This qualitative study is reported according to the Consolidated Criteria for Reporting Qualitative Research (COREQ) [20].

An explorative, qualitative study using focus groups was performed. In total, 3 focus groups were held with the aim of reaching data saturation (ie, the point where additional data collection does not produce new relevant themes). For each focus group, between 8 and 12 participants were included. Focus groups were held in the Sint Maartenskliniek Hospital in the Netherlands from February to March 2023.

Adult patients (≥18 years of age) were eligible for inclusion if they were prescribed at least 1 DMARD for an IRD from the outpatient pharmacy of the Sint Maartenskliniek Hospital, were proficient in Dutch, and owned an electronic device capable of accessing the digital human (ie, smartphone, tablet, laptop, or desktop) at the time of the study. Participants were purposively sampled based on age, sex, diagnosis, and digital inclination (defined as patients using or not using the web-based personal health record used by the Sint Maartenskliniek Pharmacy). Eligible patients were recruited via email, which contained study information and a form to state availability for the selected dates for focus groups. Written informed consent was obtained in person from participants prior to the focus groups. Incentives for participation were offered in the form of a 20 Euro (US $20.84) gift card, and travel expenses were fully reimbursed.

Pharmi [21], the digital human used during this study, was developed by Pharmi BV and Deloitte in collaboration with the Sint Maartenskliniek Hospital. Pharmi provides spoken answers to drug-related questions and offers additional information, such as (instructional) videos, drug leaflets, and other relevant sources, for approximately 80 different DMARDs. During the study, a prototype of Pharmi was used to provide information and offer predefined answers to common questions, similar to those found on a “Frequently Asked Questions” page. Topics covered included the drug’s mechanism of action, common side effects, instructional videos, storing instructions for antirheumatic drugs, information on specific IRDs, and logistical details such as how to obtain repeat prescriptions (Figure 1).

Participants were asked to complete a preparatory exercise at home prior to the focus group to familiarize themselves with the digital human. This exercise consisted of a set of predetermined frequently asked questions personalized to each participant’s current DMARD and diagnosis (Multimedia Appendix 1). In addition, participants were asked the following questions: (1) “How would you rate the digital human on a scale from 1 to 10?” (2) “What advantages do you see with the digital human? Name a maximum of 2” (3) “What disadvantages do you see with the digital human? Name a maximum of 2.”

Focus groups were held at the Sint Maartenskliniek Hospital and lasted 2 hours. Each focus group was led by an experienced moderator (author BJFB or CLB), and an assistant moderator (author LLH) was present to facilitate assignments and take additional field notes. All moderators were researchers at the time of the focus groups. A topic guide was used to standardize and structure each focus group (see Multimedia Appendix 1). This topic guide was based on the implementation outcomes outlined by Proctor et al [22] (ie, acceptability, adoption, appropriateness, costs, feasibility, fidelity, penetration, and sustainability), which are a frequently applied set of outcome measures that describe the concept of implementation outcomes. Service and client outcomes were not included in the topic guide, as these are intended to assess interventions after implementation. All focus groups were audio recorded, and recordings were transcribed verbatim. After 2 focus groups, data saturation was discussed to determine if more than 3 focus groups were necessary.

Transcripts were systematically analyzed using inductive thematic content analysis according to the steps outlined by Braun and Clarke [23] using Atlas.ti 9 (ATLAS.ti Scientific Software Development GmbH) and Microsoft Excel (Microsoft Corp). Transcripts were coded by 3 authors (LLH, VJBH, and CLB). First, open codes were generated for one transcript by LLH and checked by VJBH, after which discrepancies were discussed until consensus was achieved. Second, axial codes were generated by LLH and then discussed by LLH, VJBH, and CLB until consensus was achieved. Third, preliminary themes were constructed by LLH, VJBH, and CLB, after which open and axial coding for the remaining 2 focus groups was performed and themes were revised accordingly. Relevant quotes from participants were selected to support the themes, which were translated into English.

The Medical Ethics Research Committee of Arnhem-Nijmegen waived official ethical approval (case number 2023-16146) as this study was deemed exempt from the Medical Research Involving Humans Act.

A total of 540 eligible patients were approached for participation, among whom 214 (40%) responded and 31 (6%) agreed to participate. Of these, 9 patients did not participate in a focus group due to sickness (n=4, 44%), sudden unavailability (n=1, 11%) or unspecified reasons (n=4, 44%). Table 1 shows the participants’ characteristics.

A total of 6 themes reflecting factors influencing participants’ intention to use a digital human to retrieve medication-related information were identified relating to the patient, digital human, and the expected benefit from using the digital human (Figure 2).

Participants emphasized the importance of privacy and security in their intention to use the digital human. Regarding security, they expressed worries about the possibility of microphones being tapped and the risks associated with cookies from other websites, which could negatively influence their intention to use the digital human. Some participants felt privacy was guaranteed, as questions could be asked anonymously.

However, others had concerns about privacy when using the digital human in public spaces, as verbal questions and answers could be heard by others.

The anticipated benefits of using the digital human, both at the individual level and societal level, influenced participants’ intentions to use it. On an individual level, the digital human was perceived as a tool to enhance self-management by giving patients more control over their own care.

The participants viewed the digital human as a way to increase their knowledge about medication and thus their involvement in their treatment. They felt that the digital human enabled them to address issues independently without having to reach out to HCPs, thereby avoiding additional burdens on HCPs.

Furthermore, participants believed that using the digital human could improve their digital skills and preparedness for the digitization of health care.

This qualitative study showed that the intention of patients with IRDs to use a digital human for retrieving medication-related information was influenced by several factors. These factors included individual patients’ information needs and their confidence in using the digital human; the digital human’s features, including human-like interaction and technical functionalities; and the expected benefits of using it.

Over the past decade, digital humans have been developed for various chronic conditions to facilitate screening, diagnosis, and education [17]. However, using a digital human specifically for retrieving medication-related information in rheumatic diseases is a novel approach. As a result, direct comparisons with earlier studies investigating digital humans in the same context are difficult.

In this study, the intention to use the digital human largely depended on patients’ individual information needs, including their need for concise, trustworthy, accessible, and tailored information. These factors have also been described in other studies regarding the information needs of patients with rheumatic diseases [12,24,25]. Additionally, patients’ confidence in their ability to use the digital human and their eHealth literacy skills influenced their use intentions. This aligns with previous research showing that health literacy is associated with the ability to retrieve online health information [26,27] and evaluate its quality [28].

Another theme influencing intention was the ease of use of the digital human. This relation is widely documented in frameworks such as the Technology Acceptance Model, which describes “perceived ease of use” as a factor that indirectly influences the intention to use [29]. Given that the digital human used in this study was a prototype, participants’ remarks regarding ease of use should be considered during further development to maximize future intentions to use and thus adoption.

Additionally, participants emphasized the importance of the digital human’s technical functionalities, particularly its ability to provide correct answers. A mixed methods study on the use of digital humans for patients with dementia similarly found that technical problems and issues with speech recognition negatively impacted patient’s intention to use [30]. In this study, participants also indicated that a negative first experience with the digital human (ie, an experience in which a correct answer was not provided by the digital human) could deter future use intention. Therefore, ensuring an optimal first experience is critical for successful implementation and adoption by patients.

Research on digital humans applied in different contexts suggests that design features, such as facial expressions and tone of voice, influence use intentions [31]. For example, a previous study on a digital human for type 2 diabetes self-management found that its acceptability partly depended on the congruence between its verbal and nonverbal communication [32]. Similarly, in this study, participants indicated that the degree to which the digital human resembled a natural, human-like conversation partner influenced their intention to use it.

Various communication and information channels are available in health care, including frequently asked questions, phone or video calls, and 2-way text messaging with HCPs. These channels vary in interaction levels, user experiences, accessibility, and monetary or personnel costs associated with using the channel. It is conceivable that the digital human is not always the most suitable channel in every situation and for every individual (eg, a patient can experience multiple problems according to which information should be tailored). However, offering the digital human alongside existing channels allows patients to choose the option that best matches their needs and preferences. This could increase the actual usage of such channels to retrieve medication information, leading to patients gaining more medication knowledge, which is recognized as an important element of patients’ self-management [33].

A notable strength of this study is its theory-based approach to constructing the topic guide, which was informed by the Proctor framework for successful implementation [22]. The factors identified in this study are valuable for integrating the digital human into routine care. For example, integrating the digital human into digital patient health records could enhance the trustworthiness and accessibility of medication information. These factors are also relevant for further development of the digital human to increase future adoption. For example, speech recognition could be improved to increase the digital human’s ability to provide correct answers. This is especially valuable because the uptake of telehealth is low among patients with IRDs despite their willingness to use it [34].

A few limitations should be acknowledged. First, the digital human used in this study was a prototype and had several technical flaws, such as suboptimal processing of participants’ questions participants and difficulties in controlling the digital human. These issues may have influenced the range and type of technical factors participants identified as impacting their intention to use the digital human.

Second, the participant group was slightly older than the nonparticipant group. The nonparticipants’ median age was 59 (IQR 49-68) years. This might be explained by the fact that 2 out of 3 focus groups focus groups were held face-to-face during office hours, potentially making it easier for older adults with more spare time to attend. Additionally, this study had a low participation rate (6%). Most patients did not respond to the online invitation, and those who declined to participate were not obliged to provide reasons, leaving the reasons for nonparticipation unknown. Third, the participants were recruited from a single, specialized rheumatology clinic and might differ from the general IRD population in in terms of their experiences with the level of quality of care and medication information provided. This may limit the generalizability of the study’s findings.

This qualitative study highlights that the intention of patients with IRDs to use a novel digital human for medication-related information is influenced by their individual information needs, confidence in their ability to use the digital human, its features, and the expected benefits of using the digital human. These themes should guide the development and implementation of a digital human to maximize patients’ intention to use and adoption. Future research should examine the effect of applying the digital human as a tool to improve patients’ self-management of DRPs.