9315428 Cooper This project will apply to large medical databases a variety of state-of-the-art machine-learning techniques that have proven useful in other domains. These learning techniques will generate computer models that can be used to accomplish the following tasks: (1) suggesting general clinical policies that seem cost-effective and underutilized, (2) providing individual physicians with patient-specific suggestions about diagnostic and therapeutic actions that appear to be equally efficacious clinically but cost less than the planned action, (3) suggesting more effective actions than may be planned, and (4) detection of clinical situations that are unusual and potentially costly. The database to be used in the investigation contains information about patients with community-acquired pneumonia. ***
