Causal Understanding of Patient Illness in Medical Diagnosis

Abstract

First generati on AI in Med ic ine programs have clearl y demonstrated the usefulness of Alte chnique s. However . it h as al so bee n recog niz ~ d that the use of noti ons su ch as causal relationships . temporal patterns. ond aggregate disease categories in these programs has been too weak . From our study of clinician's behavior we reali1.ed that a diagno stic or therapeutic program must consider a case at various lev els of detail to integrate overall understanding with detailed knowledge . To explore these issues. we have undertaken a study of tht problem of providing expert consultation for electrolyte and acid-base disturbances. We have partly completed an implementation of ABEL. the diagnostic compo nent of the overall effort. In this paper we wncentrate on AUU .' s mechanism for de,crihing a patient. Called the patient -specific model, this description indudes data about the patient as well as the program's hypothetical interpretations of these data in a multi-level causal network . The lowes[ level of this description consists of pathophysiological knowledge about the patient. which is successively aggregated into higher level concepts and relations. gradually shifting the content from pathophysiological lo syndromic knowledge . The aggregate level of this description summarize!' the patient data providing a glob;~ perspective for efficient exploration of the diagnostic possi· bil itie s. Th~ pathophysiological level description provides the ability to handle complex clinical situations arising in illnesses with multiple etiologies, to evalu;lle the physiological validity of diagnostic pos sibilities being explored, and to organize large amounts of seemingly unrelated facts into coherent causal descriptions.