SIMPRAC: A web-based virtual patient to support learning through ...

SIMPRAC: A web-based virtual patient to support learning through reflection Technical Report Number 562 January 2005

Dr Douglas Chesher (Clinical Biochemistry - Pacific Laboratory Medicine Services Royal North Shore Hospital) Associate Professor Judy Kay (School of IT, University of Sydney) and Associate Professor Nicholas J C King (Department of Pathology,University of Sydney)

ISBN 1 86487 702 2

School of Information Technologies University of Sydney NSW 2006

Technical Report

SIMPRAC: A web-based virtual patient to support learning through reflection Dr Douglas Chesher Clinical Biochemistry Pacific Laboratory Medicine Services Royal North Shore Hospital St Leonards NSW 2065 Ph: 61 2 9926 8086, Fax: 61 2 9926 6395 Email: [email protected]

Associate Professor Judy Kay Department of Information Technology University of Sydney Camperdown NSW 2006 Ph: 61 2 9351 4502, Fax: 61 2 9351 3838 Email: [email protected]

Associate Professor Nicholas J. C. King Department of Pathology University of Sydney Camperdown NSW 2006 Ph: 61 2 9351 4553, Fax: 61 2 9351 3429 Email: [email protected]

20 November 2004

i

Contents CONTENTS............................................................................................................................................... II ABSTRACT..............................................................................................................................................III INTRODUCTION...................................................................................................................................... 1 1.1 1.2 1.3 1.4 1.5

MEDICAL PROBLEM SOLVING .................................................................................................... 1 LEARNING THROUGH REFLECTION ............................................................................................. 3 SIMULATED MEDICAL LEARNING ENVIRONMENTS...................................................................... 6 REFLECTION AND THE CONSULTATION PROCESS ........................................................................ 8 PROJECT................................................................................................................................... 10

SYSTEM OVERVIEW............................................................................................................................ 11 1.6 1.7 1.8

LOGIN AND CASE SELECTION .................................................................................................. 12 PATIENT INTERACTION ............................................................................................................ 16 REVIEW.................................................................................................................................... 29

SYSTEM ARCHITECTURE.................................................................................................................. 35 1.9 1.10 1.11 1.12

USER MODEL ........................................................................................................................... 35 DISEASE MODEL ...................................................................................................................... 37 ANCILLARY FUNCTIONS .......................................................................................................... 42 IMPLEMENTATION .................................................................................................................... 43

CONCLUSIONS AND FURTHER WORK .......................................................................................... 44 APPENDIX A

EXAMPLE USER MODEL................................................................................... 46

REFERENCES......................................................................................................................................... 50

ii

Abstract Patient simulations have been used for medical education and assessment for a number of years. Computer-based simulations, in particular, have been developed with the goal of improving medical education and continuing professional development by providing authentic, interactive environments for problem-based learning. However, there is evidence of case and context specificity when patient simulations are used, with many learners being unable to apply knowledge from one context to another. A promising means to address this limitation of simulation environments comes from the potential learning benefits of reflection. SIMPRAC is a computer-based simulation that can be accessed over the Internet using commonly available web-browsers. It enables users to interact with a virtual patient by taking a history, examining the patient, requesting and reviewing investigations, and choosing appropriate management strategies. The virtual patient can be reviewed over a number of consultations, and the patient outcome is dependant on the management strategy selected by the user. In addition to providing a simulation of the consultation process, an additional layer has been added with the goal of supporting reflection. While interacting with the virtual patient, users are asked to formulate and test their hypotheses. Simple tools are included to encourage users to record their observations and thoughts for further learning, as well as providing links to web-based library resources. At the end of each consultation, users are asked to review their actions and indicate whether they think their actions were critical, relevant, or not relevant to the diagnosis and management of the patient in light of their current knowledge. Users also have the opportunity to compare their activity to their peers or an expert in the case under study. Further work will be required to improve the simulation environment, improve the interfaces for supporting reflection, and further define the benefits of using this approach for medical education and professional development with respect to learning outcomes and behavioural change.

iii

Introduction With the rapid growth in medical knowledge, and associated changes in medical practice, it is important that there are effective means for primary and continuing medical education. Unfortunately, there is evidence to suggest that there are serious shortcomings with existing approaches to continuing medical education (1). Simulated medical learning environments are one way of providing interactive experiences, which enable learners to practice what they have learned. This approach has been shown to be more effective at changing physician behaviour than more traditional methods of continuing education (1). Learner reflection is a process that has been promoted as having great potential for improved learning for professional practice (2), although it has been underutilized in medicine (3). This thesis explores the support of learning through reflection, in the context of medical students and practitioners, working through a series of simulated consultations involving the diagnosis and management of chronic illness. When exploring approaches to medical education, it is first necessary to have an understanding of medical problem-solving, and the practice of medicine. Therefore, this chapter begins with a brief introduction to medical problem-solving. This is followed by and introduction to the concept of learner reflection, with a particular focus on its role in learning for professional practice. The chapter then goes on to look at simulated medical learning environments as a way of providing interactive experiences, which enable the learner to practice what they have learned. The need to include medical decisionmaking and patient management within the simulation environment, especially in the context of chronic illness, is also considered. Based on this introductory information, a model of the medical consultative process is described, as well as a model for reflection that is congruent with the consultative model. These models have been embodied in SIMPRAC, which aims to support learning about the diagnosis and management of chronic illness over several simulated consultations by encouraging learner reflection. 1.1

MEDICAL PROBLEM SOLVING

Before developing software to help medical students and medical practitioners learn, it is first necessary to have an understanding of the clinician’s approach to medical problems. Knowledge of this reasoning process, especially the differences between less experienced clinicians and more experienced clinicians, can then be used as a foundation for developing an environment capable of improving learning outcomes. Research in medical problem-solving has been undertaken for several decades but only recently has this been reflected by changes in the way medicine is taught. The classical text on this area of study was written by Elstein, Shulman and Sprafka, based on research they undertook in the early 1970's (4). They view clinical problems as illstructured problem domains where not all the information required to solve the problem is immediately available but must be gathered and evaluated over time. As information is gathered, the problem is progressively better defined. Furthermore, this data is probabilistic in nature and may have alternate interpretations. They suggest that hypothetico-deductive reasoning and early hypothesis generation is used by experienced and inexperienced clinicians to solve clinical problems. Hypothetico-deductive reasoning involves the initial acquisition of basic data from which only a small number

1

of hypotheses are generated. Most clinicians generate on average, four to five hypotheses with an upper bound of six to seven. Further data is then collected to test and refine the true hypotheses, and exclude the false hypotheses. Clinicians were observed to move to structured, routine data collection when they needed time to interpret information that had been provided to them. If this "routine" questioning identified information that might generate new insights, they then reverted to selective, hypothesis-driven questioning to explore the new avenue. Research by other groups, has suggested that medical problem-solving by experts involves pattern recognition or illness scripts based around knowledge of the clinical consequences of the disorder (5). According to Schmidt and others (1990), medical students initially try to understand their patients on the basis of causal pathophysiological mechanisms. Later, as they gain more experience, there is a move to simplified mental models that are sufficient for diagnosis and management, with a focus on the features that characterize the disease. For example, a student may initially understand myocardial ischaemia in terms of reduced coronary blood flow leading to myocardial hypoxia, which in turn leads to disruption of metabolic processes within the myocyte. These metabolic disturbances lead to increased levels of lactate and other intermediates that activate nocioceptors, resulting in the perception of pain. These disturbances also lead to impaired muscular contraction and reduced cardiac output causing a fall in systemic blood pressure, and increased end diastolic pressure. The latter, may in turn lead to pulmonary congestion, hypoxia, and subsequent feelings of shortness of breath. Furthermore, these metabolic disturbances can also lead to impairment in electrical conduction through the heart with depolarization and repolarization defects being apparent as changes on the electrocardiograph. In contrast, a more advanced student may just note that myocardial ischaemia may be associated with chest pain, dyspnoea (shortness of breath), hypotension, and changes on the electrocardiograph. When seeing a patient, practitioners are thought to select an appropriate script, verify the patient’s presenting features against the script, and if appropriately matched, manage the patient based on this understanding. Elstein et.al. have countered this by suggesting that problems familiar to the clinician might be resolved using this technique but more difficult cases, outside of frequent experience, need to be solved using a hypothetico-deductive approach (6). Models have been developed to describe the medical problem-solving process. Social judgement theory is one such model (7). This is a cognitive theory of human judgement, which emphasizes the uncertainty inherent in the physical, biological, and social environments (termed the ecology) in which judgements are made. The cues are the pieces of clinical information available for policy considerations. In this context, a policy refers to a decision on how the information is to be used and its importance. These cues vary in how they correlate with the clinical state of interest (ecological validity) and with how they are used by the clinician when making judgements (utilization validity). Cues may also vary in terms of their relationship to each other. Hammond and his associates have developed a quantitative methodology and a lens model equation (8) where achievement is the sum of a linear component comprised of the person’s consistency, knowledge, and task predictability and a second component involving the non-linear usage of the cues. "The methodology relies on multiple regression statistics for creating a quantitative representation of the physician’s judgement policy in the form of weights assigned to the variables, used as cues, in the

2

judgement task” (7). It should be pointed out that while this model may well be able to describe the outcome of a physician’s judgement in a given situation, it does not necessarily describe the process by which that clinician is able to arrive at the judgement. Having said that, this model still has significant practical application. Studies of variation among clinicians have suggested that a lot of differences occur as a result of the different weights that different clinicians apply to the same cues (9). As the physicians in the study by Kirwan were senior medical staff (specialist rheumatologists), there is a suggestion that the degree of variation does not decrease with increasing experience. It has been shown, that providing users with a comparison of their own cue weights to the optimal cue weights using computer analysis and graphics when solving a judgement task enhances the learning of complex judgement tasks (10). An important pointer for improved teaching comes from Hammond’s observation that this technique was more effective than providing the learner with outcome feedback or information about the nature of the task given before the trial (feed forward). Similarly, Kirwan et.al. (1983) were able to demonstrate a significant reduction in variability between physicians in the assessment of disease activity in rheumatoid arthritis when they were given feedback on their apparent cue weights compared to optimal cue weights. In contrast, providing outcome feedback made no difference to the congruence between physicians. This method of cognitive feedback via apparent versus optimal cue weights, is especially useful where there is a well-defined problem and clinical pathway. However, it may still be of use in ill-structured problems to reduce variation by allowing comparison against the group norm. In other words, this data suggests criterion-referencing should be undertaken where possible, where the user can compare their approach or activity against the optimal process. Where there are not sufficient validated criteria, then normative referencing, by enabling a user to compare themselves to a peer group, can improve learning. 1.2

LEARNING THROUGH REFLECTION

As referred to above, medicine and its practice is a complex and ill-structured domain (4). One method that has been promoted as a way to learn about practice in professions, such as medicine, is that of reflection (2). Traditional medical education has largely been based on an objectivist view of learning. This model holds that knowledge exists independent of the learner, and is something that can be acquired and transferred between individuals. In contrast, the constructivist view holds that learning is a constructive process, in which the learner builds their own representation of the knowledge space by negotiating multiple perspectives within a given context. This latter representation is internal to the individual and there is no shared reality. Furthermore, the constructivists hold that learning should occur in a realistic setting, reflecting real world situations (11-13). This view is held in the belief that some knowledge is context dependant, and cannot be learned independently of the context in which it is used (14). In terms of professional education, this means that learning should take place in the context of professional practice. Reflection and reflective practice is one constructivist approach that has been promoted as a means to learn from experience (2, 3, 15). The idea of reflective practice has been attributed to Dewey who in 1933 stated that, “reflective thinking is closely related to 3

critical thinking; it is the turning over of a subject in the mind and giving it serious and consecutive consideration” (16 p. 3). Others such as Schön (1983), Boyd and Fales (1983), and Boud et. al. (1985) have continued the development of the theory. In keeping with constructivist notions of learning, Schön holds that professional knowledge involves a degree of what he terms, “artistry” that can only be developed through practice within an appropriate context (2). He promotes the central idea of reflection as the means by which the student and practicing professional can maximise learning. Firstly, he describes the concept of reflection-in-action where one reflects on an activity as the activity unfolds. These reflections then guide the future direction of the activity. This is also referred to as, “thinking on your feet” (p. 26). Secondly, he describes reflection-on-action, where one reflects on the actions that have been taken, such that more appropriate action can be taken should a similar circumstance arise. Thirdly, he also describes reflection on the reflection-in-action. By this, he refers to the ability of an individual to think about what they were thinking at the time the activity was carried out. He believes that this last type of reflection leads to the best learning. Others have described this as reflection-for-action (15). Boyd and Fales (1983 p. 100) define reflection as: “The process of creating and clarifying the meaning of experience (present or past) in terms of self (self in relation to self and self in relation to the world). The outcome of the process is changed conceptual perspective. The experience that is explored and examined to create meaning focuses around or embodies a concern of central importance to the self.” They also state that the, “process of reflection is the core difference between whether a person repeats the same experience several times, becoming highly proficient at one behaviour or learns from the experience in such a way that he or she is cognitively or affectively changed”. That is, by using the reflective process, learners have a greater chance of being able to generalize their knowledge so that it can be applied to new situations and experiences. Boyd and Fales (1983) suggest this ability to generalize knowledge and apply it to new situations differentiates a technician from the professional. In their discussion of reflection, Boud et. al. (1985) have emphasized the affective nature of reflection. They list three elements that they consider are essential to the reflective process. These are, returning to the experience, attending to feelings, and reevaluating experience. By, “returning to experience” they mean recalling the key elements and features of the experience. Attending to feelings has two components. The first is to use positive feelings that may include recollection of positive aspects of the experience, or those aspects associated with enthusiasm for potential benefits derived from processing the events. The second component is to remove obstructive feeling, as only after these have been removed, can a rational consideration of the events be achieved. The third element, “re-evaluating experience” involves examining the experience in light of any new knowledge and integrating it into the learner’s conceptual framework. Based on the work of Boud and others, Atkins and Murphy have outlined a three stage model for reflection (17). The first stage involves an awareness of uncomfortable feelings and thoughts arising from the realization that the current level of knowledge is insufficient to manage the situation. The second stage involves critical analysis of the situation. This is a constructive process that may involve examination of feelings and knowledge. The third stage involves a conceptual change with the

4

development of a new perspective, which may or may not lead to changes in future actions. Over recent years a number of benefits have been asserted for reflective practice. These have been summarized in a recent review by Ruth-Sahd (2003) and include: the integration of theoretical concepts within practice (18, 19), enhanced critical thinking and judgment-making in complex situations (20), increased learning from experience (17), reduced anxiety associated with the learning environment (18), and enhanced selfesteem and acceptance of professional responsibility (21). Most of the studies regarding the outcome of reflective learning have involved qualitative methods. However, a quantitative study using post-intervention testing only, found no significant difference in the test scores achieved by those who had received teaching using reflective methods, compared to those who had received more traditional teaching (22). The authors did note that the teaching methods were new to the students, and for a new curriculum, it is significant that they performed no worse. Therefore, while unable to demonstrate an outcome advantage for the reflective approach, these authors were still in favour of using this approach in association with traditional methods. In their review of the literature, Atkins and Murphy (1993) have identified five skills they considered necessary for reflective practice and learning. Firstly, learners need to be self-aware so that they can honestly examine how a situation has affected them, and how they have affected the situation. Secondly, learners need to be able to give a comprehensive account of the experience. Thirdly, learners must be able to critically analyse an experience by being able to identify existing knowledge, challenge assumptions and explore alternatives. Fourthly, they must be able to synthesize and integrate information. Lastly, they need to be able to evaluate their new perspective. They go on to suggest that attention needs to be given to the development of these skills for reflective abilities to be engendered. Having described briefly the practice, process and possible benefits of reflection, and the skills required for reflection, one must consider possible strategies for supporting reflection. Journal-writing has commonly been used (15, 23) on the basis that, “it may enable practitioners to make explicit the knowledge that is implicit in their actions” (17). Other commonly used methods have been reflective teaching portfolios, and dialogues or debriefings where learners can discuss their experiences as a group (15, 23). Imel (1992) has listed a number of reflective processes that can also be used to support learners. These include: • Questioning what, why and how things are done. • Asking what, why and how others do things. • Seeking alternatives. • Keeping an open mind. • Seeking the underlying rationale or theoretical basis. • Viewing from various perspectives. • Asking, “what if…?” • Asking for the opinion and viewpoints of others. • Considering consequences • Hypothesizing. • Synthesizing and testing. • Seeking identifying and resolving problems.

5

If, as previously mentioned, medical knowledge and medical problems represent illstructured domains (4), then reflection provides a tool that enables practitioners to continuously re-structure their own cognitive model of the knowledge space based on experience and new information, which in turn leads to increased understanding. Schön (1987) saw reflective learning as having application to a wide range of professions, including the health professions. While it has been present in teacher education for some time (23), and has been proposed as having an important role to play in medical education and professional development (24), it has generally been underutilized in medical education (3). In distinct contrast, the nursing profession, which has similar difficulties with the so called “theory-practice gap” (23), has widely adopted the reflective-learning model (15, 18, 21, 25). In fact, the Australasian Nurse Registering Authorities have gone so far as to include this as a domain within their competency statements (23). Given the potential for improved learning within the medical domain when using reflection, and considering that the reflective approach results in outcomes no worse than traditional methods, the design of effective learning environments should include support for the process of reflection. 1.3

SIMULATED MEDICAL LEARNING ENVIRONMENTS

Traditional training in medicine involves a long period of formal education followed by, or in association with, an apprenticeship that involves practice on human beings (26). Unfortunately, in recent years there has been a reduction in the time available to physicians for teaching students, as well as a reduction in the availability of patients as an educational resource. A number of factors have led to this situation: many patients are now being managed outside of the teaching hospital, clinical periods for students have been considerably shortened, and increasing medical knowledge has resulted in the addition of new subject areas to the curriculum. Furthermore, in some cases, ethical considerations make some patients unsuitable for students (26-28). Having graduated, medical practitioners, like other professionals, continue learning through the whole of working life (29). This has traditionally involved modalities such as rounds, educational meetings, conferences, refresher courses, seminars, lectures, workshops, and symposia, often similar to those used in traditional basic medical education. Unfortunately, these modalities are often ineffective in mediating improvements in patient care through changes in physician behaviour (1). On the other hand, interactive continuing medical education sessions that enable the participant to practice skills can effect behavioural change (1). This combination of factors has led us to explore the potential role of simulation-based teaching systems to support long-term learning for medical practitioners. When simulating human medical diseases and conditions, the most authentic environment is achieved by the use of trained actors (4). Attempts have been made at developing alternative simulation environments, including paper-based systems (30, 31). With recent improvements in technology it has become possible to develop computer-based simulations or so-called virtual patients, although there has been considerable variation in how these have been structured (32, 33). Initial computerbased simulations were electronic versions of their paper-based counter-parts but with advancements in technology, rich multimedia environments have been developed (34). With the development of the Internet and the World Wide Web, multimedia environments have become readily accessible over a widely distributed network. This 6

enables much wider and more flexible access to these resources. However, web-based simulations have usually lacked the features and richness found on desktop systems (35) due to limitations of bandwidth, browsers and cross-platform programming. While learning theory might suggest that simulation-based teaching systems are an appropriate method of supporting learning, they do have their limitations. With previous medical case simulations, computer-based or otherwise, it has been consistently observed that there is considerable case-specificity (6, 36). That is, medical practitioners or students may be able to perform one case well but do poorly on a different but not altogether, different case. One approach to this problem has been to increase the number and breadth of the cases when using them as an evaluation tool for assessing clinical problem-solving skills (32, 37). Based on their studies of simulations using trained actors, Elstein, Shulman and Spafka (1978) found that the differences between the experts and the less expert problem solvers was to be found in the detail of their cognitive models consequent to their experience. Of particular interest, they found no statistically significant correlation between the thoroughness of data collection and the accuracy of data interpretation. Furthermore, Anderson et al. (1995b) suggest that optimal learning occurs when a combination of abstract and situation-specific training is provided, and that abstraction promotes the transfer of knowledge and thus insight from one situation to another. As they describe it, abstract training refers to the teaching of general principles, which are often a component of larger activities, and can be applied to a number of contexts. For example, in the situation of medical practice, one might provide instruction on the elements of good history-taking. For, without the skill of thorough history-taking, it is not possible to diagnose or manage complex medical problems. In the context of a simulation-based learning system, this means that we should take care to go beyond a specific simulation of one or more cases. Based on the work of Schön (1987), and as alluded to above, encouraging the user to reflect on their activity and their own thought processes, may assist with the process of abstraction and transfer. While diagnosis, involving history taking, physical examination, and review of investigations, is critical to the appropriate management of the patient, it is only part of the consultation. The other component is medical decision-making. “Medical decisionmaking occurs throughout the diagnostic and treatment process. It involves the ordering of additional tests, requests for consults, and decisions regarding prognosis and treatment” (38). For effective medical decision-making and management of the patient, there must be an understanding of the natural history and pathophysiology of the disease process (38), as well as a broad knowledge of appropriate management strategies. A limitation of most computer-based medical simulations is that they have been developed with an emphasis on medical diagnosis, and generally involve a single patient encounter. In contrast, much patient morbidity is associated with chronic diseases, such as diabetes mellitus and cardiovascular disease. These chronic disorders evolve over time and involve multiple doctor-patient encounters. These involve interactions of diagnosis and management that must be modulated or altered by the doctor and patient, depending on the development, progression and control of the disease and the emergence of superimposed pathology. Thus any simulation that intends to more accurately simulate actual practice should include this as a design element within the application model.

7

1.4

REFLECTION AND THE CONSULTATION PROCESS

In light of the importance of reflection for improved learning in medical practice, the following model of reflection, in association with the consultative process, is introduced. This model has two parts, a consultative process, and an associated layer of reflection. This section describes both of these in terms of Figure 1, and begins with a description of the consultation process, based on standard medical models. This is followed by a description of the reflective overlay, a new model, developed as a foundation for the design of SIMPRAC. Reflection

Consultation Process Initial information

Hypotheses / Diagnosis

Sufficient to manage patient?

Refine hypothesis

Seek additional data

Generate Hypotheses

How data affects hypotheses

Reflect on data gathering

Reflect on current knowledge

No

Yes

Manage patient

Reflect on management

Response to management

Reflect on response to management

Figure 1: Model of the consultation process. Reflective processes explicitly supported in SIMPRAC are shaded grey. While based on the hypothetico-deductive approach to medical inquiry, the model for the consultation process is not concerned with specific ideas regarding aetiological processes, so it also can be applied to the medical scripts approach. In the latter, the initial information cues a relatively mature set of hypotheses regarding the presence of a particular disease processes, and subsequent data gathering is aimed at confirming that diagnosis and excluding alternative explanations. For example, suppose a patient presents with severe crushing chest pain. If a hypothetico-deductive approach is taken, then one would generate a series of hypotheses regarding the chest pain. For instance, the pain may be considered to be coming from the heart, the oesophagus, other mediastinal structures, or radiating from the abdomen. Further questioning would then be undertaken to refine the hypotheses. If exploring the heart as a source of pain, one could ask about radiation of the pain to the neck or arms, associated shortness of breath, or a sensation of palpitations. The hypotheses might then be refined to include such possibilities as myocardial ischemia, myocardial inflammation (myocarditis), or dissection of the aorta. Having refined and

8

reformulated their hypotheses, the clinician continues gathering data in an iterative process. If, on the other hand, a medical scripts approach is taken, the clinician may recognize the pain as being typical ischaemic chest pain, suggesting acute myocardial infarction. Although there is not a lot of backward chaining, as occurs with the hypotheticodeductive approach, further clinical information is gathered to confirm the diagnosis and exclude other causes of chest pain. This would include serum Troponin, and electrocardiograph to confirm the diagnosis, as well as a chest X-ray to exclude a widened mediastinum, that might be seen as a result of aortic dissection. In both hypothetico-deductive and illness scripts approach, a hypothesis or diagnosis is formed, further information is obtained to support or refute the hypothesis or diagnosis, and at some point, a decision is made that there is sufficient information to institute a particular course of management in preference to other available options. This model can be used for a single consultation or a series of consultations with a number of management cycles. In terms of multiple consultations, the first consultation will generally be much longer than subsequent consultations, as the first consultation is used to formulate the initial diagnostic hypotheses, and typically involves a lot of history taking, physical examination, and ordering of investigations. This first consultation also orients the clinician to the social and medical context of the patient. Subsequent consultations are targeted at reviewing the response to treatment and looking for side-effects, as well as checking for complications based on knowledge of the natural history of the disease process. In general, this will take less time than is required for the first consultation. Nevertheless, despite the different focus of each consultation, the process is the same. That is, data is gathered and hypotheses generated, which in turn may lead to medical decision-making and management, or further data gathering. With the above model, one then needs to consider how each of the various processes can be used as a source of reflection. Firstly, hypothesis generation is itself a reflective process, in that the leaner needs to reflect on what they know of the patient, and what potential conditions may cause the observed symptoms and signs. At a deeper level, the learner may reflect on how the patient responded to their enquiries, as a clue to the underlying pathophysiology. For example, anger expressed by the patient may be an indication of underlying fear that needs to be addressed. Having developed a set of hypotheses, the clinician can then reflect on whether there is sufficient data, and whether the data supports or refutes existing hypotheses, or whether it leads to new hypotheses regarding the patient (reflection-in-action). If the decision is made that there is insufficient data, further data gathering can be undertaken (see decision diamond in Figure 1). The clinician can then reflect on the appropriateness of this data-gathering (top right box), as well as how this process relates to the hypotheses that have been generated (arrow from left box to right box). That is, how have the questions, examinations, and investigations contributed to their hypotheses? Are their more effective ways of eliciting the information they desired? Has the way they have asked their questions inhibited or directed the patient’s response? How much weight have they given each of these

9

responses when formulating their hypotheses? These are examples of reflection on reflection, which leads to the best learning (2). If the clinician decides there is sufficient data to proceed with management, they can then reflect on such things as what the various management issues are for the patient, how these can be best addressed, whether they have sufficient knowledge about best management practices (reflection-in-action, and reflection-on-action), and even such things as whether their own cultural or emotional biases are impacting on their management decisions (reflection-for-action) (see bottom left grey box). Finally, having instituted various treatments and observed the patient’s response, the practitioner can again reflect on their management strategies, or such things as the natural history of the disease processes, and how these have influenced the patient’s outcome (see bottom right grey box). This in turn may lead to further research and learning. From the descriptions above, it can be seen that a whole layer of reflection can be developed and overlaid upon the consultation processes. 1.5

PROJECT

With the above foundation, this report describes a learning environment called SIMPRAC, which engages practitioners and students in an authentic activity, and supports a reflective learning process. The authentic activity has been based around the diagnosis and management of a virtual patient. With SIMPRAC, we have: 1. Developed of a model for reflection that overlays the consultation process. 2. Developed, a medical consultation environment for the diagnosis and management of chronic disorders that: • Evolve over a series of consultations. • Include multiple pathways. 3. Developed a learning environment that supports multiple points of reflection through: • The maintenance of a user model that is actively reviewed by the user, and enables the user to compare their approach with that of an expert or with their peer group. • Iterative user hypothesis generation and refinement during diagnosis. • Feedback based on patient outcome, in response to management decisions made by the user. Chapter Two provides a description of the web-based virtual patient application from a user’s perspective. This includes overviews of both the user-patient interaction elements, and case review elements. Chapter Three describes the system architecture and includes details on the user model, and the disease model for multiple consultations. This chapter also describes the architecture for the question and answer interface used by the user during patient interactions.

10

11

System Overview SIMPRAC was developed to explore ways to improve learning through reflecting in a learning-environment that enables medical students or postgraduate medical practitioners to undertake a series of simulated consultations involving the diagnosis and management of chronic illness. 1. Login

Login / Registration

Select Profession

Select Module

Select Case

2 Patient Interaction / Consultation

History

Medical Record

Hypotheses

User Notes

Consultation Iteration

Library

Examination

3. Review

Investigation

Charts

Management

Score Actions

Elements supporting reflection

Figure 2: System overview with three functional units. From a user’s perspective, the system can be considered from three functional units, as shown in Figure 2. 1. General housekeeping functions such as logging in a user and selecting a case. 2. Enabling the user to undertake a virtual encounter with the patient. 3. Enabling the user to review their activity and progress, as well as compare their activity to other users. This chapter is primarily an overview of the system from the point of view of the user. While it is presented as a typical user interaction with a series of screen shots, it also describes in Section 0 and Section 1.8, those elements that have been included to support reflection. These elements are shown in grey in Figure 2. The design decisions that were made with respect to the user levels are also described (Section 0). The chapter has been structured to match the functional units listed above. The first section describes how a user is able to login to the application and select an appropriate case. The second section describes how the user is able to interact with the patient, as well as the tools available to the user to assist them with this interaction. Finally, the last section describes the user interface that has been developed to enable users to reflect on their actions and compare their activity with others.

1.6

LOGIN AND CASE SELECTION

The user is initially presented with a screen from which they can login to the application (Figure 3). If the user has not used the application previously, they have an opportunity to register as a new user (Figure 4).

Figure 3: The user must login to the application.

Figure 4: If this is the first time that the user has used the application then they must first register with the application.

12

Figure 5: Having registered with the application and logged onto the system, the user must then select their professional background. Learner Groups Having registered or logged in, the user is asked to select their professional background (Figure 5). The options currently available for selecting a user’s professional background are Medical Student, General Practitioner, or Expert, although it is possible to have any number of professional backgrounds encoded. These three groups were chosen during the development of the software because the initial goal was to provide continuing medical education for medical practitioners, especially general practitioners, on the diagnosis and management of chronic illnesses. Due to the need for evaluation, and the perceived applicability of this method of instruction to medical student education, the scope was broadened from just general practitioners. Furthermore, the “Expert” group was required to enable at least one expert to use the software for any given case so that other users could compare their activity to this individual. Additional specialists may register as an expert but only one is used to compare with other users. The selection of this one expert is currently hard-coded within the database for each case. It was also hypothesized that Medical Students, General Practitioners, and Specialists in the management of hyperlipidaemia, would have different approaches to the case, which would be reflected in their diagnostic and management choices. Therefore, as part of this exploratory investigation, these three groups were chosen. The users’ selection of their peer group determines the group comparisons available to the user during the consultation review. As is discussed in more detail later, a user is only able to compare their activity against the average for their peer group, or a single expert.

13

Figure 6: The user can choose from different knowledge domains. There are currently two modules. The first is a demonstration module used to help familiarize users with the interface. The second is the Clinical Biochemistry module that includes the case used during the user evaluations.

Figure 7: The design of SIMPRAC allows several modules with several cases within each module. The current system has only one case available within the Clinical Biochemistry module.

14

Modules and Cases Once the user has selected their professional background, they can then select the module of interest, and the specific case they wish to undertake. A module represents a group of cases covering a limited domain. These modules can be narrowly or broadly defined. Given the limited number of cases, arbitrary, and rather broad categories were chosen. As shown in Figure 6 on page 14, two modules have been defined. The first is the demonstration module containing a case that is used to familiarize users with the software. The second is the “Clinical Biochemistry” module that contains a single case concerning the diagnosis and management of hyperlipidaemia. This module name is defined by the author and is unrestricted, it could have just as easily have been titled, “Diagnosis and Management of Lipid Disorders” if all the cases were limited to this field of medical practice. The case involves a 29 year old female, who referred herself for management of hypertriglyceridaemia. The patient has a known history of the rare disorder, Type 1B Glycogen Storage Disease (GSD) characterised by short stature, characteristic “doll like” appearance, fasting hypoglycaemia, neutropenia, hypertriglyceridaemia, hyperuricaemia, and hepatomegaly with abnormal liver function tests. This disorder is caused by a defect in the hepatic microsomal glucose 6-phosphate transport system, and sufferers are not able to convert glucose 6-phosphate to glucose in the liver. At presentation, the hypertriglyceridaemia is at such a level that it places the patient at risk of developing pancreatitis. On the other hand, administration of an effective triglyceride lowering agent such as Gemfibrozil is contra-indicated in the presence of liver disease. Type 1B Glycogen Storage Disease is optimally managed by maintaining the fasting blood glucose level with the regular administration of cornstarch. With the exception of neutropenia, most of the metabolic derangements can be corrected by maintaining the blood glucose (39). Unfortunately, most patients have difficulty maintaining this therapeutic regimen because uncooked cornstarch is not very palatable. The first case was chosen because: • It was a within the domain of the author’s medical experience. • It is an uncommon condition, which most users would not have seen in their previous clinical practice. Therefore, it would be unlikely that any single user would have an advantage over any other user, making results comparable. • This case is typical of the complex issues facing the clinician, where one must balance the risks of treatment with the expected benefit. Moreover, these risks and benefits vary over time, and with respect to different individual’s response to treatment. It was hoped that users would: 1. Identify that the patient had massive hypertriglyceridaemia, and to a lesser extent, hypercholesterolaemia, secondary to Type 1B GSD by asking about the patient’s presenting problem, asking about the patient’s past medical history, and requesting triglyceride and cholesterol measurements. 2. Exclude other secondary causes of hyperlipidaemia, including excessive ethanol use, renal failure, liver disease, obesity, and diabetes. This could be achieved by asking about the patient’s past medical history, and ethanol intake, as well as by measuring the patient’s height and weight, and requesting relevant pathology, including electrolytes, creatinine, and liver-function tests.

15

3.

4. 5. 6. 7. 8. 9. 1.7

Consider cardiovascular risk factors such as smoking and hypertension by asking about tobacco use, as well as, past medical history, including current medications. Consider primary causes of hypertriceridaemia, as might be suggested by a family history of disease. Manage the GSD with strategies to maintain blood glucose. Especially the use of low glycaemia index supplements such as uncooked cornstarch. Consider the patients diet from the point of view of GSD and hypertriglyceridaemia Institute lifestyle changes such as regular exercise. Commence treatment with an appropriate lipid-lowering agent, with the preferred agent belonging to the fibrate therapeutic class. Recognize the side-effects of the therapeutic agents, and monitor the patient accordingly. PATIENT INTERACTION

The patient interaction element can be divided into two major sub-components: a diagnostic component, and a management component. These were implemented with the goal of providing an environment that would be sufficiently close to a standard medical consultation, such that it could be used as an “authentic” learning activity. The goal was not to make a perfect simulation, but one that would be close enough to enable the lessons learnt using the simulator to have a strong likelihood of being abstracted and transferred to actual medical practice. Diagnostic elements The virtual patient interaction begins with a short vignette, as shown in Figure 8, which introduces the problem to the user. Users then have the opportunity to ask questions and take a history from the patient, perform a physical examination, order investigations and review results, and to choose a variety of management options. Consistent with realworld clinical practice, history taking, physical examination, and the ordering of investigations can be undertaken in any sequence. To help users reflect on their diagnostic hypotheses, users are presented with a screen that asks them to state or update their hypotheses. This screen is displayed when the user chooses to perform a physical examination, request investigations, or select management options for the first time in each consultation.

16

Figure 8: At the beginning of the case the user is given a short introduction to the patient, as a primer to further interaction. The navigation frame on the left also provides access to a number of support functions including the user’s hypotheses regarding the patient, the medical record, library resources, and user notes. The review function is only available at the end of the first consultation. History Given the large number of possible questions a patient can be asked, and the large number of variations in which these questions can be phrased, the history-taking element was challenging to implement. Rather than allowing free-form questions and using natural language processing, a restricted set of allowable questions have been maintained in a database. These questions have been further categorized according to an arbitrary schema, thought to be reasonable based both on the past medical experience of the author, and categories used by various sources such as textbooks on clinical practice (40, 41). Example question categories include the history of the presenting complaint, past medical history, lifestyle and social questions, and so on. The patient’s history can be taken either by asking short questions with keywords, or by selecting from a list of questions by category (Figure 9). When asking questions using free text, a list of possible matching questions is displayed (Figure 10). If the system is not able to find a sufficiently close match, then the user is asked to rephrase the question. The response to the question is obtained by clicking on the desired question, which is presented as a hypertext link (Figure 12).

17

Figure 9: Questions can be asked by either entering free text or keywords at the top of the screen or selecting from one of the question categories provided within the drop down list box.

Figure 10: If a question is entered as free text, a series of matching questions from the database is provided. In this instance the user has asked the open-ended question, “why have you come today?” The search engine has found two matching questions from the question database and has listed these as hypertext links below the search field and question category drop-down box. To ask the question the user must then click on the desired hypertext link. If the user clicks on, “Can you tell me why you have come today?” they will be shown the response seen in Figure 12.

18

Figure 11: In addition to using the free text mode of questioning, questions can also be asked by selecting from a category. In this example, the user has selected the “Present Illness” category, and has been shown all the questions under this category. Again the user may click on the link, “Can you tell me why you have come today?” to be shown the response in Figure 12.

Figure 12: The user has clicked on the hypertext link representing the question, “Can you tell me why you have come today?” To which the patient has responded, “Last week while at the shopping centre …”

19

Examination The examination screen was heavily influenced by the design of other case simulations, especially the product known as DxR (42). To examine the patient, the user needs to select an appropriate view of the body, with the default being the full body view. They are then able to select one of the examination tools by clicking on the tool’s icon. As illustrated in Figure 13, the palpation tool has been selected. This is represented by the hand on the sixth button from the left in the top row. By clicking over the desired part of the body, in this case the right hypochondrium, the findings are displayed in the frame to the right. If an inappropriate selection is made, for example, using the sphygmomanometer on the head, where it is normally used for measuring blood pressure by being applied to the upper arm, the application reports that no information is available. Clinical information that does not apply to a single part of the body, for example the height and weight, can be obtained by selecting from a list that is displayed after clicking the “other” examinations button (Figure 15).

Figure 13: Palpation of the liver in the right hypochondrium has revealed that the liver is enlarged and extends 8cm below the costal margin. Placing the mouse over a tool button displays a text description of that button.

20

Figure 14: Measurement of blood pressure using the sphygmomanometer tool.

Figure 15: More general examinations, not involving a specific part of the body. Investigations Mimicking the clinical situation, a full range of laboratory and other investigations are available to the user. These investigations have been grouped into collections commonly found in clinical practice. Clicking on a group displays a list of investigations under that particular category. For example, if “Haematology” is clicked, then a list of haematology investigations, such as full blood count and simple coagulation studies is displayed to the user. Very few test panels such as FBC (full blood count), EUC (electrolytes, urea, and creatinine), and LFT (liver function tests) have been allowed. 21

Those that have been used are consistent with those allowed under the Australian Medical Benefits Schedule guidelines (43). As an alternative to selecting from the preconfigured lists, the user may enter the first few letters of the test name to search for a test from the entire database of tests. Tests beginning with the matching sequence of text are then displayed in a list (Figure 16). When the user selects the investigations from the list, they are then shown a screen with the investigations grouped as either having results available for review, or not available for review (Figure 17). Only those investigations such as electrolytes and creatinine, which are widely and rapidly available, are immediately available for review. All others, that in clinical practice usually require more time to be completed, will not have their results visible until the next consultation. If the results are available for review, then they can be viewed by clicking on the corresponding test name (Figure 18). The decision not to have all results available at the time of request, was a compromise between having all results available, and having no results available. In actual clinical practice, a medical practitioner will not have the results of investigations available at the first consultation, and will need to make management decisions on incomplete investigations. Often, that decision will be to do nothing until the results are available. At subsequent consultations, a practitioner may give the patient a request form for follow up investigations to be undertaken a day or two before the patient returns for their next consultation. In this way, the results are available to the clinician at the time of the consultation. From a learning perspective, it was felt that it was not necessary to try to model the full complexity of the clinical situation. Instead, the approach described above was taken.

Figure 16: Requesting investigations. In this screen the user has search for tests containing the sequence of letters, “chol”. From the displayed list they have then selected three tests that will be requested once they click on the select button.

22

Figure 17: As tests are selected they are added to one of two lists, those tests for which results are currently available, and those tests for which results will not be available until the next consultation. In this example, all the tests are available and they have been listed in alphabetical order.

Figure 18: The results of the requested investigations are available by clicking on the relevant item from the list of available results. When a test is chosen (refer to previous Figure), all the test results, for that consultation, are displayed in a two column table. The left column contains the test or test-group names, while the right column contains the result for each test.

23

Management elements One of the features of SIMPRAC is its focus on patient management that may vary over a number of consultations, based on the patient’s response to the management strategies chosen by the user. This section provides a description of the management interface, as well as showing how different strategies can lead to different patient outcomes. In order to illustrate the management issues, this section also provides a more detailed description of the medical case.

Figure 19: Management options screen where options have been matched against the search term, “diet”. The user has already chosen to use a fibrate, and a diet low in saturated fat. This user has also decided to see the patient again in thirty days. Once the user believes they have obtained sufficient information, they may then choose the appropriate management for the patient (Figure 19). In a manner similar to the process of history taking, the user enters keywords matching the desired management activity. Matching options are displayed in a list and can be selected by the user. Once the user is happy with their list of proposed management activities, they then choose a follow-up period and save the information, thereby concluding the consultation. The list of management options selected by the user is maintained across consultations, and to stop some management activity, it must be deleted from the list of selected options. The follow-up period does not influence the outcome in any way but was originally included so that a virtual time line could be maintained. Following the initial evaluations this approach was no longer maintained, as it made case construction very difficult. Having a time line, added another variable that increased the combinatorial complexity, but it was felt that this did not enhance learning in the context of chronic illness. Such an approach might be worth considering when looking at acute medical emergencies, where time dependence is critical. However, this was outside the scope of the current project.

24

In order to model multiple patient outcomes, each patient state is dependent on two variables. One variable is the consultation. That is the number of times the patient has been seen. The other is a variable denoted by the term “stream”. This concept is illustrated graphically in Figure 20. Other than the first consultation, where only one stream is defined, there may be one or more parallel streams representing the different outcomes for the patient in each consultation. The outcome for the patient in the following consultation is determined by the management options selected in the previous consultation. Any significant events affecting the patient are given to the user in the vignette for the following consultation. For example, in the biochemistry case used during the evaluation of SIMPRAC, if the user has not prescribed an effective lipid-lowering agent such as Gemfibrozil, or an HMG CoA Reductase Inhibitor at the end of consultation one, at the beginning of the second consultation they are simply informed that the patient has returned for review. This is represented by the state having stream 1 at consultation 2 in Figure 20. If the lipids are re-checked during this second consultation, the user will not see any significant changes in the results from the first consultation. If the user still does not instigate the use of an effective agent at the end of the second consultation, then at the beginning of the third consultation, the user is informed that the patient has suffered an Actions Red: No action or cornstarch alone Green: Fishoil alone or with cornstarch Blue: Statin alone or in combination with fishoil or cornstarch Grey: Fibrate alone or in combination with any other treatment

Initial Presentation

Consultation 1

Stream: 1

No change from baseline

Mild reduction in triglyceride

Predominant reduction in cholesterol

Predominant reduction in triglycerides

Stream: 1

Stream: 2

Stream: 3

Stream: 4

Consultation 2

First episode of pancreatitis [a]

No change from baseline

Mild reduction in triglyceride

Predominant reduction in cholesterol

Predominant reduction in triglycerides

Stream: 5

Stream: 1

Stream: 2

Stream: 3

Stream: 4

Further pancreatitis + lawsuit [b]

First episode of pancreatitis [a]

No change from baseline

Mild reduction in triglyceride

Predominant reduction in cholesterol

Predominant reduction in triglycerides

Side effect of fibrate [c]

Stream: 7

Stream: 5

Stream: 1

Stream: 2

Stream: 3

Stream: 4

Stream: 6

Consultation 3

Consultation 4

[a] Pancreatitis occurs if there have been two consecutive consultations without effective treatment. [b] Negligence if the patient has suffered pancreatitis and no effective treatment commenced [c] Pancytopenia develops if fibrate used on two consecutive consultations.

Thick line represents the preferred action

Figure 20: State transitions through the first case that has four consultations. Each box represents one particular state for the patient in each consultation. The transition from one state to another is determined by the management actions selected by the user. The bold lines represent the transitions expected by the case designer. The order of the streams in this diagram is dictated by their relationship to the streams in the preceding consultation. episode of pancreatitis (stream 5 at consultation 3 in Figure 20). They are also informed that pancreatitis can be a complication of hypertriglyceridaemia, which is present in this patient secondary to her underlying metabolic disorder. Alternatively, if the patient has been prescribed Gemfibrozil in the first, second, and third consultation, then at the beginning of the fourth consultation, the user will receive a letter from the patient’s

25

Immunologist informing them that the patient suffered pancytopenia as an idiosyncratic side-effect of the drug (stream 6 at consultation 4). Pancytopenia is a reduction in the number of red cells, white cells, and platelets in the blood. The possible states for the first case are shown in Figure 20. The case author’s preferred path is shown by following the thick line. If this path is followed, the user does suffer an adverse event due to an idiosyncratic drug reaction. As a result of the complexity and poor compliance of this patient, it is not possible for the patient to be cured and their biochemical parameters normalised. User Support During the patient interaction a number of elements are available to users to support them while they use the application. Each of these will be described in turn. Hypothesis As alluded to previously, when moving to the examination, investigation or management screens for the first time in a consultation, users are asked to update their diagnostic hypotheses (Figure 21). Hypotheses can be added by entering the relevant information and then clicking the add button. Alternatively, they can click on “Save” which will save their hypotheses and direct them to the screen for the currently selected

Figure 21: The hypothesis screen. In this example the user has listed two hypotheses. The first is that the patient has Type 1B Glycogen Storage Disease, and this is thought to be 100% true based on the history from the patient. The second hypothesis is that the patient has hypertriglyceridaemia secondary to their Glycogen Storage Disease. stage of the consultation (physical examination, investigations, or management selection). Access to the hypothesis screen is available at anytime though the hyperlink in the left navigation-frame and hypotheses can be edited or deleted as required. Users are encouraged to enter the hypothesis itself, the likelihood with which they believe the 26

statement is true, and the reason for holding the hypothesis. Nevertheless, only the statement of the hypothesis is compulsory when adding a new hypothesis. For the purposes of the software evaluation, it was not compulsory for the user to enter a probability or enter their reasons for their hypotheses, because early evaluations suggested that some users become frustrated when forced to enter this information. The Hypothesis Screen was included early in the design of this application, as a means for learners to reflect on their current knowledge and guide further assessment of the patient. Its use is consistent with the hypothetico-deductive approach to medical problesolving taught to medical students, and included in our model of the consultative process (see Figure 1). Medical Record The medical record component comes in two parts, and is included, primarily, as an aid to memory. The first part is a free-text area named Medical Notes, which enables the user to record their history and examination findings, much like a standard paper-based medical record. This was implemented as a simple text area that does not allow any text formatting or highlighting (Figure 22). The second part is a screen from which the results of investigations can be viewed. In this way the medical record tool can be used to re-orient a user to their previous line of thought when recommencing a case. For example, if the user does not want to complete the case in a single sitting, they can record their findings in the medical record, and review it at their next visit, before continuing with the case.

Figure 22: The medical record component is available as a separate window, and enables the user to record their observations about the patient. This component was included to encourage the learner to record the medical history of the patient. In addition to being part of normal practice during a consultation, documentation of the history makes explicit what the doctor or learner knows about the

27

patient. Reflection on the information contained in the record, is a useful way of integrating that information, and directing further enquiry. User Notes The user notes component was included as a text-based scratchpad onto which the user could record additional information and observations if they choose to do so. This component is accessible via a hyperlink from the main screen, or via a button on the Medical Notes popup screen. It was envisaged that a user might use this to record any learning objectives that may have arisen from undertaking the case. While such a device is not necessarily found as a specific artefact in normal clinical practice, it was hoped, that when used with the medical record component, it would encourage reflection on the part of the learner. For example, having taken a history and learning that the patient had Type 1B Glycogen Storage Disease, this fact would ordinarily be entered in the medical record. Reflection on this information might then trigger the learner to consider such things as: the natural history of the disease, the complications associated with the disease, the mode of inheritance and risk to the patient’s offspring, how the disease might be managed, as well as the psychological impact on the patient of having a chronic illness of this nature. These reflections could all be recorded in the user notes for later research and learning.

Figure 23: The user notes screen has been included as a scratch pad to enable the user to record their ideas or items to follow up. In this example, the user is unfamiliar with the major problems affecting the patient and has recorded a number of items that need to be reviewed. Library and Information Resources The library component enables the case author to provide links to a variety of learning resources. It currently links to PubMed and the University of Sydney Library. It was envisaged that this component would exist as a portal to generic learning resources, as well as case-specific information that was updated on a regular basis. For the current main case, there are no links.

28

1.8

REVIEW

The third main part of the interaction involves those elements designed to encourage reflection and enable users to review their activity. At the end of each consultation, each user is helped to begin the reflective process by being shown a page that lists all their questions, examinations, investigations, and management selections for that consultation. For each item, the user is required to classify each of their actions as one of: critical, relevant, or not relevant to the diagnosis and management of the patient in light of their current knowledge (Figure 24). This component forces users to review and reconsider their actions. The classification levels and the reason for choosing these levels are discussed in more detail in Section 0 on page 38.

Figure 24: At the completion of the consultation, the user must classify their actions and selections as being critical, relevant, or not relevant to the diagnosis or management of the patient, based on their current knowledge. Only after completing the scoring process, the user is shown a second screen that enables them to compare their activity to the: • average activity for their peer group (Figure 28), or • specific activity of a single expert in this field (Figure 25). The information is presented as a chart displaying the number of critical, relevant, or not relevant items selected by the user, for each diagnostic and management stage, within each consultation. This same information can be viewed as either a column chart (Figure 25), pie chart (Figure 26) or line chart (Figure 27), according to user preference. The inclusion of multiple forms of presentation was done to evaluate which form was favoured by users.

29

Figure 25: The graphical review screen presents the user with an overview of the proportion of critical, relevant, or not relevant actions they selected during the consultation. In this example, the user with username, “student8” (see legend at upper right) has compared themselves with the case “Expert” for the “History” stage of the consultation. The comparison group is selected by using the radiobuttons at the bottom left of the screen, while the stage of the consultation is chosen using the drop-down box at the middle left of the screen. Student8 has selected six critical (yellow), six relevant (purple) and six irrelevant (cyan) questions in the first consultation. In contrast the “Expert” asked seven critical (brown), seven relevant (green) and one irrelevant question (pink). In subsequent consultations, student8 and the expert have been in different streams, so comparable data is not available for the expert. For the first case, the author was also the “Expert” who attempted the case, and provided the activity record against which other users could compare. It should be noted that this need not be the case. When authoring a case, the author must classify each history, examination, investigation, and management element for each patient state. In contrast, the “Expert” interacts with the virtual patient, as would any other user. Subsequent users can then compare their own activity to the record of the “Expert”. Separating the author classification from the “Experts” activity has at least two benefits. First, it overcomes the start up problem, by providing comparative data for the first user. Second, it enables the element classification to be revised. Such revision may take place in the light of expert consensus, or even after actual use by learners. In this instance, having the author and “Expert” as the one person was done for pragmatic reasons, and while it increased the correlation between the “Expert’s” activity and number or critical and relevant items, it enabled evaluation to begin as early as possible without having to wait for another domain expert to undertake the case before other users. The author has had specialist training in the management of hyperlipidaemia and has previously managed a patient with Type 1B Glycogen Storage Disease.

30

Figure 26: The user’s activity data can also be viewed as a pie chart. In this example, student8 has compared themselves to the expert in the first consultation. The student has chosen relatively more questions that were regarded by the case By right clicking a column within a particular consultation and stage of the column chart, the user is able to retrieve specific textual information related to that consultation and stage combination. As illustrated in Figure 29, if the user right-clicks on the first consultation concerning the history questions, they are able to display a table that includes: 1. critical questions asked by the user, 2. critical questions not asked by the user, 3. relevant questions asked by the user, 4. relevant questions not asked by the user, and 5. questions asked by the user, that the case authors regarded as not relevant. This facility was included based on specific feedback from users, indicating that the provision of this information was essential for appropriate review (See Chapter 5). Once they have reviewed their activity, the user is able to move to the next consultation for the case (Figure 30). This cycle of patient interaction and consultation review is continued for a predetermined number of consultations for each case (see Figure 2 on page 11).

31

Figure 27: In addition to bar and pie charts, the user can also view their activity as a line chart. This is the same data as in the bar chart above.

32

Figure 28: In addition to comparing their activity to an expert, the user can also compare themselves with the average for their peer group by choosing the appropriately labelled radio-button at the bottom left of the screen. In this particular case, again reading from the right in consultation one (starting with the yellow column), student8 has selected six critical, six relevant and six irrelevant questions while on average, their peer group (medical students) have asked, five critical, five relevant, and eleven irrelevant questions during the first consultation. For this comparison to occur, this user had to indicate they were in the medical student peer group when they logged in (see Figure 5 on page 13).

33

Figure 29: Details of the questions that were asked are available by right clicking over the column chart for the consultation of interest. In this case, the user asked six questions that the author thought were critical questions. However, there were also three questions that the author thought were critical questions, but student8 did not ask.

Figure 30: After they have completed the consultation the user may return to the review screen, proceed to the next consultation or logout and return to the case at a later time. At the time of writing it was not possible for the user to start a new case without first logging out.

34

35

System Architecture An overview of the SIMPRAC data model is given in Figure 31. User

Case

User Case Status

Response

Actions

Consultation and Stream

Hypotheses

Question answers

Questions

Medical Record

Currrent Investigations

Examination findings

Examinations

User Notes

Current Managment

Investigation results

Investigations

Management stream map

Management

User Activity Log

Figure 31: Overview of application data model. The system can be viewed from three major perspectives. The first is the user model, the second is the disease model, and the third is the series of ancillary functions, including the library, medical record and user notes. 1.9

USER MODEL

Figure 32 illustrates the structure of the user model, and the model for a single user is shown at Appendix A, on page 46. The user is uniquely identified by their username and each user is associated with a set of status records. The status records, for each user, are defined by: 1. The user’s username 2. The case selected by the user, and 3. The professional background or peer group selected by the user. Thus a user may attempt any case from different professional backgrounds.

User

User Status

Consultation:Stream

Username

History Question

Case

Examinations

Profession / Group

Investigations

Management Options

Case Hypotheses

Figure 32: SIMPRAC User Model - Overview. As illustrated in the diagram above, a case is undertaken by the user as a series of consultations, each of which is comprised of history questions, examinations, investigations, and management options. Information is also held on the user’s hypotheses. An abridged version of the model at Appendix A is given below (Figure 33). It can be seen that this user has registered for the first case as a medical student. The first consultation begins in Stream One. This user, Student04 has asked 8 questions beginning with, “Do you have a past history of hyperlipidaemia?” In the full interaction, not shown in Figure 33, Student04 also performed 59 examinations, nine investigations, and selected four management options in their first consultation. In the consultation two they were in Stream Four and asked two questions, did not perform any examinations, requested eight investigations, and selected the same four management options. In the third consultation this user was in Stream Four, asked five questions, again did not perform any examinations, requested nine investigations, and maintained the same four management options. Finally, in the fourth consultation, the user was in Stream Six, asked six questions, did not perform any examinations, requested 10 investigations, and changed from using a fibrate to uncooked cornstarch.

36

USER Username: student04 Title: withheld Surname: withheld GivenName: withheld Email: withheld

HISTORY (8) Do you have a past history of hyperlipidaemia? Drugs and medication? .....

EXAMINATION (59) STATUS

Inspect Head. Inspect Abdomen .....

Profession: Medical Student Case: 1

INVESTIGATION (9) Full blood count Prothrombin time .....

MANAGEMENT (4)

CONSULTATION Consultation: 1 Stream: 1

Use a fibrate (Gemfibrozil) Diet low in saturated fat .....

CONSULTATION

HISTORY (2) EXAMINATION (0) INVESTIGATION (8) MANAGEMENT (4: unchanged)

Consultation: 2 Stream: 4

CONSULTATION

HISTORY (5) EXAMINATION (0) INVESTIGATION (9) MANAGEMENT (4: unchanged)

Consultation: 3 Stream: 4

CONSULTATION

HISTORY (6) EXAMINATION (0) INVESTIGATION (10) MANAGEMENT (4: 1 changed)

Consultation: 4 Stream: 6

Figure 33: Abridged user model for Student04 (See Appendix A for full details). 1.10 DISEASE MODEL The disease model is summarized in Figure 34. As can be seen from the illustration, the disease model is very simple. As the case proceeds, the patient moves along a series of states. These states are characterized by only two variables, the consultation and the stream. Each of these states is associated with three classes of diagnostic action response: 1. Answers to the questions, 2. Examination findings, and 3. Results of investigations. As discussed in Section 1.8, each response is assigned a relevance score that is used in the reflective component of the review. The management actions are treated slightly 37

differently to the diagnostic actions. In addition to having a relevance score for the patient’s current state, the management actions also determine the state transitions of the patient. The management actions are mapped to a new stream in the next consultation, and these mappings are assigned weight values independent of the relevance scores. The responses, relevance scores, and state transitions are described in more detail below. Patient State

Answers to questions

Consultation Stream

Results of investigations

Examination Findings

Management Action Map

Figure 34: The disease model Action Responses The diagnostic actions have been described in chapter three, and include history questions, physical examination, and investigation requests. Each action response is specific to a particular stream and consultation. However, in practice most of the actions have a default response. If a consultation- and stream-specific response cannot be found, then the default response is returned. Furthermore, each action has been coded with a default response that returns a negative or normal value. As an example, in the first case during the first consultation, there is only one stream (see Figure 20 on page 25). If a user requests a serum triglyceride measurement, they will be given the information that the triglyceride concentration is 12.0 mmol/L. If another case had been undertaken on another topic, such as childhood food allergies, a consultation- and stream-specific response would not be available and the result would simply be returned as “normal”. Relevance Scores All actions, both diagnostic and management actions, are assigned a relevance score by the author of the case. The current values range from two to zero where; two is critical, one is relevant, and zero is not relevant. The use of just three levels was chosen to make the user task simple and quick. Critical items were regarded as those items that were critical to the diagnosis or management of the patient to produce the optimal outcome. Relevant items were considered to be those that were important to know about, but not essential to the patient’s outcome. Non-relevant items were those that were not necessary for the correct diagnosis and management of the patient. For example, ordering a skull X-ray is not usually relevant to someone presenting with a suspected myocardial infarction (heart attack). An electrocardiograph, on the other hand, is critical to the diagnosis. While three discrete categories were used, it was anticipated that there would be some uncertainty and disagreement over the classification of some items. This reflects the uncertainty inherent in the practice of medicine. Indeed, disagreement should be encouraged, and should be regarded as an indicator of the success of using such a component to encourage users to reflect on their actions.

38

The classification of each element of the history, examination, investigations, and management is determined by the author or authors of the case. Under normal circumstances the case author will be a domain expert and the decision of what is critical or relevant will be based on their opinion. However, this need not be the case. For example, a group of medical students could prepare a case for their colleagues. Under ideal circumstances, it would be useful to be able to classify actions based on published guidelines or best evidence. Unfortunately, such resources are not always available. All default responses for diagnostic actions have been assigned a relevance score of zero. As a consequence of default responses having a relevance score of zero, any actions that return a normal response, yet are critical or relevant, must have a consultation- and stream-specific response. As an example, a question that might be asked by a user during this hyperlipidaemia case is, “Do you have a past history of hyperlipidaemia?” In all circumstances, the patient will respond, “I have had high cholesterol and triglyceride since I was a child but they have never been this high before. The triglycerides have usually been between 2 and 4 mmol/L while the cholesterol has generally been less than 6 mmol/L. They were last checked over 12 months ago.” If this is asked in the first consultation it is assigned a relevance score of one. However, if it is asked in subsequent consultations, since this question should have been asked in consultation one, it is assigned a value of zero. Assignment of a score of zero is not meant to penalise the user, rather the score is indicative that the action was likely to have been non-contributory to the diagnosis and management of the patient. Transitions between patient states A summary of the state transitions is illustrated in Figure 20 on page 25. As indicated previously, the patient’s state is determined by the consultation, and the value of the stream variable. The value of the new stream in the following consultation is calculated using a deterministic process based solely on the management options selected by the user and the stream values during the current and previous consultations (Figure 35). The mapping data required for this calculation is held in a database table and an extract is shown in Table 1. The dependence of the new stream on the value of the current and previous stream was included, so that users had an opportunity to change suboptimal management based on the feedback they received during the review sessions, at the end of each consultation. It was felt that users should be able to receive feedback, without their being an immediate adverse effect on the patient. A very simple algorithm is used to calculate the stream value for the next consultation. Each relevant or critical management action maps to a specified stream. Thus if only one action is chosen, this maps to the relevant stream. However, if as is more likely, more than one action is chosen, then a decision must be made as to which stream represents the outcome for these actions together. In order to make this decision, a weighting, defined by an integer, is allocated to each action. Each time an action maps to the same stream, the weight values are incrementally summed. When all the actions have been chosen, the stream with the highest score is the one that is chosen. That is, the

39

new stream is the one where ∑Weight New Stream has the greatest value. This approach is analogous to Multi-Attribute Utility Theory (MAUT), in which the overall evaluation of an object is defined as a weighted addition of its evaluation with respect to its value dimensions (44). MAUT has principally been applied to decision analysis and estimating user preferences (44, 45). Figure 35 is a diagrammatic representation of how a new stream is selected, based on the treatment chosen by the user. In this example, the user is in consultation two and stream Q. Suppose the user had been in stream P during the first consultation, then the stream sequence is “P:Q”. The weights given to treatments Rx-A to Rx-F are specific to the situation where the consultation is two and the stream sequence is “P:Q”. Suppose the user prescribed treatment Rx-A alone, then the Stream would be “A” in consultation three. If instead the user had chosen Rx-A, Rx-B, Rx-D, Rx-E, and Rx-F in combination, then the sum of the weights would be 0 (Rx-A) for Stream A, 5 (Rx-B, Rx-D, and Rx-F) for Stream B, and 4 (Rx-E) for Stream C. Therefore, in consultation three the stream would be B. STATE Consultation=2 Stream=Q

Rx-A

Wt=0

Rx-B

Wt=1

Rx-C

Wt=0

Rx-D

Wt=2

Rx-E

Wt=4

Rx-F

Wt=2

Stream=A

Stream=B

Stream=C

Figure 35: Mapping management actions to streams in the following consultation. Rx-A to Rx-F represent six different management options or treatments. Wt is the weight value given to each option. See text for examples. Table 1 is an extract of the actual mapping table used SIMPRAC. In contrast to Figure 35, the streams within SIMPRAC are described by numeric values. This table shows the stream weights for the case where user is at the end of the second consultation, and the stream is 1. As there is only one possible stream in the first consultation, the stream sequence for the last two consultations is “1:1”. Suppose the user chose a diet low in saturated fat, 6 hourly cornstarch, fish oil, and a fibrate (rows one to four respectively). From the table we can see that a diet low in saturated fat is mapped to Stream 1 with a weight of zero, cornstarch is mapped to Stream 1 with weight zero, fish oil is mapped to Stream 2 with weight two, and fibrate is mapped to Stream 4 with weight four. In this instance the total weight value (column five) for Stream 1 is zero (zero plus zero), the total weight value for Stream 2 is two, and the total for Stream 4 is four. Thus the new stream with the greatest value is Stream 4. Therefore, at the third consultation, the user would be presented with the patient in Stream 4.

40

Table 1: Stream mapping data for the end of Consultation Two with a stream sequence of 1:1. Weight

New Stream

2 2

Stream Sequence 1:1 1:1

0 0

1 1

2 2 2 2 2

1:1 1:1 1:1 1:1 1:1

2 4 8 0 1

2 4 3 1 1

Action

Consultation

Diet low in saturated fat Six hourly cornstarch to maintain blood glucose Use omega 3 fatty acids (fish oil) Use a fibrate (Gemfibrozil) Use a HMG CoA Reductase Inhibitor (Statin) Allopurinol Default

Using this simple algorithm has at least two benefits. First, it makes authoring cases relatively easy, provided the number of different patient states is constrained. The default mapping for all treatments is to Stream 1 with weight 0. Therefore if no effective treatment is provided, the case will enter Stream 1 at the following consultation. Using this system, the case author only needs to iterate through those treatments that they want to affect the stream selection. This must be done for each combination of consultation and stream sequence. The second benefit of the approach is that it enables the patient outcome to be directly related to the treatment modalities, thereby making explanations of patient outcomes relatively easy. An alternative to the deterministic approach would have been to enable probabilistic progression. While such an approach might have more realistically simulated actual patient progression, it would have lacked the beneficial characteristic of being able to directly relate treatment to outcomes, as noted above. Alternative models and implications One of the design goals for this application was for it to be able to be used as a template for a variety of chronic illnesses. While a more sophisticated disease model involving pathophysiological mechanisms and interactions that influenced the disease progression or control at each consultation could have been developed, it was unclear how well this would transfer to alternate clinical domains. In line with the approach taken by the simulation software developed by the NBME (37), careful consideration was given to the need to include a realistic time component. With such an approach, the review periods entered by users at the end of each consultation would influence the state of the patient. However, it quickly became apparent that this increased the number of potential patient states, and made determining what was relevant at a particular consultation unmanageable. For example, if a user wanted to review the patient after the first consultation in two months, the repeat measurement of triglycerides would be relevant or critical to the ongoing management of that patient. If, on the other hand, the user wanted to review the patient in one week, the repeat measurement of triglyceride would not be indicated. As a result of this type of complexity, it was decided not to incorporate strict measures of time within the model, other than that determined by the consultation number. With the current model, the diagnostic action responses and relevance scores are dependant on the current state, while the management response is dependant on the current and immediately previous state. Having this dependency means that the disease model is simple but the number of responses and associated relevance scores for any

41

particular case is potentially very large. This in turn raises questions regarding the scalability of the application and the ease with which additional cases can be authored. Despite the large number of diagnostic actions available to the user, the majority of responses for a particular case will use the default values. Nevertheless, the number of responses will generally increase proportionally with the number of possible patient states for a given case. Furthermore, the number of states is dependant on the anticipated effects of different management actions and the number of consultations. Therefore, when designing a case, care must be taken to limit the number of consultations, limit the number of outcomes for different management actions, or limit both. The other limitation of the existing approach is apparent where combinations of treatment lead to different outcomes from any single treatment. For example, suppose treatment X leads to an outcome defined by stream A, and treatment Y leads to an outcome defined by stream B. With the current system, there is no method of mapping the combination of treatment X and treatment Y to stream C. Alternate outcomes from interactions between treatments a not overly frequent. However, they could be catered for by use of a rule-based system that considered these exceptions. When developing the first case for the user evaluations, data was entered into the database using a very basic interface over a period of approximately one week. Development of an appropriate authoring interface would greatly simplify and accelerate the entry and configuration of additional cases. 1.11 ANCILLARY FUNCTIONS In addition to the two major models above, the application provides a number of additional functions. These include the medical record, user notes, current orders, current management, hypothesis lists, library resources, and free-text search engine for the history and management components. The medical record, user notes, current orders, current management, and hypothesis lists are associated with the user’s status record (see Figure 31 on page 35), while the library pages are maintained as static hypertext pages. Investigation orders and management options The current orders and current management hold the investigations that have been requested and management options selected by the user. Each investigation order record is uniquely identified by the user status, and the consultation in which it was requested. Users are therefore able to review the results of investigations for the past and current consultations. This is in keeping with actual practice, where investigations are episodic and the results of investigations form part of the patient’s medical record. In contrast to the investigation orders and results, the management orders are only identified by the user status, and therefore only represent the current management of the patient. Historical management records for different consultations are generally not required by the user except during the review process, where they can be retrieved from the user activity log. Under normal circumstance, the clinician would record changes to the patient’s therapy in the medical record. This can be done by the user, if they wish to do so.

42

Hypothesis log The hypothesis log records the user’s hypotheses for the current case and professional background. The log records the consultation, the stage of the consultation, and the stream in which the hypothesis was generated. In its current form, an audit trail of updates to the hypothesis statement, probability assigned to the hypothesis, or reason for holding the hypothesis is not maintained. On the other hand, records are never deleted. If the user requests that a hypothesis be deleted, it is merely inactivated and is no longer visible to the user. This was done so that this information would be available in the future, leaving open the opportunity to explore other ways of using this data to enhance learning. Question and answer module In both the history and management selection components of the application, the user has the option of entering free text questions or keywords that are then matched to those options held in the database. The free text searching is built around the Lucene free text indexing and searching package produced by the Apache Software Foundation (46). In the case of the questions database, the history_questions table contains a field for a full text statement of the question, and a second field containing keywords and alternate phrases that relate to the concept covered by that question. For example, the question, “Do you have a past history of heart disease?” is associated with the terms, “heart disease, ischaemic heart disease, coronary artery disease, myocardial ischaemia”. The Lucene package is used to index the text within these fields and is later used to search the index as users ask their questions. Information on the management options is maintained in an analogous manner, with a field for a text description of each option, and a second field to hold alternate descriptions and keywords. 1.12 IMPLEMENTATION SIMPRAC was implemented using the following software. 1. MySQL database (47) 2. Apache Turbine 2.1 framework (Apache Software Foundation) 3. Apache Tomcat 4.0.6 Servlet/JSP Container (Apache Software Foundation) 4. Lucene 1.2 (Apache Software Foundation) SIMPRAC was originally implemented using version 3.23 of the MySQL database. SIMPRAC has since been migrated to MySQL Version 4.0, however, the database schema remains unchanged.

43

44

Conclusions and Further Work With the goal of supporting learning through reflection, a web-based virtual patient that enables the user to interact with the patient over a number of consultations has been developed. SIMPRAC has been designed to enable medical students and practitioners to work through a series of simulated consultations involving the diagnosis and management of chronic illness. Learners are able to take a history from the patient, perform a physical examination, order and review investigations, as well as choose an appropriate management strategy. In addition to just simulating the consultation process, elements have been included with the goal of supporting learning through reflection. Support for reflection is provided in several ways. Firstly, learners are asked to generate and review their diagnostic hypotheses before examining the patient, requesting investigations, or choosing management options, for the first time in each consultation. Secondly, components are included to enable the learner to record the salient points of the case, as well as any learning issues they perceive to have arisen while interacting with the virtual patient. Thirdly, the case runs over multiple consultations with the patient’s outcome varying with the management options chosen by the learner. In this way, learners have an opportunity to reflect on this outcome feedback, and change their management if necessary. Finally, two formal review elements are included at the end of each consultation. The Stage 1 consultation review lists the activity undertaken by the learner in the diagnostic and management phases of the simulation. The leaner must classify each of their actions as one of, “critical”, “relevant”, or “not relevant” to the diagnosis, management and outcome of the patient, in light of their current knowledge. This consultation reflection forces users to review and reconsider their actions. The Stage 2 consultation review enables the learner to view the corresponding information for a single expert, the average for their peer group, and the case authors. An overview of the actions taken is given in chart form, with the users also being able to drill down to the individual actions. Further work will need to be undertaken in a number of areas. With respect to the simulation itself improvements are required to improve the ease of use and fidelity of the simulation. In particular, changes will need to be made to the question and answer interface to improve the natural language understanding. Notwithstanding, it is quite possible that category-based selections will remain the preferred option of some users. The investigation request screen needs to be updated to make it easer to order investigations, while the investigation results screen needs to altered to make it more closely resemble laboratory and other reports currently available to medical practitioners. A case authoring interface needs to be developed, so that new cases can be efficiently added by those with clinical knowledge but with perhaps lower levels of computer literacy. The layer of reflection will need to be undertaken in a number of ways. In the first instance it will be necessary to explore how learners respond to this additional reflective

layer. These findings can then be used to improve the various interfaces, and the level of support for reflection. Investigations will also need to be undertaken to learn how well users are able to integrate the information they learn and whether they can abstract to new situations, having used SIMPRAC. Ideally, but with even more difficulty, studies will also need to be carried out to evaluate how well performance on SIMPRAC correlates with clinical performance and behaviour.

45

Appendix A

Example User Model

USER Username: student04 Title: Withheld for confidentiality. Surname: Withheld for confidentiality. Given Name: Withheld for confidentiality. Email: Withheld for confidentiality. STATUS Profession: Medical Student. Case: 1 CONSULTATIONS Consultation: 1, Stream: 1 History Question: Do you have a past history of hyperlipidaemia? Drugs and medication? Is there any family history of illness? Other past medical history including psychiatric or surgery? Has anyone in your family or any friends had any similar problems?} Can you please describe your diet? Do you have any allergies? Are there any other health professionals involved the management of the current problem? Examinations: Inspection|Head Inspection|Abdomen HeightWeight|other Thermometer|Mouth Sphygmomanometer|Arm-Right Watch|Wrist-Right Inspection|Foot-Left Inspection|Knee-Right Inspection|Wrist-Right Motion|Ankle-Left Motion|Foot-Right Motion|Foot-Left Motion|Ankle-Right Motion|Knee-Right Motion|Knee-Left Motion|Wrist-Right Motion|Wrist-Left Motion|Elbow-Right Palpation|Flank-Right Palpation|Hypochondrium-Right Percussion|Flank-Right Percussion|Hypochondrium-Right Palpation|Flank-Left

46

Palpation|Hypochondrium-Left Palpation|IliacFossa-Left Palpation|IliacFossa-Right Palpation|Suprapubic Palpation|Umbilical Palpation|Epigastrium Palpation|Groin-Left Palpation|Groin-Right Palpation|Heart-Mitral Palpation|Heart-LSE Palpation|Heart-Aortic Palpation|Heart-Pulmonary Stethoscope|Breast-Left Stethoscope|Heart-Mitral Stethoscope|Back-UpperRight Percussion|Back-UpperRight Percussion|Back-UpperLeft Stethoscope|Back-UpperLeft Inspection|Hand-Right Inspection|Wrist-Left Inspection|Hand-Left Inspection|Forearm-Right Inspection|Arm-Right Inspection|Arm-Left Inspection|Foot-Right Inspection|Knee-Left Inspection|Leg-Left Inspection|Leg-Right Inspection|Thigh-Right Inspection|Thigh-Left Inspection|Nose Inspection|Eye-Left Inspection|Eye-Right Inspection|Mouth Inspection|Face-Right Inspection|Face-Left Investigations: Full blood count PT / INR Electrolytes urea & creatinine Test: Liver function tests Urate Cholesterol HDL cholesterol LDL cholesterol Glucose Management Options: Diet low in saturated fat Diet high in carbohydrate Allopurinol

47

Use a fibrate (Gemfibrozil) Consultation: 2, Stream: 4 History Question: Have you had any problems with the medications you are on? Can you please describe your diet? Examinations: None. Investigations: Full blood count Cholesterol HDL cholesterol LDL cholesterol Liver function tests Urate Electrolytes urea & creatinine Triglycerides Management Options: Diet low in saturated fat Diet high in carbohydrate Allopurinol Use a fibrate (Gemfibrozil) Consultation: 3, Stream: 4 History Question: Have you had any problems with the medications you are on? Change in weight? Do you have cornstarch on a regular basis? Can you please describe your child's diet? Can you please describe your diet? Examinations: None. Investigations: Full blood count Cholesterol HDL cholesterol LDL cholesterol Triglycerides Electrolytes urea & creatinine Liver function tests Urate Glucose Management Options: Diet low in saturated fat. Allopurinol. Use a fibrate (Gemfibrozil) Use omega 3 fatty acids (fish oil) Consultation: 4, Stream: 6 History Question: Drugs and medication? Have you had any problems with the medications you are on? Asked: Can you please describe your diet?

48

Do you have cornstarch on a regular basis? Do you take regular exercise? Have you had any muscle aches or pains? Examinations: None. Investigations: Full blood count Urate Liver function tests Electrolytes urea & creatinine Cholesterol HDL cholesterol LDL cholesterol Triglycerides Creatine Kinase Glucose Management Options: Diet low in saturated fat Allopurinol Six hourly cornstarch to maintain blood glucose Use omega 3 fatty acids (fish oil)

49

References 1. Davis D, O'Brien MAT, Freemantle N, Wolf FM, Mazmanian P, Taylor-Vaisy A. Impact of formal continuing medical education: Do conferences, workshops, rounds and other traditional continuing education activities change physician behaviour or health care outcomes? Journal of the American Medical Association 1999;282(9):867-874. 2. Schön DA. Educating the Reflective Practitioner. Toward a New Design for Teaching and Learning in the Professions. San Francisco: Jossey-Bass; 1987. 3. Branch WT, Paranjape A. Feedback and reflection: teaching methods for clinical settings. Academic Medicine 2002;77(12 Pt 1):1185-8. 4. Elstein AS, Shulman LE, Sprafka SA. Medical Problem Solving. Cambridge, Massachusetts: Harvard University Press; 1978. 5. Schmidt HG, Norman GR, Boshuizen HPA. A cognitive perspective on medical expertise: theory and implications. Academic Medicine 1990;65(10):611-621. 6. Elstein AS, Shulman LS, Sprafka SA. Medical problem solving: A ten year retrospective. Evaluation & the Health Professions 1990;13(1):5-36. 7. Engel JD, Wigton R, LaDuca A, Blacklow S. A social judgement theory perspective on clinical problem solving. Evaluation & the Health Professions 1990;13(1):63-78. 8. Tucker LR. A suggested alternative formulation in the development by Hursch, Hammond and Hursch and by Hammond, Hursch and Todd. Psychology Review 1964;71:528-530. 9. Kirwan JR, de Saintonge DMC, Joyce CRB, Currey HLF. clinical judgement analysis - practical application in rheumatoid arthritis. British Journal of Rheumatology 1983;22(supplement):18-23. 10. Hammond KR. Computer graphics as an aid to learning. Science 1971;172(3986):903-908. 11. Merrill DM. Constructivism and instructional design. Educational Technology 1991;31(5):45-52. 12. Spiro RJ, Feltovitch PJ, Jacobson MJ, Coulson RL. Cognitive flexibility, constructivism and hypertext: random access instruction for advanced knowledge acquisition in ill-structured domains. Educational Technology 1991;31(5):24-33. 13. Bednar AK, Cunningham D, Duffy TM, Perry JD. Theory into practice: How do we link. In: Anglin GJ, editor. Instructional Technology: Past, Present, and Future (2nd ed). 2 ed. Englewood, Colorado: Libraries Unlimited; 1995. 14. Anderson JR, Reder LM, Simon HA. Applications and misapplications of cognitive psychology to mathematics education. In. Pittsburgh: Carnegie Mellon University; 1995. 15. Ruth-Sahd LA. Reflective practice: A critical analysis of data-based studies and implications for nursing education. Journal of Nursing Education 2003;42(11):488-497. 16. Dewey J. How we think: A restatement of the relation of reflective thinking to the educative process (2nd ed). New York: Heath and Company; 1933. 17. Atkins S, Murphy K. Reflection - a Review of the Literature. Journal of Advanced Nursing 1993;18(8):1188-1192. 18. Davies E. Reflective Practice - a Focus for Caring. Journal of Nursing Education 1995;34(4):167-174. 50

19. Scanlan JM, Care WD, Udod S. Unravelling the unknowns of reflection in classroom teaching. J Adv Nurs 2002;38(2):136-143. 20. Clouder L. Reflective practice in physiotherapy education. A critical conversation. Studies in Higher Education 2000;25(2):211-223. 21. Johns C. The value of reflective practice for nursing. Journal of Clinical Nursing 1995;21(1):23-30. 22. Lowe PB, Kerr CM. Learning by reflection: The effect on educational outcomes. Journal of Advanced Nursing 1998;27(5):1030-1033. 23. Mallik M. The role of nurse educators in the development of reflective practitioners: a selective case study of the Australian and UK experience. Nurse Education Today 1998;18:52-63. 24. Shapiro J, Talbot Y. Applying the concept of the reflective practitioner to understanding and teaching family medicine. Family Medicine 1991;23(6):450-6. 25. Burnard P. Nurse educators' perceptions of reflection and reflective practice: a report of a descriptive study. Journal of Advanced Nursing 1995;21:1167-1174. 26. Issenberg SB, McGaghie WC, Hart IR, Mayer JW, Felner JM, Petrusa ER, et al. Simulation technology for health care professional skills training and assessment. JAMA 1999;282(9):861-866. 27. Bergin RA, Fors UGH. Interactive simulated patient-an advanced tool for student activated learning in medicine and healthcare. Computers & Education 2003;40:361-376. 28. Prideaux D, Alexander H, Bower A, Dacre A, Haist S, Jolly B, et al. Clinical teaching: maintaining an educational role for doctors in the new health care environment. Medical Education 2000;32:820-826. 29. Brna P, Irvine P, Duncan D. Support for Lifelong Self-Learners Using the Internet. In. Leeds: Computer Based Learning Unit, Leeds University; 2000. 30. McCarthy WH, Gonnella JS. The simulated patient management problem: a technique for evaluating and teaching clinical competence. British Journal of Medical Education 1967;1:348-352. 31. McGuire CH, Babbott D. Simulation technique in the measurement of problem solving skills. Journal of Educational Measurement 1967;4(1):1-10. 32. Melnick DE. Computer based clinical simulation: State of the art. Evaluation & the Health Professions 1990;13(1):104-120. 33. Friedman CP. Anatomy of the clinical simulation. Academic Medicine 1995;70(3):205-209. 34. Bryce DA, King NJC, Jeremy RW, Myers JH. Facilitating clinical reasoning through computer based learning and feedback. In: The Fourteenth Conference of the Australasian Society for Computers in Learning in Tertiary Education; 1997; 1997. p. 88-93. 35. Hayes KA, Lehmann CU. The interactive patient: a multimedia interactive educational tool on the world wide web. MD Computing 1996;13(4):330-334. 36. Fitzgerald JT, Wolf FM, Davis WK, Barclay ML, Bozynski ME, Chamberlain KR, et al. A preliminary study of the impact of case specificity on computer based assessment of medical student clinical performance. Evaluation & the Health Professions 1994;17(3):307-321. 37. Clyman SG, Orr NA. Academic computing in medicine: Status report on the NBME's computer based testing. Academic Medicine 1990;64(4):235-241. 38. Braunwald E, Fauci AS, Kasper DL, Hauser SL, Longo DL, Jameson JL. Harrison's Principles of Internal Medicine. 15th ed. 15th ed. New York: McGrawHill; 2001.

51

39. Chen YT, Burchell A. Glycogen Storage Diseases. In: Scriver CR, Beaudet AL, Sly WS, Valle D, editors. The Metabolic and Molecular Bases of Inherited Disease 7th Ed. 7th ed. New York: McGraw-Hill; 1995. p. 935-965. 40. Robinson MJ, Roberton DM. Practical Paediatrics. 4 ed. Edinburgh: Churchill Livingstone; 1999. 41. Macleod J. Clinical Examination. 6 ed. Edinburgh: Churchill Livingstone; 1983. 42. Myers JH, Dorsey K, Benz E. DxR Clinician. Carbondale: DXR Development Group; 2001. 43. Health Insurance Commission. Medical Benefits Schedule. Canberra: Health Insurance Commission; 2003. 44. Schafer R. Rules for using multi-attribute utility theory for estimating a user's interests. In: Adapitivität und Benutzermodellierung in interaktiven Softwaresystemen (Conference); 2001; University of Dortmund; 2001. 45. Von Winterfeldt D. Decision Analysis and Behavioural Research. New York: Cambridge University Press; 1986. 46. Apache Software Foundation. Lucene. In. 1.2 ed: Apache Software Foundation; 2002. 47. MySQL AB. MySQL. In. 3.23 ed. Uppsala, Sweden: MySQL AB; 2000.

52