Presentation Title
Information Flow in Serious Illness Conversation
Abstract
Each year, over 5 million people with serious illness worldwide engage in conversations with palliative care clinicians to determine the best course of action to improve their quality of life by anticipating, preventing and treating suffering. Understanding how these conversations function is a crucial first step towards evaluating and improving conversation quality. In this talk, we propose a computational approach using Markov models of speaker turn lengths for measuring information flow in a conversation. We demonstrate the model using 355 transcribed conversations between patients with advanced cancer and palliative care physicians, recorded between 2014 and 2016 as part of the Palliative Care Communication Research Initiative (PCCRI). We show how this approach can be used to capture and visualize normative trends of conversational discourse and reveal novel insights about information flow patterns in serious illness conversations. Among other interesting results, we show that the normative pattern of information flow in these conversations is dominated by directional flow from clinician to patient, but that this pattern is disrupted during the expression of distressing emotions such as fear and anger.
Primary Faculty Mentor Name
Margaret J. Eppstein
Faculty/Staff Collaborators
Robert Gramling, Donna M. Rizzo, Margaret J. Eppstein
Status
Graduate
Student College
College of Engineering and Mathematical Sciences
Program/Major
Computer Science
Primary Research Category
Health Sciences
Information Flow in Serious Illness Conversation
Each year, over 5 million people with serious illness worldwide engage in conversations with palliative care clinicians to determine the best course of action to improve their quality of life by anticipating, preventing and treating suffering. Understanding how these conversations function is a crucial first step towards evaluating and improving conversation quality. In this talk, we propose a computational approach using Markov models of speaker turn lengths for measuring information flow in a conversation. We demonstrate the model using 355 transcribed conversations between patients with advanced cancer and palliative care physicians, recorded between 2014 and 2016 as part of the Palliative Care Communication Research Initiative (PCCRI). We show how this approach can be used to capture and visualize normative trends of conversational discourse and reveal novel insights about information flow patterns in serious illness conversations. Among other interesting results, we show that the normative pattern of information flow in these conversations is dominated by directional flow from clinician to patient, but that this pattern is disrupted during the expression of distressing emotions such as fear and anger.