Developing a framework for optimal e-CBL use: Case-based blended learning (CBBL)
Our case-based blended learning (CBBL) approach consisted of a progressive three tier approach described in detail previously in Turk et al. (12), in which we had postulated that the transfer of declarative knowledge to procedural knowledge in theory, and then of procedural knowledge to procedural skill in practice, requires a three step multimodal approach. To summarize briefly, we developed, adapted, published, implemented, and evaluated our material for each of these three tiers, now described below.
A blended-learning (multimodal) approach using 1) textbooks, 2) interactive e-CBL cases transferred and created via the General Hospital electronic health records system (12, 13, 39, 40), and 3) simulated patient contact was designed and implemented university-wide for several subjects taught to medical students, starting with psychiatry in 2014 (12). In total, our e-CBL framework and platform eventually saw adaptation and publication of relevant e-CBL cases for over 16 fields/sub-specialties in pre- and post-graduate medical education in total, including (but not limited to) psychiatry, pediatrics, neurology, infectious diseases, dermatology, microbiology, orthopedics, traumatology, internal medicine, surgery, and clinical genetics.
Textbook/written resources
Case-relevant declarative knowledge was compiled in a textbook for each subject in which reforms were implemented (41). This material was compiled in a case-based manner, created to work synergistically with as well as prior to the e-learning cases, thus serving as a declarative knowledge base. Each textbook chapter included sections describing: Anatomy and Physiology, Pathophysiology and Disease Process, and Epidemiology and Genetics. The goal of our case-based textbook was to create a relevant foundation of declarative knowledge, allowing its (future) transfer to procedural knowledge.
Each chapter section “Pathophysiology and Disease Process” was designed to overlap and be integrated with its corresponding presentation and physical examination in the e-CBL case. This crucial synergy aimed to foster a close link in creating the intimate connection between morphology and clinical presentation.
Creation of E-CBL cases
E-CBL cases for each subject in which reforms were implemented were created using anonymized patient data from the Vienna General Hospital (12). A novel plug-in for the existing electronic health records (ERH) system was developed, allowing automatic import and anonymization of real patient data from cases identified by attending physicians. Patient data was first transferred via the ERH into an anonymized case (for data protection purposes) on a separate platform. Following the initial transfer, the data was then automatically imported into the e-learning case template on an online administrator platform (Moodle). Two-stage validation was key to this step. Firstly, content creators (i.e. medical educators, physicians, etc.) were able to assess, edit and complete the cases for use in a teaching setting on Moodle, once the administrator gave them access privileges. Secondly, the completed cases had to be unlocked individually by the administrator and the content creator, before being placed in an online course available to students.
The e-CBL case templates were structured to enable data transfer to relevant sub-sections, thus greatly simplifying the editing and review process. Additionally, the template was optimized for case presentation following common case-presentation structures in medical education and was designed to complement the textbook chapters.
Each e-learning chapter included sections with relevant information on: Presentation and Communication, Physical Examination and Diagnostic Techniques, and Therapy and Prophylaxis.
Once case data was reviewed, content creators designed a virtual case-based discussion, in the form of multiple-choice questionnaires (MCQs). These MCQs allowed for multiple answers, with specific feedback for each correct or incorrect response, as well as follow-up questions. Their focus was specific to salient points in each chapter section (presentation and communication, physical examination and diagnostic techniques, therapy and prophylaxis).
MCQs focused either on clinical reasoning, requiring students to consider differential diagnoses, or on ordering the correct diagnostic test following a working hypothesis. To give a representative illustration of the types of questions/content developed and adapted for each subject, a model example in cardiology might be: 1) “Which of the following diagnostic tests should you initially order to assess the possibility of or rule out a myocardial infarction [for first time, sudden onset, level 10/10 VAS, crushing substernal chest pain] without dyspnea?
-
A) 12 lead EKG [Correct]
-
B) Trans-thoracic echocardiography
-
C) Chest X-ray
-
D) Computed tomography”
In creating such case-based discussions, feedback was given for each possible answer (if selected), written by the content creators, i.e. why one choice is correct and not another, or if both choices seem correct, why one is prioritized over the other. This feedback allowed the students to confirm, practice, or retroactively develop procedural knowledge.
For the model question above, the feedback reviewed case-specific data in the context of declarative knowledge and/or expanded on possible alternative outcomes:
Initial testing must be performed in a timely manner, starting with the most relevant non-invasive procedure.
[Specific feedback for A)]: A 12 lead EKG should be obtained within 10 minutes of presentation for all patients with chest pain. Morphological changes such as ST-segment elevations >1mm in > 2 anatomically contiguous leads (STEMI) or a novel left bundle branch block; non-ST-elevation MI (NSTEMI) or unstable angina: ST-segment depression or T-wave inversion. Morphological changes in conjunction with notable heart parameters in blood testing warrant immediate emergency management.
[Specific feedback for B)]: Trans-thoracic echocardiography (TTE) is not considered in initial testing for myocardial infarction, as performing and reading TTEs is not the most efficient and sensitive method. Pathological features such as wall motion abnormalities, valve disease and congenital heart disease may be assessed as well as prognostic data garnered from color-flow Doppler transthoracic echocardiography. Sensitivity is highly dependent on the ability to appropriately image the apex and sensitivity may be as high as 90%. However, timely diagnosis is key in determining the emergency nature of the illness.
[Specific feedback for C)]: A chest X-ray should be routinely performed on all patients presenting with chest pain. However, the sensitivity and specificity of findings indicative of myocardial infarction are low.
[Specific feedback for D)]: Computed tomography is not considered in initial work-up of chest pain, due to the time needed to complete- and the ionizing radiation caused by the procedure. Specificity and sensitivity for the diagnosis of MI using CT angiography is above 95%, allowing the visualization of ventricular aneurysms and intra-coronary thrombi.
Simulated Patient contact
Students were required to successfully complete the e-learning course by correctly answering all questions before taking part in the seminar. The SP seminar has the aim of applying and transforming procedural knowledge into procedural skills, while also allowing students to experience, document, and reflect on difficulties that arise from newly encountered dimensions of face-to-face communication with mentally ill patients (42). Here the student is required to observe, give feedback, and perform a complete psychiatric consultation, including conducting a mental status examination and taking a psychiatric history, creating a clinical or diagnostic hypothesis, deciding on further case management, and finally suggesting relevant therapy options, thus integrating all previously acquired skills.
The SP employs the use of professional actors who have received training in embodying patients, requiring students to apply both the declarative knowledge and clinical reasoning skills learned in the textbook and e-learning cases. Actors learn their “roles” using prepared and anonymized patient case data. After the course, students are assessed by educators, peers, and themselves (43), with communication portfolios documenting their simulated consultations.
Data Collection
Anonymized OSCE scores of all medical students at the Medical University of Vienna who partook in the clinical psychiatry examination “Physician communication skills” [Ärztliche Gesprächsführung, ÄGF-C] of the 4th year OSCE in the years 2013, 2015, and 2016 were retrospectively retrieved by HLS from university student records’ electronic databases and analyzed. The data protection committee of the Medical University of Vienna, independent of anyone involved in this study, first anonymized all data before allowing access, and also approved this study.
Evaluated data sets included the OSCE scores from the respective cohorts of 4th year medical students in 2013 (before any implementation of the CBBL framework, thus considered “pre-intervention”), 2015 (considered thus “post-intervention”), and 2016 (also considered as being “post-intervention”, included for verification purposes, so that if any significant improvement were to be observed between 2013 and 2015, a similar improvement observed as well in 2016 vs. 2013 would be an indicator that such improvements were more likely to be a persistent and reproducible effect of CBBL implementation, rather than merely a coincidence).
Statistical methods
Non-linear regression analysis was performed, and datasets were tested for normality and homogeneity of variance. The Shapiro-Wilk test by Royston (44) and assessment of kurtosis and skewness using the D'Agostino-Pearson omnibus test (45) were carried out in order to determine the normality of distribution of each cohort year’s OSCE scores. A Brown-Forsythe test (46) was carried out in order to test homogeneity variance between groups.
Data pairing was considered, as test takers for 2013, 2015, and 2016 may be seen as being matched by level of progression of medical education (all test takers being in the 4th year (out of 6) of medical school in these respective years). However, considering the presence and possible effects of a host of unknown variables such as age, sex, ethnic group, and the number of exam attempts, etc., we felt that unpaired testing was to be favoured. Thus, comparison between cohort year datasets was performed using nonparametric, non-paired t-tests (Mann-Whitney (47)). Although the consensus is that non-parametric tests commonly have less power (48), as will be seen later, they were employed due to our non-normal data distribution.
Then, comparison of the medians of the three groups was performed through use of the Kruskal-Wallis H test (49), assuming non-parametric data distribution. Also due to our non-normal data distributions, Chi value generation for the Kruskal-Wallis test was performed using Murphy & Myors’ (50) transformation into an F value prior to Laken's adapted Cohen's formula (51), providing an eta value instead of a more common Cohen's D. Following that, Dunn’s multiple comparison tests (52) were used post hoc to compare groups within a non-parametric ANOVA. For effect size between groups, Rosenthal & DiMatteo’s (53) effect size calculation was performed.
In this study, statistical significance was assumed by a p-value < 0.05.