As telehealth technologies enable virtual and timely communication among care providers, frontline care providers particularly face challenges in enhancing service performance while using such technologies. In the primary care setting, care provider groups such as doctors, nurse practitioners, and nurses have been the first point of contact for people who seek health care services [1] within close proximity of patients in the location [2]. As the use of telehealth technology in electronic consultation (e-consultation) has expanded care providers’ role to managing some specialty care work [3,4] beyond locational boundaries [5], it has become visible how they broker specialty visits between primary and specialty care by using telehealth technology [6]. Telehealth technology is thus supposed to enhance frontline care providers’ virtual communication electronically. For example, care providers speak to patients via scheduled or on-demand/urgent visits; in addition, they communicate with specialists to ask questions and help patients avoid further face-to-face consultation with specialists [7]. Our main focus in this paper is on the latter case, often termed telespecialty consultation or e-consultation [8], which is the interaction between frontline care providers and specialists. In this environment, a patient typically does not see a specialist and relies solely on care providers' intervention to gain access to specialists, and thus, care providers’ need to manage each patient’s case in a timely manner while communicating with specialists. Thus, care providers are challenged to act as care moderators of the relationships between patients and specialists to manage expanded care responsibilities and improve service performance with the use of telehealth technology.

Such care providers’ moderating role requires them to adequately select and use telehealth technologies for successful virtual care outcome. Telehealth technology does not refer to a single technology artifact, but to a number of electronic information and communication technologies (ICTs) to facilitate long-distance clinical care, patient and professional health-related education, and public health administration [9]. Accordingly, prior literature has noted that care providers have increasingly used multiple technologies to manage not only the new form of health care, but also virtual communication simultaneously [10-12]. However, such use of multiple technologies for virtual services has led to mixed results. Care providers perceived the use of a single telehealth technology to be beneficial to the timely management of referrals to specialists, but at the same time, they felt burdened by the additional workload that had shifted from specialists [13]. In such processes, care providers’ increased use of other relevant technologies may create frustration and avoid adoption of new technology when certain technological features with which they are familiar do not perform as intended [14] or supplement the role of care providers [5]. More specifically, in the e-consultation context, care providers can feel constrained in sending messages to specialists if this familiar use of messaging technology is not integrated with other services or health systems [15]. Thus, it is fair to say that care providers’ prior use of various technologies and features and their expectation of telehealth technology influence their telehealth use behavior.

However, little attention has been devoted to understanding the telehealth-driven provider-to-provider communication in which individual providers’ technology use behavior as care moderators can influence virtual service performance. Several prior studies have mainly focused on the antecedents of virtual service performance [16], and yet, the mechanisms that influence care moderators’ perception on the use of telehealth technology and service performance in the process of e-consultation are unknown. Given that telehealth technology shares similarities with other health technologies on various devices such as smartphones, tablets, and desktop/laptop computers [17], care providers’ pre-existing experience and self-confidence in dealing with similar or new features from other technologies may not only affect their attitude about using telehealth technology [18], but also the way in which they use the telehealth technology for virtual communication [19]. In other words, a care provider may select and use a set of related telehealth technologies to manage online communication with specialists, which may allow them to enjoy familiar system features or may alter the intended capabilities of the telehealth technology artifacts (referred to as “adaptive use of IT” [20]).

Two aspects distinguish this study from prior studies. First, this study explicitly focuses on care providers’ postadoption behavior using a set of telehealth technologies. Thus far, extensive research has been performed at the intersection of human computer interaction (HCI), health informatics, design science, and information technology (IT) adoption strands, with the main focus on the cognitive/psychological aspect of technology use [21,22] and interactions with technology artifacts [23,24]. For example, prior research noted mechanisms through which individual care providers’ intrinsic and extrinsic motivation, gratification, and use environment influence adoption intention [25]. Additionally, exploration of how humans interact with social, organizational, and contextual environments has developed theoretical foundations to capture users’ technology adoption intention across nonhealth domains such as business, marketing, education, engineering, and agriculture (eg, [26]). This paper, however, examines an unexplored area of technology use—telehealth technologies and users’ level of flexibility—to mix, match, and use them for the success of telehealth care communication. This adds value to telehealth management and relevant research, given that the current technology market for telehealth is led by care providers, tech firms, and payers, and telehealth technologies may not share common features or include all necessary features for various care regimen [27]. For example, remote patient monitoring, as one of the important aspects of telehealth service, needs multiple technologies such as videoconferencing software, peripheral devices, telemedicine carts (filed kits) for the patient site, and remote patient monitoring kits [28]. In addition, care providers use audio and video technologies for live patient care [10]. In such a telehealth care environment, care providers play crucial roles in not only mediating as the first virtual contact for patients but also cocreating care plans with specialists. Their performance can thus affect the outcome of telehealth care services, which, in turn, influence telehealth technology adoption and use of other stakeholders in rural or medically underserved areas [29]. As a consequence, comprehending users’ (ie, care providers’) experiences with related technologies is essential to understand their telehealth technology behavior.

Second, this paper calls for an explicit focus on virtual communication between care providers and specialists. As telehealth service consists of multilateral communications among care stakeholders, primary care providers should moderate the encounters between those parties in case of virtual specialty care needed and enhance timely and quality of telehealth care. At the time of innovation, primary care providers are known to adapt their daily routines of care management to technological innovation, which leads to decreased productivity [30]. However, little is known about care providers’ ability to adapt telehealth technology for virtual communication and team-based care services [31]. From the perspective of frontline care providers who manage care processes in the location and connect to specialists remotely, the manner in which they evaluate the use of telehealth technologies and perform in this new care format can determine the success of telehealth care services in the long run. Figure 1 summarizes the unique role of frontline care providers in a telehealth care process and is the focus of this study. In this study, we limit our focus on variation in the provider-to-provider communication while making patient-side inputs fixed to better isolate care providers’ technology adaptation behavior.

Taken together, our aims are to examine whether care providers’ adaptive technology use behavior improves virtual service performance in e-consultation and to explore whether individual characteristics in relation to technology (personal innovativeness, computer self-efficacy, and habit) influence their adaptive use behavior. To this end, we examined two adaptive use behaviors: exploitative and exploratory use of telehealth technology. Exploitative technology use refers to the use of telehealth technology under the existing norms, while exploratory technology use involves the use of IT in a novel or unprecedented way. We hypothesized that these two adaptive IT behaviors enhance task performance on the part of care providers in the process of relaying patient information and managing specialists’ diagnoses and that personal traits may affect such adaptive IT use behaviors.

We recruited graduate nursing students at a university in the Southeastern United States who were studying to become family nurse practitioners and had experienced various types of technologies across different contexts. In telehealth care contexts, in particular, the demand of nurse practitioners is growing because they can collaborate with other specialists to comanage patient care cases for rural or disadvantaged population [42]. With incorporation of such trends and our research focus on nurse practitioners’ role as frontline care providers in e-consultations with specialists, we had to use purposive sampling for those who could provide unique information about telehealth care services that could not be obtained elsewhere [79]. The participants had been trained in the use of telehealth technologies in their education program and their communication with other care members in the simulation rooms, in which the nursing faculty had recorded and evaluated their electronic communication. In summary, our student sample could be legitimately used to capture their adaptive use of telehealth technologies. This nonprobability sampling met our goals to identify a representative sample of nurse practitioners who would be at the front line of telehealth care services as care providers (see [78] for the taxonomies of purposive sampling). To test our hypotheses, a Web-based survey was administered in the spring and fall semesters in 2018. In the online survey, each participant read a case scenario and responded to the questionnaire in terms of their perception on prior experience, adaptive use of telehealth technology, and care service performance (Multimedia Appendix 1).

A cross-sectional online survey was sent to 210 students who were enrolled in the nurse practitioner education program. A total of 146 participants responded to the survey, yielding a response rate of 69.52%; of these, 121 valid responses were used for data analysis. Twenty-five responses were dropped because they only reported demographic information in the surveys. Table 1 summarizes the respondent profiles. The majority were females (72.7%) and employed (full-time workers=54.5%; part-time workers=35.5%). All were well educated, with an undergraduate or higher degree, and nearly half were of white race (47.5%). Table 1 summarizes the characteristics of the survey participants.

In addition, an adequate sample size is necessary to estimate the PLS path model, which is guided by the 10-times rule and power analysis. On one hand, the 10 times rule dictates the use of “10 times the largest number of formative indicators used to measure a single construct, or 10 times the largest number of structural paths directed at a particular construct in the structural model” [83]. On the other hand, power analysis provides a threshold for the statistical power necessary to detect an effect based on the maximum number of independent variables in the measurement and structural models. In our case, we had five independent variables, three maximum arrows pointing to a latent construct, and thus needed at least 45 observations to achieve a statistical power of 80% and R² values of at least 0.25 (with a 5% probability of error) [83]. Thus, our sample size (N=121) was deemed adequate to test the research model.

As our survey was self-reported, we evaluated two possible biases carefully: nonresponse bias and common method bias. Nonresponse bias derives from the differences between participants and nonparticipants in the survey [84,85]. This bias can be assessed by comparing our sample’s characteristics with those in the population and by comparing early and late respondents. We compared the early respondents (74.82%) and late respondents (25.18%) on each of the demographic characteristics (age, gender, education, income, and occupational status) and health application experience using a t test. There were no significant differences between the early and late respondents in our sample.

Common method bias potentially threatened the veracity of our results, as data for the independent variables and dependent variable were collected in the same survey. Following a previous study [86], we designed the survey instrument’s contents and order carefully. Furthermore, we performed the Harman single factor analysis to assess the bias. The results showed that one factor explained 36.90% of the variance, confirming that no single factor accounted for the majority of covariance. Thus, nonresponse bias and common method bias did not threaten this study’s findings.

To determine each construct’s internal reliability, we first examined the item loadings and composite reliabilities. Each item loaded above 0.75 on its respective construct and was significant at P<.05. Cronbach alpha was calculated to assess composite reliability and confirmed that all items’ values were above 0.7. Convergent validity was established if the average variance extracted (AVE) was above the threshold of 0.5, which suggests that the variance explained by indicators was greater than the unexplained variance. Discriminant validity was tested by assessing the Fornell-Larcker criterion and cross-loadings. To confirm the discriminant validity, the AVE of each construct should be greater than its squared correlations with other constructs [87]. As shown in Tables 2 and 3, internal reliability, convergent validity, and discriminant validity were all confirmed.

Following structural model assessment suggestions of de Guinea [69], we evaluated the proposed path model using SmartPLS 3.0. The structural model’s quality was assessed by checking multicollinearity (variance inflation factor), path coefficients, R² (variance explained), f² (effect size), and the Stone-Geisser Q² (model’s predictive relevance). The variance inflation factor was checked and confirmed to be less than 5, indicating that multicollinearity was not a problem in the study. We report the path coefficients’ significance, R² and f² in the full model results. Effect size (f²) explains the changes when an exogenous construct of focus is included and when it is omitted from the model. As a rule of thumb, if f² is 0.02, 0.15, and 0.35, the effects are considered to be small, medium, and large, respectively [88]. Lastly, to assess the model’s predictive power, the Stone-Geisser Q² was used to indicate the sample’s predictive relevance. A Q² value>0 demonstrates that the path model has predictive relevance to a reflective, endogenous latent variable. The Q² values for three of the endogenous constructs were >0, indicating exploitative use (Q²=0.51), exploratory use (Q²=0.36), and virtual service performance (Q²=0.54). Thus, the model’s predictive relevance was confirmed.

For the path coefficients’ significance and the variance explained, R², our results demonstrated that exploitative technology use was positively associated with virtual service performance (β=0.76, P<.001), while exploratory use did not explain the variation in virtual service performance (β=0.036, P=.49). With regard to the effects of individual characteristics, computer self-efficacy was significantly associated with exploitative technology use (β=0.31, P=.05). Lastly, habit was associated with both exploitative (β=0.29, P=.04) and exploratory use (β=0.41, P=.006). Therefore, hypotheses 1, 4a, 5a, and 5b were supported but hypotheses 2, 3a, 3b, 4b, and 5b were not supported. Among the control variables, education (β=–0.12, P=.01) was shown to affect virtual service performance negatively, while income level (β=0.11, P=.008) was positively associated with virtual service performance, as shown in Table 4 and Figure 3.

The purpose of this study was to contextualize and test a research model that examined the determinants of the adaptive use of telehealth technologies in e-consultation. Using AST at the individual level, we examined the mechanisms by which exploitative and exploratory use of telehealth technologies and three personal traits (ie, personal innovativeness in IT, computer self-efficacy, and habit) influence virtual service performance.

Our results indicate that care providers are willing to use telehealth technologies exploitatively to communicate with specialists in e-consultations and perceive it as a virtual service performance enhancer. With regard to the hypotheses on adaptive use and virtual service performance (hypotheses 1 and 2), exploitative use of telehealth technologies was found to be a strong factor that explains care providers’ virtual service performance (β=0.762, P<.001), while exploratory use was not (β=0.036, P=.49). Moreover, with regard to hypotheses 3 through 5, we found that personal innovativeness in IT is insignificant for explaining adaptive use in our context (P=.13 for exploitative use and P=.27 for exploratory use, respectively); computer self-efficacy has a significant, positive effect on the exploitative use of telehealth technologies (β=0.311, P=.047); and habitual use of nonhealth technologies in daily life is associated with care providers’ willingness to engage in both exploitative (β=0.293, P=.008) and exploratory adaptive use of new telehealth technology (β=0.414, P=.006).

Despite the meaningful and practical findings in this study, our results should be interpreted with caution due to some limitations. First, we selected a purposive sample of graduate nursing students for our study. Although this is a legitimate sample, the ability to generalize the findings may be limited because our participants have experienced e-consultation in the education-focused, simulated contexts. Second, while our cross-sectional sample shows adequate responses to estimate our hypothesized path model using PLS-SEM, a larger sample with a panel structure would increase the statistical power of the findings and control for any unobserved confounding factor using panel data analysis. Particularly, time-series cross-sectional data on capturing various technology use among care providers would provide more in-depth understanding of virtual telehealth care services. Third, we did not include task adaptation behavior based on our research context’s characteristics. The education program requires individuals to adhere to the standard protocol of patient cases, such that any adaptive task behaviors are evaluated as a failure in the medical setting. Thus, we limited our focus to the participants’ adaptive use of telehealth technology only. It would be meaningful to investigate task and technology adaptation behaviors in the health care contexts in which the selection of tasks and technologies are flexible in future research. Fourth, this research considered care providers’ perception about their willingness to use telehealth technology either in a traditional way or in an innovative way and its downstream effect on perceived service performance. This is because we viewed telehealth technology as a set of related information and communication technologies and asked participants’ adaptation behavior in a general context. It would be worthwhile to revisit our research model to certain care contexts (eg, diabetes) and capture actual measures of virtual service performance. Lastly, future research can be further extended to explore other stakeholders’ attitude and behaviors about telehealth technology in e-consultation. For example, our research model can explain how adaptive use of telehealth technology by frontline care providers influences the level of patients’ satisfaction and health outcomes as well as those of specialists.

This study contributes to both the IS and health care literature on the postadoptive use of telehealth technology. First, this research contributes to the IS literature by exploring adaptive IT use behavior from an individual meeting facilitator’s perspective and attempting to identify a contextualized theory in e-consultation. Specifically, we considered e-consultations between care providers and specialists in a virtual meeting context and proposed that as virtual meeting facilitators, care providers’ willingness to use telehealth technologies is an important predictor of virtual service tasks’ success. Previous IS literature on online meeting technology use has documented that technology types, environmental factors, and user characteristics are key factors that predict outcomes from a multilevel perspective. From a group perspective, technical support for group users and the fit between tasks and technologies are important determinants of the success of online meeting technology use [89]. In terms of individuals’ behavior, extant studies have emphasized the important role that human facilitators’ characteristics play in predicting the outcomes of online meetings [43]. Given care providers’ unique position as those who relay clinical information and decisions between patients and specialists in health care and the availability of flexible options to select and use telehealth technologies for provider-to-provider communication, there is an urgent need to examine whether, and in what way, such meeting facilitators’ adaptive use of IT predicts the success of virtual service performance.

Our study is unique in that we focused explicitly on care providers’ virtual role in telehealth communication and their performance under a new definition of telehealth technology (ie, use of telehealth and telehealth-related technologies). Care providers’ role differs from that of general online facilitators in online meetings, in that they manage multilateral communication between patients and specialists and although there is a designated telehealth platform for their communications, sometimes, communication with the two different groups involves the use of additional technologies that complement or substitute the existing technology’s capabilities. Given that providers [58] and patients [90] were the main user groups of interest in previous studies for predicting telehealth technologies’ success, this study contributes to the health care literature by examining the adaptive structuration theory of individual from a care provider’s perspective and identifying salient constructs (the exploitative use of IT) in a research model to develop a contextualized theory. Second, the result that habit plays a significant role in adaptive IT use calls for more attention in IS research on how users’ habits accumulate from different life domains to affect their postadoptive use of multiple technologies [91]. Previous postadoption studies have established that habit reflects the extent to which people tend to perform behaviors automatically because of past learning [69,92-95]. However, the relationship between habit and the continued use of IT has been mostly tested in a single domain; for example, mobile phone habits predict mobile phone use [93] and mobile internet habits affect mobile internet technology use [95]. In a health care setting, however, our findings emphasize that cross-domain, habitual IT use influences care providers’ adaptive technology behaviors. Our interviews with two nurse practitioners in family medicine also reflected these positive effects of the habitual use of technology in nonwork domains. One nurse practitioner stated,

Another said:

As telehealth medical services include three different communication modalities (video-conferencing, texting, and a hybrid of the two), care providers’ existing technology habits can help them select and use these three forms of communication and manage communication with patients and other providers.

Moreover, we found a strong effect of exploitative use (expected use of telehealth technology) on virtual service performance by frontline care providers. These findings are in line with those of a previous study [20], such that two modes of IT adaptation behaviors (exploitative and exploratory use) are differentially salient across research contexts. In other words, exploitative and exploratory adaptation behaviors may not coincide under the same context because users have different coping strategies toward information technologies [96]. For example, explorative use of technology became salient in context of mobile technology in the BYOD (bring your own device) context [20]; in the contexts of enterprise resource planning and product lifecycle management system use, exploratory as well as exploitative adaptation were differentially significant, contingent upon input factors [36]. In our research context, the strong effect of exploitative use of telehealth technology may be due to the contextual characteristics of care process. Actually, health care is a controlled and highly concentrated environment such that care providers expect the technologies to function as expected by supplementing their clinical tasks [5]. Since this study explored technology-related traits as antecedents to technology adaptation behavior in provider-to-provider communication, it is much anticipated that such technology adaptation behavior can vary by different contexts and heterogenous technology users.

Lastly, education and income were shown to differentially influence virtual service performance. Prior studies have documented negative effects of demographic variables on technology use, as less educated participants may have less knowledge, whereas those who earn less income may have less opportunities to access advanced information technology [97,98]. In our research context, we can interpret that the current level of technology education from bachelor’s degree may not be on par with specifics of telehealth technology use for communication with specialists. In addition, our graduate nursing students with high income level may have been exposed to various technologies within and outside the education program or care settings. Thus, demographic characteristics of individual users need to be included when exploring technology adaptation behavior.

This study’s findings can be applied to inform health care practitioners and health app designers. Strategic IT management is necessary for care providers who serve as virtual meeting managers in a telehealth setting [81]. As telehealth medical services have garnered much attention, nurse practitioners have played an increasingly important role in supporting various online care services in which multiple technologies need to be operated appropriately. Our results demonstrate that care providers’ adaptive use of technology can help predict telehealth care performance, and therefore, more consideration should be given to the role of intermediary care providers in the care process between patients and specialists. As organizational structures influence both offline and online meeting outcomes [99], health practitioners need to focus on organization-level strategies to enhance care providers’ online facilitation by examining the gaps that they have experienced using a variety of technologies across multiple life domains and providing relevant education in the use of focal technology [31,100]. Moreover, this study can be beneficial to telehealth designers and developers in terms of the design of HITs. Prior studies have documented that telehealth apps’ design of features, icons, and terminologies is important and that care providers expect all of these to function as intended [14]. As health app developers continue to add new features to stay abreast of rapidly changing health care trends, it is important for them to consider health care consumers’ needs and users’ familiarity and comfort with the existing features that are evolving across a wide range of technologies and systems [101].

This study investigated frontline care providers’ unique role in e-consultation with specialists. By regarding the care providers explicitly as virtual meeting facilitators, we tested the association between their adaptive use of multiple telehealth-related technologies and virtual service performance. Care providers’ standard use of telehealth technologies was shown to be a salient factor that predicts success in virtual service, while the innovative use of telehealth technologies remained insignificant. Among their personal characteristics, the habits and computer self-efficacy that care providers acquired and developed in nonwork settings stimulated and enhanced their willingness to use multiple telehealth technologies in standard and creative ways.