The Acceptance Behavior of Smart Home Health Care Services in South Korea: An Integrated Model of UTAUT and TTF

With the COVID-19 pandemic, the importance of home health care to manage and monitor one’s health status in a home environment became more crucial than ever. This change raised the need for smart home health care services (SHHSs) and their extension to everyday life. However, the factors influencing the acceptance behavior of SHHSs have been inadequately investigated and failed to address why users have the intention to use and adopt the services. This study aimed to analyze the influential factors and measure the behavioral acceptance of SHHSs in South Korea. This study adopted the integrated model of the unified theory of acceptance and use of technology (UTAUT) and task–technology fit (TTF) to understand the behavioral acceptance of SHHSs from users’ perceptions and task–technology fit. Multiple-item scales were established based on validated previous measurement scales and adjusted in accordance with SHHS context. Data from 487 valid samples were analyzed statistically, applying partial least square structural equation modeling. The results indicated that the integrated acceptance model explained 55.2% of the variance in behavioral intention, 44.9% of adoption, and 62.5% of the continuous intention to use SHHSs, supporting 11 of the 13 proposed hypotheses. Behavioral intention was positively influenced by users’ perceptions on performance expectancy, effort expectancy, social influence, and functional conditions. Task–technology fit significantly influenced performance expectancy and behavioral intention, validating the linkage between the two models. Meanwhile, task characteristics were insignificant to determine task–technology fit, which might stem from complex home health care needs due to the COVID-19 pandemic, but were not sufficiently resolved by current service technologies. The findings implied that the acceptance of SHHSs needs to be evaluated according to both the user perceptions of technologies and the matching fit of task and technology. Theoretically, this study supports the applicability of the integrated model of UTAUT and TTF to the domain of SHHS, and newly proposed the measurement items of TTF reflecting the domain specificity of SHHS, providing empirical evidence during the pandemic era in South Korea. Practically, the results could suggest to the planners and strategists of home health care services how to promote SHHS in one’s health management.


Introduction
With the fourth industrial revolution and information and communication technology (ICT) advancement, a highly intelligent home environment is being realized [1]. Along with the industrial investments into smart home sectors, academic research has also investigated smart homes in terms of technological capabilities, their implications on various services, and the benefits to users' lives [2]. Technology-wise, smart homes are characterized by integrating technologies such as home automation, automatic control systems, communication networks, connection devices and services, remote access and control, and home intelligence [3][4][5]. Service-wise, smart homes provide diverse services such as management, and to examine the relationships of external constructs influencing users' intention to use and adoption.
Several studies have investigated user acceptance of home health care services such as mobile health applications and devices in Europe or China [23][24][25][26][27]. However, the primary target of those studies was mobile application services, and research on health care particularly focused on smart home environments was not sufficiently conducted. Therefore, among the various versions of technology acceptance models, this study adopts an appropriate model for technology-intensive SHHSs and establishes the measurement items reflecting the domain specificity of SHHSs. Consequently, the findings of this research could contribute to providing empirical evidence and offering additional insights into the technology acceptance of SHHSs specified in the South Korean context at the time of the pandemic era.

Theoretical Framework and Research Hypotheses
To assess the acceptance of technologies, researchers have established several models such as the technology acceptance model (TAM) [28], its well-known extensions of TAM 2 [29], and the unified theory of acceptance and use of technology (UTAUT) [30]. Notably, Venkatesh et al. proposed the UTAUT model that integrated eight models and theories of prior studies (i.e., the theory of reasoned action (TRA), TAM, the motivation model, the theory of planned behavior (TPB), a combined model of TAM and TPB, the innovation diffusion theory, the model of personal computer utilization, and the social cognitive theory) [30]. Therefore, UTAUT has been considered to have a comprehensive understanding of the acceptance procedure and a robust predictive power compared with other models [31]. It presents four determinants of behavioral intention: (1) performance expectancy (PE, adapted from the perceived usefulness of TAM), (2) effort expectancy (EE, adapted from the perceived ease of use of TAM), (3) social influence (SI, adapted from the subjective norm of TRA), and (4) facilitating conditions (FC, adapted from the perceived behavior control of TPB) [32]. Many researchers have employed the UTAUT to explain the technology acceptance behavior across diverse domains, including information technologies [24,32,33] and medical informatics [34,35].
Meanwhile, Goodhue and Thomson suggested the task technology fit (TTF) model comprehends the association between information system and individual performance; this model illustrates how the fit between technologies and users' tasks influences individual performance in information technology systems [36]. The task-technology fit is the level to which the technology features or functions match the task requirements, and both task characteristics (TAC) and technology characteristics (TEC) are the determinants of tasktechnology fit [36][37][38]. In other words, when the task requirements exceed the capabilities of technologies, or when technology features demonstrate unsatisfactory performances to complete the task, the task-technology fit would decline [24]. The TTF model has been widely applied to various contexts of information systems, including online services [39,40] and mobile technologies [41,42].
Although the UTAUT has been evaluated in diverse industry contexts and can describe around 70% of the users' usage intention [30], some limitations still exist. Even if a user considers specific information systems useful and easy to use, (s)he will not use the system when it cannot fit requirements or improve performance [43]. While the UTAUT underscores a user's perceptions of technology, the TTF elucidates a user's acceptance from the perspective of task-technology fit. In that sense, TTF can compensate for the limitation of UTAUT by explaining the fit between the task requirements and technical characteristics. Accordingly, several studies have combined the UTAUT and TTF models to explain user adoption behavior of advanced technology services such as mobile banking, online education, and wearable device services [24,33,43,44]. In that research, the factors of both TTF and UTAUT models significantly influenced the user's adoption behavior [43].
As mentioned earlier, SHHS is a technology-intensive service incorporating diverse home intelligence technologies to support real-time and long-term health monitoring, unobtrusive activity support, and disease prevention [14,15]. Thus, users' perceptions of the smart home and health care technologies are basically influential to their adoption behaviors. However, whether the technologies of SHHSs can satisfy users' requirements and support their tasks sufficiently would be at a different level to the perception of technologies. In other words, the acceptance of SHHSs needs to be approached considering both the technical complexity of smart homes and the task pertinence of home healthcare. Therefore, to establish a comprehensive understanding of the users' adoption behavior of SHHSs, embracing the TTF perspective as well as the typical UTAUT model is critical.
Accordingly, this study adopted the integrated model of UTAUT and TTF from previous research [24,35,43,44]. The integrated research framework was established as illustrated in Figure 1, and the hypotheses are listed in Table 1. banking, online education, and wearable device services [24,33,43,44]. In that research, the factors of both TTF and UTAUT models significantly influenced the user's adoption behavior [43].
As mentioned earlier, SHHS is a technology-intensive service incorporating diverse home intelligence technologies to support real-time and long-term health monitoring, unobtrusive activity support, and disease prevention [14,15]. Thus, users' perceptions of the smart home and health care technologies are basically influential to their adoption behaviors. However, whether the technologies of SHHSs can satisfy users' requirements and support their tasks sufficiently would be at a different level to the perception of technologies. In other words, the acceptance of SHHSs needs to be approached considering both the technical complexity of smart homes and the task pertinence of home healthcare. Therefore, to establish a comprehensive understanding of the users' adoption behavior of SHHSs, embracing the TTF perspective as well as the typical UTAUT model is critical.
Accordingly, this study adopted the integrated model of UTAUT and TTF from previous research [24,35,43,44]. The integrated research framework was established as illustrated in Figure 1, and the hypotheses are listed in Table 1.

Label
Hypotheses H1 The task characteristics of SHHSs * positively affect task-technology fit. H2 The technology characteristics of SHHSs positively affect task-technology fit. H3a Task-technology fit positively affects the performance expectancy of SHHSs. H3b Task-technology fit positively affects the behavioral intention to use SHHSs. H3c Task-technology fit positively affects the adoption of SHHSs.

H4
Performance expectancy positively affects the behavioral intention to use SHHSs.

H5
Effort expectancy positively affects the behavioral intention to use SHHSs. H6 Social influence positively affects the behavioral intention to use SHHSs. H7a Facilitating conditions positively affect the behavioral intention to use SHHSs. H7b Facilitating conditions positively affect the adoption of SHHSs. H8 Behavioral intention to use SHHSs positively affects the adoption of SHHSs. H9 The adoption of SHHSs positively affects the continued intention to use SHHSs. H10 The technology characteristics of SHHSs positively affect effort expectancy.
* SHHSs: smart home health care services.

H1
The task characteristics of SHHSs * positively affect task-technology fit. H2 The technology characteristics of SHHSs positively affect task-technology fit. H3a Task-technology fit positively affects the performance expectancy of SHHSs. H3b Task-technology fit positively affects the behavioral intention to use SHHSs. H3c Task-technology fit positively affects the adoption of SHHSs. H4 Performance expectancy positively affects the behavioral intention to use SHHSs. H5 Effort expectancy positively affects the behavioral intention to use SHHSs.

H6
Social influence positively affects the behavioral intention to use SHHSs. H7a Facilitating conditions positively affect the behavioral intention to use SHHSs. H7b Facilitating conditions positively affect the adoption of SHHSs. H8 Behavioral intention to use SHHSs positively affects the adoption of SHHSs. H9 The adoption of SHHSs positively affects the continued intention to use SHHSs. H10 The technology characteristics of SHHSs positively affect effort expectancy.
* SHHSs: smart home health care services.
The proposed model consists of ten constructs contextualized in the SHHS context. We included behavioral intention (BI), adoption (ADT), and continued intention (CI) to measure the progress of user acceptance as they are widely adopted predictors of the actual acceptance of technology-intensive services [24,33,[43][44][45][46].
The influence of TAC on TTF (H1) and that of TEC on TTF (H2) were adopted from the TTF model [36,37]. The influences on BI from PE (H4), from EE (H5), from SI (H6), and from FC (H7a) were adopted from the UTAUT model [30].
The relationship between the two models was hypothesized by referring to previous studies. Several scholars [24,33,43,44] have reported that TTF positively influences PE (H3a), BI (H3b), and ADT (H3c). Likewise, in the context of SHHSs, users can perceive that a SHHS could enhance their home health care performance only when the functions of SHHS can match well with home health care tasks (i.e., task-technology fit). Moreover, the characteristics of SHHSs (e.g., ubiquitous and real-time services) can support users' continuous health monitoring and thus diminish their effort cost. Based on this, the hypothesis that the technology characteristics of SHHSs positively affect effort expectancy (H10) was settled [24,33].
Meanwhile, in the original UTAUT model [30], demographic variables such as gender, age, experience, and voluntariness of use were included as moderators. However, due to the complexity of the unified model with many constructs, most of the previous studies adopting the unified model of UTAUT and TTF focused on the link between two models without moderators to explore the validity and applicability of the unified model [24,33,43,44]. Likewise, this study also aims to test the applicability of the unified model in the SHHS domain, thus any moderating variable was not considered in this research framework.

Measurement
Multiple-item scales were established to measure the constructs. Measurement items were adopted from pertinent previous studies and adjusted to be suitable for the SHHS context, as shown in Table 2. Keywords to represent the meaning of each measurement item were defined as properties. Notably, the properties of TAC and TEC were developed from literature to delineate the characteristics of SHHSs comprehensively. The tasks of SHHSs embrace the needs of real-time (TAC1) and long-term health monitoring (TAC2), disease prevention by detecting anomalies (TAC3), emergency management (TAC4), unobtrusive activity assistance (TAC5) [13,14], continuous customization in adaptive systems (TAC6), connection with diversified stakeholders (TAC7) [47,48], and finally improved privacy and life independence (TAC8) [49]. The technology characteristics for SHHSs include the technological requirements of smart homes such as ubiquitous (TEC1), real-time (TEC2), and reliable services (TEC3) [24,33], home intelligence (TEC4), home automation (TEC5), data communication based on IoT technologies (TEC6), and remote control and access (TEC7) [3, [50][51][52][53]. The properties of task-technology fit involve the sufficiency and appropriateness of SHHS functions and the fulfillment of user needs, function-wise and quality-wise [24,33,43,45,54].
The measurement items of UTAUT model constructs (i.e., PE, EE, SI, FC, BI, ADT, and CI) were adapted from the validated scales in the relevant studies [24,30,33,36,43,44,55,56]. All the items were measured with an 11-point Likert scale (0, strongly disagree; 5, indecisive; 10, strongly agree). Some research in social studies recommended an 11-point scale, as it enhanced scale sensitivity, normality, and understandability [57,58]. Accordingly, this study also adopted such a scale to capture the level of user perception in fine granularity.

Data Collection
Screening questions were introduced at the beginning of the survey, asking about users' experiences relevant to SHHSs. Five categories (personal health care devices, wearable devices, health information app services, customized health care app services, and telemedicine services [59]) were represented, together with each one's sample products or services. A user who did not check any experience with the suggested SHHSs was automatically dropped from the survey. Demographic questions followed at the end of the survey including gender, age, residential district, marital status, and household members. Moreover, two redundant but reverse-coded questions were added in the middle of the measurement questionnaire (TTF5 and BI5) to exclude inattentive respondents.
As sampling methods, stratified sampling for gender (i.e., the same quota for men and women) and random sampling techniques were adopted. Data collection was conducted with an online survey method through a web-based platform. The survey was randomly distributed by Macromill Embrain Inc. (one of the biggest online survey companies in South Korea) using the company's respondent panels. The response data were collected between 1 February 2021 and 5 February 2021. Inattentive responses were eliminated at the data collection stage by the company according to the reverse-coded questions. The initial responses totalled 503; then, eight outliers were removed according to a normality test. Probably due to the web survey method, the number of participants aged in their 20s and 30s took up a large share (77.2%), and the number of those aged 60 or over was deficient (5 in their 60s and 3 in their 70s). So, those in their 60s and older were excluded because of insufficient samples to represent that age group. Finally, the data for analysis was reduced to 487. Table 3 represents the demographic characteristics of participants.

Statistical Procedure
Data were analyzed using SPSS (version 25, IBM Corp., Armonk, NY, USA) and Smart-PLS (version 3, SmartPLS GmbH, Oststeinbek Germany) software. This study adopted the partial least square structural equation modeling (PLS-SEM) technique to examine the theoretical framework and test the ten hypotheses. Generally, covariance-based SEM is selected when a research goal is theory testing and confirmation or comparison of alternative theories [60]. Meanwhile, the partial least square method, a variance-based SEM, is a good alternative to covariance-based SEM in cases of studies in the early stage of theory building or an extension of an existing structural theory [60][61][62][63]. Moreover, the PLS methods are especially convenient for large and complex research models to test both formative and reflective constructs compared to covariance-based SEM methods [26,62]. Therefore, the PLS technique was appropriate for our research context because the SHHSs are still underexplored from the perspective of task-technology fit and technology acceptance. Moreover, the combination of UTAUT and TTF models generated a complex research framework with many constructs. For these reasons, most of the previous studies introducing the unified model of UTAUT and TTF adopted the PLS-SEM technique due to the complexity of the research model [24,33,44,56,64], so this study also followed the proven method appropriate for this unified model.
Lastly, path coefficients and their significance were estimated by bootstrapping 5000 subsamples [61,67]. Resampling techniques, such as bootstrapping, are necessary to gain the parameters' standard errors because PLS does not assume a particular data distribution [66], and the resample size of 5000 is most frequently adopted for the bootstrapping setting [61]. Finally, path coefficient, t values, and p values were examined to distinguish the relationships between constructs in the structural model. Accordingly, the hypotheses were statistically proven with t values greater than 1.96 and p values lower than 0.01. Table 4 demonstrates descriptive statistics of the measurement items. The mean values ranged from 5.21 (ADT2) to 7.27 (BI3). Overall, TTF and ADT showed a relatively low level of mean values, and TEC and BI included the variables with comparatively high mean values. All measurement items followed a normal distribution with all their absolute values of skewness and kurtosis indicating less than 1.
First, internal consistency reliability requires Cronbach's α, Dijkstra-Henseler's rho, and composite reliability with each value larger than 0.70 for good reliability [68,69]. Table 5 exhibits the analysis of the measurement model. In terms of internal consistency, all values of Cronbach's α, Dijkstra-Henseler's rho, and composite reliability for each construct were larger than 0.70, ensuring good reliability.
Second, convergent validity is determined satisfactory when the outer loading value of each measurement item exceeds 0.70, indicator reliability is greater than 0.50, and the average variance extracted (AVE) of each construct surpasses 0.50 [60,67]. For this model, as shown in Table 5, the outer loading value of each item surpassed 0.70; the indicator reliability of each item was greater than 0.50; the AVE of each construct exceeded 0.50.
Third, discriminant validity is examined using the Fornell-Larcker criterion when the square root of AVE is greater than the highest value among the correlations between construct items [70]. Discriminant validity is also acceptable when the outer loading of each measurement item is higher on its corresponding construct than the cross-loadings on other constructs [67]. Discriminant validity was determined satisfactory using the Fornell-Larcker criterion (Table A1); the square root of AVE of each construct was larger than its correlation coefficients with other constructs. Discriminant validity was also verified as acceptable since the outer loading of each item was higher on its corresponding construct than the cross-loadings on other constructs (Table A2). Accordingly, the measurement model passed the tests of consistency and validity with satisfactory results.

Structural Model Evaluation
The structural model was evaluated by (1) multicollinearity, (2) effect size (f 2 ), (3) coefficient of determination (R 2 ), (4) predictive relevance (Q 2 ), and (5) path coefficients and their significance [65,66]. First, the model is assessed as acceptable without multicollinearity when the variance inflation factor (VIF) between constructs is less than 5 [60]. Multicollinearity of this model was tested with inner variance inflation factor (VIF) values between constructs, as shown in Table 6. All VIF values were less than 5, so the structural model was acceptable without multicollinearity. Second, the effect size between constructs means the relative impact of exogenous variables on endogenous variables. It is assessed as low-level with the f 2 value around 0.02, middle-level with its value around 0.15, and high-level with its value around 0.35 [65]. Table 6 shows the effect size (f 2 ) between constructs of this structural model. The value between TAC and TTF and that between TTF and ADT were assessed insufficient with the values under 0.02, while other values satisfied more than the low-level effect size.
Third, the coefficient of determination (R 2 ) values of endogenous variables show low-level explanation with its value around 0.25, middle-level with its value around 0.5, and high-level with its value over 0.75. In this model, R 2 values of endogenous variables showed an acceptable level except EE (TTF: R 2 = 0.263; PE: R 2 = 0.433; EE: R 2 = 0.193; BI: R 2 = 0.552; ADT: R 2 = 0.449; CI: R 2 = 0.625).
Fourth, the predictive relevance of the PLS-SEM model is evaluated by Stone-Geisser's Q 2 value; when its values of endogenous variables are over 0, the model is considered to have predictive relevance on those variables [71]. In this model, Q 2 values of all endogenous variables represented an acceptable level (i.e., Q 2 > 0) (TTF: Q 2 = 0.221; PE: Q 2 = 0.340; EE: Q 2 = 0.142; BI: Q 2 = 0.497; ADT: Q 2 = 0.398; CI: Q 2 = 0.544). Accordingly, the structural model was generally evaluated acceptable to determine pass coefficients and to test hypotheses.

Hypothesis Testing
The results of hypothesis testing with path coefficients are exhibited in Table 6 and depicted in Figure 2. Task characteristics of SHHSs exerted a slightly negative impact on the task-technology fit, but H1 was not supported with insufficient significance. Technology characteristics of SHHSs had a positive influence on the task-technology fit and supported H2. Task-technology fit of SHHSs directly affected performance expectancy and behavioral intention, thus H3a and H3b were supported. However, the impact of task-technology fit on the adoption of SHHSs was not significant, and H3c was rejected. Performance expectancy, effort effectiveness, social influence, and facilitating conditions had a direct impact on the behavioral intention to use SHHSs, supporting H4, H5, H6, and H7a. The positive impacts of facilitating conditions and behavioral intention on the adoption of SHHSs were significant, thus H7b and H8 were supported. The adoption positively affected the continued intention to use SHHSs, supporting H9. Lastly, technology characteristics of SHHSs positively affected effort expectancy, and H10 was supported.

Hypothesis Testing
The results of hypothesis testing with path coefficients are exhibited in Table 6 and depicted in Figure 2. Task characteristics of SHHSs exerted a slightly negative impact on the task-technology fit, but H1 was not supported with insufficient significance. Technology characteristics of SHHSs had a positive influence on the task-technology fit and supported H2. Task-technology fit of SHHSs directly affected performance expectancy and behavioral intention, thus H3a and H3b were supported. However, the impact of tasktechnology fit on the adoption of SHHSs was not significant, and H3c was rejected. Performance expectancy, effort effectiveness, social influence, and facilitating conditions had a direct impact on the behavioral intention to use SHHSs, supporting H4, H5, H6, and H7a. The positive impacts of facilitating conditions and behavioral intention on the adoption of SHHSs were significant, thus H7b and H8 were supported. The adoption positively affected the continued intention to use SHHSs, supporting H9. Lastly, technology characteristics of SHHSs positively affected effort expectancy, and H10 was supported.

Principal Findings
Concerning the TTF model, the insignificant and negative relationship between TAC and TTF (H1, β = −0.007) and the positive relationship between TEC and TTF (H2, β = 0.517 **) are consistent with previous research in the case of MOOC service acceptance [43]. In several studies on the adoption of mobile banking or wearable devices, the task properties showed a lower significance level of impact on TTF than technology [24,44]. Other studies presented a significant but negative relationship between TAC and TTF [36,72]. Those researches interpreted that TTF decreases as task requirements increase; tasks can become too complex and extensive for technology to support the appropriate level of performance; if technology functionality increases enough to support the task expectations, the TTF increases [36,72]. Likewise, home health care tasks can become too diverse and complex, particularly due to the COVID-19 pandemic. Task characteristics such as real-time and continuous health monitoring and emergency management are critical issues in self-quarantine systems. Meanwhile, SHHS technologies, despite their current advancement, might not sufficiently meet users' expected performance levels. This is also supported by the result that the average value of all TTF measurement items (5.66) was lower than that of TAC (6.47) and TEC (6.80).
Regarding the UTAUT model, the analysis results proved that PE, EE, SI and FC significantly affected BI, explaining 55.2% of BI variances, consistent with the previous evidence of UTAUT studies [24,30,33,43,73]. Remarkably, the most influential factor that affects users' behavioral intention is PE (H4, β = 0.468 **). When SHHS functions meet the users' expected level of performance (i.e., usefulness, relative advantages, and outcome expectation), users' intention to use SHHSs can be influenced. Meanwhile, EE had a relatively low impact on the intention of SHHS acceptance (H5, β = 0.127 **). The reason might be that users have become familiar with using high-tech services, and they may think that adopting SHHS does not require much effort [24]. Additionally, they might accept more effort-cost when SHHS technologies offer expected functions (H10, β = 0.440 **) [40]. The influence patterns of high PE and low EE are also supported by previous studies in the cases of mobile banking [33] and health care wearable devices [24]. Moreover, FC such as the technological infrastructure to ensure a seamless experience of SHHSs (e.g., compatible tools and system circumstances) positively leads to BI (H7a, β = 0.227 **) and the adoption (H7b, β = 0.400 **), in line with previous mobile banking cases [33,44].
The relationship between TTF and UTAUT models is primarily exhibited by the influence of TTF on PE. In this study, the TTF has a strong positive impact on PE (H3a, β = 0.658 **), and TTF explains 43.3% of PE variation; it is also congruent with similar studies' results [33,43,44]. Hence, the fit between users' requirements on home health care and SHHS technologies would directly influence users' expectations of home health care performance. The higher the TTF level is, the more useful the users perceive the SHHS.
Meanwhile, the linkage between the two models is also presented by the influence of TTF on BI and ADT. TTF had a significant but negative effect on the intention to use SHHSs (H3b, β = −0.209 **), and it had an insignificant impact on the adoption of SHHSs (H3c, β = 0.085). Users might agree that smart home technologies for home health care are highly developed, and the requirements for home health care services have become diverse and complex. However, the public might consider that those technologies have not been sufficiently popularized for the home health care services to address their needs and to be adopted. This is also supported by Gartner's hype cycle for emerging technologies in 2018; SHHS relevant technologies (i.e., connected home, IoT platform, and blockchain for data security) was anticipated to reach a mainstream market in five to ten years [74]. So, the insufficient TTF could lead to a negative influence on BI or an insignificant impact on the adoption of SHHS. In a similar instance, the previous two studies on mobile banking adoption, which integrated UTAUT and TTF models, were conducted in 2010 [33] and 2014 [44]. Though the mobile banking technology developed further in 2014, the significant influence of TTF on the adoption in the former was changed as insignificant in the latter. It might be considered that the maturity and popularization of the market service have not been sufficiently achieved compared to technological development and advanced needs.

Contributions and Implications
This research has both theoretical and practical implications. On a theoretical level, firstly, this study supports the applicability of the integrated model of UTAUT and TTF, which recently was explored in technology-intensive industries, to the domain of SHHS. As mentioned in the introduction, the acceptance of SHHS needs to consider both the technical complexity of smart homes and the task pertinence of home healthcare. Therefore, the integrated model of TTF and UTAUT was adapted to establish a comprehensive understanding of the users' adoption behavior of SHHS. A limited number of previous studies combining UTAUT and TTF models engaged with mobile banking, online education, and wearable devices [24,33,43,44], and the relationships between the two models were differently examined by service. Our research model included the possible links from the previous research (i.e., TTF to PE, BI, and ADT; TEC to EE; ADT to CI) and examined their validity in the SHHS case. Hence, the findings explained behavioral acceptance from the perspective of users' perception of task-technology fit and assured the importance of integrating TTF elements into acceptance theories when assessing determinants of user acceptance towards SHHS [24,43].
Second, this study establishes the measurement items reflecting the domain specificity of SHHS. The constructs of task characteristics and technology characteristics should reflect the specificity of relevant industry domains. This research developed the distinct measurement items for SHHS derived from the literature, thus it suggested the full questionnaire set of a unified model of UTAUT and TTF specialized for the user acceptance of SHHS. It can be utilized and adapted to other contexts of SHHS when evaluating the users' adoption behaviors.
Third, this study provides empirical evidence for the research on user acceptance of SHHS during the pandemic era in South Korea. As noted in the introduction, much of the research on SHHS has investigated the technological perspective but lacked the users' perspectives such as their experience and acceptance of SHHS. With the increased need for home health care since the pandemic and the market expansion of SHHSs associated with rapid technological development, research on the actual users' acceptance was required.
In that sense, this study could be meaningful in providing empirical evidence of service acceptance by collecting a large number of actual SHHS users.
On a practical level, these research findings could help home health care service planners and marketers promote SHHS in an individual's health management. With the COVID-19 pandemic, user needs for home health care have become complicated. Despite the current rapid development of related technologies, technological maturity and service popularization do not seem to have been achieved sufficiently for users to realize. When their complex needs can be satisfied through the convergence of various technologies as well as smart home technology (TTF), users would feel that SHHS is useful and beneficial (PE) and this may finally lead to adoption. Therefore, it is recommended to communicate effectively with potential customers how convergent technologies of the service solve the specific needs of the user. Moreover, in terms of FC, strategies to improve technological infrastructure to ensure a seamless experience of SHHSs (e.g., compatible service with diverse usage contexts) would be required to promote user acceptance.

Limitations and Future Research
This study has several limitations. First, alternative technologies or service platforms for home health care may exist. The basic assumption of SHHS in this study is the situation in which home health care services are implemented based on smart home technology. However, users may not necessarily consider only health care services based on smart home technology to achieve the purpose of self-health management. As the needs for personal health care become diversified due to the COVID-19 situation, users could conveniently perform home health care by utilizing various means such as exercise equipment, services, and platforms. As in this study's result, even if the need for home health care tasks is high, it may not sufficiently lead to task-technology fit or acceptance of actual SHHS. Accordingly, it would be necessary to embrace various technologies and service methods for home health care in future studies.
Second, this study was conducted in South Korea, and the research findings might be difficult to generalize due to specificity to Korean users. Generally, South Korea is often considered as a testbed of global ICT brand companies, due to highly developed information technology infrastructure, the nation's innovativeness, and the people's digital friendliness [75,76]. Generally, Korean users may have a high level of technological expectation, and they tend to be early adopters. Moreover, due to the data collection method of the web survey, those in their 60s and older were excluded because of insufficient samples. Hence, the respondent group of this study might have a tendency to high-level innovativeness. They might consider that the task-technology fit level is unsatisfactory to adopt SHHSs more than another country's citizens would. Therefore, comparative studies targeting users from other countries may be required to validate the generalizability of this research model.
Third, this study did not consider any comparative study by user group or by period because the primary research aim was to investigate the applicability of the unified model in the SHHS domain. However, in terms of the user group, demographic variables such as gender, age, or the level of digital literacy can be included as moderators in the research framework. Especially, the acceptance behavior of SHHS might be influenced by the level of individual digital literacy or innovativeness considering the technology-intensive service characteristics. In terms of period, research to capture users' attitudinal changes toward SHHS could be a meaningful opportunity to validate the unified model. The empirical survey of this study was conducted in the middle of the COVID-19 period and the results could be limited to representing the context only during the pandemic period. Therefore, further studies to compare the users' behavioral changes between the periods before and after the pandemic would be valuable.

Conclusions
This study aimed to analyze the influential factors and measure the behavioral acceptance of SHHSs in South Korea. This study adopted the integrated model of UTAUT and TTF and contextualized it, involving SHHS characteristics to explain the behavioral acceptance of SHHSs from users' perceptions and task-technology fit. The results indicated that our integrated acceptance model explained 55.2% of the variance in behavioral intention, 44.9% of adoption and 62.5% of the continuous intention to use SHHSs, supporting 11 of the 13 proposed hypotheses. Users' perceptions of PE, EE, SI and FC can positively anticipate BI to accept SHHSs, having relatively strong influence from PE and FC. Furthermore, TTF has an influence on PE and BI, and indirectly on ADT and CI; it implies that the consideration of matching fit of task and technology, as well as user perceptions of technologies, is essential to evaluate the acceptance of SHHSs [34]. Meanwhile, TAC was insignificant to determine TTF, which might stem from the complicated needs of home health care due to the COVID-19 pandemic but was not sufficiently resolved by current service technologies. The findings implied that the acceptance of SHHSs needs to be evaluated with consideration of both the user perceptions on technologies and the matching fit of task and technology. Theoretically, this study supports the applicability of the integrated model of UTAUT and TTF to the domain of SHHS, and newly proposed measurement items of TTF reflecting the domain specificity of SHHS, providing empirical evidence during the pandemic era in South Korea. Practically, the results could suggest to the planners and strategists of home health care services how to promote SHHS in health management.

Institutional Review Board Statement:
The study was conducted in accordance with the Declaration of Helsinki, and approved by the Institutional Review Board (or Ethics Committee) of Sungshin Women's University (protocol code SSWUIRB-2020-005, initial approval on 6 March 2020).

Informed Consent Statement:
The ethics committee waived the need for consent because they approved that the risk to the study subjects was expected to be extremely low even if consent was exempted for the following reasons. First, no information that can identify participants (e.g., name, identifiers, titles, email address, contact, etc.) was included in the questionnaire, to keep the respondents' anonymity. Second, it was practically impossible to obtain consent from the participants because the online survey company randomly distributed the survey using their respondent panels. Meanwhile, all the required information, such as the objectives, confidentiality, anonymity of the survey, and the option not to answer any question the respondent felt infringed on their privacy, was described on the first page of the questionnaire. Data Availability Statement: Not applicable.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results. The square root of AVE of each construct (in bold) was larger than its correlation coefficients with other constructs. The outer loading of each item was higher on its corresponding construct (in bold) than the cross-loadings on other constructs.