Improving Medical Communication using Rubric-Guided Counterfactual Recommendations
Researchers have developed a new language model-guided system designed to improve communication in text-based telemedicine. The system analyzes patient-doctor interactions to identify and refine communication features like tone and completeness, aiming to increase the likelihood of positive patient feedback without altering the medical content. Recommendations generated by the system showed a mean 6.41% gain in predicted positive feedback probability and were non-negative for over 93% of cases, suggesting that small, interpretable changes can significantly enhance communication quality while maintaining physician control. AI
IMPACT This AI system could enhance patient satisfaction and adherence in telemedicine by providing actionable communication feedback to healthcare providers.