Personalized and adaptive systems for medical consumer applications

We have developed a range of systems that synthesize text at a fine grain—down to phrases within individual sentences—in real time, using natural language generation techniques from artificial intelligence (see our review in [3] of other work using these techniques), and provide information that is most personally relevant to the patient. We use the patient’s own medical record as a basis for selecting, linking, and filtering information. In our systems for cancer and diabetes, we provide a summary of the patients’ medical record plus hypertext pages of more general information. Items from the medical record are automatically linked to the appropriate general pages, so patients have direct access to the most relevant information, whereas a nonpersonalized version would require users to search or navigate through many medical terms. Our systems can also select personal reminders to add to the general information based on the patient’s medical record, as depicted in the accompanying figure. In a randomized trial, we found cancer patients valued information that included details from their own medical records more highly than general information alone [2]. We use information associated with the patients’ medical records to filter out irrelevant material and allow patients to focus on the most relevant information, for example, to offer detailed information about a treatment only to those patients who are having that treatment. This is especially valuable when producing printed material where space is limited. Personalized and Adaptive Systems for