JMIR Publications

ABSTRACT

Background:

Cancer genetic testing has grown exponentially in the past decade in its use for quantifying hereditary cancer risk, while also its use for targeting treatment and care. With this continued growth and shortage of healthcare workforces, there is a need for automated strategies that provide high quality genetics services to patients in order to reduce clinical demand on genetics providers. Conversational agents have shown promise in managing mental health, patient pain management, and chronic conditions and are increasingly being used in cancer genetic services. However, research on how these agents are utilized by patients is limited.

Objective:

Therefore, our current study aim was to assess users’ interactions with a conversational agent for pre-test genetics education prior to genetic testing.

Methods:

The conversational agent provided scripted content similar to what is delivered in a pre-test genetic counseling visit for cancer genetic testing. Outside of a core set of information delivered to all patients, users were able to navigate within the chat to request additional content in areas of interest. An AI preprogrammed library was also established to allow users to ask open-ended questions of the conversational agent. Descriptives statistics were used for quantitative measures and thematic analysis was used for qualitative responses.

Results:

Of 93 National Comprehensive Cancer Network guidelines eligible patients offered access to the conversational agent, 36 started the chat (38.7%), with 30 completed the chat (32.3%). Once a participant completed a chat, a transcript of the interaction was developed for the research team, from which we extracted the following data and compiled results. The majority of users who completed the chat indicated that they wanted to continue with genetic testing (70%, n=21); 30% were unsure (n=9) and no patients declined moving forward with testing. Those who decided to test spent an average of 10 minutes on the chat (SD=2.57), selected an average of 1.2 additional pieces of information, and generally did not ask open-ended questions. Those who were unsure spent 4.0 more minutes on average (mean=14.1, SD=7.41, P=.03) with the chat, selected an average of 2.9 additional pieces of information, and asked at least 1 open-ended question.

Conclusions:

Our results indicate that the chat met the information needs of the majority of patients considering cancer genetic testing. However, a subset of patients may need additional education or interpersonal support prior to making a testing decision. Such high-information seeking patients may need additional support from a clinical provider to make a testing decision. Genetic counseling team members should follow up with patients who have additional questions and concerns to alleviate any remaining concerns. Therefore, conversational agents have the potential to be a scalable alternative for pre-test genetics education, reducing clinical demand on genetics providers.