Sage Journals: Discover world-class research

Abstract

Prior to the COVID-19, various issues in ethics, laws, regulations and security risks limited physicians in their ability to deliver medical care via telemedicine. However, in 2020, telemedicine quickly became mainstream when the US government issued a waiver granting the use of telemedicine to reduce COVID-19 transmission. This study examined the US physicians’ acceptance of telemedicine after its sudden wide-scale implementation. The authors developed a survey based on the Unified Theory of Acceptance and Use of Technology to assess the US physicians’ intentions and attitudes toward the use of telemedicine in the delivery of medical care. This survey was distributed between March and June in 2021 across several voluntary online physician groups. The survey was validated using factor analysis. The 110 participants who completed this study were all active physicians practising telemedicine in the USA. The majority of the participants were female (93%), between 34 and 59 years of age (93%), had at least 10 years of medical experience (84%) and practised in a city (92%). The behavioural intention to use telemedicine was influenced by performance expectancy (β = 0.34, p<.001), social influence (β = 0.34, p<.001), facilitating conditions (β = 0.14, p<.05) and attitude towards telemedicine use (β = 0.36, p<.001). Attitude towards telemedicine use was influenced by performance expectancy (β = 0.60, p<.001), effort expectancy (β = 0.16, p<.05) and social influence (β = 0.16, p<.05). In addition, telemedicine experience moderated the influence of performance expectancy on behavioural intention negatively (β =−0.39, p<.001) and moderated the influence of facilitating conditions on behavioural intention positively(β = 0.27, p<.001). This study found that the physicians’ performance expectancy of and attitude towards telemedicine use were significant predictors of physicians’ acceptance. Social influence from colleagues and administrators also played an essential role in impacting physicians’ acceptance of telemedicine. Physicians’ effort expectancy merely affected their attitude towards telemedicine use. Facilitating conditions—such as the inability to perform physical exams, labs, etc.—only influenced the physicians’ intention to use telemedicine. Finally, the survey data suggested that as physicians gained telemedicine experience, their performance expectancy predicted lower behavioural intention than counterparts with less experience.

Keywords

UTAUT telemedicine physician acceptance technology acceptance COVID-19 structural equation modeling (SEM)

Introduction

Telemedicine is defined as the distant delivery of healthcare and exchange of medical information between patients and their healthcare providers via information, communication and networking technologies in order to promote health maintenance, prevention, alleviation and remedy of disease, pain and suffering (Waller & Stotler, 2018). Telemedicine enabled patient interactions with healthcare providers, exchange of medical information and patient remote monitoring from distant locations; the latter proved especially useful in rural and remote environments where medical services were limited (Craig & Petterson, 2005). Telemedicine also increased access to medical care through video consultations with healthcare providers (Waller & Stotler, 2018). Using home-based video interactions, a healthcare provider could perform an observational assessment after obtaining a thorough history of symptoms and exposure risk from the patient (Portnoy et al., 2020). Typical applications of telemedicine included general health care, dermatology, eye exams, nutrition counselling, medical consultations and prescriptions, nursing, teleradiology, psychotherapy, teleneurology and other applications (Nittari et al., 2020). As telemedicine improved access to medical services, increased healthcare delivery, reduced healthcare costs and promoted communication between health professionals and other benefits, it would be imperative to understand how readily physicians would accept telemedicine in medical care (Ekeland et al., 2010; Hjelm, 2005).

Telemedicine had remained relatively uncommon in healthcare delivery prior to wide-scale implementation, in 2020, due to the public health emergency from COVID-19. Its uncommon use was mainly because of prevalent business models, minimal cost benefits, limited financial funding, concerns about data security and compliance and inadequate legislation available for legal and ethical issues (Chen et al., 2013; Effertz et al., 2017, Mehrotra et al., 2021; Nittari et al., 2020; Peters et al., 2015; Sterling & LeRouge, 2019). Nonetheless, telemedicine was recognised as a disruptive innovation that could transform the healthcare industry by potentially reducing costs, and increasing quality of care, as well as patient satisfaction (Schwamm, 2014; Sterling & LeRouge, 2019). Despite these benefits, the healthcare industry often found little or no cost benefits from telemedicine due to inadequate insurance reimbursement. The early business processes around telemedicine were geared towards businesses rather than users due to its low utilisation, extra resources needed for staff training, additional staffing and difficulty in satisfying the requirements of regulatory and compliance agencies, for example, Health Insurance Portability and Accountability Act (HIPAA), 1996, Privacy, Security and Breach Notification Rules (Chen et al., 2013; Nittari et al., 2020; OCR, 2022; Sterling & LeRouge, 2019; Wade et al., 2010). Often, providers and patients did not utilise telemedicine services because health insurance like Medicaid (Centers for Medicare & Medicaid Services, 2020) in the United States did not reimburse telemedicine expenses well (Acheampong & Vimarlund, 2015). The number of telemedicine visits thus remained infrequent prior to 2020, as many healthcare providers refused to carry out telemedicine activities due to legal risks, such as the possibility of malpractice suits (Nittari et al., 2020). By and large, many healthcare providers could not provide telemedicine services unless they practised in rural/remote areas with limited medical access (Campbell et al., 2001) or during emergencies, such as natural disasters, injury hazards (Simmons et al., 2008), exposures to contaminations and disease epidemics (Ohannessian, 2015). As such, ‘face-to-face consultation between physician and patient remained the gold standard of clinical care’ (WMA General Assembly, 2018).

It was not until the global outbreak of the COVID-19 in 2020 that telemedicine became a mainstream method in order to reduce COVID-19 transmission risks (AMA, 2020). Many businesses shifted from face-to-face interactions to real-time audio and video communications, social media and other internet platforms through the information and communications technology (Nicola et al., 2020). Similarly, a drastic shift occurred in medicine, whereby the US legislation required all insurers, including Medicare (Centers for Medicare & Medicaid Services, 2020) and employee health plans, to reimburse medical services delivered through telemedicine in response to the COVID-19 public health emergency (PHI, 2020; Severino & Director, 2020). As a result, the number of virtual visits in the USA surged to more than 12% of all outpatient visits between March and April in 2020 (Mehrotra et al., 2021). Similarly, scholarly literature saw an explosion of papers on telemedicine. Despite all this, the physicians’ acceptance of telemedicine has received limited research interest.

The Unified Theory of Technology Acceptance and Use

To adequately predict how users would accept any technology, Davis’s Technology Acceptance Model (TAM) was the first of its kind that envisioned how perceived usefulness and ease of use could influence one’s intention to and subsequent use of the system (Davis, 1989). In order to predict users’ intention and resultant use of the technology, parsimonious explanations of end-user reactions were necessary. Further studies and expansions of the TAM resulted in two major upgrades. First, TAM was extended into TAM2 by including social factors that could influence users’ intentions and usages (Venkatesh et al., 2003). Second, through empirical validations and comparisons, Venkatesh et al. (2003) formulated the Unified Theory of Acceptance and Use of Technology (UTAUT) by combining several prominent models, along with TAM and TAM2. This yielded several core determinants of users’ intentions and usages, along with several moderators of key relationships (Venkatesh et al., 2003, Venkatesh et al., 2012). Ultimately, the goals of assessing technology acceptance were to promote successful implementations and continuous improvements (Berg, 2001).

The UTAUT was built on the following theoretical constructs (Fishbein & Ajzen, 1975):

Attitude (affect: feelings and evaluations) was referred to the ‘person’s feelings towards and evaluation of some object, person, issue or event’ (Fishbein & Ajzen, 1975). In particular, attitude towards technology use was defined as ‘an individual’s overall affective reaction to using a system’ (Venkatesh et al., 2003).

Belief (cognition: opinions and beliefs) was defined as ‘the person’s favourable or unfavourable evaluation of an object’ (Fishbein & Ajzen, 1975). A belief was further distinguished into three groups: ‘Behavioural belief’ was related to believing the consequence of a specific behaviour, ‘normative belief’ was a belief subjected to what others believe in and ‘control belief’ involved specific factors that would enhance or degrade behavioural performance (Ajzen, 1991). In UTAUT, performance expectancy (PE), effort expectancy (EE), social influence (SI) and facilitating conditions (FC) were the constructs for beliefs, which would later be expanded in this section.

Behavioural intention (conation), a special case of beliefs, was referred to the individual’s intention to perform a given behaviour, indicating the willingness and planning to exert the necessary effort for a specific behaviour (Ajzen, 1991).

Behaviours (observed overt acts) were defined as the ‘observable acts of the subject’ (Fishbein & Ajzen, 1975). For instance, the output variable in UTAUT would be the use behaviour (Venkatesh et al., 2003).

The theory maintains that positive attitudes and beliefs, high use and positive behavioural intention (BI) led to accepting the technology favourably (Davis, 1989). The two constructs that defined the use of and the intention to use the system were the use behaviour and BI, respectively. Moreover, to explain and predict these behaviours, UTAUT included the four determinants—PE, EE, SI and FC—that would influence user intentions and usage (Venkatesh et al., 2003). In addition, several studies found that the attitude towards the use of telemedicine (ATT) played a central role in influencing user acceptance of technologies (Dwivedi et al., 2019). Finally, UTAUT suggests that the four moderators with specific user characteristics—age, gender, experience and voluntariness of use—would alter the determinants’ effects on ATT, BI and use behaviour (Dwivedi et al., 2019; Venkatesh et al., 2003).

The Research Constructs

This research employed the UTAUT to examine the physicians’ acceptance and use of telemedicine. This section presents how the authors developed the research constructs in accordance with the UTAUT literature. To ensure each construct developed met the original definition, the authors developed three to four survey items from the original theory (also named ‘root scales’), so they were more relevant to physicians’ acceptance of telemedicine during the pandemic.

‘Use behaviour’ was an objective measure that indicated the extent of a physician’s use of telemedicine. Typically, use behaviours were measured and recorded via methods such as collecting system logs (Venkatesh et al., 2011). Unfortunately, the present study did not involve collecting and recording any observable data for use behaviours.

‘BI’ was a subjective measure of a physician’s intention to employ telemedicine in a clinical setting. As outlined in Table 1, physicians’ plans, predictions and desires to use telemedicine even after the end of the COVID-19 pandemic were shown to be true predictors of their intentions to use telemedicine. According to Venkatesh et al. (2003), users’ BI to use technology rather than their actual use behaviour would strongly suggest their acceptance of technology. Therefore, it would be sufficient to utilise BI to predict physicians’ acceptance of telemedicine.

Table 1.

BI to Use the System: Study Items with Root Items (Scales) Regarding Physicians’ Plans, Predictions and Desires to Use Telemedicine Even After the COVID-19 Pandemic was Over.

Study Items Root Scales (Venkatesh et al., 2003)
BI1.If I have a choice, I will continue to use telemedicine after the pandemic is over (i.e., when the risk of contracting COVID-19 is very low).	‘I plan to use the system in the next ⟨n⟩ months’.
BI2. I predict that I will be using telemedicine on a regular basis after the pandemic is over.	‘I predict I would use the system in the next ⟨n⟩ months’.
BI3. I intend to continue using telemedicine for communicating with my patients and others in the context of my practice.	‘I intend to use the system in the next ⟨n⟩ months’.

ATT was defined as ‘an individual’s affective reaction to using a system’ (Venkatesh et al., 2003). As shown in Table 2, ATT was derived from the four root UTAUT constructs. The root constructs were described as follows. First, attitude towards behaviour would reflect positive feelings healthcare providers had towards telemedicine use. Second, effect towards use would represent the extent of joy, elation or pleasure healthcare providers experienced when they conducted telemedicine visits. Third, intrinsic motivation would reflect how much healthcare providers were motivated to use telemedicine for the sole purpose of the activity (Venkatesh et al., 2003).

Table 2.

ATT: Study Items with Root Items (Scales) and Root Constructs About Whether Physicians had Positive ATT.

Study Items Root Scales (Venkatesh et al., 2003) Constructs
ATT1. I find that using telemedicine is a goodidea.	‘Using the system is a bad/good idea’.	Attitudetowards behaviour
ATT2. I find that using telemedicine can help me connect with my patients better.	‘The system makes work more interesting’.	Affect towards use
ATT3. My patients have good experiences during our telemedicine sessions.	‘The actual process of using the system is pleasant’.	Intrinsic motivation
ATT4. We enjoy using telemedicine.	‘I find using the system to be enjoyable’.	Intrinsic motivation

‘PE’ was defined as ‘the degree to which an individual believed that using the system would help him or her to attain gains in job performance’ (Venkatesh et al., 2003). As shown in Table 3, the PE items are based on two root UTAUT constructs,. Both perceived usefulness and job fit were related to how much healthcare providers believed telemedicine would enhance their clinical performances (Venkatesh et al., 2003).

Table 3.

PE: Study Items with Root Items (Scales) and Root Constructs About Physicians’ Beliefs in Getting More Work Done Quickly and Simplifying Their Workflows When Using Telemedicine.

Study Items Root Scales (Venkatesh et al., 2003) Constructs
PE1. I get more work done when I use telemedicine.	‘Using the system in my job would enable me to accomplish tasks more quickly’.	Perceived usefulness
PE2. I find that using telemedicine has simplified my workflow.	‘Using the system would make it easier to do my job’.	Perceived usefulness
PE3. I save time when I use telemedicine.	‘Use of the system can decrease the time needed for my important job responsibilities’.	Job fit

‘EE’ was defined as ‘the degree of ease associated with the use of the system’ (Venkatesh et al., 2003). As illustrated in Table 4, this study incorporated EE from the following two root constructs under UTAUT. First, as per (Venkatesh et al., 2003), perceived ease of use would show how much healthcare providers believed that the telemedicine system was “free of effort.” Second, ease of use would reflect healthcare providers’ perceptions about how easy it was to use telemedicine (Venkatesh et al., 2003).

Table 4.

EE: Study Items with Root Items (Scales) and Root Constructs About Physicians Believing that Telemedicine, with a Clear and Understandable User Interface, was Easy to Learn, Able to Perform the Right Job Easily and Flexible to Interact With.

Study Items Root Scales (Venkatesh et al., 2003) Constructs
EE1. The user interface of my telemedicine platform is clear and understandable.	‘My interaction with the system would be clear and understandable’.	Perceived ease of use
EE2. I find my telemedicine platform easy to learn.	‘Learning to operate the system is easy for me’.	Ease of use
EE3. I find it easy to get my telemedicine platform do what I want it to do.	‘I believe that it is easy to get the system to do what I want it to do’.	Ease of use
EE4. I find my telemedicine platform flexible to interact with.	‘I would find the system to be flexible to interact with’.	Perceived ease of use

‘SI’ was defined as ‘the degree to which an individual perceived that important others believed he or she should use the new system’ (Venkatesh et al., 2003). As shown in Table 5, the current study incorporated SI from the following two root constructs. First, the subjective norm would represent how much healthcare providers believed their decision to use telemedicine was contingent upon the extent to which people important to them think telemedicine should be used. Second, social factors would reflect how much healthcare providers believed the decision to use telemedicine was contingent on the organisational culture and various agreements made in social situations (Venkatesh et al., 2003).

Table 5.

SI: Study Items with Root Items (Scales) and Root Constructs About Physicians’ Decisions to Use Telemedicine Due to Important People, Such as Colleagues and Administrators, Who Encouraged or Demanded the Use of Telemedicine.

Study Items Root Scales (Venkatesh et al., 2003) Constructs
SI1. My colleagues and co-workers have encouraged and/or recommended me to use telemedicine.	‘People who are important to me think that I should use the system’.	Subjective norm
SI2. My administrators and/or my supervisors have directed me to use telemedicine more frequently during the pandemic.	‘My supervisor is very supportive of the use of the system for my job’.	Social factors
SI3. I use telemedicine because everyone around me uses it.	‘I use the system because of the proportion of co-workers who use the system’.	Social factors

‘FC’ was defined as ‘the degree to which an individual believed that an organisational and technical infrastructure existed to support the use of the system’ (Venkatesh et al., 2003). Some types of telemedicine were more easily arranged at patients’ homes. For instance, consultative visits would require video-conferencing equipment (e.g., webcams, etc.) to communicate between providers and patients, whereas remote monitoring would focus on taking periodic biomedical measurements by remotely using low-cost equipment, such as blood pressure cuffs, thermometers and digital scales (Baker & Stanley, 2018). On the contrary, facilitated virtual visits would require specialised equipment, such as a telemedicine examination camera and digital microscope, to conduct remote physical exams. Typically, these virtual visits would necessitate patients to travel to facilities equipped with proper technologies and resources (Baker & Stanley, 2018). ‘Physicians must be aware that certain telemedicine technologies could be unaffordable to patients and hence impede access’ (WMA General Assembly, 2018). Consequently, imposing these FCs might further limit people from using telemedicine platforms. Therefore, it would be interesting to see whether the lack of lab work, physical exams and specialised equipment would prevent physicians from using telemedicine, as shown in Table 6.

Table 6.

FCs: Study Items with the Root Item (Scale) and Root Construct, Checking Whether the Lack of Lab Work, Physical Exams and Specialised Equipment Would Deter Physicians from Using Telemedicine.

Study Items Root Construct and Scale
FC1. I find that the inability to perform a physical exam limits my willingness to use telemedicine. (Physical exam only, like what we learned in medical school—not with specialised equipment)	Perceived behavioural control
FC2. I find that the inability to perform lab work or other diagnostic tests (flu swab, covid test, x-ray) limits my willingness to use telemedicine.	‘The system is not compatible with other systems I use’.
FC3. I find that the inability to examine my patients with specialised equipment (such as microscopes) limits my willingness to use telemedicine.	(Venkatesh et al., 2003)

Moderators

A moderator is a variable that can affect the influence of an independent variable on a dependent variable. The UTAUT suggests four moderators: age, gender, experience and voluntariness of use. This subsection describes the moderators suitable for this study.

‘Age’ stereotypes often occur in workplaces. Older workers might be ‘much more accustomed to seeking and applying traditional (i.e., ‘non-technology’) solutions to job-related tasks whereas younger workers are much more reliant on the use of technology for job accomplishment’ (Morris & Venkatesh, 2000). Age was specified as the number of years, as indicated in Table 7.

Table 7.

Moderators: Study Items with Two Moderating Variables Available in This Study.

Study Items		Variables
AGE	What is your age?	Number
GDR	What is your gender?	Not used
EXP	Which year did you start using telemedicine if you used telemedicine before the pandemic?	Year_before
	In a typical week, how many hours per week do you see patients on telemedicine before the pandemic?	Hours_before
	In a typical week, how many hours per week do you see patients on telemedicine during the pandemic?	Hours_before
VOL	Voluntariness of use?	Not used

Notes: AGE was the physician’s age in the number of years. EXP referred to the experience, estimating the number of hours spent on telemedicine.

The justification for using ‘gender’ as a moderator would lie in the gender stereotypes, in which women might possess different ideals from men in workplaces (Venkatesh et al., 2011). However, gender differences were unlikely to occur among physicians because of their professional values and continual commitments to their profession (Venkatesh et al., 2011). Therefore, the gender moderator was dropped from the present study.

‘Experience’ (EXP) was related to how much time a user spent using the system. In the UTAUT model, measuring experience involved collecting longitudinal data once every few weeks or months (Venkatesh et al., 2003). However, because collecting data periodically was infeasible in the present study, this study instead compiled self-reported times by querying about the duration of telemedicine use before (hours before since years before) and during the COVID-19 pandemic in a typical week (hours during) (Araujo et al., 2017), as shown in Table 7. The authors gauged the respondents’ experiences by summing up their typical weekly hours of telemedicine use before (EXP_before) and during (EXP_during) the COVID-19 pandemic in the equation below.

E X P = E X P_{b e f o r e} + E X P_{d u r i n g}

‘Voluntariness of Telemedicine Use’ (VOL) was defined as ‘the degree to which the use of the telemedicine [was] perceived as being voluntary, or of free will’ (Venkatesh et al., 2003). At the onset of the pandemic, many physicians shifted from in-person visits to telemedicine visits, which continued until the time of the survey, as most physicians were employed by health systems and corporations (Primary Care Collaborative, 2022). As such, the voluntariness of use as a moderator would not have any significant variance during which the study was conducted and was thus dropped from this study.

The Figure 1 delineates the relationships between the determinants, moderators and the outcome variables presented above.

Figure 1.

Source: Venkatesh et al. (2003).

Hypotheses

Based on the above discussion, this study presents the following hypotheses.

H₁: PE would have a positive and direct effect on BI.

H₂: EE would have a positive and direct effect on BI.

H₃: SI would have a positive and direct effect on BI.

H₄: FC would have a positive and direct effect on BI.

H₅: PE would have a positive and direct effect on ATT.

H₆: EE would have a positive and direct effect on ATT.

H₇: SI would have a positive and direct effect on ATT.

H₈: FC would have a positive and direct influence on ATT.

H₉: ATT would have a positive and direct influence on BI.

Methods

The authors administered online surveys between March and June in 2021. Hosted on SurveyMonkey, these surveys were advertised on several voluntary online community support groups for physicians exchanging information about the ongoing COVID-19 pandemic. As such, any active physician with either an MD or DO degree who practised telemedicine in the United States could participate in this survey. Before filling out the surveys, all participants agreed to a written informed consent required by the Institutional Review Board (IRB). The IRB reviewed the survey methodology and granted clearance pertaining to ethics concerning human subjects and privacies. Despite 165 participants attempting to fill out the surveys, only 110 who completed the surveys could be used in this study.

Most of the participants were women (93%) between 34 and 59 years of age (93%), had at least a decade of medical experience (84%), and practised medicine in the urban (67%) and the Middle Atlantic USA (48%). Many participants practised medicine in hospitals (70%) for at least 30 hours per week (72%), specialising in paediatrics (20%), internal medicine (14%), psychiatry (11%) and other medical fields.

Most participants did not use telemedicine before the COVID-19 pandemic (88%). During the pandemic, 40% of survey participants conducted more telemedicine visits than in-person visits with patients. Most participants (66%) did not have a choice in whether they used telemedicine in their practices, while others were responsible for making the full decision on telemedicine use (20%) or involved in decision-making groups (14%).

Participation in the survey was purely voluntary; participants were free to take surveys and withdraw at any time. Additionally, all participants were required to agree to the informed consent to be eligible for the survey, which minimised any coercion or undue influence.

Measures and Covariates

Figure 1 illustrates the hypothesised relationships described above. The output construct was the BI listed in Table 1 with three items: BI1, BI2 and BI3. The mediator was the ATT listed in Table 2 with four items: ATT1, ATT2, ATT3 and ATT4. The first determinant was PE listed in Table 3 with three items: PE1, PE2 and PE3. The second determinant was EE listed in Table 4 with four items: EE1, EE2, EE3 and EE4. The third determinant was SI listed in Table 5 with three items: SI1, SI2 and SI3. The last determinant was FC listed in Table 6 with three items: FC1, FC2 and FC3.

Table 7 summarised the moderators in the present study. The first moderator was AGE, an integer value representing the number of years in age. The second moderator was EXP, an integer value calculated using Equation 1 that would estimate the amount of time per week the user partook in telemedicine activities since the beginning of their telemedicine use.

Data Collection

The online survey consisted of several pages that could take up to 30 minutes to complete. On the first page, the participants would fill out their personal information, including their age, gender and location where they practised medicine. On the next page, they would provide their work habits and telemedicine use before and during the COVID-19 pandemic. The next few pages included items derived from the UTAUT model.

Once all the survey responses had been collected, the authors scaled the survey items on a uniform Likert-7 scale between –3 and 3, depicting strongly disagree (–3), disagree (–2), somewhat disagree (–1), neither disagree nor agree (0), somewhat agree (1), agree (2) and strongly agree (3).

In order to be consistent with the responses to the other determinants, responses to the FC items (FC₁, FC₂ and FC₃) were scored on a reversed scale. For example, for item F₁ (‘I find that the inability to perform a physical exam limits my willingness to use telemedicine’ in Table 6), –3 denoted strongly agree and 3 strongly disagree.

Data Diagnostics

This section describes the statistical analysis performed to determine the statistical validity of the collected date.

Exploratory Factor Analysis (EFA)

EFA is a statistical technique used to discover the underlying structure. This analysis involves a pattern matrix consisting of loaders (correlations between -1 and -1) that were arranged in item rows and factor columns (Tabachnick et al., 2013). With the pattern matrix, one would isolate all the correlating items with strong loaders into independent factors (Tabachnick et al., 2013). As a rule of thumb, choosing strong loaders of 0.32 and above would help identify items with over 10% overlapping variance within the same factor (Tabachnick et al., 2013). The authors followed the above guideline to achieve optimal results by verifying Table 1 through Table 6 with strong loaders over 0.32, as delineated in the pattern matrix.

Sample Size Power and Precision

A well-fitted model is one such that the measured items have significant relationships, which in turn are represented by a non-significant chi-square value (χ²). On the other hand, a significant χ² (i.e., p < .05) indicates that the measured items are, in fact, independent (Tabachnick et al., 2013). However, χ² often becomes significant due to minute differences between various items in large samples (Tabachnick et al., 2013). First, as the value of χ² depends on the sample size, a model with χ²/df > 3 is considered poorly fitted (Cole, 1987). Second, the root-mean-square error of approximation (RMSEA), which indicates the lack of fit compared to a perfect model (Tabachnick et al., 2013), should not exceed 0.08 (Chen et al., 2008). Third, the standardised root-mean-square residual (SRMR) should also not exceed 0.08 (Hu et al., 1999). Therefore, several fit indices ranging between 0 and 1 were developed to determine the extent of the model fit in a large sample size (Tabachnick et al., 2013). Other fit indices, such as comparative fit index (CFI), normed fit index (NFI), Tucker-Lewis index (TLI), goodness-of-fit index (GFI) and adjusted goodness-of-fit index (AGFI), should have high values (e.g., 0.95 or above) in order to indicate a good-fitting, parsimonious model (Tabachnick et al., 2013).

Construct Validity: Convergent and Discriminant Validity

A construct is valid when both convergent and discriminant validities are valid. Convergent validity implies that the measures observed within the same constructs are all related. A common way to determine convergent validity is to observe the values of the average variance extracted (AVE), Cronbach alpha (α) and composite reliability (CR) with AVE ≥ 0.50 (Fornell & Larcker, 1981) as well as α ≥ 0.70 and CR ≥ 0.70 to be convergent valid (Henseler et al., 2009). On the contrary, discriminant validity suggests that the measures from different constructs are not related to one another. One could thus observe discriminant convergence from the correlation table, in which all the correlations with other factors should not exceed the square root of the AVE (Baker & Stanley, 2018) to ensure that different traits were indeed unrelated to each other.

Analytic Strategy

The authors conducted structural equation modelling (SEM), (Nachtigall et al., 2003) to understand various direct and moderating effects. In general, SEM is a collection of statistical techniques that enables researchers to ascertain any relationships between independent and dependent variables (Tabachnick et al., 2013). As such, the authors developed a measurement model to estimate and predict the items described in Table 1 through Table 7 by using the maximum likelihood estimator with the survey data. Subsequently, the authors employed the diagnostic methods described above, in order to validate the measurement model and ensure the model’s acceptability.

To rule out the hypotheses described in the third section, the authors built several structural models that contained a combination of effects and moderations. This resulted in nine direct paths and twenty paths for moderation. By running the structural model with composite values, which were computed from the measurement model, the authors estimated the standardised β-weights with p values. The authors then utilised these β-weights to predict the extent of physicians’ acceptance of telemedicine in accordance with the hypotheses listed in the third section.

Results

The initial task was to verify whether the items were related to each factor in Table 1 through Table 6 from the survey responses. First, the authors applied the scree test to estimate the number of factors by examining the scree plot (Figure 2)—line plots of eigenvalues in factor analyses against their factor numbers—with and without the ATT described in Table 2. Although several mechanisms were available to determine the cut-off number in a scree plot (Costello & Osborne, 2005), the authors found that cutting off factors after the curve started to flatten aligned well in this research. In other words, the apparent asymptote began around the sixth factor, as illustrated in Figure 2. Finally, to statistically explore and verify the underlying relationships between variables, the authors conducted the EFA on the survey responses with the items in Table 1 through Table 6, as shown in Table 8.

Figure 2.

Source: Costello & Osborne (2005).

Table 8.

EFA: The Pattern Matrix with Six Factors Extracted Using Oblimin Rotations and Maximum-Likelihood Estimations with R2 and Eigenvalues (λ), Resulting in the UTAUT Model with Strong-Loaders at Cut-Offs of 0.32.

	Factor 1	Factor 2	Factor 3	Factor 4	Factor 5	Factor 6
	R² = 0.93	R² = 0.96	R² = 0.93	R² = 0.89	R² = 0.99	R² = 0.82
PE₁	0.01	–0.01	0.76	0.13	0.11	0.08
PE₂	0.05	0.12	0.82	–0.01	–0.02	0.05
PE₃	–0.04	0.06	0.91	–0.00	–0.06	–0.05
EE₁	0.90	–0.09	0.03	0.01	0.02	0.06
EE₂	0.87	–0.04	0.04	0.07	0.02	0.11
EE₃	0.81	0.15	–0.00	–0.03	–0.00	–0.15
EE₄	0.83	0.08	–0.09	–0.01	0.01	–0.08
FC₁	0.07	–0.02	0.17	0.24	–0.15	0.49
FC₂	–0.07	0.14	–0.09	0.10	–0.03	0.74
FC₃	0.02	–0.01	0.06	–0.11	0.08	0.83
SI₁	0.08	0.02	0.02	–0.01	0.98	0.02
SI₂	–0.18	0.20	–0.22	0.06	0.53	0.02
SI₃	–0.13	–0.00	0.01	0.10	0.45	–0.05
ATT₁	–0.06	0.02	0.25	0.64	0.16	–0.06
ATT₂	0.11	0.05	–0.03	0.86	–0.05	0.03
ATT₃	–0.00	0.10	0.27	0.43	0.22	–0.11
ATT₄	0.04	0.26	0.05	0.53	–0.02	0.12
BI₁	–0.00	0.81	0.03	0.11	–0.02	–0.00
BI₂	0.02	0.83	0.08	–0.02	0.09	0.05
BI₃	0.03	0.94	0.03	–0.00	–0.01	0.02

Once the relationships in Table 1 through Table 6 were compared in Figure 2 and Table 8, the authors subsequently applied SEM to compute the measurement model, which quantified the relationship between survey responses and underlying factors described in Section 3. This was illustrated in Figure 3. The authors also included a correlation between EE3 and EE4 in the measurement model (Figure 3) because of their similarities in Table 4. The model appeared to be fitted due to sufficiently fit indices’ values, as shown in Table 9: χ²/df ≤ 3.0, RMSEA < 0.08, SRMR < 0.08, CFI ≥ 0.95 and TLI ≥ 0.95. In addition, the model was convergent valid with α > 0.6, CR ≥ 0.7 and AVE ≥ 0.5, as shown in Table 10 . It was also divergent valid with correlations greater than $\sqrt{A V E}$ in Table 11.

Figure 3.

Source: Epskamp et al. (2017).

Table 9.

Model Fit of the Measurement Model in Figure 2.

Model fit	Criteria
$\frac{χ^{2}}{d f} = \frac{215}{154} =$ $1.40 \leq 3.00$	Good (Cole, 1987)
RMSEA = 0.06 ≤ 0.08	Good (Chen et al., 2008)
SRMR = 0.07 ≤ 0.08	Good
CFI = 0.96 ≥ 0.95	Good (Bentler & Bonett, 1980)
TLI = 0.95 ≥ 0.95	Good (Bentler & Bonett, 1980)
NFI = 0.88 ≥ 0.95	(Bentler & Bonett, 1980)
GFI = 0.85 ≥ 0.95	(Anderson & Gerbing, 1984)
AGFI = 0.77 ≥ 0.95	(Anderson & Gerbing, 1984)

Table 10.

Convergent Validity of the Measurement Model in Figure 2.

	±≥07	CR≥07	AVE≥05	Reliability
PE	0.92	0.92	0.80	Good
EE	0.92	0.86	0.70	Good
SI	0.67	0.72	0.50	Acceptable (± > 0.6)
FC	0.76	0.77	0.54	Good
BI	0.95	0.95	0.86	Good
ATT	0.87	0.87	0.64	Good

Notes: Cronbach Alpha (±≥070), CR (CR≥.70), and AVE (AVE≥.50).

Table 11.

Discriminant Validity of the Measurement Model in Figure 2 Using the Heterotrait-Monotrait Ratio of Correlations (HTMT).

	PE	EE	SI	FC	BI	ATT	$\sqrt{A V E}$
PE	1.00						0.90
EE	0.23	1.00					0.84
SI	0.20	0.25	1.00				0.71
FC	0.32	0.11	0.10	1.00			0.73
BI	0.71	0.32	0.38	0.41	1.00		0.80
ATT	0.77	0.37	0.34	0.33	0.77	1.00	0.93
$\sqrt{A V E}$	0.90	0.84	0.71	0.73	0.80	0.93

Note: All the off-diagonal, correlational values must be less than AVE.

Finally, to determine which hypotheses from the third section were supported, the authors conducted confirmatory factor analysis on the structural models in Figure 1 using the measurement model in Figure 3. Table 12 presents the direct effects (D) between independent and dependent variables without any moderations, as well as the moderating effects using several moderators: age alone (A), experience alone (E) and both age and experience (A + E) together.

Table 12.

Direct and Total Effects (Σ) on BI and Attitude Towards Telemedicine Use.

	D	A	E	A + E
↓ Behavioral Intention (BI)
R ²	0.61	0.62	0.75	0.76
ΣPE	0.34∗∗∗	0.38∗∗∗	0.39∗∗∗	0.42∗∗∗
PE	0.55∗∗∗	0.59∗∗∗	0.59∗∗∗	0.61∗∗∗
ΣEE	0.06	0.03	0.10^∗	0.09
EE	0.12	0.07	0.16^∗∗	0.13^∗
ΣSI	0.17^∗∗	0.19^∗∗	0.1	0.12^∗
SI	0.23∗∗∗	0.25∗∗∗	0.15^∗∗	0.18^∗∗
ΣFC	0.14^∗	0.15^∗	0.12^*	0.12^∗
FC	0.17^∗∗	0.18^∗∗	0.14^∗∗	0.14^∗
ATT	0.36∗∗∗	0.33∗∗∗	0.32∗∗∗	0.29∗∗∗
AGE		0.03		0.04
EXP			0.01	0.01
PE × AGE		−0.01		−0.08
EE × AGE		0.11		0.06
SI × AGE		−0.06		−0.05
FC × AGE		−0.06		0.03
PE × EXP			−0.38^∗∗∗	−0.39^∗∗∗
EE × EXP			0.11^∗	0.11^∗
SI × EXP			0.04	0.04
FC × EXP			0.25∗∗∗	0.27∗∗∗
↓ Attitude Toward Use (ATT)
R2	0.54	0.58	0.56	0.59
PE	0.60∗∗∗	0.64∗∗∗	0.62∗∗∗	0.66∗∗∗
EE	0.16 ^∗	0.12	0.17 ^∗∗	0.13
SI	0.16 ^∗	0.18 ^∗∗	0.17 ^∗	0.20 ^∗∗
FC	0.08	0.07	0.07	0.06
AGE		0.11		0.12
EXP			0.04	0.02
PE × AGE		−0.09		−0.08
EE × AGE		0.02		0.03
SI × AGE		−0.13		−0.13
FC × AGE		−0.09		−0.06
PE × EXP			−0.04	−0.04
EE × EXP			0.07	0.06
SI × EXP			0.12	0.10
FC × EXP			0.00	0.02

Source: Venkatesh et al. (2011).

Notes: p > .1 ^∗p < .05 ^∗∗p < .01 ^∗∗∗p < .001.

D = Direct effects (no moderations).

A = Age-only moderations.

E = Experience-only moderations.

A + E = Age + Experience moderations.

Discussions

This article explored the US physicians’ acceptance of telemedicine during the COVID-19 pandemic, using the UTAUT. Specifically, this study evaluated the direct effects of physician factors—including their PE, EE, SI and FC (e.g., inability to perform lab work and physical exams)—on their BI and ATT. Furthermore, this study explored the moderating effects of age and experience on these direct effects.

Table 12 summarised Hypotheses from H₁ to H₉, for which the authors utilised standardised regressions to determine how well they were supported. These values were depicted in Table’s column D, the direct effects. The entries in Table 13 with significant standardised regressions (i.e., p < .05) implied that the corresponding hypotheses were supported and thus marked with checkmarks in the hypothesis column.

Table 13.

Summary of Findings: Direct Effects.

Hypotheses	Dependent Variable	Independent Variable	Interpretation
✓✓ H₁	BI	PE	PE had the strongest effect on BI.
H ₂	BI	EE	EE had a somewhat weak but non-significant effect on BI.
✓ H₃	BI	SI	SI had a significant effect on BI.
✓ H₄	BI	FCs	Facilitating condition had a significant effect on BI.
✓✓ H₅	BI	PE	PE also had the strongest effect on ATT.
✓ H₆	BI	EE	EE had a significant and weak effect on BI.
✓ H₇	BI	SI	SI had a significant effect on ATT.
H ₈	BI	FCs	Facilitating condition had a non-significant effect on ATT.
✓✓H₉	BI	ATT	ATT had a significant and strong effect on itself.

Source: Venkatesh et al. (2003).

Notes: ✓Significant, weak effect; ✓✓ Significant, strong effect.

The hypotheses marked with, ✓✓, on the other hand, were strongly supported in this study when their standardised regressions were greater than 0.3 (i.e., with at least a medium effect of 0.3 (Cohen, 1988; Fey et al., 2022).

Hypotheses H₁ and H₅ were both strongly supported, as physicians’ PE appeared to have the most significant effects on both ATT (β = 0.60 with p < .001 in PE ↓ ATT) and BI (β = 0.34 with p < .001 in PE ↓ BI). Similarly, H₉ was strongly supported, as physicians’ ATT exhibited a strong effect on BI (β = 0.36 with p < .001 in ATT ↓ BI). These three findings were consistent with several studies on healthcare professionals’ attitudes towards technologies (Ward et al., 2008), physicians’ acceptance of telemedicine (Garavand et al., 2022) and users’ acceptances of other technologies (Blut et al., 2021).

This article found that EE affects physicians’ attitude (β = 0.16 with p < .05 in EE ↓ ATT) but, interestingly, not their intention to use telemedicine (p > .05 in EE ↓ BI), thus supporting H₆ but not H₂. A systematic review of physicians’ acceptance of telemedicine (Garavand et al., 2022) examined several studies on UTAUT. While those studies supported the influence of PE (or perceived usefulness) on physicians’ acceptance (i.e., H₁), only slightly more than half of those studies supported a similar influence of EE (or ease of use).

This article also examined the role SI played in physicians’ acceptance of telemedicine. As such, both Hypotheses H₃ and H₇ were supported in this study. SI appeared to play an essential role in influencing physicians’ acceptance of, and attitude towards, telemedicine use (β = 0.17 with p < .01 in SI ↓ BI and β = 0.16 with p < .05 in SI ↓ ATT). These findings were consistent with other studies in technology acceptance (Garavand et al., 2022).

The authors also examined the influence of FC on physicians’ BIs. They found that H₄ was supported while H₈ was unsupported. In other words, FCs were likely to impact physicians’ BI (β = 0.14 with p < .05 in FC ↓ BI) rather than their attitude towards telemedicine use (p > .1 in FC ↓ ATT). This was reasonable; according to a case report (Pappan et al., 2021), the inability to perform labs or a physical exam on a patient with severe abdominal pain might impact whether their physicians intend to use telemedicine but not the physicians’ overall ATT with other patients. This finding confirms similar conclusions by some other studies on UTAUT.

This article also explored how physicians’ age would moderate various effects. Physicians’ age did not seem to moderate any of the effects in this study because all the standardised regression weights β related to the age moderator were non-significant (i.e., p > .1), as shown in Table 12. This result confirmed the similar conclusion drawn by another study (Venkatesh et al., 2011).

Lastly, summarised in Table 14, this article explored how physicians’ experience (EXP) with telemedicine moderated the effects on BI. EXP, the number of hours a physician weekly spent using telemedicine, appeared to moderate the PE’s effects towards BIs negatively (β = −0.39 with p < .001) but positively on FCs’ effects towards BIs (β = 0.27 with p < .001). These moderations were shown in Table 12’s column A+E. This suggested that as physicians gained experience with telemedicine, even if they expected strong performance, their intention to use telemedicine was not as robust as their less experienced colleagues. On the other hand, physicians with more telemedicine experience were more likely to be painfully aware of the limitations of telemedicine imposed by the FCs. This was in contrast to their less experienced colleagues, who might not have yet encountered a real-life situation in which FCs played a major pitfall in a telemedicine visit. In fact, when queried about the lack of a physical exam via telemedicine, one survey respondent, who had three times more telemedicine experience than most participants, quipped, ‘listening to the lungs is powerful’. One could imagine a clinical scenario in which a patient’s pleural effusion (‘water on the lungs’) could be missed during a telemedicine visit due to the lack of a physical exam.

Table 14.

Summary of Findings: Moderating Effects.

Support?	Dependent Variable	Independent Variable	Moderator	Interpretation
✓	BI	PE	Experience	As physicians gained experience with telemedicine, even if they expected strong performance, their intention to use telemedicine was not as robust as their less experienced colleagues.
	BI	EE	Experience	Because the direct effect was non-significant (p >.05 in EE → BI), this hypothesis became unsupported even though its effect was significant (β = 0.11 with p <.05 in EE×EXP → BI).
	BI	SI	Experience	Non-significant effect.
✓	BI	FCs	Experience	Physicians with more telemedicine experience were more likely to recognise the limitations of telemedicine imposed by the FCs.

Limitations

These participants in this study were physicians from online support groups exchanging information about the COVID-19 pandemic; therefore, this study could exhibit similar limitations as any online surveys (Andrade, 2020). Namely, these participants could already be savvy in technology use and online communications, which were prerequisite skills to use many telemedicine platforms. Moreover, due to their active online presence in physician interest groups, this study’s subjects were a self-selected group of physicians who could be more readily accepting of telemedicine. Finally, as the participants were physicians primarily located in suburban and urban United States, their beliefs could differ from those serving rural and poor communities.

Conclusion

This study employed the UTAUT to examine the US physicians’ acceptance of telemedicine. As a critical predictor of telemedicine acceptance, physicians’ BI was influenced by their PE, SI, FCs and ATT. Physicians’ ATT was also influenced by PE, EE and SI. In addition, more experience with telemedicine also altered the influence of PE and FCs on physicians’ intention to use telemedicine.

The authors recommended further exploration of the physicians’ acceptance and use of telemedicine in providing healthcare to patients. Such efforts could lead to meaningful implications on the direction telemedicine and digital health product developers should focus their work in the exciting field of digital health.

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The authors received no financial support for the research, authorship and/or publication of this article.

References

Acheampong

, & Vimarlund

(2015). Business models for telemedicine services: A literature review. Health Systems , 4(3), 189–203.

Ajzen

(1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50(2), 179–211.

AMA (Ed.). (2020). Telehealth implementation playbook .

Anderson

J. C.

, & Gerbing

D. W.

(1984). The effect of sampling error on convergence, improper solutions, and goodness-of-fit indices for maximum likelihood confirmatory factor analysis. Psychometrika , 49(2), 155–173.

Andrade

(2020). The limitations of online surveys. Indian Journal of Psychological Medicine , 42(6), 575–576.

Araujo

A. L. S. O.

, Santos

J. S.

, Andrade

W. L.

, Guerrero

D. D. S.

, & Dagienė

(2017). Exploring computational thinking assessment in introductory programming courses . Frontiers in Education Conference, 1–9.

Baker

, & Stanley

(2018). Telemedicine technology: A review of services, equipment, and other aspects. Current Allergy and Asthma Reports , 18(11), 1–8.

Bentler

P. M.

, & Bonett

D. G.

(1980). Significance tests and goodness of fit in the analysis of covariance structures. Psychological Bulletin , 88(3), 588–606.

Berg

(2001). Implementing information systems in health care organizations: Myths and challenges. International Journal of Medical Informatics , 64(2-3), 143–156.

10.

Blut

, Chong

, Tsiga

, & Venkatesh

(2021). Meta-analysis of the unified theory of acceptance and use of technology (UTAUT): Challenging its validity and charting A research agenda in the red ocean. Journal of the Association for Information Systems , 23(1), 13–95. https://doi.org/10.17705/1jais.00719

11.

Campbell

J. D.

, Harris

K. D.

, & Hodge

(2001). Introducing telemedicine technology to rural physicians and settings. Journal of Family Practice , 50(5), 419–419.

12.

Chen

, Curran

P. J.

, Bollen

K. A.

, Kirby

, & Paxton

(2008). An empirical evaluation of the use of fixed cutoff points in RMSEA test statistic in structural equation models. Sociological Methods & Research , 36(4), 462–494.

13.

Chen

, Cheng

, & Mehta

(2013). A review of telemedicine business models. Telemedicine and E-Health , 19(4), 287–297.

14.

Cohen

(1988). Statistical power analysis for the behavioral sciences . Routledge.

15.

Cole

D. A.

(1987). Utility of confirmatory factor analysis in test validation research. Journal of Consulting and Clinical Psychology , 55(4), 584–594.

16.

Costello

A. B.

, & Osborne

(2005). Best practices in exploratory factor analysis: Four recommendations for getting the most from your analysis. Practical Assessment, Research, and Evaluation , 10(1), 7.

17.

Craig

, & Petterson

(2005). Introduction to the practice of telemedicine. Journal of Telemedicine and Telecare , 11(1), 3–9.

18.

Davis

F. D.

(1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly , 13(3), 319–340.

19.

Dwivedi

Y. K.

, Rana

N. P.

, Jeyaraj

, Clement

, & Williams

M. D.

(2019). Re-examining the unified theory of acceptance and use of technology (UTAUT): Towards a revised theoretical model. Information Systems Frontiers , 21(3), 719–734.

20.

Effertz

, Alverson

D. C.

, Dion

, Duffy

, Noon

, Langell

, Antoniotti

, & Lowery

(2017). Sustaining and expanding telehealth: A survey of business models from selected prominent US telehealth centers. Telemedicine and E-Health , 23(2), 137–142.

21.

Ekeland

A. G.

, Bowes

, & Flottorp

(2010). Effectiveness of telemedicine: A systematic review of reviews. International Journal of Medical Informatics , 79(11), 736–771.

22.

Epskamp

, Epskamp

M. S.

, & MplusAutomation

(2017). Package ‘semplot’ . Recuperado de https://cran.r-project.org/web/packages/semPlot/semPlot.pdf.

23.

Fey

C. F.

, Hu

, & Delios

(2022). The Measurement and communication of effect sizes in management research. Management and Organization Review , 9(1), 1–22.

24.

Fishbein

M. A.

, & Ajzen

(1975). Intention and behavior: An introduction to theory and research . Addison-Wesley.

25.

Fornell

, & Larcker

D. F.

(1981). Structural equation models with unobservable variables and measurement error: Algebra and statistics . Sage.

26.

Garavand

, Aslani

, Nadri

, Abedini

, & Dehghan

(2022). Acceptance of telemedicine technology among physicians: A systematic review. Informatics in Medicine Unlocked , 30, 100943.

27.

Henseler

, Ringle

C. M.

, & Sinkovics

R. R.

(2009). The use of partial least squares path modeling in international marketing. In

Sinkovics

R. R.

, Ghauri

P. N.

(Eds.), New challenges to international marketing (pp. 277–319). Emerald Group Publishing Limited.

28.

Hjelm

(2005). Benefits and drawbacks of telemedicine. Journal of Telemedicine and Telecare , 11(2), 60–70.

29.

P. J.

, Chau

P. Y.

, Sheng

O. R. L.

, & Tam

K. Y.

(1999). Examining the technology acceptance model using physician acceptance of telemedicine technology. Journal of Management Information Systems , 16(2), 91–112.

30.

Centers for Medicare & Medicaid Services. (2020, March 17). Medicare telemedicine health care provider fact sheet . https://www.cms.gov/newsroom/fact-sheets/medicare-telemedicine-health-care-provider-fact-sheet

31.

Mehrotra

, Chernew

M. E.

, Linetsky

, Hatch

, Cutler

D. A.

, & Schneider

E. C.

(2021, February 2). The impact of COVID-19 on outpatient visits in 2020: Visits remained stable, despite a late surge in cases . The Commonwealth Fund. https://doi.org/10.26099/bvhf-e411

32.

Morris

M. G.

, & Venkatesh

(2000). Age differences in technology adoption decisions: Implications for a changing work force. Personnel Psychology , 53(2), 375–403.

33.

Nachtigall

, Kroehne

, Funke

, & Steyer

(2003). Pros and cons of structural equation modeling. Methods Psychological Research Online , 8(2), 1–22.

34.

Nicola

, Alsafi

, Sohrabi

, Kerwan

, Al-Jabir

, Iosifidis

, Agha

, & Agha

(2020). The socio-economic implications of the coronavirus pandemic (COVID-19): A review. International Journal of Surgery , 78, 185–193.

35.

Nittari

, Khuman

, Baldoni

, Pallotta

, Battineni

, Sirignano

, Amenta

, & Ricci

(2020). Telemedicine practice: Review of the current ethical and legal challenges. Telemedicine and E-Health , 26(12), 1427–1437.

36.

Ohannessian

(2015). Telemedicine: Potential applications in epidemic situations. European Research in Telemedicine/La Recherche Européenne En Télémédecine , 4(3), 95–98.

37.

Pappan

, Benkhadra

, Papincak

, Ashker

, Uchin

, Sidique

, Pirani

, & Clemenza

(2021). Values and limits of telemedicine: A case report. SN Comprehensive Clinical Medicine , 3, 317–319.

38.

Peters

, Blohm

, & Leimeister

J. M.

(2015). Anatomy of successful business models for complex services: Insights from the telemedicine field. Journal of Management Information Systems , 32(3), 75–104.

39.

PHI. (2020). State telehealth laws & reimbursement policies .

40.

Portnoy

, Waller

, & Elliott

(2020). Telemedicine in the era of COVID-19. The Journal of Allergy and Clinical Immunology: In Practice , 8(5), 1489–1491.

41.

Primary Care Collaborative. (2022). Report: Supermajority of U.S. physicians are employed by health systems or corporate entities . https://www.pcpcc.org/2022/04/25/report-supermajority-us-physicians-are-employed-health-systems-or-corporate-entities

42.

OCR. (2022, June 10). Guidance on how the HIPAA rules permit covered health care providers health plans to use remote communication technologies for audio-only telehealth . U.S. Department of Health and Human Services. https://www.hhs.gov/hipaa/for-professionals/privacy/guidance/hipaa-audio-telehealth/index.html

43.

Schwamm

L. H.

(2014). Telehealth: Seven strategies to successfully implement disruptive technology and transform health care. Health Affairs , 33(2), 200–206.

44.

Severino

, & Director

O. C. R.

(2020). Notification of enforcement discretion for telehealth remote communications during the COVID-19 nationwide public health emergency . US Department of Health & Human Services.

45.

Simmons

, Alverson

, Poropatich

, D’Iorio

, DeVany

, & Doarn

C. R.

(2008). Applying telehealth in natural and anthropogenic disasters. Telemedicine and E-Health , 14(9), 968–971.

46.

Sterling

, & LeRouge

(2019). On-demand telemedicine as a disruptive health technology: Qualitative study exploring emerging business models and strategies among early adopter organizations in the United States. Journal of Medical Internet Research , 21(11), e14304.

47.

Tabachnick

B. G.

, Fidell

L. S.

, & Ullman

J. B.

(2013). Using multivariate statistics (7th ed., Vol. 6). Pearson.

48.

Venkatesh

, Morris

M. G.

, Davis

G. B.

, & Davis

F. D.

(2003). User acceptance of information technology: Toward a unified view. MIS Quarterly , 27(3), 425–478.

49.

Venkatesh

, Sykes

T. A.

, & Zhang

(2011). ‘Just what the doctor ordered’: A revised UTAUT for EMR system adoption and use by doctors [Conference session]. 44th Hawaii International Conference on System Sciences, Kauai, HI, USA, January 4–7, 2011.

50.

Venkatesh

, Thong

J. Y.

, & Xu

(2012). Consumer acceptance and use of information technology: Extending the unified theory of acceptance and use of technology. MIS Quarterly , 36(1) 157–178.

51.

Wade

V. A.

, Karnon

, Elshaug

A. G.

, & Hiller

J. E.

(2010). A systematic review of economic analyses of telehealth services using real time video communication. BMC Health Services Research , 10(1), 1–13.

52.

Waller

, & Stotler

(2018). Telemedicine: A primer. Current Allergy and Asthma Reports , 18(10), 1–9.

53.

Ward

, Stevens

, Brentnall

, & Briddon

(2008). The attitudes of health care staff to information technology: A comprehensive review of the research literature. Health Information & Libraries Journal , 25(2), 81–97.

54.

WMA General Assembly. (2018). WMA statement on the ethics of telemedicine . WMA. https://www.wma.net/policies-post/wma-statement-on-the-ethics-of-telemedicine/

An Assessment of the COVID-19 Pandemic’s Impact on US Physicians’ Acceptance of Telemedicine: An Exploratory Study

Abstract

Keywords

Introduction

The Unified Theory of Technology Acceptance and Use

The Research Constructs

BI to Use the System: Study Items with Root Items (Scales) Regarding Physicians’ Plans, Predictions and Desires to Use Telemedicine Even After the COVID-19 Pandemic was Over.

ATT: Study Items with Root Items (Scales) and Root Constructs About Whether Physicians had Positive ATT.

PE: Study Items with Root Items (Scales) and Root Constructs About Physicians’ Beliefs in Getting More Work Done Quickly and Simplifying Their Workflows When Using Telemedicine.

EE: Study Items with Root Items (Scales) and Root Constructs About Physicians Believing that Telemedicine, with a Clear and Understandable User Interface, was Easy to Learn, Able to Perform the Right Job Easily and Flexible to Interact With.

SI: Study Items with Root Items (Scales) and Root Constructs About Physicians’ Decisions to Use Telemedicine Due to Important People, Such as Colleagues and Administrators, Who Encouraged or Demanded the Use of Telemedicine.

FCs: Study Items with the Root Item (Scale) and Root Construct, Checking Whether the Lack of Lab Work, Physical Exams and Specialised Equipment Would Deter Physicians from Using Telemedicine.

Moderators

Moderators: Study Items with Two Moderating Variables Available in This Study.

Hypotheses

Methods

Measures and Covariates

Data Collection

Data Diagnostics

Exploratory Factor Analysis (EFA)

Sample Size Power and Precision

Construct Validity: Convergent and Discriminant Validity

Analytic Strategy

Results

EFA: The Pattern Matrix with Six Factors Extracted Using Oblimin Rotations and Maximum-Likelihood Estimations with R2 and Eigenvalues (λ), Resulting in the UTAUT Model with Strong-Loaders at Cut-Offs of 0.32.

Model Fit of the Measurement Model in Figure 2.

Convergent Validity of the Measurement Model in Figure 2.

Discriminant Validity of the Measurement Model in Figure 2 Using the Heterotrait-Monotrait Ratio of Correlations (HTMT).

Direct and Total Effects (Σ) on BI and Attitude Towards Telemedicine Use.

Discussions

Summary of Findings: Direct Effects.

Summary of Findings: Moderating Effects.

Limitations

Conclusion

Footnotes

Declaration of Conflicting Interests

Funding

References