Clinical problem solving--the role of expert laboratory systems.

The objective was to determine whether or not a laboratory based computer diagnostic program could aid the clinician in solving problems, outside his or her field of expertise, by expertly interpreting ¿Emergency Room' haematological and biochemical data and providing a list of possible diagnoses. The program, which uses Fuzzy Sets and pattern recognition as its Inference Mechanism coupled with a data base comprised of haematological and biochemical responses to disease collected over a period of 10 years in a teaching hospital, analysed data published in two leading journals--the 'Clinical Problem-Solving' section of the New England Journal of Medicine and the 'Lesson of the Week' feature of the British Medical Journal. It was found that the computer program often presented diagnoses not thought of by the clinician. With such a system, sometimes as few as three routine investigations suggested the diagnosis. The diagnostic accuracy could be improved with a more structured approach to ¿Emergency Room' laboratory investigations. It is concluded that the computer, programmed to recognize a disease by the pattern of its response to routine haematological and biochemical investigations, could contribute significantly to diagnosis.

Epidemic polyarthritis in northeastern Australia, 1978-1979.

In a study of epidemic polyarthritis in central and northern Queensland, clinical disease was observed most frequently in urban-dwelling adult females. Clinical and subclinical infections were detected year round, although the majority of clinical infections occurred between the months of January and June. In the communities studied, there were approximately 50 subclinical infections for each clinical infection. There was no detectable association between any HLA-A or HLA-B phenotypes and epidemic polyarthritis.

Automated differential leucocyte counters: an evaluation of the Hemalog D and A comparison with the Hematrak. I. Principles of operation; reproducibility and accuracy on normal blood samples.

The precision and accuracy of the Hemalog D, which uses cytochemical identification of cells in suspension, and of the Hematrak, which uses image recognition of cells on a stained blood film, have been compared counting normal blood samples. The Hemalog D showed superior precision for all cell types, as would be expected since 10,000 cells are counted per sample; however, the precision for monocytes was worse than expected for the number of cells counted. The precision of the Hematrak was equivalent or superior to that of a manual count of the same number of cells but showed the poor precision inevitable when only 100 cells are counted. With respect to accuracy, both automated counters showed statistically significant differences from manual counts and from each other in counting neutrophils, lymphocytes and eosinophils, bu the differences were not sufficiently great to be of practical importance. The Hematrak counted monocytes accurately (though imprecisely) whereas the Hemalog D overestimated monocytes on average by 2.3%, or 40% of the mean monocyte percentage. This was consequent on the counting of esterase positive neutrophils as monocytes, and the difference from the manual count was sufficient to be of some practical importance. The Hemalog D counted basophils both accurately and precisely. The precision of manual and Hematrak basophil counts was poor; the accuracy of the Hematrak basophil count was dependent on the quality of the stain and that of the manual basophil count was dependent on the quality of the stain and the attentiveness of the technologist. For other cell types in blood samples from normal volunteers, the Hematrak was versatile and accuracy was not greatly affected by the use of May-Grünwald-Giemsa rather than Wright's stain, nor by the use of hand spread rather than machine spread films.