190 likes | 331 Views
Agent-mediated Patient-Clinician Communication. Timothy Bickmore College of Computer & Information Science Northeastern University. Motivation: Rapid Acceptance. Natural, intuitive modality provides rapid acceptance Emulate human face-to-face conversation
E N D
Agent-mediated Patient-ClinicianCommunication Timothy Bickmore College of Computer & Information Science Northeastern University
Motivation: Rapid Acceptance • Natural, intuitive modality provides rapid acceptance • Emulate human face-to-face conversation • Focus on nonverbal communicative behavior • gaze, posture, gesture, etc.
Acceptance “Who would you rather receive discharge instructions from?”
The Clinical Issue:It sucks to be a hospital patient • Poor patient-provider communication • Poor patient activation • Poor intra- and inter- institution hand-offs • Loneliness • Sleep deprivation • Pain and other discomfort • Disorientation, confusion & delirium • Poor provider understanding of total patient state • Etc etc etc.
Hospital Buddy • Persistent hospital bedside presence • Sensors to make the agent aware of events in the room
The Communication Goal:Improve the inpatient experience • Improve patient-provider communication & activation • Maintain agenda of patient questions & issues • Prompt patient when provider enters room • Patient evaluates interaction afterwards • Improve provider awareness of patient state • Patient self-report & sharing • Provide companionship • Improve sleep • Track sleep • Intervene with patient and providers • Reduce medical device false alarms
Lessons Learned to DatePilot study (N=3) Patient 1 Patient 2 Patient 3 • Storytelling • Discuss events in the hospital room (waking, provider visit, meal, etc) • Self-report (pain, stress, etc)
Lessons Learned to DatePilot study (N=3) “The best thing about the system … when you don’t have anyone here with you…it was actually nice to have her. I mean it kept me company.” “I’m glad you came to me with this option, and have a chance to use it, to me, it helped me last night… the downtime, being lonely sometimes, this gives you something to do, something to hear.”
Lessons Learned to DateSensor Calibration Study • 12 general medicine patients, rigged with • Portable EEG • Heart & Respiration monitors • Wrist and ankle accelerometers • Actigraph • Audio recording • For 24 hours each • Required • Patient who met long list of eligibility requirements & would be in the hospital for at least another day • Sleep tech available • All equipment working • Multiple IRB approvals • Duration: ~1 year!
Lessons Learned to Date • Having a champion within the system is essential. • Clinicians generally love to help and are willing to embrace a new technology, IF • You’ve gone through channels • You have credible clinicians on your team • You start with benefits to them • You include them in the design process • Avoid integration with hospital systems (avoid IT in general) • Develop stand-alone systems as long as possible
Unexplored New Questions • Best medium for all? • Agents may not be best • Do high literacy patients want companionship in any form? • How can we include the family? • What about illiterate patients?