A pilot study of breast cancer patients: Can machine learning predict healthcare professionals' responses to patient emotions?

An increasing amount of evidence demonstrates the importance of patient oriented interaction in the delivery of health care. The reduction in the incidence of negative emotions both enhances patient satisfaction and improves primary outcomes. The Verona coding definitions of emotional sequences (VR-CoDES) is a reliable system for identifying and coding patient emotions and the corresponding HCP responses. Coded transcripts can be used to instruct and guide healthcare professionals (HCPs) and trainees in an objective and evidence driven fashion. Notwithstanding this powerful potential, the use of VR-CoDES remains limited for practical reasons. Firstly, the coding system itself is complex and training is required to ensure consistent transcript annotation. Moreover the process of annotation is time-consuming and laborious even for a trained expert. Recent advances in machine learning, and its use in text analysis in particular,could be an indispensable benefit. The present paper describes the first work in the literature that explores this possibility. In particular, using 86 consultations between radiotherapists and adult female breast cancer patients we evaluate a range of state of the art classifiers in terms of their ability to predict the responses of HCPs to emotional cues and concerns. We demonstrate highly promising performance, with the best classifiers achieving prediction accuracies of over 82%.