Augmented intelligence for next-best-action in patient care

Abstract

A next-best-action system includes real-time operation and off-line training and processes. During off-line training databases of medical conditions and associated diagnostic factors, tests and treatments are used to create models used for later natural language processing of input electronic text. Diagnostic factors are ranked according to the probability that they indicate conditions and are stored in a matrix. Treatments corresponding to conditions are also stored in a matrix. During real-time operation electronic text is received from patient history of an EHR system, from a transcribed conversation between a physician and a patient, or from input that the physician makes in the EHR system or in an overlaid diagnostic user interface. The electronic text is processed by an NLP pipeline that derives clinical diagnostic factors and test results for the patient. The factors and test results are matched against the matrix of factors to produce a list of likely conditions, along with an ordered list of factors and tests that are either unknown or not yet performed for that patient.