1]. Problem-based learning (PBL) has been widely adopted as a way of integrating knowledge acquisition, teamwork, and problem-solving skills using active learning techniques within small-group settings . However, PBL has remained relatively unchanged since it was introduced with paper cases—the principal medium around which PBL activities are organized. Even the use of the Internet has, at least so far, had limited impact on the design of the PBL case or the activities that are structured around it.
There are a number of educational technologies that share certain characteristics with PBL. For instance, virtual patients are on-screen learning resources that typically present a clinical problem for learners to solve or manage, and in doing so involve aspects of both PBL and simulation . Although virtual patients can come in many forms, one of the more common forms is the branched case where learners select the best available course of action from predefined options [ ]; each decision the learner makes can have consequences and lead to different outcomes.
St George’s University of London (SGUL) replaced paper PBL cases for midstage undergraduate teaching with decision-PBL (D-PBL) cases. These are Web-based branching virtual patients that require learners to discuss, debate, and make patient management decisions to negotiate the case . This paper describes the efficacy of this approach using results from a controlled trial comparing end-of-year face-to-face exam performance for learners in PBL groups who worked with linear or branched cases. The paper closes with a consideration of the implications of this approach to teaching tomorrow’s doctors.
PBL involves a combination of students working in small groups with a facilitator and independent research. Each group works through a predefined patient case, discussing the information provided and implied, exploring possible diagnoses, suggesting investigations and treatments, and identifying the research they need to undertake to be able to resolve the case . Learners then undertake their research before the next face-to-face session. The case may be resolved in the second face-to-face session or a second round of research may be identified which is then resolved in a third and final session. Groups of learners are typically given 1 PBL case per week. The pattern is repeated each week with different cases over a semester, term, or year. Although each PBL group will approach a case in different ways, the group facilitator directs the proceedings using the predefined PBL case outline (that the learners do not see) that sets out the key issues, anticipated learning objectives, and other intended features of the case, that allows them to keep learners from straying too far from the intended learning outcomes for each case.
PBL has been deployed in many different contexts  and although the efficacy and effectiveness of PBL continues to be debated [ – ], research into PBL indicates that its benefits come from key components of the PBL activity that tend to be obscured if PBL is considered as a single indivisible intervention [ , ]. These components include the construction of the PBL case, the behavior of the facilitator, and the curriculum context for the activity [ – ]. A key limitation of using paper-based cases is that they can only unfold in a single direction giving learners little or no opportunity to influence the outcome of the case.
There are other teaching modalities in medical education that employ cases, including simulation, case-based learning, and virtual patients, reflecting the role of the patient case as “the primary, vicarious means of shaping clinical judgment for new learners and experienced practitioners alike” . Simulation differs from PBL in that the former is directed more to skills training and teamwork, but there are also similarities [ , ]. Virtual patients are interactive computer simulations of “real-life clinical scenarios for the purpose of medical training, education, or assessment” [ ]. Although virtual patients can take many forms, they are intended usually for individual study [ ]. The use of virtual patients for PBL has not previously been the focus of PBL developers although a number of schools have placed their paper-based PBL cases online as static documents [ ].
The study presented here explored the efficacy of D-PBL in terms of student exam performance. Our working hypothesis was that D-PBL experiences would lead to better exam performance than linear PBL experiences. A randomized controlled trial in crossover design was run to compare aggregate learner performance in those learners who had D-PBL with those who had linear PBL with supplementary PBL “triggers” to compensate for the absence of D-PBL. The trial analyzed students’ performance on questions related to D-PBL decision points compared with their performance on questions not related to D-PBL. For each comparison cohort the learning activity was different, but the learning objectives were the same.
SGUL is located in South London in the United Kingdom and runs a multiple stream medical education program. The undergraduate entry stream undertake a 5-year program and the graduate entry stream undertake a 4-year program. SGUL has employed PBL since the 1990s  and at the time of this study the 140 graduate entry learners (who were the participants in this study) completed 2 years of full PBL and then completed the 2 remaining years of clinical attachments.
A project was initiated in 2007 to rewrite the second year graduate entry paper PBL cases as branching virtual patients (D-PBL) and to modify the facilitation model to stimulate debate around predefined choices, and their consequences, at key points in the D-PBL virtual patient case . The objective was to make PBL more engaging for learners with prior PBL experience by adding structured options and alternative outcomes to the linear PBL activity model, and encouraging robust debate around structured patient management options and the consequences of pursuing a particular course of action. This was called decision-PBL (D-PBL).
A D-PBL virtual patient case is designed as a series of interconnected nodes, each of which is presented as a Web page. Each node represents a step in the unfolding of the case. Some nodes are connected in chains; others have multiple nodes linked to them that allow learners to choose which path they will follow. Learners can only take 1 path through a case and they need to deal with the consequences of their decisions as the D-PBL case unfolds. The small-group facilitator encourages learners to consider the options and to debate different courses of action at each D-PBL decision point. Tutors report increased student engagement at option points and students express the value of simulating the experience of managing cases in real life, including the stresses that can lead to poor decisions [, ].
The use of a Web-based virtual patient delivery system was essential to creating and running D-PBL activities. The virtual patient cases were to be delivered to multiple PBL classrooms in the same time frame, the scenarios were required to respond to the choices of different learner groups dynamically and independently, and the paths taken by the individual groups were tracked and logged centrally. The latter would permit later analysis and research into the paths that groups took to better understand the effectiveness of the option steps and to iteratively improve the cases.
To minimize the impact on the curriculum as a whole, D-PBL was designed to only differ from traditional linear PBL in a few key areas. D-PBL (like PBL) involved small-group (6 to 8) learners, face-to-face sessions at the start and end of the week, independent research and study between face-to-face sessions, scaffolding by a case with triggers based around problems, and 1 case worked on per week. The differences were the replacement of static paper cases with dynamic Web-based virtual patient cases and the periodic debates and decision making when alternative paths were presented (within face-to-face sessions).
The 5-week Life Protection module (covering immunology, pathology, hematology, and microbiology material) in the second-year SGUL graduate entry stream was selected as the context for the study. All 5 cases within the Life Protection module were rewritten both as linear PBL cases and as D-PBL cases using the same virtual patient software; the only difference being the addition of different routes through the case for the D-PBL versions.
The faculty committee for each module reviewed the completed D-PBL cases to make sure that the modified cases still fell within the range of the existing learning objectives (options were based on management issues that were already described in the learning objectives). Then the linear PBL cases were supplemented by trigger questions that would cover the same area of learning emphasized by the options. For example, in the anaphylaxis case shown in, which focused on problems caused by a rapid intravenous injection of adrenaline. The 4 options would appear only after the students had read and discussed the text (which has been shortened in this figure, for reasons of clarity). This scenario, based on a real-life incident, describes a patient presenting in Accident and Emergency with anaphylactic shock. The scenario is written to elevate stress (“Toni is in poor condition, you have to think quickly what to do”), which in turn can lead them to follow their instinctive response to give a rapid intravenous injection, as happened in real life. This proved fatal. Students have previously noted that serious consequences such as this are particularly memorable [ ].
In this way, the study sought to ensure that the D-PBL provided different learning opportunities rather than additional learning opportunities. The development of the D-PBL cases has been discussed more fully elsewhere , although the additional prompts were not included in the linear PBL in the earlier study. Students agreed that the group would take a majority view on their chosen direction through the case.
A preliminary analysis of D-PBL in 2007 compared student engagement between learners in paired cohorts who had either a linear version of the case (without options) or a branched version (with optional routes through the case; D-PBL). Cohorts alternated between branched and linear over the first 4 weeks, with a fifth case delivered in the branched version to both cohorts. It was noted that students had performed markedly better than in previous years to a question that was related to 1 of the D-PBL decision points in the week 5 case. However, the D-PBL deployment was not designed as an experiment and there was no control group or any other means to systematically assess the impact of using D-PBL on summative assessment in this first year of using D-PBL. Therefore, a study was developed and run in the following academic year (2008) analyzing the exam performance of students with questions that had been previously encountered in either a D-PBL or linear PBL format.