The MARVIN Chatbots to Provide Information for Different Health Conditions

A Master Research Protocol to Adapt and Evaluate an Artificial Intelligence Based Conversational Agent to Provide Information for Different Health Conditions: the MARVIN Chatbots

This research is a continuation of a usability study with the MARVIN chatbot. The investigators aim to adapt the MARVIN chatbot to open it to other health domains (e.g. breast cancer) and populations (e.g. pharmacists). Therefore, this protocol constitutes a master research protocol that will englobe different research projects with individual chatbots. The investigators adopt an adaptive platform trial design, which will allow flexibility in handling multiple interventions adapted to different populations while retaining the characteristics of a platform trial design allowing early withdrawal of ineffective trial arms based on interim data (implementation outcomes) and introduction of new trial arms.

From 2020, the Principal Investigator's team has developed an innovative chatbot named Minimal AntiretRoViral INterference (MARVIN), for people with HIV. The primary objectives with this master protocol are to: 1) develop or adapt the MARVIN chatbots for different healthcare contexts through co-construction; 2) evaluate the global usability of individual chatbots in their respective use scenarios with a small sample of research participants; 3) evaluate the implementation and user outcomes (e.g., fidelity, appropriateness) of individual chatbots in their respective use scenarios with a larger sample of research participants. This includes also documentation of factors that inhibit or enhance the utilization of the chatbots and outcomes; and 4) establish and evaluate, through the previous objectives, different partnerships with research participants and research team members to co-construct chatbots that will respond to their needs and encourage continued use of the tool.

Chatbot, Intelligent Conversational Agent, Co-Construction, Adherence, Antiretroviral Therapy, mHealth, Implementation Science, Care, Breast Cancer, HIV, Self-Management

The investigators adopt a platform trial design, which is defined as "a study designed to evaluate multiple interventions in the context of a single disease in a perpetual manner, with therapies allowed to enter or leave the platform on the basis of a decision algorithm.". Platform trials can also integrate elements of basket trial design, which aims to study a single intervention for different diseases. In the current case, with different healthcare settings and diseases or conditions under study, the investigators would have a mix of platform trial and basket trial designs for an adaptative platform trial design. Such will allow flexibility in handling multiple interventions adapted to different populations while retaining platform trial characteristics, such as early withdrawal of ineffective trial arms based on interim data (implementation outcomes) and introduction of new trial arms. Meanwhile, there is no initial fixed duration or sample size for each arm.

Co-construction of the chatbot, Usability study, Implementation, Evaluation of outcomes and Continuous improvements

Co-construction of the chatbots, Usability study, Implementation, Evaluation of outcomes and Continuous improvements

Co-construction of the chatbots, Usability study, Implementation, Evaluation of outcomes and Continuous improvements

The UMUX-Lite is a 2-item questionnaire, answered on a 7-point Likert scale (1 = strongly disagree to 5 = strongly agree), that was deemed appropriate for evaluating medical technology. The items ask whether the chatbot meets the user's needs and the user's perceived ease of use. Scores range from 0 to 100, with more than 68 being considered high usability according to the literature.

The AES contains 6 items rated with different 5-point Likert scales (e.g., 1=very difficult to 5=very easy). It is a validated measure of acceptability and usability of computer-based interventions for health populations. Scores range from 0 to 30, and according to the literature, scores above 24 are considered high acceptability.

The AES contains 6 items rated with different 5-point Likert scales (e.g., 1=very difficult to 5=very easy). It is a validated measure of acceptability and usability of computer-based interventions for health populations. Scores range from 0 to 30, and according to the literature, scores above 24 are considered high acceptability.

Appropriateness relates to the relevance or compatibility of the innovation to address a particular issue or problem. The compatibility of an information technology innovation is the extent of which it is considered consistent with users' values, needs, and past experiences. The Compatibility Subscale is a validated tool that contains three items and evaluates on 7-points Likert scales (1 = extremely disagree to 7 = extremely agree) how an IT innovation "fits" with the user's work style. A modified version could be used by healthcare professionals, such as pharmacists.

Appropriateness relates to the relevance or compatibility of the innovation to address a particular issue or problem. The compatibility of an information technology innovation is the extent of which it is considered consistent with users' values, needs, and past experiences. A validated tool that could be used by patients is the Intervention Appropriateness Measure (IAM). It also contains 4 items scored on a five-point Likert scale (1=strongly disagree to 5=strongly agree) of agreement evaluating how the innovation is suitable for users.

Inclusion Criteria for all objectives: being 18 years or older; being fluent in English and/or French; being able to understand the requirements of study participation and provide informed consent during the duration of the study; having access to a smartphone, tablet, or computer at home/at workplace; having access to an internet connection at home or data plan on their device. Inclusion Criteria Specific to objectives 2 and 3: accept to use a Facebook Messenger-based Chatbot; accept to use or create a personal Facebook account; accept Facebook's privacy and data security policies. Exclusion Criteria: not meeting the inclusion criteria any reason, in the opinion of the investigator, which would make the candidate inappropriate for participation in an investigative study involving a chatbot (e.g., cognitive deficit)