Digital natives' privacy concerns in eHealth platforms Open Access

The COVID-19 pandemic catalyzed the rapid expansion of digital healthcare platforms worldwide, which was driven by heightened health demands and restrictions on physical movement (Wan, Zhang, & Yan, 2020; Pereira, Tavares, & Oliveira, 2023). eHealth refers to the delivery of healthcare services and health-related information through Internet-connected electronic devices such as smartphones and computers (Mannan, Ahamed, & Zaman, 2019; Verma, Kumar, & Sharma, 2020). As a form of telemedicine, eHealth reduces the need for physical visits and overcomes geographical barriers by enabling users to access services remotely (Pereira et al., 2023). Traditional healthcare systems were overwhelmed during the pandemic, and access to in-person services was restricted due to lockdown procedures (Mustapha, Khan, Qureshi, Harasis, & Van, 2021).

The pandemic undoubtedly amplified the use of eHealth platforms, making them the preferred option for healthcare delivery due to their convenience and safety (Lin, Carter, & Liu, 2021). As a result, digital health technologies became essential in maintaining the continuity of healthcare services and helping to reduce the strain on overworked practitioners. According to Statista (2024), the global digital health market experienced a significant increase in the number of users, from 1.17 billion in 2020 to 1.79 billion in 2023. The digital health industry continues to expand even in the post-pandemic era, with a projected 36.14% increase in revenue from 2024 to 2028 (Statista, 2024).

The rise of Healthcare 4.0, which combines cutting-edge technologies like cloud computing, big data analytics, artificial intelligence (AI), and the Internet of Things (IoT), has transformed digital healthcare by enabling the collection, storage, and sharing of real-time data to enhance service delivery (Wong & Hazley, 2020). Younger generations, especially millennials and Gen Z, are more likely to embrace these technologies because of their familiarity as digital natives (Munsch, 2021; Pramudita, Achmadi, & Nurhaida, 2023). Having grown up with technology, they are naturally more adept at using eHealth platforms than previous generations due to their upbringing with it. Although they are highly adaptable to new technologies, this demographic is increasingly aware of the complex privacy risks involved in the online sharing of personal health data, particularly in terms of collecting, storing, and managing data (Lin et al., 2021; Sureani, Awis Qurni, Azman, Othman, & Zahari, 2021). While there is extensive literature on privacy concerns in digital environments, there is limited focus on how digital natives perceive and respond to privacy issues in the context of eHealth platforms, specifically those in developing countries (Adu, Mills, & Todorova, 2020). This gap is crucial, as privacy concerns can significantly affect users' behavioral disposition to engage with digital health services (Smith, Dinev, & Xu, 2011; Prakash & Das, 2022).

Health-related information is among the most sensitive forms of personal data and is subject to confidentiality (Esmaeilzadeh, 2018, 2024). The increasing use of eHealth platforms raises concerns over their management of information privacy, particularly given the fragmented nature of healthcare databases across different facilities in Malaysia (Ministry of Health Malaysia, 2021). Understanding the privacy concerns of digital natives, who are both frequent users and are critically engaged with digital health technologies, is essential for enhancing security, transparency, and long-term user trust in existing eHealth platforms.

This study aims to examine how privacy concerns among digital natives influence their attitudes toward eHealth platforms. Due to their continuous exposure to technology, digital natives are often aware of privacy concerns and tend to take proactive measures to protect their personal data (Adu et al., 2020). This heightened awareness reflects a growing need for privacy, which in turn influences their interactions with digital services, including eHealth platforms. Considering the scope of the study, the following research questions are posed:

Studies on information privacy have introduced several conceptual instruments to understand privacy concerns (Kim, Kim, Peterson, & Choi, 2023). Among them, scholars widely reference the CFIP model (Smith, Milberg, & Burke, 1996), IUIPC model (Malhotra, Kim, & Agarwal, 2004), and IPC model (Dinev & Hart, 2006). The CFIP is particularly recognized for its adaptability in addressing privacy issues across a wide range of organizational settings, both online and offline. Conversely, IUIPC and IPC primarily focus on Internet-based settings. This study adopts the CFIP as it offers a generalized framework for evaluating organizational privacy practices. Although the IUIPC model demonstrates solid measurement validity, CFIP has gained wider recognition as a de facto standard across multiple disciplines (Bélanger & Crossler, 2011). According to Wang, Sun, Dai, Zhang, and Hu (2019), the CFIP effectively addresses consumers' cognitive and attitudinal responses, making it suitable for contexts like digital health platforms. Unlike banking or e-commerce, healthcare often lacks transparency in data practices, leading to heightened privacy concerns due to involuntary data sharing (Princi & Krämer, 2020; Bessenyei et al., 2021). Table 1 summarizes recent studies (2020–2024) that applied the CFIP framework in healthcare contexts.

Table 1 highlighted mixed findings regarding the relationship between CFIP dimensions and behavioral outcomes. Some studies confirm significant relationships (e.g. Hwang & Lin, 2020; Gimpel, 2021; Prakash & Das, 2022), while others found weaker correlations or no significant effects (e.g. Princi & Krämer, 2020; Pereira et al., 2023). This inconsistency may stem from contextual factors (e.g. type of healthcare technology or demographic) or model operationalization. Moreover, most studies focused on intention rather than attitude and overlooked sub-dimensional distinctions within CFIP.

With robust psychometric validation (van der Schyff, Flowerday, & Lowry, 2020), CFIP consists of four key dimensions: information collection, unauthorized secondary use, improper access, and data errors (Smith et al., 1996). This study adapts CFIP into two second-order constructs: information collection concern and information management concern, in accordance with Stewart and Segars (2002). Tseng, Hung, Hwang, and Chang (2020) and Adu et al. (2020) further validated the model in healthcare context. The information collection concern reflects the information collection dimension, while the information management concern encompasses the dimensions of unauthorized secondary use, improper access, and data errors.

Stewart and Segars (2002) highlighted that users are concerned with the extent of information collection and how the information is managed by organizations. While users may exert some control over data disclosure at the collection stage, this control diminishes once data is submitted, placing the responsibility for secure storage and management on healthcare providers (Adu et al., 2020). In turn, privacy concerns tend to escalate when consumers become aware of poor data management practices, particularly those involving unauthorized access or secondary use (Bessenyei et al., 2021).

Guided by the Antecedents–>Privacy Concerns–>Outcomes (APCO) framework (Smith et al., 2011), information privacy orientation (IPO) was posited as the antecedent and attitude toward eHealth platforms as the outcome. The APCO framework presents a progressive and structured framework for understanding the relationships between antecedent factors and privacy concerns. IPO refers to an individual's efforts to control access to their data (Mpinganjira & Maduku, 2019; Adu et al., 2020). Dienlin and Metzger (2024) emphasized a strong association between IPO and privacy concerns. Given the sensitivity of health information, individuals with strong IPO are likely to show heightened privacy concerns, especially in relation to data collection and management in eHealth platforms. As a result, users may become more protective of their personal data and resist unnecessary disclosure. This highlights the relevance of examining whether IPO influences privacy concerns, particularly regarding information collection and management, where personal health data is highly sensitive and private. Therefore, the following hypotheses are proposed:

In reference to Table 1, prior studies have examined the influence of these concerns on behavioral dispositions. For example, CFIP dimensions influence user intention in healthcare settings (Dhagarra, Goswami, & Kumar, 2020; Tseng et al., 2020). Individuals with high privacy concerns were reluctant to disclose specific health details, including confidential health information and chronic health diseases (Esmaeilzadeh, 2024). In the context of digital natives, users are more cautious and remain highly aware of potential risks in data disclosure and data management by healthcare organizations, despite generally being willing to share health information for healthcare services (Adu et al., 2020). A notable gap in the literature is that many studies focus on intention, often overlooking attitude, a critical precursor that influences behavior (Princi & Krämer, 2020). Integrating attitude as an outcome of privacy concerns provides a clearer understanding of how these concerns influence attitudes toward eHealth platforms. Although digital natives may engage with eHealth platforms, concerns over information collection and management may result in negative attitudes. Hence, this study hypothesized:

Two demographic variables, age and gender, are included as controls due to their potential influence on privacy-related behavioral dispositions (Esmaeilzadeh, 2018, 2024). Literature suggests that women generally exhibit higher privacy concerns than men (Hsu, 2006; Fox & James, 2021), while older individuals express greater concerns due to health conditions and lower digital literacy (Adu et al., 2020; Henkenjohann, 2021). This heightened concern may stem from their greater susceptibility to chronic and sensitive health conditions. Although digital natives are often considered technologically adept, differences in technological immersion between millennials and Generation Z (Pramudita et al., 2023), are still likely to influence attitudes toward digital health systems. Figure 1 illustrates the proposed research model.

The survey was distributed via web-based platforms, and a quantitative research design was adopted consistent with prior studies. Data analysis was conducted in two stages: (1) preliminary data cleaning using SPSS and (2) measurement and structural model assessment using Partial Least Squares-Structural Equation Modeling (PLS-SEM) via SmartPLS.

Purposive sampling was employed with predefined criteria; participants were Malaysian Internet users aged 18 or older with prior experience using eHealth platforms (Schindler, 2022). Given that this study focuses on privacy concerns in eHealth platforms, Internet users were considered an appropriate population, as access to these platforms necessitates the use of Internet-connected devices. The sample size was estimated using G*Power software, which indicated a minimum requirement of 68 respondents (Hair, Hult, Ringle, & Sarstedt, 2017).

Table 2 summarized the sources of measurement items used in this study. Detailed measurement items can be found in Supplementary File. The reliability and validity of the survey instruments were ensured through adaptations from prior rigorously tested studies in similar contexts. A bilingual survey with forward and backward translation was used to ensure comprehensibility of the items. Language and academic experts reviewed the items according to Yusoff (2019), and their feedback was used to revise wording and improve clarity before the pilot test. All items were reflectively modeled and assessed using a 5-point Likert scale (1 = strongly disagree to 5 = strongly agree), consistent with prior CFIP studies (Smith et al., 1996; Stewart & Segars, 2002).

A total of 383 responses were collected. The dataset underwent systematic data cleaning and preparation to ensure validity, including removal of responses failing filter questions, straight-lining patterns (SD = 0) (Tan, Chan, Tan, Wong, & Chaichi, 2023), missing values, and coding inconsistencies (Schindler, 2022). Straight-lining, identified as a threat to validity, was removed to enhance data quality. Responses with a standard deviation (SD) lower than 0.25 were also removed as minimal variance across items may indicate response bias (Collier, 2020).

The Mahalanobis distance approach (p < 0.001) was applied to identify multivariate outliers (Ringle, Sarstedt, Sinkovics, & Sinkovics, 2023). Nine outliers were excluded based on this criterion (Hair, Anderson, Babin, & Black, 2010), leaving 339 valid responses for analysis. Figure 2 details the screening steps and number of responses removed at each stage.

Table 3 summarizes the respondents' demographic profile. All respondents are digital natives, with a slight majority from the millennial generation (ages 27–42, 51.3%) and the remainder from Generation Z (ages 18–26, 48.7%). Given that both generations are known to spend the most time online (Malaysian Communications and Multimedia Commission, 2022), there is no substantial difference in the number of respondents. The number of female respondents (58.4%) is slightly higher than male respondents (41.6%), consistent with prior research where females are most likely to participate in questionnaires (Prakash & Das, 2022; Pereira et al., 2023). In terms of education, the highest responses hold a bachelor's degree (57.5%), followed by a diploma (17.7%), postgraduate (13.6%), and high school degree (7.4%). The remaining 3.8% hold other qualifications, such as matriculation or professional certificates.

The measurement model was assessed for reliability and validity using four criteria: outer loadings, reliability, convergent validity, and discriminant validity. Firstly, the items' outer loadings were examined. Hair, Risher, Sarstedt, and Ringle (2019) recommended that item loadings should exceed 0.708, indicating that the construct accounts for over 50% of the indicator's variance. This cut-off value led to the removal of five items from the assessments: IPO1, SU1, CL6, CL7, and ER5. Additionally, item SU2 was discarded because of its high cross-loading on both the SU and IA constructs.

Table 4 indicates excellent composite reliabilities with IPO (CR = 0.877), CL (CR = 0.929), SU (CR = 0.929), IA (CR = 0.911), ER (CR = 917), and AT (CR = 0.919). All values are lower than the maximum threshold of 0.95, avoiding potential compromise of construct validity (Hair et al., 2010). The convergent validity for the constructs was evaluated based on average variance extracted (AVE). Table 4 shows that all first-order constructs had acceptable AVE values, exceeding 0.5 (Hair et al., 2010). For the higher-order construct IMC, both CR (0.944) and AVE (0.532) exceeded the acceptable thresholds, confirming reliability and convergent validity.

The discriminant validity was examined using the heterotrait-monotrait ratio of correlations (HTMT). Table 5 depicts the HTMT values of the constructs, all of which are below the threshold of 0.85 recommended by Henseler, Ringle, and Sarstedt (2015). This indicates a strong discriminant validity. As summarized in Tables 4 and 5, all constructs were within the acceptable threshold.

To further describe the CFIP constructs, Table 4 presents the mean values for each dimension. Item means above 3 on a 5-point Likert scale indicate that respondents express concern for information privacy (Hwang, Han, Kuo, & Liu, 2012). The mean for information collection is lower (3.42), suggesting moderate concern about how eHealth platforms collect personal information. Following this, respondents express stronger concerns on other dimensions, including unauthorized secondary use (4.78), improper access (4.66), and errors (4.54), where the mean values approaching 5. These results are consistent with prior studies where information collection typically has a lower mean than the other CFIP dimensions (Tseng et al., 2020; Adu et al., 2020).

Collinearity issues are determined by examining the VIF value where Hair et al. (2019) suggested a VIF value lower than 3 in order for the constructs to be free from collinearity issues. Table 6 shows all VIF values for all constructs were lower than the suggested threshold value; therefore, no collinearity issues were reported.

The structural model assessments were applied according to a bootstrapping procedure suggested by Hair et al. (2019). The procedure ran using a 5,000 resampling technique. Two control variables, gender and age, were included to examine their influence on digital natives' attitudes toward eHealth platforms. The results revealed that gender differences do not influence the attitude towards eHealth platforms. This is consistent with prior studies in the healthcare privacy concern context (Hwang et al., 2012; Esmaeilzadeh, 2024), which reported that gender had an insignificant influence. Hwang et al. (2012) argued that individuals disclose comprehensive information in exchange for healthcare services, regardless of their gender.

In contrast, Generation Z had a stronger significant influence on attitude toward eHealth platforms compared to millennials (β = 0.375, p < 0.001). Although this contradicts Henkenjohann (2021), a plausible reason is that both cohorts are technologically savvy and aware of privacy issues in health informatics (Rahman et al., 2021; Prakash & Das, 2022). While millennials have adapted to technological advancements, Generation Z has grown up with them, making them more conscious of organizational information collection and management practices. In turn, this significantly influence their attitude toward using health applications (Pramudita et al., 2023).

Using the same bootstrapping procedure, the results revealed that IPO positively influenced concerns about information collection (β = 0.239, p < 0.001) and information management (β = 0.418, p < 0.001), supporting H1a and H1b. The results showed that information collection concerns negatively influenced digital natives' attitudes toward eHealth platforms (β = −0.209, p < 0.001), providing empirical support for H2a. In contrast, H2b was not supported, as information management concerns were positively related to attitude (β = 0.458, p < 0.001), contrary to the proposed hypothesis. These results support the empirical validity of assessing CFIP as a reflective construct using a two-factor model, which is consistent with previous studies (Stewart & Segars, 2002; Tseng et al., 2020; Adu et al., 2020).

To further evaluate the strength of the relationships, the effect size (f²) values were examined. The results indicate small effect sizes between IPO and CL (f² = 0.061), and medium effect sizes between IPO and IMC (f² = 0.211). Meanwhile, CL and AT demonstrated a small effect size (f² = 0.056). The model explains 24.2% of the variance in digital native attitude toward eHealth platforms. Overall, the findings support the proposed model, with three out of four hypotheses supported. A summary of hypothesis testing, VIF values, f² and R² values is presented in Table 6, while Figure 3 displays the structural model assessment.

To reinforce the validity of the main findings, robustness tests were performed in PLS-SEM by re-estimating the structural model with alternative model specifications (Hair, Howard, & Nitzl, 2020). Following Neumayer and Plümper (2017), robustness checks through the inclusion or exclusion of control variables are crucial for assessing the stability of quantitative research findings. Accordingly, three alternative model specifications were tested sequentially: Model 1 included gender as a control variable, Model 2 included age as a control variable, and Model 3 was estimated without any control variables. The results, presented in Table 7, consistently replicated the main structural model reported in Table 6. Specifically, H1a, H1b, and H2a remained supported, while H2b consistently remained unsupported due to its significant positive effect, which was opposite to the hypothesized direction. No collinearity issues were detected (Hair et al., 2019). These findings suggest that the hypothesized relationships are robust, as they hold across alternative model specifications (Hair et al., 2020). Detailed robustness test results are provided in Supplementary File.

This study aims to investigate the determinants of privacy concerns among digital natives in Malaysia and to examine how these concerns reflect their attitudes towards eHealth platforms. Drawing on the guidelines provided by Stewart and Segars (2002), Tseng et al. (2020), and Adu et al. (2020), the research model conceptualizes CFIP as two sub-factors, namely information collection concerns and information management concerns.

As expected, the findings confirm the positive relationship between IPO and concerns about information collection and information management, supporting hypotheses H1a and H1b. This corroborates the findings from previous studies that IPO is the determinant for privacy concerns (Xu et al., 2011; Dienlin & Metzger, 2024). The association between the disposition to protect privacy and privacy concerns has been established on many occasions, indicating that individuals who have a strong desire for privacy tend to be more cautious when disclosing personal information (Mpinganjira & Maduku, 2019). The nature of health information, which is considered sensitive and confidential, is prone to triggering negative stigma in society if the data is violated (Esmaeilzadeh, 2018; Gimpel, Manner-Romberg, Schmied, & Winkler, 2021).

According to Taylor et al. (2015), the social-psychological factor, IPO, is propelled by the risk avoidant behavior and intellectual factor. Xu et al. (2011) emphasized that there is a certain risk associated with disclosing personal information. This motivates individuals to engage in protective behaviors to safeguard their personal health information (Kim et al., 2023). Individuals with a high privacy orientation are highly likely to be well-aware and knowledgeable about the potential risks and consequences, which prompts them to adopt a critical and discerning stance toward both information collection and information management (Prakash & Das, 2022). This reflects the characteristics of digital natives who are digitally fluent and attentive in privacy issues (Rahman et al., 2021). The unruly practice of data collection across healthcare facilities will cause them to be critical and vigilant about health information collection in eHealth platforms (H1a). The same reasoning applies, as frequent issues arising from irresponsible data management by organizations have made digital natives more vigilant toward data management in eHealth platforms (H1b). For instance, in the third quarter of 2023, Surfshark identified Malaysia as the eighth most breached country, citing the country's history of data mismanagement issues (Yeoh, 2023). Continuous media attention on security failures, privacy threats, and the poor governance of personal data has intensified the awareness of privacy threats (Smith et al., 1996). For digital natives, who have ubiquitous access to information, such coverage plausibly strengthens their desire to protect personal data (Xu et al., 2011).

Additionally, the study found that concerns about information collection negatively influence the attitude towards eHealth platforms, thereby supporting H2a. The findings provide empirical support for a prior study (Taylor et al., 2015). One probable reason is the lack of an integrative system between the public and private sectors in healthcare settings, thereby contributing to data redundancy each time users visit or sign in to different healthcare platforms (Reegu et al., 2023). This occurs when the same piece of information is stored in different standalone health information systems (Razak, Ithnin, & Osman, 2020), as highlighted by Hassan's (2020) observation of the government's alarming practice of collecting the same data for various agencies during the pandemic. Unfortunately, the disorderly and ineffective data collection procedure by healthcare providers has become a norm among digital natives (Rahman et al., 2021), causing them to express concerns about information collection and subsequently influencing their attitudes towards eHealth platforms.

Unexpectedly, concerns about information management were found to have a positive significant effect on attitude. The finding indicates that consumers are more likely to be positively inclined toward eHealth platforms when they have concerns regarding information management. In general, information management refers to the practice of an organization storing, organizing, and maintaining data (Stewart & Segars, 2002). Although this finding contrasts with the proposed direction, it is similar to the results of Adu et al. (2020), who discovered that the dimensions of unauthorized secondary use, improper access, and errors positively impacted information management concerns. A possible reason is due to a loss of control, as argued by Adu et al. (2020). Compared to the information collection process, consumers remain in control of the information that is disclosed to receive proper medical treatment (Princi & Krämer, 2020). However, the disclosure of information exposes consumers to the risk of data mishandling.

Although the mean values of the three dimensions (see Table 4) suggest that consumers are concerned about information management, they often remain vulnerable and disadvantaged by healthcare providers. In practice, they are compelled to accept potential risks of data mismanagement in exchange for healthcare services, particularly when their health conditions are deteriorating. Even with their worries and anxiety, consumers still need to seek healthcare services from the platforms, thereby explaining the results reflected in this study. Hence, H2b is not supported.

This study offers a number of theoretical and practical contributions. From a practical standpoint, one key implication lies in the urgent need for government-led intervention in health data infrastructure. The government should allocate sufficient resources and develop a strategic framework for implementing an interoperable health data sharing system across public and private healthcare facilities. This can effectively reduce excessive data collection, minimize data fragmentation, and ensure more organized and coherent data management (Akhtar, Khan, Qayyum, Qureshi, & Hishan, 2022).

Despite prior efforts, Razak et al. (2020) reported that only 15.2% of 139 public hospitals in Malaysia have adopted system-based records, with the majority still relying on paper-based documentation. This indicates that the national rollout of hospital information systems (HIS) is delayed considerably (Rahi, Khan, & Alghizzawi, 2020). Inconsistent HIS implementation leads to data redundancy and compromises the overall integrity of health information (Reegu et al., 2023). To address this issue, the adoption of blockchain-based data storage and sharing, overseen by the government, has been proposed to safeguard the integrity and confidentiality of health records (Akhtar et al., 2022; Reegu et al., 2023).

Furthermore, healthcare facilities and platform developers that manage digital health platforms should prioritize investment in robust cybersecurity infrastructure (Bessenyei et al., 2021; Esmaeilzadeh, 2024). Strengthening security measures is essential to prevent unauthorized access and mitigate risks associated with the secondary use of personal health data. Weak security protocols potentially expose sensitive health information to unauthorized access and mismanagement (Reegu et al., 2023). Thus, the implementation of stronger cybersecurity measures, such as encryption, multi-factor authentication, and access controls, can enhance the trust and promote wider adoption of the platforms.

This study provides theoretical contributions to digital natives' privacy concerns within the healthcare context by highlighting privacy orientation as a potential antecedent of CFIP. It offers insights into how digital natives in Malaysia experience privacy concerns. Furthermore, the finding supports the application of the two-factor CFIP structure modeled by Stewart and Segars (2002), which aligns with prior studies (Hwang et al., 2012; Adu et al., 2020). Mean analysis (see Table 4) further supports that the collection dimension is conceptually distinct from the other three dimensions (secondary use, improper access, and errors), suggesting that digital natives' concerns about information collection differ significantly from concerns related to information management after disclosure (Tseng et al., 2020).

First, there may be some sampling bias, as most respondents were digital natives from Selangor, one of the most developed states in Malaysia. Thus, the findings might not fully reflect the experiences or perspectives of people in less urbanized or rural areas. Future research could include diverse group of respondents to better understand how people in different parts of the country engage with digital health platforms.

Second, digital health infrastructure in Malaysian healthcare is still in its early stages, and eHealth platforms are not yet widely implemented or accessible across all healthcare facilities (Rahi et al., 2020). As a result, this may influence the perspectives of the general population (Verma et al., 2020). Future research could explore potential differences in privacy concerns between these two sectors.

Another limitation is the model's low explanatory power, explaining only 24.2% of the variance. While values between 10% and 50% are generally acceptable in social science research (Ozili, 2022), Hair, Ringle, and Sarstedt (2011) regard 24.2% as weak. This indicates that the model may overlook salient variables, and future studies should incorporate additional determinants to enhance its explanatory strength and provide additional insights.

In conclusion, this study highlights the complex role of privacy concerns in influencing digital natives' attitudes toward eHealth platforms. By addressing the identified limitations and extending future research, scholars and healthcare stakeholders can better understand and manage privacy challenges to support wider adoption of digital healthcare solutions in Malaysia.

All authors contributed to the study review and design. Material preparation and data collection was performed by Umi Umairah. Data analysis was performed by Umi Umairah, Ponmalar and Zuha Rosufila. The first draft of the manuscript was written by Umi Umairah and all authors commented on previous versions of the manuscript. Noor Fadhiha revised the final version of the manuscript. Lastly, all authors have read and agreed to the final manuscript.

Informed consent was obtained from all subjects involved in the study.

This study sincerely acknowledges the reviewers for their constructive criticism and suggestions.

The supplementary material for this article can be found online