This study adopted a quasiexperimental design involving the recruitment of patients with CKD from the nephrology departments of 2 regional hospitals in northern Taiwan that are similar in size and scope. Each hospital was assigned to either an experimental or comparison group. Health care professionals initially screened patients based on the eligibility criteria and referred interested parties to the research team. This process continued until 30 participants were selected from each hospital. However, due to the COVID-19 pandemic, only 14 participants in the experimental group and 16 participants in the comparison group completed the study. Participant attrition was primarily attributed to the lengthy intervention period and pandemic-related disruptions. Participants in the experimental group received both intervention and usual care provided by nurse practitioners, whereas those in the comparison group received only usual care. During recruitment, the researchers explained the use of the interactive robot and the basic interface operations required. After obtaining written informed consent, the participants were provided with an operation manual and guidance regarding the use of the interface. The flowchart of this study is presented in Figure 1.

The inclusion criteria required participants to be adults aged 20 years or older with a diagnosis of CKD. The exclusion criteria included dialysis, transplantation, severe mental illness, dementia, or significant sensory impairments. Patients undergoing dialysis were excluded because of significant differences in care standards and intensity compared with CKD management [23].

An interdisciplinary team including 2 nephrologists and nurse practitioners developed a 12-unit health coaching program using the interactive robot Kebbi Air S (NUWA Robotics Corp). The Kebbi robot, known for its versatility and interactivity, has been used across various domains, including education, health care, and customer service [24]. Prior studies have investigated its effectiveness for language learning and social assistance in health-related contexts and the impact of its expressions on participant engagement [25-27]. The program was designed using government-developed guidelines to ensure standardization and validity [21] after incorporating clinical experiences, patient interactions, and insights from 14 development meetings. These physicians and nurse practitioners acted as virtual health coaches through the robot, delivering content through dialogue, animations, and interactive tasks using artificial intelligence and natural language processing to facilitate dynamic patient engagement. The intervention was conducted using participant-staff 1-on-1 sessions during which staff guided participants in familiarizing themselves with the interface and provided assistance as needed. The robot was programmed to emulate health care professionals’ tone and demeanor, offering positive reinforcement (praise, encouragement, and motivational gestures) to sustain patient engagement. Correct answers on standardized quizzes were rewarded with applause, while supportive hints and explanations were provided after incorrect responses to facilitate learning. Virtual health coaches presented information tailored to the 5 stages of CKD, responded to patient inputs, and provided feedback. The program covered critical areas of CKD-specific HL, including functional, communicative, and critical literacy as well as CKD-specific knowledge, including basic knowledge, prevention, lifestyle, dietary intake, and medication. Table 1 provides an overview of the course learning objectives and corresponding outcome variables. Figure 2 illustrates an example of the program’s interface.

The research team selected an experimental hospital and collaborated with attending nephrologists to define the goals, procedures, and protocols of the study. Upon receiving the hospitals’ approval to conduct the study, recruitment posters were distributed to encourage patients with CKD to participate, while an orientation session introduced participants to the study’s objectives and collaborative framework. Data collection involved 1-on-1 interviews, enabling in-depth interactions and ensuring that participants were fully engaged with the intervention. Patients were encouraged to interact with the robots to become familiar with the interface. The experimental group participated in a 12-unit intervention, with each unit lasting approximately 20 minutes to 30 minutes. Participants completed 3 to 4 units per session across 3 to 4 learning sessions. The comparison group received usual care, including personalized education, from nurse practitioners. Structured questionnaires were used for the pre and postintervention assessments.

The effectiveness of the intervention was assessed using a validated questionnaire [5,28]. The CKD-specific HL scale and CKD-specific disease knowledge scale were adapted from validated tools specifically designed to assess HL and disease knowledge in patients with CKD. The CKD-specific HL scale includes 17 items for functional, communicative, and critical literacy in a multiple-choice test format. The CKD-specific disease knowledge scale has 13 true or false items across 5 domains. Higher scores indicated better HL and knowledge. The Cronbach coefficients were 0.78 and 0.76 for the CKD-specific HL scale and the CKD-specific disease knowledge scale, respectively.

Data were analyzed using SPSS version 23.0 (IBM Corp). This study primarily aimed to evaluate the intervention’s effectiveness at enhancing CKD-specific HL and disease knowledge. No significant differences were found after comparing the pretest scores of the outcome variables between the experimental and comparison groups using t tests, eliminating the need to control for the pretest scores. However, due to missing data for some variables, generalized estimating equations (GEEs) were used to analyze the pre and posttest results between the 2 groups while also controlling for age differences. The GEE approach was selected because it effectively handles repeated measures and correlated data, making it well-suited for our study design. It is robust and accommodates missing values, providing reliable estimates and enhancing the validity of our findings.

This study was approved by the Joint Institutional Review Board of Shin Kong Wu Ho-Su Memorial Hospital (approval number 20211202R) and Cardinal Tien Hospital (approval number CTH-110-3-5-002). All participants were fully informed and provided their consent before the interviews began.

The demographic characteristics of participants in the experimental (n=14) and comparison (n=16) groups were not significantly different in gender (χ²₂₈=0.27; P=.61), marital status (χ²₂₈=0.05; P=.82), and educational level (χ²₂₈=0.20; P=.65), although the experimental group was significantly younger.

As shown in Figure 3, the results of the t tests indicated that the posttest scores for CKD-specific HL (t₂₇=–4.44; P<.001) and disease knowledge (t₂₈=5.58; P<.001) were significantly higher in the experimental group than in the comparison group.

Table 2 shows how both groups’ results changed over time. After age adjustment, the GEE analyses demonstrated significant improvements in CKD-specific HL (P=.02) and disease knowledge (P=.001) in the experimental group compared with the comparison group. A significant group ×time interaction was observed for both CKD-specific HL and disease knowledge. Specifically, the experimental group showed improvements in HL (coefficient=2.51, Wald χ²₁=5.89; P=.02) and disease knowledge (coefficient=1.66, Wald χ²₁=11.75; P=.001).

The group differences in patterns of change over time shown in Table 2 also demonstrated that the experimental group had significant improvements in the scores for communicative literacy (coefficient=1.26, Wald χ²₁=6.17; P=.01) and critical literacy (coefficient=0.99, Wald χ²₁=5.68, P=.02) compared with the comparison group, but functional literacy did not significantly improve. Regarding CKD-specific disease knowledge, the experimental group showed significant improvement in medication (coefficient=0.80, Wald χ²₁=13.32; P<.001) compared with the comparison group, but there were no significant changes in areas such as basic knowledge, prevention, lifestyle, or dietary intake.

This study demonstrated preliminary evidence to support interactive robot-assisted virtual health coaching as an intervention tool for patients with CKD. Notably, participants in the experimental group showed significant improvements in both HL and disease knowledge compared with those in the comparison group. The experimental group was significantly younger than the comparison group, which may have influenced their receptiveness to the technology. However, prior studies indicate that older adults generally exhibit similar attitudes toward robots to those of younger individuals, challenging the stereotype of their lower robot-related receptiveness [29]. Furthermore, older adults often prefer human-like robots, such as the Kebbi robot, highlighting the importance of robot design and appearance in fostering acceptance [30,31]. These findings indicate that technology-enhanced educational interventions can effectively enhance HL and disease knowledge, which are associated with improved CKD-related outcomes.

The improvements observed in CKD-specific HL and disease knowledge were consistent with the benefits of technology-supported health education programs. Previous studies have shown that interactive digital interventions, particularly personalized coaching, can boost patient engagement and improve health outcomes [32,33]. This study builds on previous research by demonstrating the effectiveness of interactive robot-assisted virtual health coaching in managing chronic diseases. Our multidisciplinary approach included health professionals such as nephrologists and nurse practitioners, whose clinical experiences were integrated into the course content and represented by virtual health coaches [34]. Previous research has shown that avatar-based education has a positive effect on patient engagement and HL, which aligns with our findings [35].

A literature review of digital health coaching for older workers highlighted how virtual health can support healthy aging [36]. Although said review focused on a different group, it showed that digital interventions could effectively enhance health knowledge and well-being, especially when adopting a user-centered approach. This aligns with the findings of our study, which designed the intervention using health care professionals’ clinical experiences with patients. By integrating these real-world interactions, the digital tools were tailored to address the practical needs and circumstances of patients with CKD, making the content more relevant and applicable. These results underscore the importance of health care professionals’ insights when developing effective digital health interventions.

Participants in the experimental group showed significant improvements in communicative and critical literacy following the intervention, whereas functional literacy was not significantly changed. This pattern is consistent with the findings of a previous study in the Netherlands, which found that communicative HL and critical HL were more influential on chronic disease self-management, whereas functional literacy was less impactful [37]. Furthermore, a recent cross-national study involving geriatricians from Europe and Japan highlighted that virtual health interventions could enhance HL, emotional well-being, and social engagement among older adults, emphasizing the benefits of personalized digital coaching [38]. Our findings suggest that interactive virtual health coaching successfully supported patients in comprehending and using health information, making informed decisions, and critically evaluating health messages. However, the lack of improvement in functional literacy may be attributed to the program’s emphasis on enhancing communication, lifestyle modifications, and decision-making skills rather than basic skills such as reading medical documents or interpreting health indicators. Additionally, certain robot interface-related limitations may have impeded the delivery of more comprehensive health information necessary for developing functional skills. Future iterations should include more comprehensive content on functional skills to address all aspects of HL, thereby enhancing its overall effectiveness.

In addition to literacy, this study evaluated the effectiveness of the intervention on CKD-specific knowledge. The results showed a significant improvement in medication use in the experimental group, whereas no notable changes were observed in other areas such as basic knowledge, prevention, lifestyle, or dietary intake. This suggests that the program’s structured content was particularly effective for teaching patients about medication. The materials, designed by health care professionals based on their clinical experiences, incorporated visual aids and interactive discussions, making the content more engaging and easier to understand. This aligns with the findings of Occa et al [39], who highlighted that interactive features and visual informativeness help improve patient knowledge by enhancing cognitive absorption. Similarly, Hassan and Davies [40] found that virtual health tools improve medication adherence by boosting patient HL and reducing barriers, emphasizing convenience and empowerment. The limited improvement in other CKD-related knowledge areas may have been due to the small sample size, which may have affected the statistical significance of our findings. Additionally, some participants reported that similar topics had already been covered in educational brochures, which might have reduced their motivation to engage fully. A more comprehensive approach that places greater emphasis on prevention, lifestyle changes, and dietary guidance is required to address these areas better.

This study has important implications for integrating digital health technologies into CKD education. Interactive robots, combined with a virtual health coaching program, have exhibited promise in enhancing HL and disease knowledge, particularly in resource-limited settings where health care resources are scarce and chronic disease management is challenging. By leveraging health care professionals’ expertise and adhering to official health education guidelines, this study emphasizes the importance of creating standardized and engaging digital education tools that effectively meet patient education requirements. Furthermore, prior studies have demonstrated the potential of robots as impactful didactic tools in health science education, yielding promising results for enhancing cognitive therapies and facilitating health care intervention [41,42]. Expanding the use of such innovations may contribute to more effective patient education, improved disease knowledge, and enhanced HL, ultimately leading to improved CKD management.

Despite these promising findings, this study had several limitations. First, participants in the experimental group were significantly younger than those in the comparison group, partly due to recruitment challenges during the COVID-19 pandemic, which hindered the enrollment of a sufficient number of older participants. Although age differences were controlled in the GEE analysis, which helped mitigate their impact, they remain a potential source of bias. Future research should further explore age-related differences, potentially incorporating multivariable analyses, to better understand and minimize their effects, ensuring the findings’ broad applicability. This study primarily aimed to evaluate intervention effectiveness. Though digital literacy gaps related to age differences were not the primary focus, they represent an important area for future research.

The relatively small sample size limited the findings’ generalizability, and 1-on-1 recruitment made larger sampling challenging. Statistical power was not explicitly considered during the study’s design due to the lack of feasibility data, which accounts for the absence of a formal power calculation. Future studies should prioritize power analysis and incorporate larger sample sizes to confirm these results and enhance external validity. Additionally, the quasiexperimental design limited our ability to draw causal conclusions, suggesting that future research should use randomized controlled trials to strengthen causal inferences.

Another limitation involved the Taiwanese CKD population’s unique characteristics, including dietary habits, medication use (eg, traditional Chinese medicine), and lifestyle factors, which may limit the findings’ applicability to other populations. These cultural differences introduce potential biases, and future research should consider these factors to improve the results’ generalizability.

This study demonstrated that interactive robot-assisted virtual health coaching could potentially enhance CKD management by improving disease knowledge and HL. To our knowledge, this is the first study to use an interactive robot for CKD health education, presenting a novel approach that alleviates the burden on health care professionals. Developed by a multidisciplinary team, the program delivers standardized, credible education while engaging patients in their care. The findings emphasize the scalability of digital health solutions, particularly in resource-limited settings, offering a valuable tool for nephrology practices. Future research should validate these results across diverse populations, examine the long-term effects on HL and disease outcomes, and refine the intervention based on patient feedback to optimize its content and usability. Integrating digital tools into clinical practice could significantly advance patient education and chronic disease management.