AI platform enables oncologists to practice conversations with terminally ill patients
Click Here to Manage Email Alerts
Check out the February cover story here: "Maintaining bond with patients crucial after cancer-directed treatments are no longer beneficial"
“I want you to be honest with me — how long do I have left?”
Conversations like this one, with a patient named SOPHIE, are among the most difficult clinicians must have with their patients.
Although a poorly handled conversation of this kind may leave a patient with cancer feeling upset or confused, patients generally don’t critique their physician’s conversational missteps.
SOPHIE does.
“You often expressed less positive emotion than is recommended,” SOPHIE might inform the clinician. “You should try to end on a more positive note next time.”
SOPHIE (Standardized Online Patient for Healthcare Interaction Education) is an online virtual “patient” a team of computer scientists, palliative care specialists and practicing oncologists at University of Rochester created. SOPHIE uses a realistic, computer-generated avatar and artificial intelligence to provide scenarios through which clinicians can practice communication with terminally ill individuals.
“Many people have tried to improve conversations at the end of life; there have been many research studies and many methods developed, but not all clinicians have access to such programs or opportunity to practice new communication skills,” SOPHIE co-creator Ronald M. Epstein, MD, professor of family medicine, oncology and medicine, co-director of the center for communication and disparities research, and co-director of mindful practice in medicine at University of Rochester School of Medicine and Dentistry, said in an interview with Healio | HemOnc Today. “Changing communication requires a lot of practice. You don’t change your communication style overnight. So, the question is, how do you get experienced clinicians to practice?”
‘SOPHIE never gets tired’
Ehsan Hoque, PhD, associate professor of computer science at Rochester’s Hajim School of Engineering and Applied Sciences, brought Epstein on to the SOPHIE project. Hoque is a pioneer in research demonstrating the value of automated systems in helping physicians to improve their social and interpersonal skills. He played a key role in developing LISSA (Live Interactive Social Skills Assistant), a tool designed to assist individuals on the autism spectrum in communicating effectively. Hoque had been looking for a way to similarly improve end-of-life communication but found many available tools either were costly or required travel to workshops.
“SOPHIE can be used in the office, which is more convenient,” Epstein said. “Also, physicians can use it as many times as they need to until they really master the skills. The SOPHIE platform offers promise in that regard, because SOPHIE never gets tired.”
Hoque knew of Epstein’s work on facilitating better communication between terminally ill patients and their physicians, and it impressed him. Additionally, he was interested in the potential data from the patient-physician transcripts that Epstein and his associates had amassed and begun to analyze.
“Ron and his colleagues designed an experiment that included hundreds of interactions between [patients in the final stage of cancer] and their oncologists,” Hoque told Healio | HemOnc Today. “We started analyzing the texts, as well as the tone of voice, to understand what really matters during this conversation. We looked at successful conversations and what those oncologists did differently compared with conversations where the communication broke down.”
Breaking the habit of ‘lecturing’
Hoque and his research team created algorithms that could be applied to the transcripts of the conversations. Postdoctoral researcher Mohammad Rafayet Ali and former PhD student Taylan Sen — both employees of Hoque’s Rochester Human-Computer Interaction Lab — published a paper on this process. They used algorithms to develop measures of effective end-of-life conversations.
The researchers found “lecturing” by the clinician to be one of the hallmarks of a less effective interaction.
“In this context, lecturing means they took longer turns than the patient, used technical jargon or didn’t show empathy,” Hoque said.
To address the problem of lecturing, SOPHIE provides feedback to prevent the clinician from overwhelming the patient with excessive information or esoteric language.
“If SOPHIE notices that your turn is lasting more than a certain number of seconds, she will let you know that you’re lecturing and that you might want to give the patient a chance to talk,” Epstein said. “We all glaze over when we hear a lot of language that we’re not used to, but it’s so easy when you’re an expert to forget that people don’t have the same knowledge base and may need additional time to assimilate information.”
SOPHIE also notes the use of “hedge words,” open-ended questions, speaking rate, appropriate vocabulary/reading levels and positive language. At the end of the practice conversation, SOPHIE gives the physician a report card assessing their skills in each metric.
“SOPHIE provides little suggestions and tips so that the doctor is left feeling more empowered,” Epstein said. “We don’t want them to feel discouraged or that there is no point in trying. We want them to feel that they can improve, and we want to encourage them to improve.”
Emphasis on clarity
Another aspect of a successful conversation is clarity. According to a University of Rochester press release, up to 68% of patients with late-stage cancer walk away from conversations with their oncologists underestimating their disease severity, overestimating life expectancy or both. Such a lack of clarity can prevent these patients and their families from making informed, realistic decisions about how to proceed with their treatments.
“A patient may come for an appointment and ask the doctor how long they have to live, and then the doctor might say something like, ‘Well, everybody is different and we don’t really know’ or ‘I had a patient last week whose case was something like yours, and he really beat the odds and did well — anything’s possible,’” Epstein said. “What would you take away from that? That’s just hedging; you’re not answering the question directly.”
Some examples of hedge words include “really,” “believe,” “could” and “possible.”
The report card lists frequently used hedge words, as well as words of empathy the physician used.
In one sample dialogue, SOPHIE asks the clinician if she will likely live to see her grandson — who is in middle school at the time of the conversation — graduate high school.
A clinician who scored average in explicitness but above average in empathy said, “SOPHIE, that would be really great — I would really love for you to be able to see his graduation, but I don’t think it’s very likely at all, unfortunately. I’m really sorry.”
SOPHIE is currently being validated and not yet available for use. Hoque said the researchers plan to evaluate it further in a randomized, controlled trial, and it will then be made available for first-year medical students.
The current version of SOPHIE is designed for use through a computer browser, enabling physicians to access it from anywhere in the world. Epstein said this availability could help physicians in low-income countries who lack access to such training.
“So, a doctor in Bangladesh or Rwanda or Haiti could interact with SOPHIE,” he said. “SOPHIE could also be taught to speak other languages. It’s leveling the playing field in terms of communication skills worldwide.”
- Reference:
- Ali MR, et al. IEEE Transactions Affective Computing. 2021;doi:10:1109/TAFFC.2021.3054717.
- For more information:
- Ronald M. Epstein, MD, can be reached at Department of Family Medicine, 1381 South Ave., Rochester, NY 14620; email: ronald_epstein@urmc.rochester.edu.
- Ehsan Hoque, PhD, can be reached at 3013 Wegmans Hall, Rochester, NY 14620; email: mehoque@cs.rochester.edu.
Collapse
Read more about
Click Here to Manage Email Alerts