Intelligent, Adaptive Systems for Health Care Informatics

The Care Augmenting Software Partners (CASPERs) project envisions a multi-agent system that assists patients and their health care providers by enhancing care coordination and addressing challenges of health literacy. CASPERs are intended to support a diverse, evolving team in formulating, monitoring and revising a shared care plan that operates on multiple time scales in uncertain environments and to support enhanced patient-physician communications.  The project comprises three complementary efforts, (1) GoalKeeper, (2) Yoda/Augie, and (3) SimplyPut. 

Health Care Context for CASPER: Children with Complex Conditions: The care team for children with complex conditions is diverse, including not only physicians but also the child's primary caregiver (usually a parent), other types of medical providers (e.g., therapists) and caregivers in community settings (e.g., teachers). The care team members vary in their extent and duration of involvement as well as expertise, and they seldom all meet together. They thus may not be aware of each other's goals or patient/parent goals and will have different views of a patient’s status. Teamwork and effective communication of health information are essential to good care, but not well supported by existing health care systems.

GoalKeeper:  Our work on GoalKeeper aims to enhance care plan management and care coordination by supporting “dynamic care plan” formation and tracking of progress towards care goals and of execution of care plans.  Systems challenges include eliciting goals and plans from the care team, providing effective mechanisms for updating patient status related to goals, and determining when and with whom to share information as patient status changes.  Fundamental CS challenges include designing effective information sharing mechanisms for multi-agent systems and designing appropriate human-computer interaction (HCI) methods.

Yoda/Augie:  Our work on YODA and Augie aims to improve patient-physician communication within the context of office visits. YODA focuses on supporting patients in preparing for appointments and follow up. Augie complements this effort by tracking the patient-physician dialogue during a visit and intervening appropriately to improve patient comprehension, engagement and activation.  Challenges include determining conversation topics and patient comprehension, interruption management, and HCI design for home and office settings.

SimplyPut: Our work on SimplyPut addresses the problem that a significant portion of health information (online and in print) is cannot be adequately understood by most patients.  The SimplyPut crowdsourcing platform aims to support the production of purpose-driven summarizations, personalized to match varying intended-reader characteristics, and including simplified descriptions of procedures patients need to perform.  Challenges include creating a general framework to support different summarization types, leveraging diverse communities with varying levels of medical expertise, and integrating crowdsourcing with text-processing algorithms to produce such customized summaries at scale.