Though the performance characteristics of test sequences for HIV infection are among the best to be found in clinical laboratory practice, large scale screening combined with low prevalence results in significant personal and social costs in terms of employability and insurability, and hence carry major liability implications. A major component problem is the partial pattern matching often involved in interpretation of the Western blot confirmatory test. In these tests, reporting protocols continue to be idiosyncratic and ate sometimes completely incorrect. Furthermore, a small proportion of test results, but significant in terms of absolute numbers, can only be characterized as indeterminate at any one time. The positive or negative predictive power of such a result is very sensitive to test operating characteristics in a particular laboratory setting and to the a priori probabilities determined by the history of risk behavior. We plan to explore the development of software, for the clinical laboratory or the practitioner's office, to propose epidemiologically sound predictive inferences from all reported combinations of results from HIV test sequences, incorporating adjustments for risk group and regional seroprevalence. This package is to include, at the least versions for MS-DOS and Macintosh platforms.Awardee's statement of the potential commercial applications of the research:Interpretive report generation software is now routine in several areas (e. g. coronary disease risk) of the clinical laboratory and is seen with increasing frequency in the offices of practitioners. When several variables with complex interactions and significant consequences are involved, such programs will become virtually mandatory for consideration of clinical, liability and accreditation issues.National Institute of Allergy and Infectious Diseases (NIAID)
