordsE-health Mental health Virtual agent Relational agent Conversational agent Virtual reality Post-traumatic stress disorder Therapy system Ontology Question system
Post-Traumatic Stress Disorder (PTSD) is a mental disorder following one or more traumatic experiences. It is characterized by recurring intrusive memories, avoidance of reminders of the trauma and a persistent negative mood . Cognitive Behavioral Therapy (CBT) with exposure is one of the most widely used treatments for PTSD. It relies on active recollection of the memory of the trauma to reduce the automatic fear response and facilitate cognitive restructuring [2, 3, 4]. Despite the existence of well-documented treatment, many PTSD patients don’t seek help. Stigma on mental health-care is high, especially amongst veterans , and issues such as travel times and cost can form further barriers to care. A stand-alone home-therapy system can therefore fill an important gap by providing a treatment which is easily accessible, privacy sensitive and cost-effective. Although many e-solutions for mental health are being developed, fully autonomous systems offering exposure therapy for PTSD are rare. The multi-model memory restructuring (3MR) system for home therapy for PTSD is one such system [6, 7]. In this paper we present 3MR version 2 (3MR_2), which incorporates several new elements such as a virtual agent and a question system for trauma recollection.
The original 3MR system was designed for in-clinic use, where the patient works together with a therapist in a face-to-face setting. Its goal was to support this therapy by facilitating trauma recollection and storytelling. The system was developed using expert input and included three main functionalities, namely a timeline, diary and 3D world editor. Memories could be added to the timeline to form an overview, while pictures, text and maps could be added to a diary, and a basic 3D version of the memory could be created in the 3D editor [6, 7]. The patient could work on these environments with the therapist, but also at home as part of homework assignments. By creating a visual representation the therapist would also get a better understanding of the patient’s experiences. This combination of exposure platforms was novel to 3MR, but several other technology-driven methods exist for exposure therapy. The most common is virtual reality (VR), which has had some very promising results [8, 9]. The main drawback of traditional VR therapy for PTSD is that the virtual environments are pre-created, and therefore difficult to match with personal memories. Although feasible for a specific group of veterans, most other patient groups generally do not share similar memories. In order to better personalize VR therapy, several platforms have been developed wherein therapists can build a virtual environment for their patients [10, 11, 12]. However, the original 3MR system shifts this task from therapist to patient, changing the main therapeutic component from the experience to the creation of a virtual world, which requires triggering the memory in an active way. As patients create their own 3D world, it facilitates the creation of a very personal autobiographic virtual environment. This shift also makes these environments suitable for home-use, as well as use in a stand-alone therapy where no therapist is present. The new 3MR_2 system expands on the concept of personal memory recreation by further developing both the diary and the 3D editor.
As 3MR_2 is designed for home-use instead of the original in-clinic use with a therapist, two additional main functionalities are necessary. Firstly, because the system is fully autonomous, it requires some form of procedure to ensure patient safety . With the rise of technology in mental-health care, new ethics guidelines also need to be in place. Patient safety is a particular concern for systems that display some level of automation. Many systems have safety checks not in the system itself, but in the procedure surrounding its use. These checks can take the form of exclusion criteria, but also regular email or phone contact with a clinician [14, 15, 16, 17]. In this way, it is still a human who provides the safety support. An alternative is to include all safety checks in the system itself without any human in-the-loop. These checks can take the form of questionnaires and crisis management options within the system . A combination of these two solutions is to facilitate monitoring by a clinician through the system. In this situation, the system itself monitors the patient but a therapist uses the information gathered to ensure patient safety. An example of this situation is given by Robinson et al., where a clinician monitors distress scores in an application for students at risk for suicide . A similar approach is taken by the 3MR_2 system; questionnaire scores can be monitored by clinicians, who make the call to interfere if patient safety is in question.
Secondly, a home-therapy system requires some form of interaction and guidance. The first goal of this interaction is to inform the patient and provide the rationale behind the assignments. The second goal is to assist the patient with memory recollection in a personalized way. Virtual agents have been gaining popularity in health applications as a way to add a social and personal aspect to systems. The addition of virtual agents has been shown to have a positive effect on attention , adherence [20, 21] and likability [22, 23] of applications. To assist with memory recollection during exposure, some knowledge of traumatic events is necessary. Ontologies provide a good way to add domain knowledge to computer systems [24, 25]. Through combining multiple-choice and open questions, knowledge can be gathered from the patient while still retaining a natural interaction between patient and system . A combination of these paradigms has been shown to result in a question system which can elicit greater detail in the responses . In the 3MR_2 system, a virtual agent is added to provide assistance and personalization, employing an ontology and a structured dialogue to pose the correct questions while assisting in memory recollection.
In the rest of this paper, we will present the 3MR_2 system in more detail. We will first expand on the monitoring function and the additions to the diary and virtual environment already present in the original 3MR. Secondly, we will describe the virtual agent and specifically the question system it employs to assist the patient during exposure. This paper will conclude with the results of a usability study of the system with former PTSD patients.
The 3MR_2 system offers therapy for PTSD. It includes the exposure environments introduced in the original 3MR, namely a timeline with memories which can be both described in a digital diary and re-created in a 3D WorldBuilder. The 3MR_2 system introduces monitoring within the system via questionnaires, improvements to the virtual environment design and a virtual agent to guide users through therapy and assist in memory recollection. It is specifically aimed at either victims of childhood sexual abuse (CSA) or war veterans. Differences consist of wording in some texts (e.g. ‘When you think back on your deployment’ vs. ‘When you think back on the period of the abuse’), the possible content of the virtual environment (e.g. models of tanks in the war version, children’s beds in the CSA version) and the concepts in the ontology on which the question system is based.
The WorldBuilder is 3D tool in which patients can build a virtual environment, recreating their memories. Where the original 3MR only used a birds-eye view, two perspectives have been added to the new WorldBuilder. Using a collection of 3D–models, patients can make a top-down sketch of their recollections. In addition to this top-down edit view, patients can view the scene from a birds-eye perspective, which allows for zooming and viewing the scene from different angles. When building the scene, the patients can select 3D models from a menu on the right and drop them in the scene, rotating and moving them to put them in place. The interface is designed such that people with basic computer skills can use the tool. A brief instruction video is included as well, to further familiarize users with the interface. The CSA version is developed for indoor scenes, the scale of items reflecting the size of a large to smaller room. The version for veterans on the other hand is designed for outdoor scenes. Both versions have a selection of textures for the walls and ground, including both indoor textures such as wooden or carpet floors, and outside textures such as dirt and grass. Additionally, buildings are offered. The CSA version uses European-style buildings specifically, while the veteran version also includes a broad range of eastern-styled buildings. Similarly, the CSA version includes regular vehicles such as a car and bike, while the veteran version includes a range of army vehicles such as trucks and tanks. Regarding furniture, the CSA version is equipped with a range of objects to build bedrooms, studies and bathrooms, while the veteran version only includes basic furniture. The human models in both versions come in different ethnicities, both including a range of male, female and child models. The veteran version also includes a selection of Dutch army models. Finally, both versions are equipped with a range of general objects such as toys and books for the CSA version, and crates and roadblocks for the veteran version. Aside from some additions in the model library, the new 3MR_2 system has several other novel functions. These are the function to watch the environment through the eyes of the different people in-scene, and the option to create different scenes of the same memory. These scenes can be used to represent a timeline of events.
The virtual agent is not present while patients work on the 3D WorldBuilder, but does have a couple of functions related to this environment. First, if patients wish to take a break during the exposure they can select the option to have a relaxation exercise that is led by the virtual agent. Additionally, the virtual agent asks a number of questions after the 3D world has been finished. Several questions are general and focused on the experience, such as what the patient hears and feels when looking at their world. However, the virtual agent also notices some significant events in the 3D environment. It notices when objects move between scenes, when people models are laying down instead of standing, and when explosion models are added. On these occasions, it specifically asks the patient what this event meant for them.
The goal of exposure within PTSD therapy is that patients confront their memories and experience that thinking about these moments is possible. It is important that people think back in detail, so personalized and detailed questions are very helpful. To do this, the virtual agent employs an ontology-based question system. Past research has shown that such an ontology-based question system is able to elicit more detailed descriptions in memory recollection . The ontology in the 3MR system represents knowledge about traumatic memories, differing slightly for either the veterans or CSA victims. Four different versions exist of the military ontology, one for Afghanistan, one for Bosnia, one for Libya and one not related to a specific deployment. These locations are chosen to best represent the Dutch missions.
Usability was evaluated in two stages. An initial usability test with healthy participants was performed for the virtual agent and diary environment. Based on this test, improvements were made to the system. An additional study was done with former PTSD patients studying both the usability of the system and its usefulness for recollecting traumatic memories.
First usability test
Three healthy participants were recruited for an initial usability test, all were researchers or students at the Computer Science department, two had a background in psychology, one in computer science. They performed the first therapy session, in which a positive memory is described. This session did not yet include questions from the question module or the 3D WorldBuilder. No explicit instructions were given. Participants were asked to ‘think aloud’, and an experimenter was present to note all comments and what went wrong. Based on these usability tests, small improvements such as button placement were made to the system. Two instructional videos were made using the usability input, one describing the general system, and one specifically for the 3D WorldBuilder.
Second usability test
A second usability test was performed with former PTSD patients. Its goal was threefold, firstly to study the general usability of the system and its components. The second goal was to study how much the system elements contributed to therapy according to users. The final goal was to study how useful and appropriate the questions generated by the question module were in recollecting traumatic memories. The design of this study was approved by the ethics committee of Delft University of Technology.
4 participants were recruited via practicing therapists. All participants had in the past followed therapy for PTSD. Participants 1 and 2 were war-veterans (both male), participants 3 and 4 had experienced childhood sexual abuse (both female).
The first measure was how useful the questions generated by the system were for remembering the trauma. All questions were rated on an analog scale ranging from works against recollection to helps a lot recollecting. The center point was marked as question has no effect. All questions were rated immediately after asking. The second measure was how useful and understandable the functions in the program were. Two questions were posed on an analog scale, firstly asking how useful the function was from detrimental to helps a lot. The second asked how understandable the function was, ranging from confused me to very understandable. The final questionnaire was the Dutch version System Usability Scale , applied to the whole 3MR program, answered on a 5 pt. Likert scale.
All statistical analyses were carried out with R version 3.3. Before the analysis, the usefulness scores for the questions were transformed to range from −50 to 50, with 0 as the neutral point so deviation from 0 could be tested. Multilevel analyses taking participant as a random intercept were conducted on the usefulness scores for the questions, the usefulness scores of the system components and the usability scores of the system components. The analysis of the usefulness of the questions only included a fixed intercept. In the analysis on the usefulness and usability of the system, system component was included as a fixed effect to study if this factor influenced the result.
Results & discussion
The analysis of the usefulness scores revealed that on average, participants found the system useful (Mean = 16.32, SD = 16.17, F(1,20) = 13.98, p = 0.001). No difference was found between the different system components (F(5,15) = 1.34, p = 0.30). However, ratings varied significantly between participants (SD random intercept = 6.12, 95% CI [1.54, 24.36]). A closer inspection showed that only participant 2 found functions to be detrimental to the therapy, namely the virtual agent and the instruction video. This might partly be explained by the fact that he had worked with a system without either virtual agent or question module in the past, and was therefore used to working with the diary alone. This might have resulted in impatience with the virtual agent and questions. Nevertheless, this preference for a system with less guidance might also exist in people without prior experience with 3MR.
The usability scores reveal a similar pattern to the usefulness scores. On average, the system was rated as well usable (Mean = 21.98, SD = 20.71, F(1,20) = 7.26, p. = 0.014). No significant difference was found between system components (F(5,15) = 0.61, p = 0.69), while participants did significantly vary in their ratings (SD random intercept = 14.96, 95% CI [6.78, 32.99]). The rating on the System Usability Scale ranged between 73 and 75 for participants 1, 2 and 3, which can be labeled as above average. The exception was participant 3 who gave a rating of 55. This is probably caused by a bug only this participant experienced in the 3D WorldBuilder (the scale of the 3D models was wrong).
Conclusion & discussion
In this paper we described the 3MR_2 system, a therapy system for PTSD patients. The system contains two exposure environments, a digital diary and a 3D WorldBuilder in which memories can be recreated. During a 12-session therapy, a virtual agent guides and assists patients with their therapy tasks, employing an ontology-based question module. A human therapist is involved only to monitor progress.
Initial evaluations revealed that the system was usable by both non-patients and former PTSD patients. These evaluations did reveal small usability concerns, which were then resolved. They also exposed some differences in personal preferences, one participant strongly preferred working at his own pace without much guidance. The current system is less appropriate for patients with such a working style, which should be kept in mind for future use. The questions generated by the question module were deemed useful. The evaluations presented in this paper were only concerned with usability and the appropriateness of the generated questions, but did not look into therapeutic effectiveness. A benchmark study is currently being set-up to test if the 3MR_2 system is successful in significantly reducing PTSD symptoms.
Although the 3MR_2 system is specifically designed to treat PTSD, components are also relevant for other domains. Firstly, providing safety and reducing human resources are very important in e-mental-health [34, 35, 36]. The 3MR_2 gives an example of limited human monitoring, which achieves both goals. Secondly, the exposure environments present in 3MR_2 provide a novel view on technology-assisted exposure. Many studies have been done with VR environments for exposure therapy , but the concept of patients themselves recreating these worlds is novel, and might solve the difficulty of building worlds relevant to different patients . Although embedded in a home-therapy in 3MR_2, both the diary and the 3D WorldBuilder could also be used as tools in regular therapy. Thirdly, the 3MR_2 system incorporates a virtual agent that can pose questions aimed at memory recollection. Although exposure to memories is specific to PTSD therapy, other health applications for this technology do exist. One example is expressive writing, a therapeutic tool aimed at writing about negative memories [37, 31]. Another possible application might be found in the field of life review therapy, wherein reminiscence is used to alleviate mental health symptoms, for instance in older adults with depression .
The 3MR_2 system provides opportunities to reduce the barriers to care. However, it is not suitable for all PTSD patients. Depression is a common comorbid disorder to PTSD, and suicide rates among PTSD patients are high . Because patients work alone with the 3MR_2 system, it might be less suitable for patients with a history of suicidality. Similar concerns arise for patients which have substance problems , or which severely dissociate . Another limitation to the 3MR_2 system is that patients cannot follow therapy while ignoring the virtual agent. Some people prefer more guidance than others, and some might not respond well to following a set course. Where a therapist might be able to steer the patient back on track, this is more challenging for a virtual agent.
Despite these challenges, we believe that the 3MR_2 system is a valuable addition to the spectrum of PTSD treatments. Through the use of new technologies, it offers a novel type of therapy which is convenient to patients and costs very little in therapist resources. Given the societal impact of PTSD, it may have a great positive effect on society.
This work is part of the programme Virtual E-Coaching and Storytelling Technology for Post-Traumatic Stress Disorder, which is financed by the Netherlands Organization for Scientific Research (pr. nr. 314-99-104).
- 1.American Psychiatric Association, Diagnostic and Statistical Manual of Nental Disorders, Fifth Edition, American Psychiatric Association, 2013.Google Scholar
- 6.van den Steen, M., Brinkman, W.-P., Vermetten, E. and Neerincx, M.,Design and Usability Evaluation of a Multi-Modal Memory Restructuring System for the treatment of combat-related PTSD," in ECCE 2010 Workshop on Cognitive Engineering for Technology in Mental Health Care and Rehabilitation, 2010.Google Scholar
- 7.Brinkman, W.-P., Vermetten, E., van den Steen, M., and Neerincx, M., Cognitive Engineering of a Military Multi-Modal Memory Restructuring System. J. Cyber Ther. Rehabil. 4(1):83–99, 2011.Google Scholar
- 9.Motraghi, T.E., Seim, R.W., Meyer, E.C., and Morissette, S.B., Virtual Reality Exposure Therapy for the Treatment of Posttraumatic Stress Disorder: A Methodological Review Using CONSORT Guidelines. Journal of Clinical Psychology. 0(0):1–12, 2013.Google Scholar
- 10.Baños, R. M., Guillen, C. B. V., García-Palacios, A., Quero, S., Bretón-López, J. and Alcañiz, M., An adaptive display to treat stress-related disorders: EMMA's world, British Journal of Guidance and Counselling, 2009.Google Scholar
- 12.Riva, G., Carelli, L., Gaggioli, A., Gorini, A., Vigna, C., Algeri, D., Repetto, C., Raspelli, S., Corsi, R., Faletti, G., and Vezzadini, L., Neuro VR 1.5 - A free virtual reality platform for the assessment and treatment in clinical psychology and neuroscience. Studies in health technology and informatics. 144:57–60, 2009.PubMedGoogle Scholar
- 15.Klein, B., Mitchell, J., Abbott, J., Shandley, K., Austin, D., Gilson, K., Kiropoulos, L., Cannard, G., and Redman, T., A therapist-assisted cognitive behavior therapy internet intervention for posttraumatic stress disorder: Pre-, post- and 3-month follow-up results from an open trial. Journal of Anxiety Disorders. 24(6):635–644, 2010.CrossRefPubMedGoogle Scholar
- 16.Kordy, H., Backenstrass, M., Hüsing, J., Wolf, M., Aulich, K., Bürg, M., Puschner, B., Rummel-Kluge, C., and Vedder, H., Supportive monitoring and disease management through the internet: An internet-delivered intervention strategy for recurrent depression. Contemporary Clinical Trials. 36(2):327–337, 2013.CrossRefPubMedGoogle Scholar
- 19.Ortiz, A., del Puy Carretero, M., Oyarzun, D., Yanguas, J.J., Buiza, C., Gonzalez, M.F., and Etxeberria, I., Elderly users in ambient intelligence: Does an avatar improve the interaction? In: Lecture notes in computer science, 2007.Google Scholar
- 22.M. M. Morandell, A. Hochgatterer, S. Fagel and S. Wassertheurer, Avatars in assistive homes for the elderly a user-friendly way of interaction?, in 4th symposium of the workgroup human-computer interaction and usability Engineering of the Austrian computer society, 2008.Google Scholar
- 23.N. Yee, J. N. Bailenson and K. Rickertsen, A meta-analysis of the impact of the inclusion and realism of human-like faces on user experiences in interfaces, in CHI 2007, 2007.Google Scholar
- 25.Noy, N. F. and McGuinnes, D. L., Ontology development 101: A guide to creating your first ontology, 2001.Google Scholar
- 26.S. A. Friederichs, A. Oenema, C. Bolman, J. Guyaux, H. M. van Keulen and L. Lechner, Motivational interviewing in a web-based physicalactivity intervention: Questions and reflections, Health Promotion International, 2013.Google Scholar
- 27.M. L. Tielman, M. van Meggelen, M. A. Neerincx and W.-P. Brinkman, An ontology-based question system for a virtual coach assisting in trauma recollection, in Int. Conf. on Intelligent Virtual Agents, 2015.Google Scholar
- 28.F. Weathers, B. Litz, T. Keane, P. Palmieri, B. Marx and P. Schnurr, The PTSD checklist for DSM-5 (PCL-5), 2013.Google Scholar
- 30.Kaplan, D.M., Smith, T., and Coons, J., A validity study of the subjective unit of discomfort (SUD) score. Measurement and Evaluation in Counseling and Development. 27:195–199, 1995.Google Scholar
- 33.J. Brooke, SUS: A “quick and dirty” usability scale, P. Jordan, B. Thomas, I. McClelland and B. Weerdmeester, Eds., Taylor & Francis, 1996, pp. 189–194.Google Scholar
- 36.Donker, T., Blankers, M., Hedman, E., Ljótsson, B., Petrie, K., and Christensen, H., Economic evaluations of Internet interventions for mental health: a systematic review. Psychological Medicine:1–20, 2015.Google Scholar
- 39.M. Pompili, L. Sher, G. Serafini, A. Forte, M. Innamorati, G. Dominici, D. Lester, M. Amore and P. Girardi, Posttraumatic stress disorder and suicide risk among veterans, The Journal of Nervous and Mental Disease, 2013.Google Scholar
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.