Computer-Aided Prognosis

Abstract

To design a scientific trial of treatment for a group of patients, or to plan individual treatment for a single patient, a thoughtful clinician must review the details of past experience. For the therapeutic trial, the clinician wants to divide his population into subgroups with similar prognoses, so that different modes of treatment can be appropriately allocated and compared.¹ For the individual patient, the clinician wants to know what was done and what happened during the treatment of previous patients whose clinical situation resembled the one now at hand.² These procedures require that the clinician remember, classify, and sort the many details of information describing the diversity of previous clinical experience. To estimate prognosis and choose treatment for a patient with a particular disease, the clinician would have to begin by recalling all the characteristics of previous patients with this disease; he would then select a subgroup of