Introduction

Despite being safe and effective^1,2, the HPV vaccine is underutilized in the United States. For those who did not receive the HPV vaccine as an adolescent, CDC guidelines recommend vaccination up to age 26. Therefore, addressing barriers for young adults who are unvaccinated is just as important as addressing barriers for parents with children. Motivational interviewing (MI) is a construct developed by Miller and Rollnick³ to effect behavior change, and over time it has been used for a wide variety of populations and health conditions, supported by a growing body of evidence in the literature. MI has shown promising results in addressing vaccine hesitancy, including HPV vaccination⁴.  With the development of artificial intelligence (AI) chatbots we see the possibility to merge MI – a complex, conversational framework – and a technology platform that can alleviate the burden and time commitment of carrying out such conversations for physicians.

Methods

A flowchart script was created to model dialogue for an AI chatbot conversation pathway utilizing a motivational interviewing framework. The script serves two purposes: 1) to provide a script to a human operator for a simulation study (“Wizard of Oz protocol”) of MI chatbot for HPV counseling, and 2) to provide a blueprint for the development of a MI chatbot to discuss HPV vaccine to young adults.

The initial framework was developed from the Transtheoretical Model of behavior change (TTM)⁶, guiding the conversation from pre-contemplation, to contemplation, to preparation and action stages. Within these stages of change, the content of discussion drew upon the Health Belief Model of behavior change (HBM)⁷ to emphasize discussion topics that may best influence behavior change. With this framework and content rooted in public health theory, documented motivational interviewing techniques were applied in the dialogue itself. A breakdown of these theories and their applications within the script is outlined in Figure 1.

Results

The complete version of our script model is available here. It utilizes key MI elements, such as OARS (Open ended questions, Affirmations, Reflective listening, and Summarizing) statements, asking permission, asking to weigh pros and cons, and a readiness scale. It uses a framework of TTM to move the user through the stages of change, and identifies elements of the HBM that may be influencing the user’s readiness for behavior change. Other MI strategies include asking permission, normalizing, and asking people to evaluate the beneficial and harmful aspects of their behavior⁸. Also, the readiness scale (Figure 2) is utilized to both elicit change talk from the user, and to evaluate where the user may fall on the TTM stages of change.

Conclusion

We intend to work with informatics researchers to test the dialogue model in a Wizard of Oz (WoZ) experiment in a dental clinical environment at the School of Dentistry (SOD) at University of Texas Health Science Center at Houston (UTHealth). The WoZ experiment⁹ entails having live participants interact with a natural language interface of the chatbot, where an operator feeds the agent utterances from the script. Informatics researchers have experimented with chatbots in HPV vaccine counseling using a WoZ experiment^10,11.  We will work with informaticians to embed this dialogue model into a computational model for the development of chatbot to perform MI for HPV vaccine uptake.

Also, this script can be used in the training of dental students and faculty on implementing MI in their discussions of HPV vaccination with patients. At UTHealth’s SOD, the pediatric clinics have already initiated brief questions on HPV vaccination status within our EHR. For patients who are unvaccinated or still considering vaccination, MI could be implemented during their dental appointment if faculty, students, and residents receive the appropriate training. This script can serve as a framework for that training for our future piloting and implementation in dental education.

References