Why virtual patient platforms fall short of expectations...
Virtual patients are nothing more than digitized patients (mostly constructed clinical information rather than real) to supplement medical teaching. They are a necessary part of the armamentarium used in most medical schools, simply because the access to real clinical material for student practice and rehearsal is limited and decreasing. Virtual patients can be as simple as a pdf of a case study or set of powerpoint slides circulated for discussion. Usually though, virtual patient platforms are digitized cases with some embedded interactivity that allow a stepped accrual of relevant clinical information leading to pattern recognition, eventual diagnoses and some degree of virtual intervention.
Virtual patient platforms have been available for 30 years or more. Sadly, the basic design and application of these virtual platforms have remained relatively unchanged over the last 30 years. The basic design of such platforms consist of 3 parts: a] the digitized information of the clinical case, deconstructed into convenient 'boxes' or paragraphs of information, b] a process of pattern recognition based on revealed 'boxes' of information, and most importantly c] a process that allows the user to extract or reveal the 'boxes' of information. Almost all virtual patient platforms in the market utilize graphical interfaces such as, drop down menus, or clickable action buttons.
There are 2 major constraints in such a design for the virtual patient platform. Firstly, the cases, although often rich and detailed, are, because of the way they are written, fixed and invariable. As a result, the cases lose their novelty and very quickly become predictable. The second major constraint is the use of graphical interfaces. Graphically interfaces are used mainly because they are easy to implement. However, they are not a 'natural' means for doctor and patient to interact. This plus the fact that the revealed information often is in the form of a relatively indiscriminate 'data-dump' (depending on the degree of resolution of the data developed). These constraints may be expected to reduce the student's real learning when using the platform, and to discourage continued usage for practice and rehearsals.
But virtual platforms do not need to be limited by these design constraints. AI technology allows us to break free of such constraints. We propose that future virtual platforms need to be able to a] build populations of randomly generated patients with internally consistent and coherent histories, and b] enable these randomly generate patients to be conversationally engageable with the student/user.
We have tried to follow these principles in designing the Virtual Integrated Patient (VIP) platform, and have a working platform that is currently being used in a number of teaching domains. More information about the VIP can be found here (www.virttualintegratedpatient.com)