Computer-based Virtual Environment Simulations for Differential Diagnosis in Medical Education


Computer-Based Virtual Environments, Differential Diagnosis, Medical Education, Situated Learning, Preparation for Future Learning, Problem-Solving Before Instruction


Differential diagnosis (DD) is generally defined as “the distinguishing of a particular disease or condition from others that present similar clinical features” (“Merriam-Webster (Dictionary),” 2019). DD management  assigns symptoms to diagnoses and diseases, which allows targeted therapies, disease prognoses, and patient care organization (Battegay, 2017). DD construction is a highly-demanding cognitive task where many factors influence the medical decision-making. Furthermore, an accurate diagnostic performance depends on the diversity and number of disease patterns a physician is bearing in mind (Eva, 2005; Kassirer, 2010). The underlying mental process of DD is clinical reasoning (CR) which describes the thinking and decision-making process associated with clinical practice (Higgs & Jones, 2008; Thampy et al., 2019). CR is an essential skill that must be applied in every moment of patient attendance (Norman, 2005) and is fundamental for a timely diagnosis of disease. It includes the core dimensions of knowledge, reflective inquiry, and metacognition (Higgs & Jones, 2008). Consequently, CR It is a complex cognitive process that uses formal and informal thinking strategies to gather and analyse patient information, evaluate the significance of this information and weigh alternative actions (Simmons, 2010).

Empirical studies reveal that undergraduate medical students struggle on counseling patients in the clinical practice of medicine (Prince et al., 2005) and that beginning clinical students cannot construct a DD the way more experienced physicians do (Benbassat & Bachar-Bassan, 1984). To cope with their inexperience, clinical students often conduct more comprehensive physical examinations and tests than needed for a correct diagnosis. Additionally, this goes with a higher error rate in CR (Nendaz & Perrier, 2012). Evidence suggests that the most common medical error results from inadequate CR (Graber et al., 2005). Finally, this can lead to more harm and discomfort to the patients and more costs for the health system (Schwartz et al., 2012). However, the struggling of students to perform well in clinical practice and to generate an adequate DD may not primarily be due to a deficit of knowledge. It rather is a problem of the capability to transfer this acquired knowledge to clinical practice. This lack of transfer from the university setting to clinical practice in medical education may stem from current methods of instruction (Collard et al., 2016; Norman, 2009). In such settings the focus is on first providing direct instruction on basic knowledge without an adequate attention to situate this knowledge in disciplinary practice. Additionally, providing training in CR as early as possible in medical education could improve reasoning skills in future doctors because it provides a scaffold for future learning, whereas retraining reasoning can be challenging (Phillips, 1995; Ypinazar & Margolis, 2006).

Accordingly, to overcome this problem I propose that (a) the acquisition of clinical knowledge and CR skills must be situated in disciplinary practice and (b) learning must be situated in an actual disciplinary problem.


Whilst we acknowledge that there are several options to integrate situated learning, we aim to explore the use of medical computer-based virtual environment (CVE) simulations with virtual patients. By setting differential diagnosis as the primary anchor for learning, we will start the learning process with an actual medical problem. CVE scenarios with simulated patients can represent aspects of situated learning approaches and allow the implementation of preparation for future learning, drawing on foundations of the learning sciences. However, empirical research reveals mixed results for comparisons between CVE patient simulations and other types of interventions, or between variants of the same modality (Kononowicz et al., 2019). In the present project we will first explore the effectiveness of problem-solving in virtual patient simulations on clinical knowledge and CR skills acquisition and transfer. Second, we will explore the impact of applied computer-based virtual patient scenarios during a semester course on later courses in the medical study trajectory. Third, we will evaluate the use of virtual patients as an assessment tool. With the present project we will (a) fill the gap in the learning sciences and education research of when and how computer-based virtual environment simulations are effective in medical education, (b) examine the extent to which medical computer-based virtual environment simulations can enhance the acquisition and learning transfer of clinical knowledge and clinical reasoning skills, and (c) derive principles for when and how to best implement such virtual patient scenarios for DD into a medical curriculum.

Figure 1. Proposed Implementation Approach.

Note. Initial situation, reasoning, solution approaches and implementation plans of the present project. Abbreviations: CR: Clinical reasoning.

Figure 2. Implemented CVE Platform

Note. The platform we will use simulates a complete patient encounter with animated avatars, human physiology and pathophysiology. Figure retrieved January 23, 2021 from

State of the project

To address the suggested goals, this project is divided into two main work packages (WPs) which are related to each other.

Work Package 1 Sequencing of Problem-Solving in CVE Patient Scenarios and Direct Video Instruction

In WP 1 we will implement Bransford & Schwartz’s (1999) framework of preparation for future learning (PFL). Conceptually, this framework aims to understand how students use their past knowledge to solve novel problems when they cannot use routine practices efficiently or effectively. This is particularly important because medical students cannot possibly experience all situations they will face in their professional life during their studies. Accordingly, a critical aim of medical education must be to prepare medical students for their future learning to ensure a trajectory of adaptability throughout their career as doctors. Related to preparation for future learning is the instructional design of problem-solving prior to instruction (PS-I) (Kapur & Bielaczyc, 2012). This approach combines a preparatory problem-solving phase, in which students engage with a novel problem, with a subsequent instruction phase. Recent meta-analytic evidence suggests that problem-solving preceding instruction, on average, results in better conceptual understanding and transfer outcomes than instruction-first learning approaches for comparisons carried out in the domain of medicine (Sinha & Kapur, in press). However, there also are limitations of the PS-I approach where the I-PS sequence (instruction prior to problem-solving) might be more appropriate (Sinha & Kapur, 2019).

According to examine the effect of CVE simulations on PFL and to explore when CVE simulations have to be implemented to best enhance isomorphic and transfer outcomes, we will take up the PS-I approach. We will combine problem-solving in CVE with a virtually simulated patient – which situates learning in disciplinary practice – with another learning activity and vary their sequence. By applying the PS-I approach, we aim to determine the sequence of two learning activities which leads to best acquisition and transfer outcomes. In the experimental condition we will set CVE problem-solving (including feedback) prior to direct instruction (CVE → I). The comparison condition will follow the typical sequence where CVEs are used as a learning tool after direct instruction (I → CVE). Furthermore, we will supplement the learning activity sequencing exploration by adding two more conditions, each either going through the I or CVE learning activity only. By doing so, we aim to investigate the effects of the individual learning activities.

The covered topic will be neurology (migraine, subarachnoid hemorrhage). Direct instruction will be provided through a monologue video lecture where the diagnostic approach

to headache will be explained. As represented in Figure 3, all groups will first go through a pre-intervention multiple-choice quiz which contains questions about content knowledge. The subjects of the two learning activity sequencing groups then will go through the two learning activities but in the reversed order. The two remaining groups will either go through the I or CVE learning activity only. Once this intervention is completed, the groups will repeat the learning scenario in the test mode (no feedback). Hence, this will be an isomorphic assessment of CR skills. Additionally, the pre-intervention quiz questions will be answered again, which will be an isomorphic assessment of clinical knowledge. This whole step will be repeated in a scenario with a different related and unrelated diagnosis. By doing so we can assess near and far transfer of clinical knowledge and CR skills.

We are conducting this study in spring semesters 2020, 2021 and 2022 with third-year medical students at ETH Zürich. Hence, data analysis in progress.

Figure 3. Study Design to Investigate the Sequencing of Direct Instruction and Problem-Solving in CVE

Note. Abbreviations: AMI: acute mesenteric ischemia; CVE: computer-based virtual environment; CVE Learning: CVE patient scenario with instant feedback; CVE testing: CVE patient scenario with no feedback or hints; I: direct video instruction (lecture); SA: subarachnoid hemorrhage.

Work Package 2 – Influence of CVE Problem-Solving Exposure on Performance in a Consecutive Semester Course

In WP 1 we will assess isomorphic testing outcomes and transfer of clinical knowledge and CR skills in an experimental design with short-period learning activities (within two days). In contrast, with the present WP we suggest a longitudinal study in which we will assess the transfer of clinical knowledge and CR skills from a basic semester course to an advanced one.

There are two related courses in the trajectory of the medical curriculum at ETH, both focusing on DD and CR. However, the first one is a basic course, whereas the second consecutive one is a more advanced course. Consequently, students who completed the basic semester course will attend the advanced one-week course two semesters later. Both courses were affected by the current CoVid-19 situation (“Coronavirus COVID-19,” 2020).

An important part of the basic course were interactive group discussions in which different patient cases were discussed under the guidance of a specialist after each lecture (spring semester 2019). These case discussions could not take place any more from mid spring semester 2020 onwards because of CoVid-19 issues. Instead, they were replaced by patient problem-solving on a CVE platform. This naturally resulted in two conditions, where in one condition patient problem-solving in CVE took place but not in the other. On the other hand, the one-week advanced course took place completely online on CVE and Zoom (“Zoom Video Communications,” 2020) in spring semester 2020. During this online course in spring semester 2020, the participants acquired clinical knowledge and CR skills through CVE patient problem-solving and consolidation lectures. Furthermore, the students had to solve several testing scenarios and multiple-choice quizzes. The same was true for spring semester 2021, where the schedule of the course was the same than in the previous year.

With the present WP we aim to investigate whether the CVE application in a basic semester course has an impact on an advanced course later in the medical study trajectory. Hence, we will compare the outcomes in the advanced course among two cohorts to each other. Please refer Figure 4 for an illustration of the trajectory schedule.

We collected data for this WP in spring semesters 2020 and 2021 with third-year medical students at ETH Zürich. Hence, data analysis in progress.

Figure 4. Schedule of Work Package 2

Note. Cohort 1 and 2 represent different conditions of CVE application in a semester course. Cohort 1 represents the No CVE condition where no CVE simulations were applied during the semester course at all. Cohort 2 represents the With CVE condition, where students were exposed to CVE patient problem-solving during the semester. Abbreviations: CVE: computer-based virtual environment; SS: spring semester.


Battegay, E. (2017). 1 Grundlagagen der Differenzialdiagnose. In Differenzialdiagnose Innerer Krankheiten – Vom Symptom zur Diagnose (21st ed., p. 38). 980

Benbassat, J., & Bachar-Bassan, E. (1984). A comparison of initial diagnostic hypotheses of medical students and internists. Academic Medicine, 59(12), 951–956.

Bransford, J. D., & Schwartz, D. L. (1999). Rethinking transfer: A simple proposal with multiple implications. Review of Research in Education.×024001061

Collard, A., Brédart, S., & Bourguignon, J.-P. (2016). Context impact of clinical scenario on knowledge transfer and reasoning capacity in a medical problem-based learning curriculum. Higher Education Research & Development, 35(2), 242–253.

Coronavirus COVID-19. (2020). Retrieved August 3, 2020, from ETH Zürich website:

Eva, K. W. (2005). What every teacher needs to know about clinical reasoning. Medical Education, 39(1), 98–106.

Graber, M. L., Franklin, N., & Gordon, R. (2005). Diagnostic error in internal medicine. Archives of Internal Medicine, 165(13), 1493–1499.

Higgs, J., & Jones, M. (2008). Clinical decision making and multiple problem spaces. Clinical Reasoning in the Health Professions, 3–17.

Kapur, M., & Bielaczyc, K. (2012). Designing for Productive Failure. Journal of the Learning Sciences.

Kassirer, J. P. (2010). Teaching clinical reasoning: Case-based and coached. Academic Medicine, 85(7), 1118–1124.

Kononowicz, A. A., Woodham, L. A., Edelbring, S., Stathakarou, N., Davies, D., Saxena, N., … Zary, N. (2019). Virtual patient simulations in health professions education: systematic review and meta-analysis by the digital health education collaboration. Journal of Medical Internet Research, 21(7), e14676.

Merriam-Webster (Dictionary). (2019). Retrieved September 11, 2019, from

Nendaz, M., & Perrier, A. (2012). Diagnostic errors and flaws in clinical reasoning: mechanisms and prevention in practice. Swiss Medical Weekly, 142(4344).

Norman, Geoff. (2009). Teaching basic science to optimize transfer. Medical Teacher, 31(9), 807–811.

Norman, Geoffrey. (2005). Research in clinical reasoning: past history and current trends. Medical Education, 39(4), 418–427.

Phillips, D. C. (1995). The good, the bad, and the ugly: The many faces of constructivism. Educational Researcher, 24(7), 5–12.

Prince, K. J. A. H., Boshuizen, H. P. A., Van Der Vleuten, C. P. M., & Scherpbier, A. J. J. A. (2005). Students’ opinions about their preparation for clinical practice. Medical Education, 39(7), 704–712.

Schwartz, A., Weiner, S. J., Weaver, F., Yudkowsky, R., Sharma, G., Binns-Calvey, A., … Jordan, N. (2012). Uncharted territory: measuring costs of diagnostic errors outside the medical record. BMJ Quality & Safety, 21(11), 918–924.

Simmons, B. (2010). Clinical reasoning: concept analysis. Journal of Advanced Nursing, 66(5), 1151–1158.

Sinha, T., & Kapur, M. (2019). When Productive Failure Fails.

Sinha, T., & Kapur, M. (2020). When problem-solving followed by instruction works: evidence for Productive Failure.

Thampy, H., Willert, E., & Ramani, S. (2019). Assessing Clinical Reasoning: Targeting the Higher Levels of the Pyramid. Journal of General Internal Medicine.

Ypinazar, V., & Margolis, S. (2006). Clinical simulators: applications and implications for rural medical education.

Zoom Video Communications. (2020). Retrieved April 16, 2020, from

Christian Fässler

Prof. Dr. Jörg Goldhahn

PD Dr. Christian Schmied

Prof. Dr. Elena Osto

Prof. Dr. Manu Kapur