Laura Barrett, MLS, at the University's George Smith Library of the Health Sciences in Newark, explains how the online diagnostic clinical decision software was chosen and how it works: "We looked at available programs to see which were most appropriate for UMDNJ, and we chose DXplain, developed by Massachusetts General Hospital, and QMR, produced by the University of Pittsburgh."
Both were designed by teams of doctors; the databases were developed by programmers, and both used thousands of journal articles and cases drawn from each institution's patient population. They were chosen by UMDNJ in part because those populations closely resembled the patients using University facilities and affiliated hospitals.
Barrett says the "programs are meant to assist physicians and dentists in making diagnoses in the same way that consulting colleagues or doing literature searches help."
QMR is only available on local networks in computer centers on each campus, but anyone connected to the University computer network can access DXplain. It is an option once you have chosen UMDNJ libraries databases from the initial menu. DXplain contains information mainly on adult diseases, but includes some pediatric and ob/gyn problems as well. It is based on 2,000 diseases, 5,000 symptoms and some 17,000 relationships between the two.
DXplain needs the age, gender and at least three symptoms to do an adequate search. The more symptoms you list, including lab results, the more accurate the diagnostic options you will see. The possible diagnoses will be listed in two categories of disease: very common and common, rare and very rare. Within each group, the illnesses will be ranked from likely to least likely.
There are other functions like "question the user," in which the program asks for more symptoms to help narrow the field, and it can explain why it has asked for a particular one.
These databases have been available at UMDNJ since January. Barrett says that although some faculty were initially skeptical about the programs, by August they had been accessed more than 2,000 times.