Medical Arguments in an Automated Health Care System

People use arguments to justify their claims. Computer systems use explanations to justify their conclusions. User acceptance of a system depends partly on how well the system explains its reasoning and justifies its conclusions. We have built WOZ, an explanation framework that justifies the conclusions of a clinical decision-support system. WOZ’s central component is the explanation strategy that decides what information justifies a claim. The strategy uses Toulmin’s argument structure to define pieces of explanatory information and to orchestrate their presentation. WOZ uses explicit models that abstract the core aspects of the framework such as the explanation strategy. In this paper, we present the use of arguments, the modeling of explanations, and the explanation process used in WOZ. WOZ exploits the wealth of naturally occurring arguments, and thus can generate convincing medical explanations. The Proposal – Explanations as Arguments Clinical practice guidelines encapsulate bodies of medical evidence and other knowledge necessary for decisionmaking about specific health problems. In the past few years, we have seen considerable interest in decisionsupport systems that automate clinical practice guidelines. Such systems can be used by clinicians to provide patient care that is informed by the guidelines. These systems especially help inexperienced or non-specialist clinicians to follow advanced evidence-based best practices relevant to their patients. In our laboratory at Stanford Medical Informatics we have developed the EON architecture – a set of software components with which developers can build robust guideline-based decision-support systems [Musen et al. 1996]. Based on EON, we have built ATHENA DSS, a decision-support system for managing primary hypertension at the VA Palo Alto Health Care System [Goldstein et al. 2000]. ATHENA DSS interprets the hypertension guideline model using specific patient data, and provides patient-specific recommendations for clinical management (Figure 1). For example, as shown in Figure 1, ATHENA DSS can recommend adding a specific drug to the patient’s regimen because the patient Figure 1. ATHENA DSS recommendations screen. The screen presents patient-specific recommendations. Clinicians can ask the system to explain recommendations such as “Add