Chaired Panel 4 Emily Conradi

A Workable Model for Virtual Patient Design

Dr Jonathan Round, Emily Conradi, Dr Terry Poulton, Arnold Somasunderam e-Learning Unit, CMHCE, St George’s University of London, [email protected] ABSTRACT St George’s University of London have created a generic ‘model’ for virtual patient (VP) design, simple enough for clinicians to use, yet flexible enough to simulate real clinical decisions. This method of VP creation is disseminated to educators within our institution through regular VP workshops. For each VP an ideal pathway is described, with 3 or 4 critical points or ‘nodes’ that the patient must pass, in order to progress through the case. These might be, for example, the restoration of cardiac output after an arrest, ward transfer, or referral to another doctor. In order to navigate between nodes, a map of different interconnected possibilities is designed, typically with 3-4 steps and 3-4 choices at each step. Choices at each step mimic some of the choices that would be available for a real patient. Many of these would not allow progress to the next node. The online activity modeling system ‘Labyrinth’, developed by the University of Edinburgh, allows these ideas to be quickly and easily transferred into a digital virtual patient. Labyrinth offers an easy to use VP creator and player that conforms to the latest technical standards. The finished VP can be accessed online or through an institutional VLE, with optional add-ons such as timing and scoring. As a result, a simple educational model can be used to create ergonomically designed VPs. The next step will be to determine how to incorporate the generated VPs into our medical curriculum. A trial study to replace our current PBL-based curriculum with a VP curriculum is being piloted. VPs for the clinical years are to be developed for mobile devices, to provide the students with ‘just-in-time’ learning.

KEYWORDS

SOLSTICE 2007 Conference, Edge Hill University

1

Virtual patient, medicine, healthcare, higher education, decision-making, gaming

WHAT IS A VIRTUAL PATIENT? A virtual patient (VP) is defined as: “an interactive computer simulation of real-life clinical scenarios for the purpose of medical training, education, or assessment.” (Ellaway, Candler et al., 2006). In its simplest form a VP allows the user, usually via a computer, to make a choice based on some clinical information. The user is then given feedback dependent on their choice. More complex VPs will offer more choices, and then link pages together, so that the information and choices available at any stage depend on the choices made earlier in the scenario. Although virtual patients come from a medicine and healthcare perspective, VPs are essentially problem-solving exercises – the user must apply knowledge to progress through the scenario (Henderson, 1998). If the case is reflective of real situations and real choices, they can be excellent tools to practice using knowledge, reasoning, and decision-making skills, applicable in many fields. Practising knowledge and skills using virtual patients can offer some advantages over learning through real-life practise including: • • • • • •

Repetition; Consistent feedback; A potential for a greater exposure of scenarios; Mistakes do not carry real world repercussions; The ability to investigate alternative courses of action; Learning can be undertaken in a time and place convenient to the learner.

Simulators and games such as VPs have proven effective and popular e-learning tools (Aldrich, 2005; Quinn, 2005) that can underpin and extend current practice in teaching and learning. In particular Virtual patients have already been used with success for student learning and training in healthcare (Bergin & Fors 2003). Professional and vocational education needs to be as close to real practice as possible, whilst still offering educational opportunities and activities. It is emulating real practice that makes the design of VPs a challenging task.

HOW TO MAKE A VIRTUAL PATIENT Approaches to virtual patient design VPs can be time-consuming and expensive to produce. There are four distinct approaches to virtual patient design:

SOLSTICE 2007 Conference, Edge Hill University

2

•

The linear approach. Here the user is prevented from going down any wrong paths by immediate correction. This is clinically unrealistic, where there are often several ways to tackle a problem and mistakes are often not immediately obvious. This approach might be used for testing knowledge of a protocol, but will not engage students in the same way as more complex, multi choice VPs. An example of a linear case of this sort is paper-based problem-based learning case, where the case only proceeds in a single direction.

•

The Hi-Fi approach. This approach demands a large amount of time, money and effort to model all of the possible choices the VP writing team can think of. Effort is then spent on linking the case to other media and on the appearance of the case. This is an expensive process – Dr J.B McGee from the University of Pittsburgh has estimated the cost at over $100,000 per case, making the creation of a large VP bank impossible for institutions, even collaboratively. Patient simulators such as those used to train emergency medicine staff and anaesthesiologists1 are becoming more available, although their purchase price and running costs are large. These offer the ‘ultimate’ in hi-fi virtual patients. Even so these might not be suitable for a large range of history based of investigation based scenarios.

•

The algorithm method. Here formulae are developed that mimic physiologic processes in the body and in disease states, so that changes made by the user (typically administration of drugs or fluids) alter the output of the formulae and produce changes on the display, typically of biophysical variables. However this approach can only be used to create VPs where alterations in vital signs or biochemistry are the main elements of the case, such as anaesthesia, metabolic medicine or critical care pharmacology. Most of clinical medicine cannot be tackled in this way, as it is descriptive and history based. An example is the Virtual Center for Renal Support (Prado et al., 2002).

•

The Lo-Fi method. Here effort is spent on creating a large, but limited number of choices. Users are allowed to make around 2-3 wrong choices before finding out their mistake. They are given the option, after making a wrong choice, of making the correct choice, as long as the choice was not dangerous. These cases are often not as media-rich or as interactive as hi-fi cases. The interaction will focus on a specific set of options rather than a much broader set of choices. An example of a lo-fi VP is Sarah-Jane, developed at St George’s University2.

The problem of choice

1

Example of a hi-fi patient: http://anesth.utmb.edu/simcenter/ [last accessed 23/03/07] Example of a lo-fi patient: http://www.elu.sgul.ac.uk/virtualpatients/examples/sarah_jane/SJP_h_21_NT_HM.html [last accessed 23/03/07] 2

SOLSTICE 2007 Conference, Edge Hill University

3

Virtual patients are dependent on choice. If a VP is created with 3 choices at each level, within just 4 levels there have been over 100 situations which the user may have entered.

Fig: 1 The problem of choice

Most clinical encounters will involve over 20 choices (e.g. should I ask about the immunisations, should I do a neurological examination, should I take a full blood count?) so perfectly realistic simulation becomes near impossible to model. At SGUL the lo-fi approach is seen as developing VPs that offer enough engagement for the user to actively practise clinical reasoning and knowledge, whilst remaining practical to produce in terms of time and budget restrictions – an effective compromise between realism and practicality. The lo-fi approach also emphasises decision making – something very difficult to teach in other parts of a curriculum – rather than knowledge acquisition. As a result, a workable 10-step model for VP creation has been developed and tested, that can be transferred to all virtual patients.

The 10-step model for VP creation 1. Decide on a suitable patient Here you want a scenario that needs evaluation, ideally involving several steps, which might be in history taking, physical assessment, investigation and can go on to management steps. Examples might include a man with chest pain, a vomiting baby, a woman with post menopausal bleeding etc.

SOLSTICE 2007 Conference, Edge Hill University

4

2. Set key nodes Key nodes are stages in the case that act as a gateway to the next part of the case. This is a construct to limit the exponentially expanding number of choices during the case back to one. There will typically be 3-5 nodes in a case, and they will represent the start of a stage of the patient’s management – for instance they might be triage in A+E, completion of resuscitation, admission to the ward, cardiac catheterisation and discharge home for a patient with a myocardial infarction. 3. Create an ideal pathway This does not have to be the only way through the case, but will give the number of steps that will need to be programmed. There should be 3-4 between each node. This limitation is again a device to restrict the potential number of situations between each node, as for each correct choice there will need to be some other choices, each of which will then lead onto other choices. The steps to this stage are shown in figure 2 below:

Figure 2: Nodes interconnected with an ideal pathway

4. Put interconnected, branching boxes between the nodes. These will represent the situations that the patient goes through and the choices connecting them. It is important that they are placed in the emerging VP empty, again another device to manage the number of situations that the patient will go through in the case. You will need at least 100 pages in total to represent a reasonable VP (see Figure 3).

SOLSTICE 2007 Conference, Edge Hill University

5

Figure 3: An empty node map of a VP

5. Decide what the boxes represent and what choices connect them. Here you will need to think what might be reasonable in a real situation, and use this as the basis for naming the choices. The pattern of empty boxes (stage 4) might not reflect real life, and the pattern can, to some extent be adjusted with boxes added or removed where essential. Further connection may also be made at this stage, including alternative routes through the case. 6. Prune the case. Dead end branches need an explanation and redirection back to an earlier node or to the start. 7. Name each box. A logical and short naming system is needed. In our cases 0 starts the case, and the next stage or step is 1, then 2 etc. Each situation at each stage is assigned a letter alongside: 2a, 2b etc. Dead end explanation pages are given the suffix ‘_e’ beside the page they are associated with: 4b_e for instance. 8. Create a spreadsheet to define the VP This is the way the case writer communicates with a technologist to describe the text at each situation, the choices and the names of the situations that selecting a particular choice will direct the user to. This part of the process is the most laborious, requiring imagination to create the narrative describing the clinical state of the patient in each situation (see Figure 4).

SOLSTICE 2007 Conference, Edge Hill University

6

page name link to link name

0 1a 1b

finish curry run to A+E

Text Welcome. You have now been a paediatrician since 4 pm, when your shift started. Its day 1 of the job, and you are trying to remember all that stuff from the paed course. Anyway, it wasn't too busy and now you have discovered the mess and the takeaway menus. A Lamb Rogon Josh has arrived and you are half way through, when the crash bleep goes off "Paediatric arrest in A+E". What would you like to do? Excellent. Curry before your patients. Way to go. Just remember, sometimes it really is an arrest, and its your job to be there. After a quick detour into an elderly care ward, you make it to A+E. As you are catching your breath, one of the A+E nurses tells you that a six month old girl has been brought in by ambulance with a 25 minute fit, and now she's 'desatting'. She is called Shani, you are told. You notice that she has an IV cannula - at least you don't have to put in one of those. However she is still fitting, shaking all limbs. What would you like to do?

1a

2a

call mdu

1b

2b 2c 2d

Assess ABC Examine Shani Take a history Figure 4: Example of the VP excel spreadsheet

9. Insert into a user interface This can be done simply by creating individual html pages, for instance in Dreamweaver or even MS Word, creating an XML schema, developing a suitable Flash player, or using another application, such as Labyrinth (see below: ‘How to make the VP digital’). 10. Enrich with media The skeleton of the case is now complete. Depending on time and resources, the VP can now be complemented with other features – clinical photography, video, or sound. It can be linked to other sources of information, such as on-line course materials and relevant websites. This approach will typically produce a case of 10 steps, containing 120-150 pages. Approximately 10 hours is required to create such a case for one person.

Making the VP play The e-Learning Unit at SGUL has adopted a set process for developing the files supplied by clinicians and creating the finished virtual patient. Two open source software systems are used:

SOLSTICE 2007 Conference, Edge Hill University

7

• •

Visual Understanding Environment (VUE) developed by the University of Tufts; Labyrinth developed by the University of Edinburgh.

VUE is an information management application that provides an interactive, concept mapping interface, allowing a case to be quickly drawn out to emulate the node map.

Figure 5: Image of a VP node map drawn in VUE

Labyrinth is an experimental educational pathway authoring and delivery system. It has an easy to use interface, and requires no prior programming knowledge. It incorporates the latest technical standards for electronic virtual patient simulations developed by MedBiquitous, a consortium of international medical schools and technical developers. VUE and Labyrinth are complementary tools, with files created in VUE easily imported into Labyrinth. Each box in VUE becomes a page in the case and each arrow becomes a link. The text in each box becomes both the page title and also the text of the associated link. For example, the following VUE would lead to the following output:

Figure 6: Example VUE map with corresponding output in Labyrinth

SOLSTICE 2007 Conference, Edge Hill University

8

Once the structure of the VP is in place via the VUE node map, the narrative from the clinicians excel spreadsheet is added to each page. Additional resources such as images, video and sound files can be added to the case if desired. Labyrinth allows the VP to have scoring attached, with choices having different positive or negative weightings. There is also a timing function that can be activated if desired. The user can review their pathway at any point, to look back and reflect on the choices they have made. Finished cases can be played within Labyrinth, displaying the VP as a series of web pages that can be accessed online. Labyrinth generates the virtual patient through an XML schema that can be exported out and used with other supporting VP players. Alternatively, Labyrinth can act as a web service, allowing other VP players to access cases in realtime.

Figure 6: Example of a virtual patient played in Labyrinth using a St George’s skin

The e-Learning Unit at SGUL are now working towards creating a packaged VP that can be inserted into an institutional VLE, or downloaded and played offline on devices such as a mobile phone or PDA.

The VP Workshops In response to a high level of interest in virtual patients, the Centre for Medical Education at St George’s University of London have begun to run full day workshops for the creation of virtual patients. These workshops are run every couple of months to educators within the medical school and Faculty of Health and Social Care Sciences. The workshop gives an overview of virtual patients and how they can be used in education. The delegates are then split into small groups, usually according to specialty,

SOLSTICE 2007 Conference, Edge Hill University

9

and begin to create their own patient, using the methodology outlined here. By the end of the day the groups will have produced a whiteboard drawing of the nodal map of their case, and begun the spreadsheet containing the accompanying narrative. This content is taken away by the e-Learning Unit, and from it an interactive electronic virtual patient is produced, for the educators to use in their teaching.

Fig 7: Image from a VP workshop

The next step is adapting these workshops to make the creation process more efficient, particularly in terms of stages 7 and 8 of the model. Currently the process relies on whiteboards and laptops with Microsoft Excel. The use of interactive whiteboards would speed the process, as the node map could be saved electronically directly from the board, and made available to the technologist immediately, rather than having to be effectively ‘copied’ into an electronic format. These node maps could be uploaded directly into Labyrinth, making the step from concept to creation almost instantaneous. SGUL are also looking to expand the workshops externally, due to a high level of interest from partner institutions and the e-learning community. This will begin with a workshop to the Higher Education Academy in the summer 2007.

APPLICATIONS FOR VIRTUAL PATIENTS There are numerous applications for VPs in medical and healthcare curricula. Several uses for VPs have been suggested, for different courses and at different stages of learning. Currently the number of freely available VPs in medical education is low. With the advance of easier models for VP design, shared practice, and pooling of resources

SOLSTICE 2007 Conference, Edge Hill University

10

within the e-learning community (for example see the EC funded eViP project3), the number of VPs available for teaching will rise. At first glance, VPs appear to work from existing knowledge and test clinical decisionmaking, so VPs might be most suitable for the ‘clinical’ years in medicine. This is reflected in the emphasis of most of the VPs so far developed. SGUL are now working towards developing VPs for mobile devices to provide students with self-directed, ‘justin-time’ learning, to revise and practice their abilities when they are not in the formal learning environment of their university. There is however, no reason why a VP approach might not work as the ‘case’ in problem based learning. We are piloting a replacement of our current 4yr PBL-based medical curriculum with a VP curriculum, in order to engage the learner with a more interactive and less linear learning scenario. VPs will be specifically adapted for this purpose, to reflect the necessary basic clinical science teaching, and less emphasis on the ability to apply clinical reasoning. Cases are written with a set of learning objectives in mind, linked closely into the module. There is also interest in using VPs for inter-professional working. VPs can allow the student to gain an insight of what it is to work within an inter-professional team, by allowing the user to interact with the same patient from a variety of professional perspectives. Alternatively, the role of different professions throughout the course of the patient’s story can be easily demonstrated. VPs are also suitable for use in distance learning and disseminated learning courses. Finally, we are exploring the use of virtual patients as tools for both formative and summative assessment, particularly in later years of education, with a scoring system reflective of the pathway chosen by the candidate.

CONCLUSIONS Creating realistic choices in virtual patients makes VP creation difficult. Using methods to limit the number of choices, cases that are close to real clinical encounters can be created. As a result, a simple educational model can be used to create ergonomically designed VPs. VPs have the potential to be used in a variety of settings, including teaching of clinical reasoning, communicating knowledge and assessment, and problem-based learning. The 10-step VP development model outlined here can be broadly adapted to create VPs for a number of different educational applications. For more information on the eViP project, virtual patient workshops, or to experience a virtual patient first hand, please visit our virtual patient website: http://www.elu.sgul.ac.uk/virtualpatients/ [last accessed 22/03/07].

3

For more information on the eViP project please visit the SGUL e-learning site: http://anesth.utmb.edu/simcenter/ [last accessed 23/03/07]

SOLSTICE 2007 Conference, Edge Hill University

11

REFERENCES Aldrich, C., 2005. Learning by Doing. San Francisco, US: Pfeiffer. Bergin, R. & Fors, U., 2003. Interactive Simulation of Patients – an advanced tool for student-activated learning in medicine & healthcare. Computers and Education, 40/4 361-376. Ellaway, R., Candler, C., Greene, P. and Smothers, V., 2006. An Architectural Model for MedBiquitous Virtual Patients. Baltimore, MD: MedBiquitous. Henderson, J., 1998. Comprehensive, Technology-Based Clinical Education: The "Virtual Practicum”. International Journal of Psychiatry in Medicine, Vol 28(1); 41-79. Prado, M., Roa, L., Reina-Tosina, J., Palma, A., Milan, J.A., 2002. Virtual center for renal support: technological approach to patient physiological image. Biomedical Engineering Volume 49, Issue 12, Page(s): 1420 – 1430. Quinn, C. N., 2005. Engaging Learning: designing e-learning simulation games. San Francisco, USA: Pfeiffer.

SOLSTICE 2007 Conference, Edge Hill University

12