University of Pittsburgh
Pittsburgh, Pennsylvania
James H. Harrison, Jr, MD, PhD

Drug therapy problems related to pharmacokinetic variability in patients and dosing errors are common, but often do not come to the attention of clinical laboratory services for follow-up and quality assurance. Furthermore, these cases are not easily identified for teaching purposes during medical student clinical pharmacology rotations and thus an excellent teaching resource is underutilized.

We have developed a simple rule-based expert system that scans clinical laboratory drug level data to identify data patterns indicating patients who will likely benefit from follow-up. Rules for data pattern identification were developed from a review of drug level data for patients receiving digoxin, phenytoin or theophylline, and generalized to cover eight additional drugs. The expert system was implemented with an interactive data display, report printing and simple statistical analysis on an Apple Macintosh computer.

Data for scanning was downloaded as ASCII files from the clinical laboratory information system. In three- and six-month retrospective studies from two separate teaching hospitals (1944 and 2560 patients, respectively), 29% and 31% of patients under therapy with the targeted drugs were identified as having patterns appropriate for review. For the past four years, the system has formed the basis for the 4th year clinical pharmacology elective (20 students per rotation) in the School of Medicine at Tulane University. Data was downloaded twice weekly and paper reports were distributed to the students for follow-up and presentation. In addition to providing useful clinical feedback and excellent cases for instruction, the students' reviews provided information to the hospital continuous quality improvement initiative that resulted directly in development and modification of hospital policy.