Pediatric Informatics: Computer Applications in Child Health (Health ...

174 M.J. Feldman
ED-12: “Practical opportunities for the direct application of…critical analysis of
data…[and] computer-based exercises where students either collect or utilize data
to test and/or verify hypotheses or to address questions about biomedical principles
and/or phenomena… the ability to collect, analyze, and interpret data”
An Association of American Medical Colleges Report in 1998 also recommended
certain relevant competencies for medical students, including the ability to “identify
and locate, when possible, the crucial pieces of missing clinical information, and
determine when it is appropriate to act on incomplete information” and to “make
critical use of decision support, demonstrating knowledge of the available sources
of decision support which range from textbooks to diagnostic expert systems to
advisories issued from a computer-based patient record.” 44 Because many MDDSS
allow a student to observe how a change in clinical data can incrementally affect
diagnostic considerations, such systems can be helpful as adjuncts to teaching concepts
of medical decision making. Students can learn, interactively, which clinical
data strongly raise consideration of particular diagnoses (high PPV clinical findings).
Many medical school faculty and students already use MDDSS programs.
12.7 Case Study: A Teenager with Rabies 45
As an illustrative example (not intended to be a comprehensive analysis or detailed
comparison of the three systems), a published case of rabies in a teenager where
the history of a bat bite is omitted is used to compare the three MDDS systems
described. This was the first reported case of a patient who survived rabies that had
not been treated by either vaccine or postexposure prophylaxis. Once the history of
a bat bite was revealed, the diagnosis was straightforward, but without that critical
history, the diagnosis can be challenging and MDDSSs could be useful.
Non-demographic findings in the initial presentation from the published case report
were extracted, omitting the history of the bat bite. There were 17 non-demographic
findings that were entered into each system as allowed. For DXplain, 14 of 17 findings
were enterable (using its controlled vocabulary); for GIDEON, 4 of 17 findings were
enterable; and for Isabel (since it accepts narrative text entries rather than a controlled
vocabulary) all 17 findings were entered. In addition, the case was run a second time
in Isabel using the entire textual description as published in the report.
For DXplain: Rabies was not in the 50 top diagnoses displayed (see Fig. 12.5).
Many of the entered findings had low PPV for rabies resulting in a low score (see
Fig. 12.6). Most suggested diagnoses were CNS lesions or other noninfectious
neurologic conditions though encephalitis, an infectious disease concept related to
rabies, does appear on DXplain’s list. Given that this patient has fever, a user may
want DXplain to focus more on infectious etiologies. The “FOCUS” feature (see
Fig. 12.7) eliminates diagnoses that do not contain “fever,” and causes rabies to appear
as a Rare Diagnosis. The “Disease Description” function (see Figs. 12.8–12.10)
allows the user to see the findings in DXplain’s KB for any disease. 21