Conceiving an Application Ontology to Model Patient Human Papillomavirus Vaccine Counseling for Dialogue Management

"With advancements in speech technology and technologies that can support dialogue systems, machines could perform natural conversations and realistically mirror human-to-human interactions in a consistent manner."

In the United States (US) and other parts of the world, uptake of the human papillomavirus (HPV) vaccine uptake falls short of coverage targets, leading to preventable cancer deaths in both women and men. Many patients prefer face-to-face interaction and counseling to learn more about the HPV vaccine, which can be a challenge for physicians - for instance, due to limited time or a penchant for technical jargon that could impede communication. The hypothesis of the present paper is that automating the vaccine counseling session, with conversational agents, may provide a method for standardising and formalising dialogue with health consumers and provide an efficient means to communicate health information that could ultimately improve HPV vaccine uptake.

The editors of the Journal of Dialogue Systems have defined a dialogue system as "a computational device or agent that (a) engages in interaction with other human and/or computer participant(s); (b) uses human language in some form such as speech, text, or sign; and (c) typically engages in such interaction across multiple turns or sentences." Since the 1990s, health dialogue systems, whether telephone- or computer-based, have been used in a variety of health-based applications. Health dialogue systems with the power of speech recognition have been found to be able to mimic the face-to-face interaction between provider and patient and automate that experience.

An ontology, as described here, is "a formalized encoding of knowledge that enables machines and computerized agents to understand domain information....Ideally, if machines and systems can harness ontologies to understand and reason about domain information, they can possibly wield knowledge about counseling discourse with patients and perhaps lead towards theoretical plan-based dialogue management, which requires reasoning to implement..."

This paper introduces an application ontology for health information dialogue called Patient Health Information Dialogue Ontology, which is based on work and experiences in developing a dialogue script for vaccine counseling that was later executed in a Wizard of OZ experiment (see Related Summaries, below). Wizard of OZ protocol simulates dialogue interaction between human and machine (e.g., robot), and there is a remote operator speaking on behalf of the machine. The development of the script included refinements from domain experts in public health and from healthcare providers who interact with patients on a daily basis. In addition, the script was framed on the Health Belief Model (HBM), a behaviour model that leads to change in action - in this case, vaccine uptake.

The ontology's class level hierarchy is segmented into 4 basic levels - Discussion, Goal, Utterance, and Speech Task. The ontology also defines core low-level utterance interaction for communicating HPV information. The paper discusses the design of the ontology and the execution of the utterance interaction.

The crucial part of the PHIDO ontology is the Discuss Health Topic. The goal of this task is to confirm that the user understands the information that is communicated, answer any questions the user may have, and address any concerns about the communicated health information. In the sequence of this Speech Task type, one piece of information (e.g., "HPV vaccine might cause some minor discomfort and pain or soreness at the injection site.") is spoken by the agent (Health Information), and, later, the agent inquires if the user understands this (Confirm Health Information). From here, the ontology helps the application handle the expected utterance of the participant user - repeating of information (Request System Repeat), clarify the user's utterance (Unintelligible >precedes>Request Repeat), or attend to user's issue with the information (Disconfirmation). Discuss Health Topic incorporates a question and answering transition if the participant user has a follow-up question (Question).

Four of PHIDO's Communication Goals are based on HBM constructs: Communicate Benefit, Communicate Effectiveness, Communicate Harms, and Communicate Uncertainty. In addition, the Communication Goal called Pursuit Before Exit is based on research that explored the use of presumptive nudging and tone to encourage patients, who may be hesitant, to adhere to vaccination. This approach involves "pursuing" the patient if he or she simply rejects adherence to vaccination schedule.

This application ontology is intended to be used in a prospective dialogue engine for embedded and mobile devices that will automate a counseling session for HPV vaccine. Currently, the focus is on vaccine counseling, but the researchers foresee the possibility that PHIDO could cover patient-centric communication of health information for a variety of topics, while being grounded in some behavioural theory. The researchers' outline various next steps, including developing the software engine that will utilise the ontology and automate the dialogue interaction of a software agent.