Grant Details
Grant Number: |
1R01CA290559-01A1 Interpret this number |
Primary Investigator: |
Daskivich, Timothy |
Organization: |
Cedars-Sinai Medical Center |
Project Title: |
Natural Language Processing Based Consultation Feedback to Improve Patient Decision Making and Physician Risk Communication |
Fiscal Year: |
2025 |
Abstract
PROJECT SUMMARY
Prostate cancer is a highly prevalent disease requiring complex decision making due to its multiple treatment
options, each with tradeoffs of survival and sexual and urinary function. As a result, shared decision making
(SDM)—a decision-making process where physicians collaborate with patients to select the option that best
matches their values and preferences—is the standard of care for treatment decision making. For SDM to be
effective, physicians must clearly communicate risks and rewards of treatment options, and patients must retain
that information so that they can weigh these tradeoffs when making a treatment decision. Yet, in pilot work, we
found that physicians often fail to adequately communicate tradeoffs to men with prostate cancer. Furthermore,
patients often feel that they do not have enough knowledge to be fully informed participants in SDM. We recently
developed a highly accurate natural language processing (NLP)-based feedback system capable of delivering
point-of-care, real-time feedback to patients and providers about what was said about key risks and rewards of
treatment options during prostate cancer treatment consultations that has potential to enhance both patient
decision making and physician risk communication. At the patient level, receiving a summary of conversation
about tradeoffs has the potential to reinforce awareness of key tradeoffs, enhance perception of risk of those
tradeoffs, reduce decisional conflict, improve quality of SDM, and ultimately improve appropriateness of
treatment choice. At the physician level, receiving feedback on what was said will provide awareness of the
quality of their own communication and promote accountability for the quality of information. In this grant
application, we propose a cluster randomized controlled trial to test whether providing NLP-based feedback with
artificial intelligence (AI) summaries of NLP-extracted information on key tradeoffs to patients and physicians
after prostate cancer treatment consultations improves patient decision making and physician risk
communication compared to standard-of-care counseling. In Aims 1/2, patients will be randomized within
clusters of physicians to: (1) a control arm, in which they will receive standard of care counseling along with
American Urological Association-endorsed educational materials on treatment decision making (for patients) and
SDM (for physicians) or (2) an experimental arm, in which patients and physicians will receive NLP+AI-based
feedback on what was said about key tradeoffs during the consultation. A follow-up phone visit will permit further
discussion. In Aim 3, we will analyze variation in quality of risk communication in subgroups of race, tumor risk,
and provider specialty in the control arm to identify groups at risk for poor communication. We hypothesize that
providing patients and physicians with key information extracted using the NLP system will improve decisional
conflict, shared decision making, and appropriateness of treatment at the patient level and quality of risk
communication at the physician level. We also hypothesize that risk communication will differ by race, tumor risk,
and physician specialty, which will permit targeting of communication interventions to at risk groups in the future.
Publications
None