Recent developments in natural language processing (NLP) techniques for working with clinical texts have largely on extracting medical problems, diagnoses and treatments from hospital notes in physical medicine, such as progress notes, discharge summaries and lab reports. This has been often been done for purposes of decision support: modeling the patient as a set of problems to be solved and identifying the correct course of treatment according to the prevailing medical model.
Narrative medicine, however, places importance on the meaning of illness as experienced by the patient, or as reflected by the clinician's personal experience of working with the patient. These narratives may make rich use of emotive language that may be missing from traditional clinical notes.
Patient narratives consist of a series of unfolding events involving interacting protagonists and their roles. Computationally, these can be modeled as narrative event chains: sets of partially ordered events related by a common protagonist. In psychotherapy, cognitive analytic therapy (CAT) is a model that specifically considers these narrative events in order to reformulate the patient¹s problems in terms of the event sequences and reciprocal roles that lead to and maintain maladaptive personal relations. In this presentation, I present exploratory work on developing a computational framework for extracting and visualizing narrative event chains from CAT narratives, and consider the potential uses and benefits of such a framework.