AI-Enhanced Chatbots for Real-Time Symptom Analysis and Triage in Telehealth Services

Abstract

Despite the vast amount of data generated by patients' interactions with chatbot platforms used in telehealth services, it is not as widely used as it could be in triage and is mostly used for automated routing or conversational agents. We will consider an extension of these platforms to provide real-time AI-enhanced symptom analysis and triage functionalities. One of the unique challenges specific to utilizing real-time chatbot history during the current conversation is the ability to decide whether or not enough symptom-related information has already been provided in the chat history by the patient for an optimal symptom analysis and triage decision. Our evaluation shows that by employing our available machine learning and data preprocessing libraries, we obtain high accuracy across a wide variety of machine learning model families using less than 5% of the collected data. Slow adoption of telehealth services has been transformed into unprecedented rapid deployment due to public health restrictions imposed during the recent pandemic. By completing this shift, we are now forced to create innovative ways to scale the obvious shortcomings of current infrastructure and to enable chatbots with enhanced capabilities to support primary care services.