Using critiquing for improving medical protocols: harder than it seems

M. Marcos
G. Berger
F. van Harmelen
A. ten Teije
H. Roomans
S. Miksch

Medical protocols are widely recognised to provide clinicians with high-quality and up-to-date recommendations. A critical condition for this is of course that the protocols themselves are of high quality. In this paper we investi-gate the use of critiquing for improving the quality of medical protocols. We con-structed a detailed formal model of the jaundice protocol of the American Associ-ation of Pediatrics in the Asbru representation language. We recorded the actions performed by a pediatrician while solving a set of test cases. We then compared these expert actions with the steps recommended by the formalised protocol, and analysed the differences that we observed. Even our relatively small test set of 7 cases revealed many mismatches between the actions performed by the expert and the protocol recommendations, which suggest improvements of the protocol. A major problem in our case study was to establish a mapping between the ac-tions performed by the expert and the steps suggested by the protocol. We discuss the reasons for this difficulty, and assess its consequences for the automation of the critiquing process.

(PDF paper, 88Kb)

@InProceedings{AIME01,
  author =       "M. Marcos and G. Berger and F. van Harmelen and 
                  A. ten Teije and H. Roomans and S. Miksch",
  title =        "Using critiquing for improving medical protocols: 
                  harder than it seems",
  booktitle =    "Proceedings of the Eighth European Conference on Artificial
Intelligence in Medicine (AIME'01)",
  year =         2001,
  pages =        "431-441",
  publisher =    "Springer Verlag",
  series =       "LNAI",
  volume =       2101,
  keywords = {Medical Knowledge Representation},
  urlPaper = "http://www.cs.vu.nl/~frankh/postscript/AIME01.pdf"
}

<- Back to list of papers