Predictive Models in Medicine: Some Methods for Construction and Adaptation
Staal A. Vinterbo · 2000
Health care personnel make predictions routinely every day. They group patients according to disorders, render prognoses of the health status of a given patients at a future point in time, or classify laboratory specimens. With the emergence of integrated hospital information systems, the potential of using computerized predictive models to support tasks like these is significant. This thesis concerns itself with problems of realizing predictive models based on knowledge of associations, e.g., findings -- diagnosis associations. The theoretical context of the work contained in this thesis lies in the intersection between artificial intelligence, machine learning in particular, statistics and medicine. Although predictive model results reported in the literature often surpass the results of control groups of relevant health care professionals, few are found in clinical practice. The reasons for this are multiple and touch upon issues concerned with ethics, sociology, psychology and te...