Computer must be a partner

0ne of the most exciting attempts to improve the delivery of health care has been in the area of computer-aided decision making. Diagnosis involves data collection and decision making. Medical students are trained well to collect complete descriptions of a patient. The signs and symptoms of diseases are also carefully detailed. But, the difficult process of assimilating information about a disease and using it to make an ideal diagnosis is not taught well. Diagnosis, largely taught and practiced as an art, is usually uncertain. Scientists still do not completely understand how to generate appropriate hypotheses. Yet, the process is receiving increasingly more attention. Decision analysis, borrowed from operations research, is now applied to medicine to create computer-aided medical decision-making systems. However, slow progress is expected in this field. Collaborations between computer and health care professionals have been productive, yet plagued by misunderstandings. Successful programs focus on narrow problems in the complex field of medical diagnosis, but the key to future success lies in developing a formal theory of medicine. Precisely understood processes in health care seem the best possibilities for partial automation. In the meantime, gaps in the current theory of medicine limit the applications of computers to medical decision making. Many computer programs have been written to aid medical decision making. The programs, which can be classified by their methodological emphasis, fall into three roughly defined categories: formula-based, flowchart-based, and knowledge-based programs. Though there are many kinds of formula-based programs, only Bayesian and pattern- recognition systems will be discussed here.