Developing Intelligent Interviewers to Collect the Medical History: Lessons Learned and Guidelines

Abstract

Background: Physicians spend a lot of time in routine tasks, i.e. repetitive and time consuming tasks that are essential for the diagnostic and treatment process. One of these tasks is to collect information on the patient’s medical history. Objectives: We aim at developing a prototype for an intelligent interviewer that collects the medical history of a patient before the patient-doctor encounter. From this and our previous experiences in developing similar systems, we derive recommendations for developing intelligent interviewers for concrete medical domains and tasks. Methods: The intelligent interviewer was implemented as chatbot using IBM Watson assistant in close cooperation with a family doctor. Results: AnCha is a rule-based chatbot realized as decision tree with 75 nodes. It asks a maximum of 44 questions on the medical history, current complaints and collects additional information on the patient, social details, and prevention. Conclusion: When developing an intelligent digital interviewer it is essential to define its concrete purpose, specify information to be collected, design the user interface, consider data security and conduct a practice-oriented evaluation.