The long-term goal of this project is to develop technology that enables cost-effective, customizable, automated in-home conversational assistants that can help patients man- age their treatment and assist in monitoring their health using natural spoken dialogue over the telephone. Such a system would help patients and/or their caregivers manage their medical care, provide reminders, answer questions, and engage in dialogue to collect information for monitoring a patient's current state. The system would not make medical decisions, but would help patients follow the instructions that they have been given by their doctors, and provide status reports back to medical support teams. We are proposing to determine the feasibility of this technology by constructing a proto- type system that can automatically interview patients with congestive heart failure and produce reports for their doctors and nurses. Working from a substantial technology base that we have built over the past decade, we will develop a robust prototype system that supports natural spoken dialogue to assess their current status of patients with congestive heart failure. We will then evaluate the system's performance by formal experimentation. The key feasibility requirement is that the information obtained from the automated dialogue system is as accurate as the information obtained in typical interviews by nurse practitioners. To establish this claim, we will test the prototype system on a set of volunteers patient who are currently under treatment for heart failure problems. We will also gather data that will allow us to explore other important issues related to the coverage of the information attained and usability. If we can show that our spoken dialogue technology works in this experiment, we will be in a position to perform clinical trials on the medical effectiveness of using such systems. It has been shown that more frequent contact with in-home patients can significantly improve medical outcomes for a range of patients with different diseases. The cost of such intervention, however, prevents its widespread use. If we can build an automated system that can improve the efficiency of monitoring patients, then many more patients can benefit from this approach. Furthermore, because of the broad range of medical conditions for which such systems could be used, the impact of this technology on health-care delivery would be far-reaching and dramatic. ? ? ?
