Embedded Expert Systems

In medical applications, the requirement for manual input of data by busy clinicians is a significant barrier to the use of diagnostic expert systems (Graber and VanScoy, 2003). In the 1970s, several research groups took this barrier seriously. They began to explore methods to embed diagnostic expert systems— providing diagnostic support and, more generally, decision support—to clinicians without asking them to enter patient findings. Their research demonstrated that it is possible to obtain needed findings directly from existing clinical information systems (Evans, 1991, Gardner et al., 1999, Rind et al., 1993, Rind et al., 1995, McDonald, 1976c, McDonald et al., 1992a, McDonald, 1976b, McDonald et al., 1992b, Overhage et al., 1996, figure 13.12 An anthrax-case detection rule in the Clinical Event Monitor, a rule-based expert system.The rule is written in CLIPS, a language and expert system shell developed by the National Aeronautics and Space Administration.

(del rule detect anthrax (event Radiology report) (report (patient id ?patient id)) (fveetext ''freetext)

(test (str-index "wide mediastinum" (lowcase ?freete\t)))

(if (> (cCheckGrams ?palient_id "GPR" 10)-1) then (event ANTHRAX)

(printout t "Found gram positive rods and widened mediastinum in chest x-ray" crlf)

(printout t "=== Detected a possible ANTHRAX case =====" crlf)

Wagner et al., 1997) and from alterations in work flow, such as replacing unstructured paper records with computer-generated encounter forms (from which findings can be optically scanned or manually extracted) (McDonald, 1976a, McDonald et al., 1992a, McDonald, 1976b, McDonald et al., 1992b). The research also demonstrated how diagnostic expert systems can be embedded in clinical information systems, such as physician order entry systems (Tierney et al., 1993, Dexter et al., 2001) and electronic medical records (Gardner et al., 1999, Warner et al., 1997). Several of these systems functioned in the domain of hospital infection control (Hripcsak et al., 1997, Hripcsak et al., 1999, Kahn et al., 1996a, Kahn et al., 1996b, Kahn et al., 1993).

The secret to successful deployment of a diagnostic or other expert system in medicine is designing a system in which the benefits to clinicians (from decision support and increases in efficiency) outweigh any additional effort required by clinicians to enter data by a considerable margin.

Was this article helpful?

0 0

Post a comment