Role and use of expert systems within the clinical laboratory.

Artificial intelligence can be used, mainly in the form of knowledge based systems, for different parts of clinical laboratory activity. Most of the iterative tasks for control and validation, or for maintenance or repair of automated instruments can be assisted by specialized software. There are now specific programmes (i) for control and validation of the request, where the goal is to modify the requesting behaviour of the physician, using protocol driven requests and immediate feedback, incorporating simultaneously a check of test request redundancy, (ii) for helping in equipment trouble-shooting, (iii) for technical and biological validation because too many QC data are issued from large equipment to be checked and used properly by the technologist whilst the pathologist can be assisted for the final revision of reports and (iv) for assistance in the interpretation of laboratory reports, one of the most valuable aspects of the pathologist's participation in the patient care process.

Laboratory medicine in France. A jeopardized situation.

The expenses for health care in France have risen considerably during the present decade, ranking third after USA and Canada in the Western world. In spite of the very low cost of laboratory medicine (2.4% of the total expenditure in 1995), clinical laboratories have undergone a severe squeeze, due to two limiting factors; a decrease in the ordering of laboratory tests from private physicians and a reduction in the total expenses for laboratory services from the Social Security. Consequently, there has been unemployment of technical and secretarial staff and severe restriction in investment for buying new equipment. However, hospital laboratories will manage to assume their challenge in developing robotics, automation, molecular pathology techniques and expert systems. Private laboratories, in spite of their efforts to follow the technological advances in automation, will survive thanks to consolidation of regional networks that operate in a cooperative rather than competitive mode. Therefore, the challenge will be not in the adaptation of clinical laboratories, but in the limitation of overspending at the national level and in modification of the behaviour of irresponsible citizens accustomed to spending freely on health care services.

Automated technical validation--a real time expert system for decision support.

Dealing daily with various machines and various control specimens provides a lot of data that cannot be processed manually. In order to help decision-making we wrote specific software coping with the traditional QC, with patient data (mean of normals, delta check) and with criteria related to the analytical equipment (flags and alarms). Four machines (3 Ektachem 700 and 1 Hitachi 911) analysing 25 common chemical tests are controlled. Every day, three different control specimens and one more once a week (regional survey) are run on the various pieces of equipment. The data are collected on a 486 microcomputer connected to the central computer. For every parameter the standard deviation is compared with the published acceptable limits and the Westgard's rules are computed. The mean of normals is continuously monitored. The final decision induces either an alarm sound and the print-out of the cause of rejection or, if no alarms happen, the daily print-out of recorded data, with or without the Levey Jennings graphs.

VALAB: expert system for validation of biochemical data.

In large laboratories that use "high-throughput" equipment, it is now possible to use artificial intelligence techniques to aid decision making and validation of data. This paper describes an artificial intelligence project, VALAB, that has been carried out in our laboratory. VALAB, an expert system that permits real-time validation of data, is designed to be equivalent to validation by the laboratory director. The decision produced by the expert system is based on several factors, including correlation between repeated laboratory results, physiological association between different variables, the hospital department from which the test was ordered, and the patient's age and sex. In 200 abnormal chemistry profiles randomly selected, VALAB's ability to detect abnormal cases (i.e., sensitivity = 0.75) was exceeded by only one of seven laboratory experts. However, all seven experts outperformed VALAB's measured specificity of 0.63. The VALAB system incorporates greater than 4000 rules. Operational since November 1988, it has validated greater than 50,000 medical patients' reports in real time.

The performance of the knowledge-based system VALAB revisited: an evaluation after five years.

In 1988, inundated by the tedious work of validation of laboratory reports in a large hospital biochemistry laboratory, we designed VALAB, a knowledge-based system specially dedicated to this iterative function. Coping at first with a few biochemical tests, the program has been progressively expanded to forty-five common chemical tests. Simultaneously some new rules have been introduced to "weight" the conclusion in different circumstances and rules taking into consideration some clinical data have also been written. Moreover the program moved to other disciplines, pH and blood gases, haematology and coagulation. Accordingly the evaluation protocol has been modified, incorporating a new step, the consensus decision of the pathologists, operating within the initial protocol and based upon the various criteria of epidemiology. These major changes and improvements have led us to check and describe again the performance of this updated VALAB knowledge-based system.