Research paperEvaluation of a virtual agent to train medical students conducting psychiatric interviews for diagnosing major depressive disorders
Introduction
Clinical diagnosis relies on both signs (i.e., externally, observable phenomena - expressions) and symptoms (i.e., patient's subjective complaints - experiences) (Nordgaard et al., 2013) in order to perform semiology1 extraction (i.e., “clinical evaluation of signs and symptoms, leading to the identification of a disorder”) (Micoulaud-Franchi et al., 2018). In most medical specialties, physicians use tools to measure signs (e.g., blood pressure monitor, medical imaging), and clinical interviews to collect subjective symptoms. In the field of affective disorders, however, the vast majority of signs investigated, such as body movements, language and discourse, are expressed as patients progressively disclose their symptoms, so the psychiatrist has to rely on his/her ability to conduct an appropriate clinical interview in order to disentangle and collect both the signs and symptoms of their patients (Shea, 2016; Silverman et al., 2015). In this context, the psychiatric interview should be conversational, contextually adapted and empathic (Nordgaard et al., 2013). Notably, Shea (Shea, 2016) suggests that a first psychiatric interview should follow three main phases:
- 1
The introduction and beginning of the interview, which aims to lower the patient's anxiety of coming to see a psychiatrist, and expose the objectives of the present interview;
- 2
The main part of the interview, during which objectives are to help the patient express his symptoms by guiding him through the different dimensions of depressive disorders;
- 3
The ending of the interview, where the psychiatrist presents the diagnosed disorder and proposes an adapted solution, while taking into account the patient's feelings and representations and giving him hope about future recovery.
Therefore, two major skills have to be acquired by future psychiatrists. First, the ability to extract semiology based on their knowledge of clinical signs and symptoms for each category of mental and affective disorder as listed in the DSM-5 (American Psychiatric Association, 2013). However, as importantly, the psychiatrist needs to create an empathic relationship with the patient (Bhugra et al., 2017; Plakun, 2015) in order to facilitate the procedure. Empathy is arguably the most important psychosocial characteristic of a physician engaged in patient care (Colliver et al., 2010), as it helps build patient trust (Deladisma et al., 2007), increases patient satisfaction and compliance, improves medical care outcomes and may reduce medical malpractice lawsuits (Kim et al., 2004). Based on the literature, we distinguish two types of empathy: verbal empathy, referring to the physician's ability to ‘help the patient express his/her symptoms’ (Shea, 2016); and nonverbal empathy, corresponding to the physician's ability to stay neutral (Nordgaard et al., 2013) and to show empathic listening (Plakun, 2015).
Medical education consists primarily in passive learning through lecture-based classroom and clinical observation, which has demonstrated poor performances in remembering (Tolks et al., 2016). Complementarily, new techniques are now being used to improve students’ empathic skills (for a review, see Batt-Rawden et al., 2013), mainly provided by role play with standardized patients (SPs), i.e. actors trained to act as patients. However, even if these initiatives have been effective in improving medical students’ empathy, they are sometimes not feasible in terms of schedule and resources to train and employ. Additionally, assessment methods need to evaluate students’ abilities to conduct an empathic interview with a patient. New tools are therefore needed to provide future psychiatrists with active, practical and experiential training and assessment while remaining feasible with time and resource constraints, standardized, and common to all medical schools (Bhugra et al., 2017).
In this context, computerized tools are regarded as a promising solution to provide new tools for training and assessment in medical education. Notably, embodied conversational agents (ECAs), defined as “virtual digital representations of a computer interface in the form of human-like faces”(Cassell et al., 2000), are now being developed for use as virtual patients (VPs) in medical training (Cook et al., 2010). Notably, in their recent study, Maicher's team (Maicher et al., 2019) developed a VP to train medical students’ information-gathering skills. Results with 102 students showed that the VP was comparable to human raters to evaluate information-gathering skills. However, the interaction was based on typed text, thereby precluding all forms of nonverbal and empathic interaction. In a randomized controlled trial with 70 first-year medical students, Foster and colleagues (Foster et al., 2016) found that students interacting through a text-based interface with a depressive VP giving feedback about empathic responses were later able to be more empathetic in an interaction with an SP. Another study (Kleinsmith et al., 2015) found that students were more empathetic to VPs than to SPs. They considered that they were being judged less, felt less stressed that they were not dealing with real patients, and had more time to think about their answer. Similarly, in (Deladisma et al., 2007), students felt less nervous than when talking with a real SP after interacting with a life-size VP suffering from abdominal pain. Taken together, these findings show that VPs can provide problem-oriented, standardized, repetitive and safe practice that simulate cases not possible for human actors (e.g., facial paralysis), while providing situations less stressful for students and with no consequences for patients. However, until now, only a few VPs have been developed and tested, they simulate only short and non-realistic (mostly text-based) interviews, and none of them has focused on both semiology extraction and the assessment of empathy during the interview. Therefore, their applicability to training for conducting psychiatric interviews is limited. Despite the high prevalence of depression (Bromet et al., 2011), only one study focused on a VP suffering from this disorder. Moreover, depression symptomatology exhibits both cognitive and motor dimensions (Kaplan and Sadock, 1988), suggesting its appropriateness for simulation. For these reasons, the objectives of this study were to design and validate a realistic psychiatric interview with a VP simulating major depressive disorders, and to assess medical students’ skills in conducting an interview, in terms of semiology extraction of depression and empathic communication.
Section snippets
Participants
Thirty-five students were recruited from June 2016 to July 2017. They all were fourth-year medical students2
Semiology extraction and verbal empathy: scores and errors
Globally, students had very good scores and made few errors (Table 1). Scores were significantly lower for semiology MCQs than for empathy questions (t(68) = 3.489; p < .001). Furthermore, students made significantly more errors during semiology MCQs than in empathy questions (t(68) = 8.064; p <.001).
In addition, results showed significant differences regarding semiology MCQ scores and errors depending on the student's specialty department (i.e., neurology and psychiatry) (Fig. 3), trainees in
Discussion
This study is the first to validate the use of a virtual patient (VP) simulating a realistic psychiatric interview to train and assess medical students’ semiology extraction and empathic skills in the field of affective disorders. The findings are encouraging and pave the way for new training modalities in psychiatric education.
The students managed to interact appropriately with the system, as overall they had good scores and made few errors. Interestingly, while both groups of students showed
Funding
This project was supported by the French State within the framework of the national grants LABEX BRAIN (ANR-10-LABX-43), EQUIPEX PHENOVIRT (ANR-10-EQPX-01), and IdEx Bordeaux (ANR-10-IDEX-03-02).
CRediT authorship contribution statement
Lucile Dupuy: Data curation, Formal analysis, Investigation, Supervision, Validation, Writing - original draft, Writing - review & editing. Jean-Arthur Micoulaud-Franchi: Conceptualization, Investigation, Methodology, Supervision, Validation, Writing - original draft, Writing - review & editing. Hélène Cassoudesalle: Conceptualization, Investigation, Methodology, Validation, Data curation. Orlane Ballot: Data curation, Formal analysis, Investigation, Visualization. Patrick Dehail:
Declaration of competing interest
None.
Acknowledgments
We thank the medical students who took time to test our VP and give their precious feedbacks. We also thank Léa Chapolin, Caroline Louria and Anthony Latrille who helped with the video analysis, Emilien Bonhomme who participated in the development of the virtual patient, and Mr. Ray Cooke for proofreading.
References (39)
- et al.
The WPA- lancet psychiatry commission on the future of psychiatry
Lancet Psych.
(2017) - et al.
Do medical students respond empathetically to a virtual patient?
Am. J. Surg.
(2007) - et al.
Understanding empathy training with virtual patients
Comput. Hum. Behav.
(2015) - et al.
Making psychiatric semiology great again: A semiologic, not nosologic challenge
L'Encéphale
(2018) - et al.
Validation of the French version of the Acceptability E-scale (AES) for mental E-health systems
Psychiatry Res
(2016) Psychotherapy and psychosocial treatment: recent advances and future directions
Psychiatr. Clin.
(2015)- et al.
Impact de la formation théorique et clinique sur les attitudes de stigmatisation des étudiants en médecine envers la psychiatrie et la pathologie psychiatrique
L'Encéphale
(2018) Diagnostic and Statistical Manual of Mental Disorders (DSM-5®)
(2013)- et al.
Teaching Empathy to Medical Students: An Updated, Systematic Review
Acad. Med.
(2013) - et al.
Cross-national epidemiology of DSM-IV major depressive episode
BMC Med
(2011)