Critical conversations: a user-centric approach to chatbots for history taking in the pediatric intensive care unit

In this article, we describe the potential utility and design of chatbots to improve history taking in the pediatric intensive care unit (PICU). The fast-paced, high-stakes environment of the PICU often forces clinicians to obtain only enough information to make immediate clinical decisions. Specific barriers to comprehensive history taking include insufficient time, frequent interruptions, caring for a wide range of conditions, need for timely interventions, and language differences. We propose that chatbots could play a critical role in improving history taking in the PICU by collecting information related to a patient's current presentation and exploring areas that are commonly neglected, such as social histories. To explore the use of chatbots in the PICU setting, we will first describe the current scope of chatbots as medical history taking aids. Next, we will outline specific considerations for the development of chatbots for the PICU, including methods for involving users, such as patients, caregivers, and clinicians directly in the design, mitigating false information, and establishing safeguards for chatbot behavior. Finally, we will review methods to evaluate chatbots. The overall purpose of this perspective article is to 1) propose the PICU as a novel environment where chatbots could improve history taking and diagnostic reasoning and 2) delineate specific user-centric design and evaluation methods.