Introduction

Older adults frequently seek medical advice, yet in some countries, the scarcity of medical resources prevents them from accessing immediate healthcare services (Strassmann et al. 2020; Turner et al. 2021). Virtual health agents are a form of digital health that facilitates convenient access to online medical resources, covering a wide range of areas such as disease self-management, monitoring, and health education (Milne-Ives et al. 2020). Users can assess their health status at home and receive health education (Winkler et al. 2023). Although virtual health agents could provide the online health services, people still have doubts about their effectiveness and do not fully accept them (Boustani et al. 2021; Kim et al. 2019). Therefore, exploring how to improve users’ trust and intentions to use virtual health agents can help alleviate the burden of social medical resources and provide better online health services for older adults.

Virtual health agents could bring convenience to users, and many previous studies have focused on the factors influencing users’ trust and intentions. Some studies used the Technology Acceptance Model (TAM) to predict the acceptance of virtual agents, with perceived ease of use and perceived usefulness as two key variables (Beldad and Hegner, 2018; Li et al. 2022). While the Technology Acceptance Model demonstrates high reliability and validity in predicting users’ intentions to use, many external factors also influence users’ intentions and behavior (Ghazizadeh et al. 2012). Social presence not only influences users’ trust in virtual health agents (Lee and Sun, 2023), but also affects the doctor-patient relationship between them (Liu et al. 2022). A stronger social presence makes virtual health agents exhibit more human-like characteristics, leading to more personalized conversations and making users feel like they are interacting with a real doctor (Lei et al. 2021). Furthermore, researchers have found that subjective norms influence users’ willingness to use technology (Choe et al. 2022; Schepers and Wetzels, 2007).

Virtual health agents establish a strong doctor-patient relationship with users through listening and communication (Philip et al. 2020). However, many previous studies have focused on the influence of technological capabilities on users’ acceptance, neglecting the doctor-patient relationship attributes of virtual health agents. Narrative medicine is a medical practice that involves the ability to recognize, absorb, interpret, and be moved by stories (Charon, 2006). Doctors with narrative abilities can not only listen to patients’ stories but also encourage patients to narrate their illness experiences. Thus, doctors can better understand patients’ health status and make accurate diagnoses (Charon, 2001). Therefore, this study aims at exploring how perceived medical narrativity influences older adults’ intentions to use virtual health agent.

This study focuses on enhancing older adults’ trust and their intentions to use virtual health agents from the doctor-patient relationship perspective. To achieve this, the study proposed a hypothesized model integrating the TAM and narrative medicine theory. Subsequently, this study conducted a questionnaire survey among 230 older adults from China. The results confirmed the influences of social presence, perceived narrativity, and subjective norms on intention to use virtual health agents.

Literature review and research hypotheses

Virtual health agents

Virtual agents are an emerging technology which combines artificial intelligence (AI) technology and a computer-generated entity to perform tasks (Sin and Munteanu, 2020). Virtual health agents mainly undertake three types of assignments, including health intervention, monitoring and assessment, and guidance and agency (Winkler et al. 2023). Health intervention refers to the agents managing symptoms of disease or conducting virtual rehabilitation training for users. Monitoring and assessment refers to the agents monitoring the users’ health status and conducting assessments. Guidance and agency aims to support users in developing health habits to cope with daily life and enhance their health autonomy. Although virtual health agents can provide health advice and enhance self-management skills, older adults do not accept virtual health agents. On the one hand, users question the professionalism of virtual health agents and do not trust them (Laranjo et al. 2018). On the other hand, the low penetration rate of virtual health agents and insufficient communication among patients affect users’ attitudes. Therefore, older adults often have concerns about using virtual health agents.

Technology acceptance model

In order to explain how users accept computers, Davis et al. (1989) proposed the TAM based on the Theory of Reasoned Action (TRA). Subsequently, the TAM has been applied in various domains and has shown high effectiveness in explaining user acceptance (Marangunić and Granić, 2015). The main purpose of the TAM is to explain the impact of external factors on internal beliefs, attitudes, and intention to use technology (Davis et al. 1989). In the TAM, perceived ease of use and perceived usefulness are two key factors to explain and predict users’ intentions to use virtual health agents (Davis, 1989).

Perceived ease of use refers to users’ beliefs that using virtual health agents does not require too much effort (Davis, 1989). Declining cognitive and information processing capabilities make older adults face more difficulties in using new technology, leading to technology anxiety (Cheng et al. 2023). This means older adults need more time to accept a new technology. Older adults would find virtual health agents convenient if they can use the agents smoothly. Therefore, perceived ease of use can enhance users’ perceived usefulness of virtual health agents (Marangunić and Granić, 2015).

H1a: Perceived ease of use has a positive influence on perceived usefulness.

Perceived usefulness refers to users’ perception of the anticipated benefits brought by virtual health agents (Bhattacherjee, 2001). Users believe that virtual health agents can provide them with accurate diagnoses and offer appropriate advice to improve their physical health. Perceived usefulness is crucial for users’ acceptance of a new technology. Goyal et al. (2022) found that perceived usefulness allows users to perceive the value of health consultation websites, especially in increasing their usage intentions at the initial stage. Thus, based on the TAM and the empirical research results, we hypothesize that:

H2a: perceived usefulness has a positive influence on intention to use virtual health agents.

Trust in virtual health agents

Trust is an important factor in determining users’ acceptance of a health technology (Philip et al. 2020). Trust is defined as “the attitude that an agent will help achieve an individual’s goals in a situation characterized by uncertainty and vulnerability” (Lee and See, 2004). It plays a role in navigating uncertain situations and helps individuals overcome uncertainties (Kuen et al. 2023). The purpose, process, and performance of virtual health agents can influence users’ trust (Schroeder et al. 2021). When using virtual health agents, users need to continually disclose their health status. This process involves risks and uncertainties due to the lack of transparency in the consultation process (Kuen et al. 2023). Older adults are not only concerned about the privacy of their information but also question the professionalism and accuracy of diagnoses made by virtual health agents (Beldad and Hegner, 2018). On the one hand, older adults need to provide their personal information to virtual health agents and illustrate their medical conditions. They are concerned about whether their personal information will be misused. On the other hand, only when older adults adopt the medical advice given by virtual health agents can they judge the accuracy of the diagnoses (Dhagarra et al. 2020). Trust encourages users to establish the doctor-patient relationship with virtual health agents and influences users’ compliance with medical plans (Petrocchi et al. 2019). Older adults would like to share their medical history with virtual health agents, follow the agents’ medical advice, and adhere to the agents’ health guidance (Philip et al. 2020). Therefore, trust is the most important factor influencing the intention to use virtual health agents (van Bussel et al. 2022).

H3: Trust has a positive influence on intention to use virtual health agents.

The main point of the TAM is that users’ beliefs affect their attitudes toward technology, thereby influencing their intentions and ultimately leading to corresponding behaviors (Kuen et al. 2023). Perceived ease of use and perceived usefulness are two beliefs closely related to intention to use. Many studies have combined trust with the TAM in different ways, especially in contexts with high uncertainty. Some studies defined trust as a belief or attitude. Kim et al. (2023) consider trust as a belief and divide it into four attributes: reliability, safety, security, and resilience. Benamati et al. (2010) argued that the relationship between “beliefs-attitudes-intentions to use” in the TAM is reflected as “trusting beliefs-trusting attitudes-intentions to use” in trust. They found that users trust technology that could bring benefits, and the impact of perceived ease of use on trust is not significant. However, Wu and Song (2021) argued that, compared to younger people, older adults lack sufficient experience with new technology. For older adults, the impact of perceived ease of use on attitudes is equally important. Therefore, we propose the following hypotheses:

H1b: Perceived ease of use has a positive influence on trust.

H2b: Perceived usefulness has a positive influence on trust.

Medical presence

Social presence is “a psychological state in which virtual social actors are experienced as actual social actors in either sensory or non-sensory ways” (Lee et al. 2021). Social presence not only influences users’ trust in virtual health agents (Lee and Sun, 2023), but also affects the doctor-patient relationship between them (Liu et al. 2022). During health consultations with virtual health agents, older adults experience a sense of medical presence. Medical presence enables older adults to establish a psychological connection with virtual agents, and make them more willing to share their experiences of being sick (Lee et al. 2006).

In the theory of computer-mediated communication, social presence is a factor to evaluate communication effectiveness. To measure and validate the role of social presence, Biocca et al. (2001) proposed social presence can be divided into three dimensions: co-presence, psychological involvement, and behavioral engagement. Subsequently, Biocca and Harms (2002) further elaborated these three dimensions as three levels: co-presence level, subjective level, and intersubjective level. Media that enable face-to-face communication have the highest social presence, followed by media that combine audio and video. Media that only use audio communication have the weakest social presence. Social presence affects the degree of interpersonal attraction between individuals and virtual agents in terms of task attraction and social attraction, thereby enhancing trust in virtual agents (Lei et al. 2021).

Medical consultations are social activities that require high involvement and should use virtual agents with high social presence (Fulk et al. 1990). Older adults need to participate in medical decisions, discuss their medical conditions with doctors, and negotiate medical plans together. The immersion and high involvement of older adults in health consultation with virtual health agents can create a sense of medical presence. Higher medical presence can establish trust that lasts for a period of time in users (Torre et al. 2020). Therefore, users may be more inclined to trust virtual health agents that can provide a higher medical presence.

H4a: Medical presence has a positive influence on trust.

Previous studies have shown that social presence has a positive impact on intention to use virtual agents (Silva et al. 2023). Virtual health agents with higher medical presence can provide users with useful health information, meet users’ daily medical needs (Lee et al. 2021), and provide users with a better medical service experience (Silva et al. 2023). Krieger et al. (2021) found that people are more willing to share their illness experiences and seek health guidance with virtual health agents that have higher medical presence. Users feel warmth in communication with virtual agents (Silva et al. 2023). Therefore, we propose the following hypotheses:

H4b: Medical presence has a positive influence on intention to use.

H4c: Medical presence has a positive influence on perceived usefulness.

Perceived medical narrativity

Narrative is a presentation of connected events and characters within limited space and time, containing explicit or implicit information related to the theme (Kreuter et al. 2007). Characters and their experiences constitute the story in the narrative, and the extent to which the story is conveyed is perceived narrativity (Ma, 2023). Narrative is a familiar and fundamental social communication method. People exchange information through stories to understand what is happening around them (Kreuter et al. 2007). Patients come to visit doctors with their own stories, hoping that doctors can provide a diagnosis based on their detailed narratives. However, this process is often interrupted by doctors, which leads to some conflicts.

Narrative medicine applies narrative theory from literary studies to medical practice, addressing the shortcomings of modern medicine in terms of singularity, humility, accountability, and empathy toward patients. Virtual health agents with narrative medical characteristics can respect the uniqueness of each user, show empathy towards users in listening to their health status, and provide tailored medical advice. Perceived medical narrativity is the story of illness, recovery, and healing, that users perceived in health consultations. Virtual health agents with higher narrative medical characteristics make users feel understood and eliminate the need to expend additional effort in explaining their conditions (Charon, 2006). Therefore, older adults may perceive virtual health agents with higher narrative medical capabilities as easier to use.

H5c: Perceived medical narrativity has a positive influence on perceived ease of use.

Narrative medicine has five features, namely temporality, singularity, causality/contingency, intersubjectivity, and ethicality. Intersubjectivity takes place between two individuals. Phenomenology suggests that intersubjectivity includes not only cognitive behaviors of perception and interpretation but also individual transformations caused by interpersonal relationships (Charon, 2006). Intersubjectivity also exists in medical presence which generated in user-agent interaction. When interacting with others, people want to be close enough to ensure they can not only perceive others but also be perceived by others (Biocca and Harms, 2002). The medical narrative characteristics of virtual health agents can improve the sense of medical presence. Therefore, we hypothesize that:

H5a: Perceived medical narrativity has a positive influence on medical presence.

Narrative medicine regards patients as real individuals rather than diseases, emphasizing the need for healthcare professionals to listen to patients’ disease stories (Ahlzén, 2019). Narrative communication contributes to establishing positive doctor-patient relationships. Virtual health agents with narrative capabilities can construct the complete and causal disease stories during interactions with patients. This could provide some benefits throughout the diagnostic process. Before diagnosis, narrative communication helps users overcome resistance to examinations and disease monitoring. Simultaneously, users can express their true situation under the guidance of virtual health agents to avoid misdiagnosis. After diagnosis, narrative communication involves users in medical decision-making, offering optimal choices and coping strategies (Kreuter et al. 2007). The accuracy of diagnosis and the effectiveness of medical advice allow users to perceive the value of virtual health agents.

H5b: Perceived medical narrativity has a positive influence on perceived usefulness.

Subjective norms

Subjective norm, also known as social norm, which refers to “a person’s perception that most people who are important to him think he should or should not perform the behavior” (Schepers and Wetzels, 2007). In this study, the subjective norm for older adults’ use of virtual health agents is defined as older adults’ perception that important people in their lives expect them to use these agents.

In the Theory of Planned Behavior (TPB), attitudes, subjective norms, and perceived behavioral control can influence users’ intentions (Ajzen, 1991). Some studies after Ajzen suggested that norms have a much lower impact on intentions compared to behavioral beliefs and control beliefs. However, subsequent studies have argued that subjective norms are an important factor influencing behavioral intentions. These studies have developed the Focus Theory of Normative Conduct (FTNC) based on the TPB to explain specific influence paths (Randazzo and Solmon, 2018). Cialdini et al. (1990) categorized subjective norms into descriptive norms and injunctive norms. Rimal and Real (2005) proposed the Theory of Normative Social Behavior (TNSB) to further demonstrate the relationship between descriptive norms and injunctive norms. Subjective norms can lead to compliance effects and internalization effects. The compliance effect makes users believe in the opinions of individuals who are important to them, even if they do not originally hold those views. The internalization effect refers to users tending to base their judgments on information provided by individuals who are important to them. The internalization effect influences users’ perceptions of technology usefulness and thereby affects their acceptance of the technology (Schepers and Wetzels, 2007). Therefore, attitudes and opinions of older adults’ family members or peers toward virtual health agents can shape older adults’ technology usage experiences. If others perceive virtual health agents positively in terms of user experience and trust their ability to offer accurate medical advice, older adults are more inclined to sense the medical presence of the agents, finding them useful and ease to use. Therefore, we hypothesize the following:

H6a: Subjective norms have a positive influence on medical presence.

H6b: Subjective norms have a positive influence on perceived usefulness.

H6c: Subjective norms have a positive influence on perceived ease of use.

Figure 1 shows the hypothesized relationships among the research constructs.

Fig. 1
figure 1

Hypothesized model.

Full size image

Methods

Measurement

The questionnaire consists of two parts: the main content and demographic information. The main content includes 7 constructs, namely perceived medical narrativity, medical presence, intention to use, subjective norms, trust, perceived ease of use and perceived usefulness. Demographic information consists of 2 items, including age and gender. All items of the constructs were adapted based on mature scales in the existing literature to ensure the reliability and validity of the questionnaire. All items applied a 5-point Likert scale, ranging from 1 (highly disagree) to 5 (highly agree). The 3 items measured “perceived medical narrativity” were inspired from the scales used by Ma (2023) and Busselle and Bilandzic (2009). The 3 items measured “medical presence” were adapted from the scale used by Jin and Youn (2023). The 3 items measured “intention to use” were adapted from the scale used by Jin and Youn (2023). The 5 items measured “subjective norms” were adapted from the scales used by Miao et al. (2017) and Chen and Tung (2014). The items measured “perceived ease of use” and “perceived usefulness” were all adapted from the scale used by Srivastava et al. (2010). The 4 items measured “trust” were adapted from the scale used by van Bussel et al. (2022) and Lei et al. (2021). Table 1 provides the 7 constructs, 25 items and their respective references.

Table 1 Constructs, items and reference.
Full size table

Sampling and data collection

To ensure the diversity of the sample source, this study employed both offline and online survey methods for questionnaire distribution. Both survey methods were conducted from November to December 2023. For the offline survey, convenience sampling was utilized to recruit participants from a hospital and a community. For the online survey, this study commissioned the Credamo platform to conduct the survey. Credamo is a platform that provides professional survey services, with registered users from various provinces and regions across China (Wang et al. 2024). We set recruitment criteria and published the questionnaire on the platform, allowing eligible users to choose whether to participate. Older adults over the age of 55, regardless of whether they had previously used virtual agent technology, were eligible to participate in this survey. Participants needed to carefully read the preface before filling out the questionnaire. The preface explained the main purpose of the research, defined “virtual health agents” and their main uses, and included pictures as examples. Considering the older adults might not understand how to fill out the questionnaire, after the preface, we provided an example question to illustrate how to make a choice. Subsequently, participants needed to select the most appropriate option for each question based on their feelings. A total of 267 older adults participated in this survey (87 offline and 180 online). To ensure the objectivity and validity of the data, we removed the questionnaires with excessive repetitive answers and those completed in an unusually short time. As a result, 230 valid questionnaires were obtained for this study (65 offline and 165 online), with a valid rate of 86.14%. Table 2 shows the demographic information of participants.

Table 2 Demographic information of participants (N = 230).
Full size table

To better reflect the sociological characteristics of the participants, we conducted follow-up interviews to gather detailed information on their educational and occupational backgrounds. To provide compensation to each participant, we obtained their consent to collect contact information (phone number or email) after the initial questionnaire survey. Thus, we contacted all the participants to finish the follow-up interviews. 97.0% of the offline participants and 88.5% of the online participants agreed to take part, resulting in a follow-up rate of 90.1% with a total of 209 older adults. In terms of education, 44.9% had completed elementary school or below, 40.8% had received a middle school education, 10.2% had completed high school (including technical school), and 4.1% had received a college education or higher. In terms of occupation, 6.1% of participants work in public institution (civil servant, teacher, or doctor), 53.1% are employed in public service agency (security guard, cleaner, nanny, or waiter), 12.2% are self-employed, 10.2% are workers, and 18.4% of the older adults are retired.

Data analysis

This study adopted the two-step procedure proposed by Anderson and Gerbing (1988), and analyzed the data by Spss19.0 and Mplus 8.3 software. We conducted a confirmatory factor analysis (CFA) to ensure that the observed variables could reflect the latent variables. Cronbach’s α, composite reliability (CR), and standardized item loadings were used to assess the reliability of the measurement model. Cronbach’s α and CR greater than 0.7 indicated good internal consistency reliability, and values higher than 0.6 were acceptable (Barclay et al. 1995; Hair et al. 2009). The standardized item loading of every observed latent were recommended to be higher than 0.6 (Hair et al. 2009). Average variance extracted (AVE) and discriminant validity were used to assess the validity of the measurement model. An AVE value higher than 0.5 indicated good convergent validity (Hair et al. 2009). Discriminant validity refers to the differences among the traits represented by the latent variables. The square root of AVE (sqrtAVE) of every latent variable should larger than the Pearson’s correlation among other latent variables, indicating good discriminant validity (Hair et al. 2009). Model fit refers to the consistency between the hypothesized theoretical model and the actual data. We used the ratio of chi-square to degree of freedom (χ2/DF), the root mean square error of approximation (RMSEA), the standardized root mean square residual (SRMR), comparative fit index (CFI) and Tucker-Lewis index (TLI) as indictors. The criteria for good model fit were: χ2/DF did not exceed 3, CFI and TLI were not lower than 0.9, and RMSEA and SRMR were not higher than 0.8 (Bentler and Bonett, 1980; Browne and Cudeck, 1992; Iacobucci, 2010). Subsequently, we used Mplus software and employed maximum likelihood estimation to test the 14 hypotheses in the research model.

Results

Measurement model

To test the measurement model, this study assessed the reliability and validity of 7 constructs and 25 items. Table 3 shows standardized item loadings, Cronbach’s α, CR value and AVE value. The standardized item loadings of three items ranged from 0.6 to 0.7, while the rest were above 0.7. These results indicated a satisfactory reliability of the measurement model. All the items could reflect the constructs, and the internal consistency of the constructs was good. The AVE value of each construct was above 0.5. Therefore, the measurement model indicated a satisfactory convergent validity.

Table 3 Reliability and convergent validity for the constructs.
Full size table

Table 4 not only shows the mean values and standard deviation of all the constructs, but also represents the discriminant validity of the constructs. The sqrtAVE values were larger than any of the Pearson’s correlations. Thus, the constructs in this study were not highly correlated.

Table 4 Correlation analysis and discriminant validity of the constructs.
Full size table

Structural model

First, we assessed the fit of the structural model. Table 5 shows the relevant model fitting indices. Specifically, χ2/DF = 1.635, RMSEA = 0.053, SRMR = 0.040, CFI = 0.950, and TLI = 0.940. All the indices were within acceptable ranges, indicating a good fit between the data and the hypothesized model.

Table 5 Model fitting indices.
Full size table

Next, we tested our previous hypotheses. Table 6 and Fig. 2 shows the standardized coefficient and the significance of each hypothesis. Perceived ease of use was positively related to perceived usefulness (H1a: β = 0.914, p < 0.001) and trust (H1b: β = 0.816, p < 0.001). Perceived usefulness was positively related to intention to use (H2a: β = 0.558, p < 0.001), but not to trust (H2b: β = −0.345, p = 0.167). Trust was positively related to intention to use (H3: β = 0.571, p < 0.001). Medical presence was positively related to trust (H4a: β = 0.620, p < 0.001) and perceived usefulness (H4c: β = 0.435, p < 0.001), but not to intention to use (H4b: β = −0.162, p = 0.134). Perceived medical narrativity was only positively related to perceived ease of use (H5c: β = 0.679, p < 0.001), but not to medical presence (H5a: β = 0.184, p = 0.124) and perceived usefulness (H5b: β = −0.343, p = 0.074). Subjective norms were positively related to medical presence (H6a: β = 0.482, p < 0.001) and perceived ease of use (H6c: β = 0.220, p < 0.05), but not to perceived usefulness (H6b: β = −0.017, p = 0.855).

Table 6 Results of structural model.
Full size table
Fig. 2: Assessment of the structural model.
figure 2

*p < 0.05, **p < 0.01, ***p < 0.001.

Full size image

Discussion

The attributes of a doctor-patient relationship contribute to enhancing older adults’ intentions to use virtual health agents. Perceived medical narrativity, medical presence, and subjective norms are three factors related to establishing a doctor-patient relationship. This study investigated how these three external factors, through the Technology Acceptance Model, influence older adults’ trust and intentions to use virtual health agents. The results supported the nine hypotheses proposed earlier and provided the following key discoveries.

Medical presence could positively predict trust and intention to use

Medical presence can improve intention to use virtual health agents in two ways. First, it can enhance older adults’ trust in these agents, thereby increasing their intention to use them. Second, medical presence is positively associated with perceived usefulness, which also influences intention to use. Previous studies have suggested two possible relationships between medical presence and intention to use. Silva et al. (2023) indicated that social presence can directly influence users’ intentions to use virtual agents. Zhang et al. (2022) found that social presence can enhance users’ trust in online health consultations, thereby increasing usage intentions. The above studies focused on measuring users’ post-acceptance intentions (i.e., whether users are willing to use the agents after initial use) (Li et al. 2022). In contrast, this study focuses on measuring older adults’ initial intentions to use virtual health agents, which may account for the differences in research findings.

Medical presence can bring positive health consultation experiences for older adults. It makes older adults gradually develop trust in virtual health agents (Huang et al. 2023). Empirical studies have shown that social presence can influence users’ trust. This may be because virtual health agents with higher social presence can bring more social cues during communication. The social cues include verbal, auditory, visual cues, and invisible cues (such as waiting time) (Tastemirova et al. 2022). Communication lacking social cues is more likely to mask dishonest behavior. Conversely, more social cues make older adults perceive the entire process as transparent, leading them trust virtual health agents (Lee et al. 2021). On the other hand, virtual health agents with high medical presence tend to exhibit behaviors that better align with the expectations of older adults, enabling them to perceive the agents’ usefulness (Gefen and Straub, 2004).

This study focuses on older adults’ initial intentions to use virtual health agents. The research validates the influence of external factors, internal beliefs, and trust on intention to use. Medical presence makes older adults immerse in the experience of health consultation. Its importance lies in enabling older adults to trust virtual health agents and strengthening the doctor-patient relationship.

Perceived ease of use and medical presence are crucial mediators between certain variables

The mediating role of perceived ease of use

Perceived medical narrativity and subjective norms positively influence trust and intention to use virtual health agents, with perceived ease of use acting as a key mediator. Perceived medical narrativity, a key variable introduced in this study, have been less discussed in previous research regarding its connection with perceived ease of use. As for subjective norms, previous studies have indicated that they influence perceived usefulness. However, the relationships between subjective norms and perceived ease of use have rarely been studied (Schepers and Wetzels, 2007; Yi et al. 2006). Choi and Chung (2013) found that subjective norms impact both perceived ease of use and perceived usefulness. However, the results of this study indicate that subjective norms only enhance older adults’ perceived ease of use of virtual health agents, without leading them to perceive these agents as more useful.

The theory of narrative medicine can explain the relationship between perceived medical narrativity and perceived ease of use. In health consultations, virtual health agents with higher narrative medical abilities make older adults feel they are easier to use. In the context of medical diagnosis, doctors often lean heavily on examination reports and their experience, paying less attention to patients’ narratives. Consequently, patients often feel that doctors do not truly understand them. Virtual health agents with narrative characteristics can listen to older adults’ disease stories, encourage them to narrate, and let them hear their own voices (Charon, 2005). Virtual health agents show empathy towards older adults in listening and communicating, making them feel understood. Older adults do not need to spend too much effort explaining or repeating their conditions, which makes them feel understood and perceive the agents as ease to use.

Subjective norms from family members and other patients can influence older adults’ perceived ease of use, but they do not affect perceived usefulness. According to the social information processing theory, social cues from others can result in actual behavioral responses after being processed individually (Crick and Dodge, 1994). Obtaining subjective norms regarding virtual health agents from different sources allows older adults to filter information related to usefulness and ease of use based on their existing experiences. Subsequently, older adults engage in causal analysis of events mentioned in the information and make their own conclusions. This process is influenced by their previous memories and experiences and may change their original perception of virtual health agents. Thus, subjective norms from other patients can alter older adults’ perception of ease of use. However, in health consultations, the usefulness of virtual agents is related to the effectiveness of provided health interventions and guidance, which older adults can perceive only after adopting and implementing the recommendations. This explains why subjective norms cannot enhance older adults’ trust and usage intentions through perceived usefulness.

The mediating role of medical presence

Beyond the mediation of perceived ease of use, subjective norms also enhance trust and the intention to use through the mediating role of medical presence. This suggests that older adults are inclined to use the virtual health agents that provide a more immersive experience. Subjective norms can further enhance older adults’ sense of immersion, making them feel as if they are communicating with real doctors even in virtual scenarios. To explain this result, this study introduces the concept of narrative perspective. Narrative perspective refers to “the physical and psychological point of perception presented in a story” (Kim et al. 2020). Compared to a third-person narrative perspective, a first-person narrative can evoke stronger empathy and mental transportation (Pachucki et al. 2022). Older adults’ subjective norms regarding virtual health agents mainly come from family members and friends with similar illnesses, who often use first-person narratives to describe their usage experiences. The first-person narrative style allows older adults to experience stronger narrative engagement. It makes older adults immerse in others’ usage experiences (Samur et al. 2021). Therefore, attitudes and usage experiences from family members and friends significantly influence older adults.

Furthermore, the results suggest perceived medial narrativity do not enhance older adults’ trust and usage intentions through medical presence. Kim et al. (2020) indicated that narrative perspective can influence social presence. Virtual health agents often use third-person pronouns such as “other patients” or “he/she” in their narratives. Nan et al. (2017) indicated that, compared to a first-person narrative style, using a third-person narrative style has a weaker effect on persuading people to change health behaviors. While third-person narratives can enhance the objectivity of information delivery, they are not as effective as first-person narratives in enhancing cognitive and emotional identification (Samur et al. 2021). Virtual health agents primarily focus on providing medical narratives for disease self-management but are unable to share personal experiences like other patients. This third-person narrative style does not enhance older adults’ medical presence.

Difference between perceived ease of use and perceived usefulness

The impact pathways of perceived usefulness and perceived ease of use on intention to use are different. Perceived usefulness directly influences older adults’ intention to use technology, while perceived ease of use influences older adults’ intention to use virtual health agents in two ways. On the one hand, perceived ease of use enhances perceived usefulness for older adults, thereby increasing their usage intentions. On the other hand, perceived ease of use helps older adults trust virtual health agents, ultimately boosting their intentions to use them. Previous studies based on TAM mostly have explored the relationships between attitudes, perceived ease of use, and perceived usefulness (Wu and Song, 2021). Ha and Stoel (2009) found that only perceived usefulness influences college students’ attitudes toward virtual agents. Wu and Song (2021) focused on older adults, discovered that both perceived usefulness and perceived ease of use influence user attitudes. They also found perceived ease of use has a more significant impact on older adults’ attitudes. Considering the importance of trust in doctor-patient relationships, this study specifies attitudes as older adults’ trust in virtual health agents. Moreover, this study hypothesizes trust as a mediator which influences the relationships between perceived ease of use, perceived usefulness, and usage intention. The results show that perceived ease of use influences older adults’ trust in virtual health agents, thereby affecting their intentions to use them. However, perceived usefulness does not have a similar effect.

This may be because perceived ease of use helps older adults establish doctor-patient relationships and emotional connections with virtual health agents, whereas perceived usefulness is limited to fulfilling technical purposes. Therefore, perceived usefulness cannot influence the intention to use by fostering trust. Lewis and Weigert (1985) argued that trust includes three dimensions: cognitive, affective, and behavioral. Affective trust is based on emotional involvement and emotional connection, usually involving shared identities and values (Duenas and Mangen, 2023). There are two processes that occupy older adults’ cognitive resources when using virtual health agents. One is continuously providing information to the agents, and the other is understanding the feedback information from the agents. Therefore, older adults’ perceived ease of use of virtual health agents is influenced by these two aspects. When older adults feel that they do not need to spend too much effort explaining their health status and can understand the health information provided by the agents, they will identify with doctor-patient relationships and trust the agents.

The impact of perceived usefulness on trust is insignificant, but perceived usefulness directly influences older adults’ usage intentions. This indicates that improving usefulness is the key factor for older adults to accept virtual health agents, and merely improving the accuracy of information provided by virtual health agents cannot enhance the emotional connection between older adults and the agents. Social exchange theory posits that “social behavior is the result of an exchange process”, where individuals seek to maximize benefits and minimize losses during exchanges (Surma, 2016). This behavior also applies when users use online communication platforms, as seen in information exchanges. In other words, when older adults encounter outcomes that surpass their expectations during interactions with virtual health agents, they become more inclined to utilize the services.

The participants in this study were older adults, a demographic that has received relatively less attention in previous research. Older adults prioritize the usefulness of virtual health agents, making the provision of expected useful information their primary goal in communication. Furthermore, older adults are not as familiar with emerging technology as younger people, so they are more concerned about whether they can learn to use virtual health agents. If older adults can quickly master this technology, they will develop an emotional connection with it. This is reflected as forming a strong doctor-patient relationship in the communication with virtual health agents.

Conclusion

Major findings

This study reveals that doctor-patient relationships between older adults and virtual health agents stem from both technological characteristics (perceived medical narrativity and medical presence) and social factors (subjective norms). The results suggest that developers can enhance older adults’ usage intentions in two ways. On the one hand, developers should improve the usefulness of technology and make older adults believe that virtual health agents can improve their health status. For example, enhancing the specialization of agents in providing medical information. On the other hand, developers should foster a trust-based doctor-patient relationship between older adults and virtual health agents. Thus, developers can improve medical presence of virtual health agents and equip the agents with medical narrative abilities to enhance their ease of use. Lastly, subjective norms from society can influence older adults’ medical presence and perceived ease of use. The study expands the TAM and demonstrates the importance of virtual health agents possessing medical narrative capabilities. The findings not only enlighten technology developers to embed social factors into technology usage, but also guide designers in designing virtual health agents that better meet the needs of older adults.

Research limitations

Although this study provides some key findings, we must acknowledge its limitations. First, the study lacks comprehensive sociological information about the participants. Certain characteristics of the participants, such as education and profession, were not considered. These factors may influence users’ intention to use virtual health agents and should be included and discussed. Incorporating these sociological characteristics would provide a more accurate representation of the participants’ profiles. Additionally, the model proposed in this study could benefit from further refinement. While the construct of “perceived medical narrativity” in this study has five features, we used a scale that evaluates the overall perception of medical narrativity rather than one that separately assesses these five aspects. Future research could address these by measuring each aspect independently, thereby enriching and refining the existing observational tools.