UTAUT 2 provides a more thorough understanding of the factors that influence users' willingness to use new technology than is possible with other technology acceptance models. A systematic review has shown that UTAUT 2 is an efficient theory, with the minimum explained variance of behavioral intention being 35% and the maximum value being 94% (9). Although the UTAUT 2 model was not specifically developed for healthcare, it is perceived to be a robust integrative theory focused on medical providers and users (10). Through induction and summary, we have defined the model variables; the measurement items and model construction of each research variable in the model have also been proposed, as seen in Table 1. The hypothesized research model is depicted in Figure 1.

This is a cross-sectional study investigating hospital users' willingness to use telemedicine services. At present, telemedicine service centers are mainly situated in the most developed cities in China, such as Beijing, Shanghai, and Guangzhou (23). Similar facilities are not available in many less developed cities and regions. This research setting was a tertiary pediatric specialized hospital in Shanghai, a mega-city with the largest population (nearly 25 million) in China. This survey was initiated in March and concluded in May 2020 at Shanghai Children's Hospital, which is the first established specialized children hospital in China. The target population are the end-users of telemedicine system, such as pediatricians and family members of pediatric patients. Family member refers to the parent or guardian of the child patient. Considering the low awareness and utilization rate of telemedicine services, the target group participants filled out the electronic questionnaire by scanning the code on their mobile phones through posted flyers on the nurse stations of Shanghai Children Hospital. The electronic questionnaires were answered anonymously and were collected on the spot. The inclusion criteria for participants included the following: (1) ability to fill out the questionnaire independently, with clear consciousness and no obvious cognitive impairment; (2) willingness to voluntarily participate in this study; and (3) age range of 18–60 years. The exclusion criteria included the following: (1) having a mental disorder that prevented normal communication and (2) refusing to participate in this investigation. Prior to the survey, trained investigators explained telemedicine services to the respondents to assist them in understanding the meaning of the survey questions. A sample of 456 valid questionnaires was eventually collected. This study was approved by the Ethics Review Committee, Shanghai Children's Hospital, Shanghai Jiao Tong University (file number 2021R077-E01).

All of the items were based on the UTAUT 2 model measuring the acceptance of new information technologies, which was adopted from Venkatesh et al. (18) and Gao et al. (22) with necessary validation and wording changes tailored to the telemedicine service and healthcare context. The items are shown in Supplementary Material. The questionnaire was administered in Chinese through a web hosting service after being translated by a professional translator. To ensure that the content did not lose its original meaning, a back-translation was made from the Chinese instrument to English, again by a professional translator, and compared to the original (24). All questions were measured using a 7-point Likert scale ranging from 1 (completely disagree) to 7 (completely agree). A draft set of survey questions was refined by employing cognitive interviews and a pretest. Interviewees (n = 5) who were postgraduates majoring in health informatics or end users of telemedicine services were asked to verbalize the mental process entailed in providing answers. The wording of the questions that were difficult to understand or that generated ambiguity was subsequently modified based on the feedback from 50 respondents in the pretest.

Once the primary collection was completed through a structured electronic questionnaire, the data were coded, cleaned, labeled, and verified with regard to missing values. A two-step approach was employed for structural equation modeling (25). The reliability and validity of the measurement model was examined in the first step, and the structural model was tested in step two. Confirmatory factor analysis was performed to validate the measurement model, and the hypothesized paths were examined using structural equation modeling. The moderating effects of three end-user traits among the hypothesized paths within the core research model were tested using logistic regression analysis and multiple group analysis. Additionally, independent t-tests were conducted to test differences in perception and intention to use between physicians and family members of patients. According to the degree of consistency between the theoretical model and the actual data, the theoretical model is evaluated to achieve the goals of quantitative research on actual problems. Structural equation modeling overcomes the shortcomings of multiple regression analysis method. It not only explains the relationship between variables but also allows the existence of measurement error of the variables. It can realize the estimation of factor structure and relationship as well as the simultaneous estimation of the degree of model fitting. All analyses were conducted using SPSS Statistics, version 24.0 (IBM Corp), and SPSS Amos, version 24.0 (IBM Corp) software.

In this study, a total of 456 respondents completed the survey, including physicians (53.5%) and family members of patients (46.5%). As shown in Table 2, 18% of the respondents were men, and 82% were women. The largest age group was 30–40 years (42.3%). The group with the highest level of education had bachelor's degrees (51.3%).

The 23-item scale appeared to be internally consistent. The Cronbach α for the 7 subscales ranged from 0.890 to 0.957, indicating that the measurement scale had good reliability (see Supplementary Table 1). Convergent validity was adequate when the factor loading and the composite reliability (CR) were above the recommended threshold of 0.7 and the AVE was greater than 0.5 (26) (see Supplementary Table 2). Discriminant validity was also confirmed when the square root of the AVE for each construct was larger than the corresponding inter-construct correlations (27). Hence, this measurement model achieved acceptable levels of reliability and validity. Additionally, confirmatory factor analysis indicated a good fit, as the fit indices for the measurement model exceeded the critical level of 0.80, and the chi-square/degree of freedom equals to 2.714, which was below the suggested value of 3.0 (28). The indicators were as follows: goodness-of-fit index (GFI) = 0.904, normed fit index (NFI) = 0.957, comparative fit index (CFI) = 0.972, root mean square residual (RMR) = 0.062, root mean square error of approximation (RMSEA) = 0.061. It can be seen that the variables had good discriminant validity. According to the degree of consistency between the theoretical model and the actual data, the theoretical model was evaluated to achieve the goals of quantitative research on actual problems. In sum, our results indicated the appropriateness of the measurement model.

A two-independent samples t-test was carried out focused on physicians and family members of patients in relation to the perception of influencing factors on the acceptance of telemedicine services. Overall, there was a positive perception of telemedicine services in a Chinese pediatric hospital among the participants with all mean scores larger than 5. The findings indicate that family member of pediatric patients reported a more positive and optimistic perception toward telemedicine services vs. pediatricians in all subscales of measurement. Table 3 reported on the means and standard deviations of perception of telemedicine services. The mean scores of facilitating conditions (t = −2.19, p = 0.03), hedonic motivation (t = −2.65, p < 0.01), price value (t = −3.26, p < 0.01) and intention to use telemedicine (t = −2.99, p < 0.01) were significantly higher on the demand-side than on the provider-side. Respondents were the most positive about the performance expectancy (mean = 6.09, SD = 1.08). However, respondents reported the lowest mean score (mean = 5.24, SD = 1.48) on the hedonic motivation.

In this study, model verification of the parameters of the initial hypothetical model was carried out to analyse the relationship between the variables and the mechanisms of influence (see Table 4). A P value <0.05 was considered statistically significant. To explain the variance of the constructs, the R² values were examined. With an R² value of 0.737, our model explains 73.7% of the variance in behavioral intention determined by three variables: facilitating conditions (H4: β = 0.280, p = 0.008), hedonic motivation (H5: β = 0.131, p = 0.040) and price value (H6: β = 0.302, p = 0.016). Given the significance of the model path coefficient (β), H4, H5, and H6 are accepted, while the other variables of the model, such as performance expectancy (H1), effort expectancy (H2), and social influence (H3), are rejected. Table 6 clearly shows that facilitating conditions, hedonic motivation, and price value are the three main factors affecting pediatric hospital users' behavioral intention to adopt telemedicine services. They all had a positive impact on behavioral intention, and among the three factors, price value seems to be the most powerful influencing factor.

First, we conducted a separate test by establishing a logistic regression model to measure the relationship between gender, age, education level, experience, and willingness to use telemedicine service (Table 5). Participants with no prior experience and lower education may be more willing to adopt telemedicine service. Second, we performed an individual estimation for age (younger or older), gender (male or female) and experience (none or experienced), education level (lower or higher) and then conducted a multi-group analysis to determine whether the moderating effects of hypothesized paths were different between sub-groups. The chi-square differences between the unconstrained model and constrained model are presented in Table 6. In the unconstrained model, all paths were unconstrained between the two sub-groups. In the constrained model, each path was constrained as equal and was hypothesized to be moderated across two sub-groups. We found that the path between PE and BI was significantly stronger for males than for females (Δχ² = 3.82, p = 0.05) and for older than for younger individuals (Δχ² = 8.63, p < 0.001). The moderating effect of age and gender on the relationship between FC and BI was not significant, but the results for individuals who were younger (β = 0.35, p = 0.01) and female (β = 0.30, p = 0.02) showed a statistically stronger relationship between FC and BI. The moderating effect of experience on the relationship between FC and BI (Δχ² = 6.02, p = 0.01) was significant, and the path coefficient for not experienced participants (β = 0.45, p < 0.001) was larger than that for those who had already used telemedicine services (β = −0.03, p = 0.83). In the age results, although the path from PV to BI was stronger for the younger group, the standardized path coefficient was not statistically significant. In the gender results, the path from EE and HM to BI was stronger for the female group, and the standardized path coefficient was statistically significant. In addition, the multi-group analysis indicated that the difference in chi-square values was statistically significant between EE and BI (Δχ² = 4.67, p = 0.03) but not between FC, HM and BI. With respect to the new moderator, the moderating effect of education level on the relationship between PE (Δχ² = 5.24, p=0.02), EE (Δχ² = 6.27, p = 0.01), and HM (Δχ² = 5.36, p = 0.02) toward adopting telemedcine service was significant and participants with lower education showed a stronger relationship.

In our study, price value is the most important influential factor, particularly for younger users, which aligns with previous research on the same topics (32, 33). However, some studies based on the UTAUT 2 model did not include PV as a predictor for the acceptance of a new technology (22, 34). Meanwhile, it is noteworthy that the path coefficient between PV and behavioral intention had some differences among countries. Comparing the path coefficients of 0.130 in the United States (33), 0.147 in France (12), 0.320 in Iran (32), and 0.302 in China, one may conclude that price value may have a greater impact in less developed countries. Payer reimbursement was the leading influencer of anticipated future use of telemedicine (35). To date, telemedicine services have not been covered by medical insurance payments in China, and the prices vary from province to province. However, insurers in the United States have expanded their coverage and reimbursement of various types of telemedicine services. Scholars believe that if the price of certain technological services is very low or free of charge, it probably will not have a strong influence on behavioral intention (4, 36). Physicians and patients are more likely to take advantage of telemedicine if their perception of its value is higher, such as saving them money and time by avoiding an out-of-town trip to the hospital (11). Telemedicine charges and reimbursement standards should be constructed to meet the actual local needs, and the labor value of telemedicine service providers should be reasonably compensated, including material and spiritual, monetary and non-monetary compensation incentive strategies.

Facilitating conditions were also a significant predictor of the intention to use telemedicine in this study, which was consistent with previous studies (37). Training can not only increase the confidence of telemedicine users but also strengthen collaboration between patients and physicians. In terms of external conditions, the infrastructure, system interface, image quality, network signal and transmission speed of telemedicine systems all need to be ready and periodically maintained, especially in rural areas. In terms of internal conditions, the service process should be optimized to shorten the waiting time. The content of cooperation, service processes, rights and obligations, risks and responsibilities between medical institutions should be determined in a formal agreement before the launch of telemedicine services. Considering the moderating effects of age, gender and experience, female and young users without experience had more positive behavior intention of adopting telemedicine services. Male and older experienced users need more organizational and technical infrastructure supports for the implementation of telemedicine services.

Hedonic motivation also had a positive impact on behavioral intention (38). The adoption of a product based on hedonic motivation is not long term. Once adopters acquire experience with it, its effectiveness outweighs all its other attributes (18). Telemedicine has just taken off in China in recent years. Early adopters pay more attention to its convenience but easily overlook the importance of medical quality. However, telemedicine cannot be merely considered a pleasure-oriented service but is perceived as more of a utilitarian solution (18). Hospitals should provide telemedicine services on the premise of ensuring medical safety and being closely connected with offline medical treatment.

From our results, pediatricians and family members of patients were consistently agreed performance expectancy and effort expectancy as important factors of adopting telemedicine service. Even though these positive effects on behavior intention have not been validated in this study. This may be because users have not yet changed their traditional perceptions and are not fully aware of the usefulness of telemedicine services. Pediatric patients have difficulty operating the system and communicating with physicians due to their limited understanding. In a tertiary hospital in China, clinicians are often constrained by their busy work schedules and therefore have limited energy for learning how to work with new technologies. To solve this problem, hospitals could equip technical assistants responsible for communication and system operation. Individuals could also be impacted by people who are important to them when adopting a new technological product (39). We speculate that healthcare and medical treatment for minors is a very personal and private issue; therefore, SI had limited impact in our study, and previous studies found the same results (40).

In addition, certain attributes may influence users' decisions. In healthcare studies, gender and age were found to be moderators of acceptance (41, 42). Regarding experience, users and non-users showed significant differences regarding PE, EE, SI and FC in the acceptance of mobile health monitoring services (1, 43). The results of moderating effects of age, gender and experience deepen our understanding of underlying differences in telemedicine services adoption behavior. We found that gender moderates the relationships of PE→ BI and EE→ BI. This finding is similar to what Duyck et al. (44) suggested: PE is a stronger factor for males than for females. The relation between EE and BI is stronger for women, which is consistent with previous research (15). Participants with no prior experience and lower education with telemedicine may be more willing to try this new type of technology. The results suggest that hospital administrators should focus more on increasing publicity to enhance user awareness of telemedicine, strengthening homogeneous management of online and offline medical quality, so that acquiring positive feedback from respondents who have used it.

The study has some limitations that can be regarded as opportunities for future research. First, this study recruited a cross-sectional sample of participants in Shanghai; it may not universally reflect the willingness of all users to adopt telemedicine services in Chinese pediatric hospitals, especially in rural areas. Further research in different regions will provide more accurate evidence if the results depend on socioeconomic factors. Second, telemedicine involves the privacy protection and information security of personal health data, and future research should incorporate perceived risk and trust factors to further extend the theoretical model. Third, females were over-represented in the respondent groups, and respondents younger than 40 years old accounting for 70% in the context of pediatric hospital. Further studies should collect data following the gender and age ratios of the general population.