Outcomes Analysis using Information/Control Theory, Sequential Hypothesis Testing, and Parallel Processing

Medical outcomes analysis can be used for treatment plan recommendation, physician training, organization and budgeting of health care programs, and for resource allocation. Objectives are to apply the following disciplines to outcomes analysis: Information theory, optimization/control theory, sequential hypothesis testing, and parallel, distributed processing. Utility/loss measures are presently based on physical/social activities and qualitative descriptors of symptom/problem complexes. For dynamic assessment of diagnosis and treatment, utility measures should also include quantitative physiological test data at a sequence of sampling times. Medical procedures can then be chosen to optimize a dynamically time varying, utility- based functional that uses updated posterior probabilities of different diagnoses and patient condition at each time increment, as well as probabilistic prognoses (predictions) for alternative tests and treatments. Dual control theory is relevant because treatments (controls) are often used to both improve level of well-being and to augment diagnostic information (observations). Monetary costs and information-theoretic measures for evaluation of diagnostic data can be incorporated into the utility-based functional to be optimized. Parallel, distributed processing can generate relevant hypotheses (alternative diagnoses) via associative memory and can be used for efficient sequential updating of probability distributions. Anticipated Benefits: Benefits include better resource allocation and cost control at individual, facility, community, and national levels. Commercial applications include software packages and customized PC-based work stations containing parallel, distributed processing boards for medical training, physicians, advisories, insurance/actuarial decision making, and medical consulting services.