The Blog to Transform Advanced Care
Advancing Care through Innovation, Observation and Collaboration.
Machine learning could shed light on the themes of serious illness communication, according to research published in December in Patient Education and Counselling.
Researchers at the University of Vermont’s Conversation Lab used machine learning algorithms to analyze the transcripts of conversations, with the goal of finding common themes and storylines. Previously, this method had been used to evaluate the language in fiction manuscripts to identify different types of stories.
“One of the ways in which we make meaning and share who we are is by telling stories about ourselves and what matters to us,” says senior researcher Robert Gramling, M.D., D.Sc., who serves as director of the Conversation Lab and the Miller Chair of Palliative Medicine at the UVM Larner College of Medicine.
Dr. Gramling says that the team took an ecological approach to their work, viewing different conversations as distinct species.
“No one type is good or bad,” says Dr. Gramling. “It’s just how [the conversation] functions in the environment that it is in.” In this case, Dr. Gramling explains, the environment is the context of a patient’s illness and what suffering means to them.
The team examined 354 transcripts of conversation involving 231 patients in New York and California, collected by the Palliative Care Communication Research Initiative.
Dr. Gramling et al. found that the conversations in their study tended to follow a few common trajectories:
- Over the course of these conversations, patients tended to shift from discussing the past to talking about the future
- Conversations tended to shift from a focus on sadder subjects to happier topics over time
- Symptoms tended to a major subject early in discussions, with a shift toward treatment options and ultimately prognosis over time
In the future, Dr. Gramling’s team hopes to develop large, multi-site studies that will allow them to observe conversations involving people from a variety of backgrounds. By scaling up, researchers can gain a better understanding of the way that patients and clinicians talk about serious illness. This understanding can foster innovation as we seek to improve serious illness communication.
“Particularly in serious illness, the ability to measure what is happening and what people experience is the way we will provide feedback and develop our systems,” says Dr. Gramling. To learn more about this research, entitled “Story Arcs in Serious Illness: Natural Language Processing features of Palliative Care Conversations,” please click here.