Design
The study had a mixed design, with comparisons of before and after the intervention and between the intervention and control groups (figure 1).
Population
The population was recruited during the 2019-2020 academic year among the 131 fifth-year undergraduate students at the University of Versailles Saint-Quentin-en-Yvelines-Paris Saclay (the year of compulsory psychiatric training).
Intervention
The intervention consisted of one day (8 hours) of teaching of psychiatry by simulation with a simulated patient. The scenarios, decided within a group of 10 hospital-university teachers (from university fellow to professor in general medicine, child and adolescent psychiatry, or adult psychiatry), had to be: 1) addressed in the curriculum of the official national program (41), 2) a pathology frequently encountered in general practice, and 3) realistically performed by a team of psychiatry teachers not trained in acting (eliminating the scenario of schizophrenia, which is challenging to play (15)). The four scenarios presented a drug suicide attempt in the context of borderline personality disorder and alcohol addiction assessed in the emergency room by a psychiatrist and bereavement associated with post-traumatic stress disorder, hypomania, and a refusal to go to school by a 14-year-old adolescent assessed in a general-practice setting.
Each simulation session included a briefing (10 min), the simulation (10-20 min), a structured debriefing (45 min), and a theoretical synthesis slide presentation (20 min). Learners were divided into groups of eight (2 actively participating, 6 watching the live video broadcast in an adjacent room). Three teachers were involved, one playing the role of the patient, one as a potential facilitator, and one staying with the learners.
No randomization was used to assign the intervention or control group status. Students in the intervention group were recruited voluntarily and accepted an optional teaching unit on the condition that they actively participated in one scenario. The control group, recruited from non-participating students, received the same usual psychiatric instruction as the simulation group in the form of a compulsory two-day interactive seminar with the technique of the flipped classroom (42). All fifth-year undergraduate students were provided with the pedagogical written content of the simulation sessions to ensure that any differences between the groups were related to the simulation teaching technique itself.
Measures
1) Knowledge
Theoretical knowledge was measured using multiple choice questions (MCQs) three times: for all students, two months before the simulation teaching (45 questions before the compulsory psychiatry seminar) and two months after (50 MCQs during the psychiatry examination, covering the entire national program of psychiatry for the university grade) and for the simulation group, before and after the teaching (28 questions). All scores were scaled from 0 to 20.
2) Confidence (Supplementary Information 1)
Confidence was assessed by the specifically created Confidence in Psychiatric Clinical skills Questionnaire (CPCQ): 12 items, rated on a four-level Likert scale, explored confidence in theoretical knowledge, clinical skills (clinical reasoning and psychiatric interviewing), communication and interpersonal skills (with the patient, the patient’s proxies, and other professionals), and the management of psychiatric disorders. The individual mean score was used in the analyses.
The change in professional practice was evaluated with one question: “How much do you think this teaching of simulation psychiatry will improve your future practice? “ It was rated on a 4-level Likert scale ranging from “very unimportant” (coded 1) to “very important” (coded 4).
3) Satisfaction (Supplementary Information 2 and 3)
General satisfaction was rated out of 10. A 10-item questionnaire, rated on a four-level Likert scale, explored various aspects of satisfaction, such as the preference for simulation over another pedagogical modality, the perceived realism of the situation, the importance of being actively involved, etc. In addition, learners who underwent a clerkship in psychiatry were asked to compare it to the simulation. Questions about the scenarios and free comments were collected.
Satisfaction with the briefing and debriefing was assessed using the student version of the DASH(43). This scale, with excellent internal consistency (0.82-0.95) (44–46), explores the climate, structure of the debriefing, ability to engage in exchange, and strengths and areas for improvement. The mean across all items (6 overall assessments, 23 behavioral assessments) was used.
Statistical analysis
Comparisons between simulation and control groups
First, age and participation in a clerkship in psychiatry (a potential confounding factor for confidence (39)) were compared between the two groups using chi² tests and scores on the pre-requisite exam using Student’s t-test. Analyses of covariance (ANCOVA) was then carried out with the mean CPCQ score and the psychiatry final exam score as dependent variables, the group as the independent variable, and the covariates that differed significantly between the two groups (clerkship in psychiatry).
Pre/post-simulation comparisons
The average CPCQ and knowledge test scores just before and after simulation were compared using paired sample Student t-tests. Satisfaction was measured post-stimulation.
Psychometric characteristics of the CPCQ scale
Construct validity was explored by exploratory factor analysis using oblim rotation and maximum likelihood factorization as the factorization method. Two criteria were used to determine the number of significant factors: first, Catell’s scree test, i.e. factors present to the left of the eigenvalue curve deflection (47), and second, Kaiser’s criteria, i.e. factors for which the eigenvalue is > 1(48). The internal consistency of each identified factor was evaluated using Cronbach’s α coefficient(49), with an acceptability threshold set to 0.7 (50). These analyses were carried out on the largest sample for the same time of measurement (final exam) and by bringing the two groups together.
Test-retest reliability was assessed by the intra-class correlation coefficient (ICC), calculated using a mixed model with a random double effect. It was defined as poor for an ICC < 0.4, acceptable between 0.4 and 0.59, good between 0.6 and 0.74, and excellent between 0.75 and 1(51). The two times of measurement chosen to calculate it were those for which the least possible change was expected, i.e. just after the simulation and two months later.
Number of required subjects
The number of required subjects was calculated for knowledge (score out of 20 on the usual psychiatric examination). According to the results of the previous year, the average score was 13.3, with a standard deviation of 1.9. To show a mean difference of 2 points with an alpha risk of 5% and a statistical power of 90% required at least 19 subjects per group.
Ethics statement
The research was authorized on 20/12/2010 by the Ethics Committee of the University of Paris-Saclay (CER-Paris-Saclay-2019-061). All participants signed written and informed consent.