Amruta Inc and Mary Washington Healthcare present Emotion Recognition in Patient Feedback
Health systems collect patient feedback to improve their safety and quality outcomes as well as operational performance. Understanding sentiment and emotions that prevail in patient and caregiver feedback, along with explanations for that prevalence are beneficial for improving the safety and quality outcomes, operational performance, as well as value based payments and profitability of health systems.
The sentiment analysis and emotion recognition require automated text processing and artificial intelligence (AI) based natural language processing (NLP). Predicting the sentiment and emotions is accompanied by causal explanations. This explainable AI leads to informative and actionable insights that clinicians, engineers, and administrators can use to improve health systems' safety, quality, and profitability.
Recognizing the sentiment and emotion in patient feedback involves understanding the semantics, context, syntax, and ever-evolving slang and abbreviations. While the sentiment is positive, neutral, or negative, emotion can be happiness, trust, surprise, anticipation, sadness, anger, fear, or disgust. Besides predicting the sentiment and emotion types, we also predict their respective intensities. Separately, we cluster the feedback to identify actionable topics present in the patient feedback.
Knowing the words that are causing the emotion in our predictions (explainable AI) and the keywords in topics involving nurse, physician, and staff treatments and facility conditions provide actionable insights to be proud of as well as to improve upon. While the NLP, AI, and machine learning methods associated with the sentiment and emotion predictions are challenging to develop, building upon the available literature and methods enabled us to make rapid progress in providing actionable insights from the patient and caregiver feedback. Our next steps are to intervene and improve patient satisfaction and the profitability of health systems using continuous insights.