Generating Personalized Hospital-Stay Summaries for Patients
MetadataShow full item record
Comprehending medical information is a challenging task, especially for people who have not received formal medical education. When patients are discharged from a hospital, they are provided with lengthy medical documents that contain intricate terms. Studies have shown that if people do not understand the content of their health documents, they will neither look for new information regarding their illness, nor will they take actions to prevent or recover from their health issue. In addition, the discharge notes and education materials are usually developed using a “one size fits all” approach and by keeping the “general” population in mind. However, in reality, not many people can benefit from such documents. The primary objective of this research is to provide hospital-stay information to patients in a form that they can understand. In this thesis, I present a framework for automatically generating textual summaries of hospital stays. First, I developed a metric for determining the complexity of medical terms and supplement difficult terms with explanations. Second, based on the research on patient-centered care and patient education, I identified the metrics that can be used for capturing a patient’s familiarity with health information and their level of motivation to get involved in their own care. Third, I developed a mechanism for synthesizing the scores from these metrics, as well as my findings from the analysis of a collection of interviews where patients have recounted their health experiences. Fourth, I developed an algorithm that generates the summaries. The summaries my framework produces were found to cover around 80% of the health concepts that were considered as important by a doctor or a nurse. An online survey conducted among 150 participants verified that my algorithm’s interpretation of the personalization parameters is representative of that of a larger population.
SubjectNatural Language Generation, Personalized Summaries