International Journal of Information Management xxx (xxxx) xxxx
Contents lists available at ScienceDirect
International Journal of Information Management journal homepage: www.elsevier.com/locate/ijinfomgt
Big data management in healthcare: Adoption challenges and implications Peng-Ting Chena,*, Chia-Li Linb, Wan-Ning Wuc a b c
Department of Biomedical Engineering, National Cheng Kung University, No.1, University Road, Tainan City 701, Taiwan, ROC Department of Marketing Management, Shu-Te University, No.59, Hengshan Rd., Yanchao, Kaohsiung 824, Taiwan, ROC Department of Information Management, National Sun Yat-sen University, 70 Lienhai Rd., Kaohsiung 804, Taiwan, ROC
ARTICLE INFO
ABSTRACT
Keywords: Healthcare information system Medical big data Organizational barrier Adoption strategy Multiple Criteria Decision Making (MCDM)
The computerized healthcare information system has undergone tremendous advancements in the previous two decades. Medical institutions are paying further attention to the replacement of traditional approaches that can no longer handle the increasing amount of patient data. In recent years, the healthcare information system based on big data has been growing rapidly and is being adapted to medical information to derive important health trends and support timely preventive care. This research aims to evaluate organization-driven barriers in implementing a healthcare information system based on big data. It adopts the analytic network process approach to determine the aspect weight and applies VlseKriterijumska Optimizacija I Kzompromisno Resenje (VIKOR) to conclude a highly appropriate strategy for overcoming such barriers. The proposed model can provide hospital managers with forecasts and implications that facilitate the withdrawal of organizational barriers when adopting the healthcare information system based on big data into their healthcare service system. Results can provide benefits for increasing the effectiveness and quality of the healthcare information system based on big data in the healthcare industry. Therefore, by understanding the sequence of the importance of resistance factors, managers can formulate efficient strategies to solve problems with appropriate priorities.
1. Introduction Traditional and manual approaches for recapitulating medical data limit the capability of data storage and analysis in hospitals and clinics. To overcome this limitation, many medical institutions have exerted considerable efforts to combine the use of large data resources and innovative technologies (Becker, 2017). Electronic healthcare records (EHRs), which use big data analytics for major evaluations of diseases and performance of epidemiological analyses, can be regarded as a breakthrough in medical information management (Hännikäinen, 2017; Perera et al., 2016). Despite endeavors to construct successful big data systems, numerous medical institutions have encountered early failure when adopting these new systems. The present study aims to investigate the organizational barriers that prevent medical institutions from implementing a successful big data system by identifying and evaluating the barriers and providing managers with strategic solutions to these obstacles. Big data is generally acknowledged as an explication for the management process of various organizations across industries (Ziora, 2015). Hence, a large number of institutions have experienced early failure in the application of medical big data systems because of predictable and unpredictable hurdles. Barriers to such systems may ⁎
originate from various factors, such as organizations, doctors, patients, or governments. This study specifically focuses on identifying the internal barriers that stem from the adopters, which are particularly indicated as the organizational side. On the basis of the barrier framework of innovation resistance theory (Sheth & Ram, 1987), we determine the barriers of medical big data. In particular, we explore the healthcare industry of Taiwan through semi-structured interviews with important experts directly associated with the adoption of medical big data systems, including physicians, medical staff, and scholars. We also conduct a literature review to identify the previously indicated barriers in different studies to facilitate the comparison of inherent barriers in practice and those mentioned in theories. Growing scrutiny on Taiwan by various governments and the public makes investigating the country worthwhile. More than 15,000 clinics in Taiwan adopt EHRs, which represent an enormous medical database covering the country’s entire population. Every Taiwanese patient visits his or her doctor for consultation 15 times per year on average, and the reproduction of laboratory tests and prescriptions is considerable. Therefore, the construction and adoption of EHR systems in the field of big data are of utmost importance and necessity. The remainder of this paper is structured as follows. Section 2
Corresponding author. E-mail address:
[email protected] (P.-T. Chen).
https://doi.org/10.1016/j.ijinfomgt.2020.102078 Received 13 September 2018; Received in revised form 15 December 2019; Accepted 17 January 2020 0268-4012/ © 2020 Elsevier Ltd. All rights reserved.
Please cite this article as: Peng-Ting Chen, Chia-Li Lin and Wan-Ning Wu, International Journal of Information Management, https://doi.org/10.1016/j.ijinfomgt.2020.102078
International Journal of Information Management xxx (xxxx) xxxx
P.-T. Chen, et al.
reviews previous studies on topics such as the healthcare environment in Taiwan, hospital information systems (HIS), and medical big data and its inherent barriers to establish the research background of this study. Section 3 illustrates the applied methods, including expert interviews, Analytic Network Process (ANP), and Vlse Kriterijumska Optimizacija I Kompromisno Resenje (VIKOR). Section 4 explains the practical analysis and results of the collected data. Section 5 provides implications for managers of medical institutions and the limitations and direction for future research. Finally, Section 6 presents the conclusions.
Nevertheless, adopting health information technologies has been considered a challenge in the context of hospitals and clinics. 2.3. Big data in healthcare Boyd and Crawford (2012) describe big data as a high-volume, highvelocity, and/or high-variety information asset that demands cost-effective and innovative forms of information processing, thus enabling enhanced insight, decision making, and process automation. Big data has been applied in many fields to gradually replace the traditional database management system for facilitating the adequate management and analysis of massive data (Boyd & Crawford, 2012). Archenaa and Anita (2015) divide the life cycle of big data into five major phases, namely, data collection, cleaning, classification, modeling, and delivery. Big data for healthcare, commonly known as medical big data, plays a particularly crucial part in the decision-making process of many hospitals and medical clinics (Chen, 2018; Wang, Liu, & Hong, 2016). Martin-Sanchez and Verspoor (2014) state that medical institutions create information assets with high volume, velocity, value, and diversity through healthcare data resources. This type of data should be processed through suitable approaches to enhance the decision-making process (Martin-Sanchez & Verspoor, 2014). Medical big data is diverse in type and form; typically unstructured and complicated, they require a solid management system capable of enhancing their effective use (Chen, 2018). The HIS can solve the decision problem of medical diagnosis and healthcare for medical staff and patients on the basis of the domain expertise, algorithms, and a considerable amount of highquality data (Holzinger, Kieseberg, Weippl, & Tjoa, 2018). Given that big data is a complex and large-sized dataset, this type of database is difficult to process using traditional data-processing tools. The data mining process of big data can extract valuable information from largesized datasets. In addition, association rule mining is useful in the datamining process and can establish a set of association rules for determining the association between items. The study proposes a new algorithm of neutrosophic association rule for generating association rules and adopts the algorithm to process the indeterminacy, membership, and non-membership functions of items. The research finding indicates that the new algorithm can increase the number of generated association rules and enhance decision quality (Abdel-Basset, Mohamed, Smarandache, & Chang, 2018). Many organizations seek to utilize big data to serve humanity well and solve various problems, such as proliferating medical costs, unemployment, natural disasters, and terrorism (Jee & Kim, 2013a). Machine learning (ML) is a popular field in computer science, and novel algorithms provide many new opportunities in the application of recommender systems, speech recognition, and autonomous vehicles. However, with the lack of high-quality big data in the healthcare domain, the development of healthcare information faces considerable challenges. Nevertheless, interactive ML (iML) can improve the insufficiency of high-quality big data through the advantages of reinforcement, preference, and active learning. Although iML is not yet well adapted, it can be trained by agents or humans. Therefore, computationally difficult problems can be solved by human-in-the-loop based on the assistance of a human agent (Holzinger, 2016). Moreover, big data faces many challenges related to its complexity, security, and patient privacy. Although automatic ML can solve decision problems without human intervention, humans can occasionally aid in solving computationally difficult problems based on human-in-the-loop. The present study proposes new experimental insights for improving computational intelligence via human intelligence in the iML. It provides the ant colony optimization (ACO) framework to explain multi-agent approaches in the loop with human agents. ACO can also be applied in the healthcare field (Holzinger, Plass et al., 2018). The visualization technology can adopt computational intelligence to simulate medical imaging and achieve gene and protein simulations for cancer development and immunity. This study proposes the MapReduce framework
2. Literature review 2.1. Healthcare environment in Taiwan Along with the dynamic development of the economy, Taiwan initially embarked on its healthcare reform in the early 1980s (Liu, Hwang, & Chang, 2011). The government has exerted considerable efforts to learn from the experiences of many countries with professionally and continuously advanced healthcare systems, such as the United States and Japan, prior to establishing its unique system. Wu, Majeed, and Kuo (2010) report that the formation of the National Health Insurance (NHI) model in 1995 was the first attempt of the country to transform the healthcare industry by providing Taiwanese citizens with maximum benefits (Wu et al., 2010). Lee, Liang, and OuYang (2002) state that the NHI model benefits patients by covering preventive medical services, prescription drugs, dental services, Chinese medicine, and home nurse visits (Lee et al., 2002). Compared to 2001 when only 97 % of the population was involved in the model, nearly 100 % of the population was enrolled in the NHI scheme by 2016 (Tang, Jiang, You, & Cheng, 2017). 2.2. Hospital information system Many areas in health information systems, in which HIS is a typical illustration, have been widely adopted in the hospital care environment (Haux et al., 2016). HIS is a factor of health informatics that predominantly emphasizes the administrative demand of hospitals. It was developed to establish a paperless environment that can cover every aspect of hospital operation, such as clinical, administrative, and financial systems (Hung, Hsu, Su, & Huang, 2014; Nilashi, Ahmadi, Ahani, Ravangard, & Ibrahim, 2016). An enormous and overwhelming number of daily outpatient visits must be handled by medical institutions, resulting in numerous outpatient registrations, physical orders, patient checkouts, and operational work demanding test data and result entries at many testing locations. Hence, data are uninterruptedly transmitted between clients and server hosts. With this overwhelming load, translating medical data can be obstructed during peak times in hospitals; a downturn in computer response rate can also restrict the advancement of HIS. Internet of Things (IoT) offers a convenient living environment by connecting people and objects. IoT technology enables the movement of people from computer-based centralized schemes to distributed environments via smart wearables, smart homes, smart mobility, and smart cities. The present study analyzes the applicability of IoT in medicine and healthcare on the basis of the holistic architecture of the IoT eHealth ecosystem. It considers the integration of patient-centric healthcare devices, fog computing, and cloud to process these complex data in terms of speed, variety, and latency. This architecture of fog-driven IoT can offer diverse applications in assisted living, mobile health, e-medicine, implants, early warning systems, and population monitoring in smart cities (Farahani et al., 2018). Many hospitals have exerted considerable efforts to cultivate administrative efficiency, scale down costs, and provide patients with high-quality and user-friendly services (Chang, Chang, Wu, & Huang, 2015; Lorence & Spink, 2004; Yucel, Cebi, Hoege, & Ozok, 2012). Accordingly, most hospitals are developing cloud-based HIS as an optimal alternative. 2
International Journal of Information Management xxx (xxxx) xxxx
P.-T. Chen, et al.
based on the fusion algorithm to simulate medical imaging. The MFusion and M-Update functions can attain good performance evaluation and process and visualize more than 40 GB of data within 600 s. The current study also indicates that computational intelligence can improve efficiency for healthcare research via simulations and visualization (Chang, 2017). The technology of data analytics and visualization can identify weak spots of the tumor and simulate the development of malignant tumors. This work proposes that visualization technology can help detect the presence of cancerous cells in genes, realize the simulations of malignant tumors, and evaluate the status of such tumors. It likewise indicates that visualization technology not only solves the current healthcare research limitation but also applies to different interdisciplinary studies (Chang, 2018). According to Sultan (2014), cloud or utility computing, which is also known as the platform for accessing big data, has opened many possibilities for organizations engaged in healthcare provision (Sultan, 2014). Specifically, big data derived from clouds can provide opportunities for cost-saving and innovative solutions to healthcare problems. Furthermore, cloud computing has altered healthcare IT, particularly within the EHR domain (Ali, Shrestha, Soar, & Wamba, 2018; Chen, 2018; Gagnon et al., 2016; Kordzadeh, Warren, & Seifi, 2016). EHRs are easy to use, have tangible benefits, and are consistent with professional and social norms. Despite its increasing popularity, serious and real concerns related to big data remain, such as security, outages, vendor lock-in, and regulation issues (Ali et al., 2018; Sultan, 2014, 2015). Sultan (2015) additionally states that selecting the right cloud provider is a real challenge for many healthcare providers who wish to work with big data (Sultan, 2015). Big data and predictive analytics (BDPA) tools and methodologies can improve operational and strategic capabilities and increase corporate financial performance. BDPA is applied in diverse fields, such as business analytics, information systems, supply chain management, industrial engineering, and other business and engineering disciplines. It can likewise effectively handle data-intensive processes and become a critical support for supply chain management (Hazen, Skipper, Ezell, & Boone, 2016). Big data has become a critical force for seeking competitive advantage. Several researchers utilize content analysis to investigate the essential factors that influence the adoption of organizational big data. The present study integrates technology–organization–environment (TOE), diffusion of innovation theory (DOI), and the institutional approach. It proposes the big data organizational adoption framework and determines the 26 factors of the TOE framework through a literature review of business intelligence and analytics during the 2009–2015 period (Sun, Cegielski, Jia, & Hall, 2018). Data analysis is a time-consuming activity for humans. Therefore, sophisticated cognitive systems can be used to handle massive amounts of data. The researchers explain the past, present, and future directions of the development of domain big data and cognitive computing. The study also adopts cognitive computing for the systematic literature review through databases, such as Scopus, Database systems and Logic Programming (DBLP), and Web of Science (Gupta, Kar, Baabdullah, & Al-Khowaiter, 2018).
been widely applied to investigate consumer behavior and influential factors in terms of resistance, instead of acceptance, toward innovations (Chen & Kuo, 2017). This theory has been an effective and significant alternative for many researchers who aim to uncover the factors of resistance. In contrast to the current research, the majority of previous studies apply the technology acceptance model (TAM) or related theories, such as TAM2, unified theory of acceptance and use of technology (UTAUT), theory of reasoned action (TRA), and theory of planned behavior, to predict and explain user reaction to health IT (Holden & Karsh, 2010; Yarbrough & Smith, 2008; Taylor & Todd, 1995; Venkatesh, Morris, Davis, & Davis, 2003). However, some scholars state that resistance from adopters or users is an important source of innovation failure (Sheth & Ram, 1987). This finding indicates that the successful adoption of innovations is impossible without penetrating user or adopter behavior regarding innovation resistance. Therefore, the application of innovation resistance theory plays a pivotal role in uncovering the resistance of medical institutions in terms of adopting medical big data (Chen, 2018). 2.4.1. Expertise barrier Scholars remark that technical specialization is the most important element of successful innovation (Sheth & Ram, 1987). Hence, the lack of technological expertise may lead to the failure of adopting technological innovations in organizations (Schaeffer, Booton, Halleck, Studeny, & Coustasse, 2017). Almost all organizations have a tendency to gain a strong level of technical specialization and a concomitantly low level of technical versatility. Sheth and Ram (1987) use large companies in the pharmaceutical industry as an example. As a consequence of strongly specialized technologies, most of these companies only have a few innovations to their credit (Sheth & Ram, 1987). In terms of medical big data systems, efforts in data analysis would be pointless unless medical institutions acquire the capability to analyze all data gathered in a warehouse. Schaeffer et al. (2017) note that many foreseen and unforeseen barriers have inhibited hospitals from developing medical big data technology because of the significant shortage of technological expertise (Schaeffer et al., 2017). In particular, smallsized institutions without technological expertise cannot sufficiently utilize and adopt a medical big data system, resulting in the degradation of system performance. The competency of medical employees is also believed to be the key factor in the successful implementation of medical big data systems. According to the survey by eHealth Initiative (2013), two-thirds of the respondents believe that the shortage of welltrained staff is a hindrance for big data development in medical institutions. Jee and Kim (2013a, 2013b) note that finding talents in big data analytics is difficult, while the demand for qualified staff in big data adoption is always enormous (Jee & Kim, 2013a). 2.4.2. Operation barrier Together with the expertise barrier, the operation barrier is another obstacle to innovation in the “specialization trap” discovered by Sheth and Ram (1987). This barrier refers to the issue of overspecialization and resistance to change in organizations, particularly those associated with production and operation instead of research and development. Many organizations encounter this type of barrier especially when they own highly specialized operations, which results in a feeling of unwillingness to change their operations for the adoption of any new innovation. For example, appliance firms may be unwilling to switch from metals to plastics in their production process because such a switch requires considerable changes in the operation and alterations in the established patterns in manufacturing and employee training (Huq, 2005). Given the enormous amount of data on a daily basis, ranging from terabytes to exabytes (Al-Jarrah, Yoo, Muhaidat, Karagiannidis, & Taha, 2015), traditional relational data have limitations in terms of realizing big data storage (Kong, Feng, & Wang, 2015). Thus, the popularity of big data is increasing with regard to data creation and management. However, not every adopter feels confident in relation to
2.4. Barriers of big data to healthcare Although big data undoubtedly introduces enormous benefits for medical institutions if adequately applied in the healthcare context, its adoption entails challenges from many sides, especially the internal side of an organization. The outcomes of the survey by eHealth Initiative (2013) indicate that 84 % of healthcare executives believe that the adoption of medical big data is a considerable challenge for medical institutions. In their innovation resistance theory, Sheth and Ram (1987) divide the barriers to the adoption of innovation into five main dimensions, namely, expertise, operation, resource, regulation, and market access. In the healthcare context, these barriers are inherent as organization-driven obstacles that inhibit the application of medical big data systems (Sheth & Ram, 1987). Innovation resistance theory has 3
International Journal of Information Management xxx (xxxx) xxxx
P.-T. Chen, et al.
big data application in medical systems because it is a demanding process that may change an entire established operation.
receptive customers (Sheth & Ram, 1987). These barriers become less significant when the market share of the innovation is large. For example, Coca-Cola and Pepsi-Cola are large beverage companies that cause a considerable market access barrier for small companies in the market. In certain global markets, the government attempts to protect local companies from the severe competition of large foreign rivals, an endeavor that is also considered part of the market access barriers (Holmes, McGrattan, & Prescott, 2015). Big data to healthcare has also entailed various challenges in relation to acceptance and application in medical institutions (Roski, BoLinn, & Andrews, 2014). Every medical product must be examined by the FDA under strict quality and safety control regulations before being launched in the market. The market access barrier occurs because the FDA requires a huge investment of R&D resources and clinical trial time. Consumers’ limited awareness of regulations, particularly those that protect patients’ privacy and security, is seriously important as well (Roski et al., 2014). In recent studies, Raghupathi and Raghupathi (2014) emphasize that regulating the privacy and security of patients are the most notable problems, while Jee and Kim (2013a) focus on data security and compliance programs. Current enterprises collect big data from inbound and outbound data sources. Inbound data are generated from business operations, such as human resource management, manufacturing, supply chain management, marketing, and other business activities. Outbound data are generated from customers, transactional histories, product reviews, surveys, and market analysis. The present study proposes the new concept of big data reduction at the customer end and adopts early data reduction operations to achieve multiple objectives, such as delegating data sharing control to customers, enabling secure data sharing, preserving the privacy of customers, enhancing trust between customers and enterprises, and lowering service utilization costs (Rehman, Chang, Batool, & Wah, 2016). A privacy-preserving e-health system combines cloud storage, IoT, and big data. Patients’ physiological data are monitored by the IoT network and then aggregated to EHRs. The large volumes of EHRs are outsourced to the cloud platform, and patients can distribute the IoT group key to medical nodes with an authenticated mechanism in the e-health system. The IoT messages, which can be encrypted through the IoT group key and transmitted to the patients, can be batch-authenticated by the patients. Encrypted EHRs can be shared among patients and different data users in a fine-grained access control manner (Yang, Zheng, Guo, Liu, & Chang, 2018).
2.4.3. Resource barrier In general, inadequate funds discourage the success of any business venture (Sheth & Ram, 1987). A considerable amount of cash can be the most important factor to ensure the early success of innovation adoption. In the healthcare context, previous related studies also indicate that the time and cost associated with the application of medical big data are regarded as the major causes of failure in establishing a big data warehouse (Chute et al., 2013; Jee & Kim, 2013b; Jonathon Northover, 2014; Raghupathi & Raghupathi, 2014; Shapiro, Mostashari, Hripcsak, Soulakis, & Kuperman, 2011; Wills, 2014). In particular, Jee and Kim (2013a) reveal that integration costs are commonly high, while Shapiro, Mostashari, Hripcsak, Soulakis, and Kuperman (2011)) observe that the cost of improving interfaces is remarkable. Furthermore, the development of medical big data systems is expensive because of the lack of data standardization (Raghupathi & Raghupathi, 2014), the sheer volume of data, and the shortcoming of system connectivity (Wills, 2014). Medical institutions are burdened by the problem of allocating suitable levels of capital and necessary human resources to transform raw data into useful information exists (Halamka, 2014). 2.4.4. Regulation barrier Some recent studies discuss the market access barrier based on institutional theory (2001, Alston, Eggertsson, & North, 1996; North, 1990; Scott, 1992). Many institutions can constrain interaction behavior in business activities owing to the condition of uncertainties in human interaction. Moreover, an analytical framework is proposed to explain how institutions and institutional change influence the performance of economies in a given time (North, 1990). Several empirical studies on institutional change also propose to analyze the issues from the evolution of the securities markets in 17th-century England to the property rights issue of airports in modern America. The researchers analyze institutions and institutional change in various places worldwide across periods. This stream of research explains the incentive roles of transaction costs and property rights in the economic arena and contributes significantly to the new economics of institutions (Alston et al., 1996). The school of organization theory also introduces various approaches for solving different organizational problems and defining organizational, rational, natural, and open organizational systems, environments, strategies, and structures, organizational knowledge can be applied in various fields as well, such as the business strategies, international business, business management, industrial engineering, cognitive psychology, anthropology, and sociology (2001, Scott, 1992). Regulations can exist in various forms, and most industries are subject to at least one such regulation. Sheth and Ram (1987) categorize regulations for organizations into four major types: industry self-regulation (codes of business practice and ethics), government regulation (e.g., FDA and DOJ), regulation of utility monopolies (e.g., water, gas, and electricity), and regulation of patents and trademarks (Sheth & Ram, 1987). In the healthcare industry, the changes and control of regulations and policies cause difficulties in the implementation of medical big data technology (Jonathon Northover, 2014; Raghupathi & Raghupathi, 2014). Although developing medical big data without the restriction of regulations is impossible, the creation and use of medical big data have not been clearly and properly established in terms of regulations. Various examples of barriers to medical innovation are identified, ranging from the recognition of new medicines to the augmentation of cancer treatments (Sheth & Ram, 1987).
3. Research methods 3.1. Expert interviews This study aims to determine the resistance factors toward big data development in medical institutions through interviews with experts in the industry. Bellamy, Bledsoe, and Traube (2006) define expert interviews as a research method involving specialists with exclusive knowledge and experiences resulting from the actions, responsibilities, and obligations of the specific functional status within an organization or institution (Bellamy et al., 2006). Froschauer and Lueger (2009) emphasize that expert interviews introduce a wide range of benefits, such as a good understanding of interviewees, fast recruiting and scheduling process, rich collected data, and direct interaction with interviewees (Froschauer & Lueger, 2009). Dorussen, Lenz, and Blavoukos (2005) state that expert interview is an appealing data collection technique because it enables researchers to bridge the gap between case studies and the comparison of various countries on the basis of generalized and publicly available data. In addition, expert interviews can provide researchers with control over the dimensions that are central to comparative research (Dorussen et al., 2005).
2.4.5. Market access barrier Innovation creators must overcome a wide range of barriers from the market to approach potential users in the market. Market access barriers generally refer to all impediments that limit innovations from reaching
3.2. Analytic network process Saaty proposed the concepts of the ANP approach in 1996 to solve 4
International Journal of Information Management xxx (xxxx) xxxx
P.-T. Chen, et al.
the limitation that the Analytic Hierarchy Process (AHP) approach is too ideal to evaluate decision problems of the real world (Saaty & Begicevic, 2010; Saaty, 2006, 2007). The ANP approach can determine the weights of aspects in terms of the dependence and feedback relationship in real-world decision problems (Lin & Kuo, 2018; Lin, 2015; Lin, Shih, Tzeng, & Yu, 2016; Liu, Tzeng, & Lee, 2012). Several researchers adopt the ANP approach for strategy evaluation and alternative selection, such as the framework of strategic assessment and selection based on ANP and TOPSIS techniques (Chang, Abdel-Basset, & Ramachandran, 2019). The ANP approach is conducted by following five main steps (Niemira & Saaty, 2004; Saaty, 2006). 3.2.1. Step 1: constructing the structure of an evaluation system as a network Initially, the problem is precisely described and divided into clusters, which include the alternative cluster. The interdependence between nodes and clusters is then clarified. Hence, a network comprising cycles connecting its nodes and loops with related components is created. 3.2.2. Step 2: developing pairwise comparisons to determine the relative importance of factors Comparisons are conducted by utilizing the nine-point fundamental scale of Niemira and Saaty (2004), in which a score of 1 illustrates the same importance between factors, and a score of 9 shows the extreme importance of one criterion over another (Niemira & Saaty, 2004).
Fig. 1. Ideal and best compromise solution.
methods in multi-criteria decision-making. The basic idea of VIKOR is to define the ideal solution (positive ideal solution) and negative ideal solution. The so-called ideal solution refers to the best among all evaluation factors in the alternatives, whereas the negative ideal solution is the worst alternative in the evaluation factors. The priority between alternatives is then ranked by comparing the estimated values of the alternatives with the ideal proximity. When calculating the proximity of each option to the desired solution, the scores of the evaluation factors must be summed. In the aggregate approach, the VIKOR summation method is developed by the Lp metric of the eclectic programming approach (Yu, 1973; Zeleny, 1982). The characteristics of this method include maximization of “group benefits” and minimization of “objection to the individual regret” to allow acceptance of the compromise solution by policymakers. Fig. 1 illustrates the concept of compromise solution of VIKOR (Opricovic & Tzeng, 2007; Tzeng, Tsaur, Laiw, & Opricovic, 2002; Tzeng, Lin, & Opricovic, 2005). The VIKOR approach can be used to evaluate alternative ranks and assess the best and suitable alternatives. This approach can also solve discrete decision problems with a non-commensurable and conflicting condition (Opricovic & Tzeng, 2004, 2007; Opricovic & Tzeng, 2002, 2003; Tzeng et al., 2005; Tzeng, Teng, Chen, & Opricovic, 2002; Tzeng, Tsaur et al., 2002). Several researchers adopt VIKOR for solving complex decision problems, such as e-government website evaluation, by using a neutrosophic VIKOR approach (Abdel-Basset, Zhou, Mohamed, & Chang, 2018). F* is the ideal solution, f1* denotes the ideal value of the first evaluation criterion, and f 2* stands for the ideal value of the second evaluation criterion. When conflict exists between the evaluation factors, the first criterion to achieve the desired value must sacrifice the performance of the second criterion, and vice versa. Therefore, two conflicting evaluation criteria must mutually agree. In this circular arc, F c is the closest feasible solution to the ideal solution among all feasible solutions, such that F* is the best compromise solution after the compromise solution. f1 = f1* f1c is the extent to which the first criterion is compromised, and f2 = f 2* f 2c is the extent to which the second criterion is compromised. The VIKOR calculation steps are as follows:
3.2.3. Step 3: conducting a consistency test Given that the comparisons are conducted according to personal evaluation, the judgement consistency must be measured. Saaty formulated the consistency index (CI) and consistency ratio (CR) as follows (Saaty & Bennett, 1977; Saaty & Vargas, 1980):
CI =
CR =
n
max
n
1
CI RI
(1) (2)
In Eq. (1), max is the maximum eigenvalue, and n is the number of factors to be evaluated. In Eq. (2), RI is the consistency index of a randomly generated reciprocal matrix from the fundamental scale with forced reciprocals. 3.2.4. Step 4: establishing and calculating the supermatrix The local priority vectors are placed in suitable columns of a matrix known as a supermatrix to achieve global priorities in a system with interdependent relationships. First, an unweighted supermatrix comprising every eigenvector extracted from the pairwise comparisons is created. Second, the eigenvector obtained from the cluster-level comparison regarding the control criterion is adopted to the unweighted supermatrix as an aspect weight for the acquisition of the weighted supermatrix. The columns of the supermatrix have a sum of at least 2 if clusters in a network are interdependent. 3.2.5. Step 5: finalizing the best solution In this step, the weighted supermatrix is assembled to obtain a longterm solid range of weights. Therefore, the weighted supermatrix is increased to the power of 2k, in which k is an arbitrarily large figure. The result of this process is known as the limit supermatrix (Niemira & Saaty, 2004). The best solution is the alternative with the highest overall priority after the calculations are performed by utilizing the matrix operations (Liang et al., 2013).
3.3.1. Step 1: finding the ideal and negative ideal solutions As shown in Eqs. (3) and (4), the ideal and negative solutions of each alternative are determined. f k* is the ideal solution, whereas f k is the negative ideal solution. f k* denotes the performance evaluation
3.3. Vlse Kriterijumska Optimizacija I Kompromisno Resenje VIKOR belongs to one of the best compromise programming 5
International Journal of Information Management xxx (xxxx) xxxx
P.-T. Chen, et al.
value, k indicates the alternative k, and i refers to the evaluation factors, which will be obtained through the questionnaire. I1 denotes the benefit evaluation factors set, and I2 pertains to the set of cost evaluation factors. f k* = {(maxfik |k
I1), (minfik |k
k
k
I2); or setting the aspired level for i criterion},
4.2. Analytic network process Saaty (2005) states that the ANP approach is a pragmatic tool that facilitates the modeling of the problem as a network of factors and alternatives categorized into clusters. These factors and alternatives can relate to one another in many possible ways in relation to feedback and interdependence within and between clusters. After ANP implementation, this research reveals the causal relationship between the dimensions and resistance factors according to their weights (Fig. 2). This research uses ANP to calculate the weight of factors. Each dimension has only one principal component. Thus, we normalize and transpose the total influence matrix to obtain the relative weight matrix (Table 2) and the weighted supermatrix after providing the relative importance weight (Table 3). Finally, after we multiply the weighted supermatrix by itself, the dependent relationship between the principal components gradually converges and the limit supermatrix is obtained (Table 4). The supermatrix of component weights is the optimum weight (Table 5). This research transposes the total impact matrix. Each row factor is divided by the sum of the row to obtain the weighted matrix. Finally, the supermatrix of the relative weights between factors can be acquired. On the basis of the supermatrix results, the weight values of the five aspects are used to multiply the weight value of each factor. Table 6 displays the weight and rank of the resistance factors toward medical big data development.
k
(3) f k = {(minfik |k
I1), (maxfik |k
k
k
I2); or setting the worst level for i criterion},
k
(4)
3.3.2. Step 2: calculating the overall benefit and individual regret of alternatives Eq. (5) is used to calculate the difference between each dimension and indicators. The overall benefit Sk is the sum of the difference of all factors and ideal solution, and wi denotes the relative weight of the evaluation factors. Eq. (6) can be used to obtain the distance ratio Qk , individual regret of each factor, and negative ideal solution
Sk = dkp = 1 =
n
n
wi Srik , i=1
Qk =
dkp =
wi = 1 ,
(5)
i=1
= max {rik |i = 1, 2, ..., n}.
(6)
k
3.3.3. Step 3: calculating and sorting the comprehensive benefit of the alternatives Through Eq. (7), the overall benefit Sk and individual regret Qk are used to establish the comprehensive benefit Rk and rank the alternatives. S* and S denote the ideal and negative ideal solutions of each dimension and indicator, respectively. Q* and Q are the ideal and negative ideal solutions of individual regret, respectively. v is the decision-making mechanism coefficient. When v is higher than, close to, or less than 0.5, the decision-making method is based on the majority, agreement, or rejection, respectively.
Rk = v (Sk
S *)/(S
S *) + (1
v )(Qk
Q *)/(Q
Q *),
(8)
Q* = minQk , Q = maxQk
(9)
k
k
k
4.3.1. Finding the ideal and negative ideal solution In this study, the VIKOR correlation values are obtained by the questionnaires, and the factor scores range from 0 to 10. F* is both the ideal solution and the negative ideal solution. In this study, we set up the positive ideal solution ( fi*) as 10, the negative ideal solution ( fi ) as 0, and the weighted value as the value obtained after the ANP calculation. As shown in Table 7, the participants represent relatively different scores for the dimensions. In particular, among physicians, resource and operation barriers have the highest and lowest scores, respectively. In terms of medical staff, regulatory and resource barriers gained the highest and lowest scores, respectively. Finally, with regard to scholars, the highest score is demonstrated by the regulatory barrier, whereas the lowest score belongs to the operation barrier.
(7)
S * = minSk , S = maxSk k
4.3. Vlse Kriterijumska Optimizacija I Kompromisno Resenje
In Eq. (7), the smaller the Rk value, the better. 4. Practical analysis and results
4.3.2. Calculating the overall benefit and individual regret of alternatives In this study, ANP is used to obtain relative weighted values. Table 8 presents the results of the importance index of medical big data. In terms of Svk , physicians have the highest value with 0.257, whereas scholars represent the lowest value with only 0.144. With respect to Qvk , physicians have the highest value at 0.282, whereas scholars have the lowest value at 0.171.
4.1. Expert interviews In this research, we conduct 32 interviews with important practitioners in the healthcare industry, including physicians, medical staff, and scholars. Among the participants, 31 experts come from large and renowned medical institutions throughout Taiwan, such as Taipei Veterans General Hospital in Taipei, China Medical University Hospital in Taichung, and National Cheng Kung University Hospital in Tainan. Only one interview with a big data specialist and researcher at the University of Toronto in Canada is conducted on the telephone due to geographic distance. The interviews range in length from 45 min to 60 min and are conducted at the institutions where the participants work. Table 1 represents the demographic diversity of experts who participated in the interviews. Subsequently, we analyze the communication of the participants and summarize the main themes that they shared. Five main barrier dimensions are generally available, namely, expertise, operation, regulation, resource, and market access barriers. In each dimension, the participants reveal various, yet specific, resistance factors. These factors are eventually divided into four groups comprising the most frequently mentioned factors for each dimension. These dimension and resistance factors are subsequently analyzed through the use of the ANP method.
4.3.3. Calculating and sorting the comprehensive benefit of alternatives In accordance with the final step of the VIKOR method, we calculate the Rvk of each strategy. Table 9 delineates the values of Rvk under different v values ranging from 0 to 1. We use v = 0.5 to pursue the maximum overall benefit and minimum individual regret. Given that a small Rvk is preferred, when v = 0.50, the Rvk values of physicians, medical staff, and scholars are 0.269, 0.203, and 0.157, respectively. 4.3.4. Strategy sequence The study uses v = 0.5 to evaluate the appropriate alternatives and establish the indicator that pursues the maximum overall benefit and minimum individual regret. Table 10 strategizes the Rvk of the participants when v = 0.5, thus confirming the strategy sequence. The outcome shows that compared with physicians and medical staff, scholars have the largest concern regarding medical big data development and 6
International Journal of Information Management xxx (xxxx) xxxx
P.-T. Chen, et al.
Table 1 Demographic diversity of participants. Institution
Location
National Cheng Kung University Hospital Chi Mei Medical Center Sin-Lau Hospital Anonymous medical institution Taipei Veterans General Hospital Anonymous institutions and companies China Medical University Hospital Kaohsiung Veterans General Hospital Show Chwan Memorial Hospital Fear Eastern Memorial Hospital Anonymous big data company University of Toronto Total (32) Ratio (100 %)
Number of participants
Tainan
Taipei Taichung Kaohsiung Changhua New Taipei Hsinchu Toronto
Physician
Medical staff
11 2 1 2 1 2 1 1 1 1
1
Scholar
2 2
1
1
23 72 %
1 1 3 9%
6 19 %
Fig. 2. ANP model of the main dimensions and resistance factors.
provide the most solid suggestions toward the removal of barriers for the implementation of big data systems in medical institutions.
Table 2 Relative weight matrix. Dimensions
EB
OB
RB
LB
MAB
Expertise Barrier (EB) Operation Barrier (OB) Resource Barrier (RB) Regulation Barrier (LB) Market Access Barrier (MAB) Total
0.183 0.209 0.204 0.191 0.213 1.000
0.195 0.196 0.205 0.192 0.213 1.000
0.195 0.209 0.191 0.191 0.213 1.000
0.194 0.209 0.204 0.180 0.213 1.000
0.195 0.209 0.204 0.192 0.200 1.000
5. Discussion and implications In this study, the results of ANP and VIKOR analyses show that the main barriers of medical big data application are operation and market access barriers. Operation barrier mainly comes from data collection and quality, which may be caused by the non-parametric model of
Table 3 Weighted supermatrix. Dimensions
Principal components
EBP1
OBP1
RBP1
LBP1
MABP1
Expertise Barrier (EB) Operation Barrier (OB) Resource Barrier (RB) Regulation Barrier (LB) Market Access Barrier (MAB)
Lack of communication and cross-domain analysis (EBP1) Difficult to obtain patient information (OBP1) Insufficient supporting measures and heavy work (RBP1) Limited use of data and access (LBP1) Limited bonus and lack of incentives (MABP1) Total
0.183 0.209 0.204 0.191 0.213 1.000
0.195 0.196 0.205 0.192 0.213 1.000
0.195 0.209 0.191 0.191 0.213 1.000
0.194 0.209 0.204 0.180 0.213 1.000
0.195 0.209 0.204 0.192 0.200 1.000
7
International Journal of Information Management xxx (xxxx) xxxx
P.-T. Chen, et al.
Table 4 Limit supermatrix.
Expertise Barrier (EB) Operation Barrier (OB) Resource Barrier (RB) Regulation Barrier (LB) Market Access Barrier (MAB)
Principal components
EBP1
OBP1
RBP1
LBP1
MABP1
Lack of communication and cross-domain analysis (EBP1) Difficult to obtain patient information (OBP1) Insufficient supporting measures and heavy work (RBP1) Limited use of data and access (LBP1) Limited bonus and lack of incentives (MABP1) Total
0.192 0.206 0.202 0.189 0.210 1.000
0.192 0.206 0.202 0.189 0.210 1.000
0.192 0.206 0.202 0.189 0.210 1.000
0.192 0.206 0.202 0.189 0.210 1.000
0.192 0.206 0.202 0.189 0.210 1.000
medical big data, resulting in the obstacles caused by the hidden nodes of data network and exponential growth of computational complexity (Al-Jarrah et al., 2015). Therefore, the government and related agencies still need to support and manage data acquisition and storage (Kong et al., 2015). In addition, the major problem of market access barrier is the restriction on value-added application of medical big data. Big data projects developed by governments and the healthcare industry have similar goals (Jee & Kim, 2013a). However, numerous barriers to achieving valuable practical applications exist. The effective application of medical big data includes a complete IT infrastructure, analysis tools, processes and interfaces, and balance between medical big data benefits and patient protection (Roski et al., 2014). Therefore, the criterion and regulations for the use, data sharing, moral privacy, and management of medical big data will become key with regard to whether medical big data can bring tangible benefits. This research also brings forward the different views of three groups of experts on medical big data, namely, physicians, medical staff, and scholars. The scholars are considered specialists who can provide the most strategic and reliable consultancy for medical institutions pursuing the successful implementation of big data systems. Although the physicians and medical staff who directly work inside hospitals or clinics are commonly considered to have a thorough command of the internal development of medical institutions, a variety of possible implications supports this conclusion. First, during the interview process, the physicians and medical staff from hospitals and clinics reflected their mistrust in medical big data. According to them, medical big data has more considerable disadvantages than do traditional data collection approaches. They can likewise perceive the difficulties and demanding workloads when the system is new to them because they work directly with the big data system. Therefore, the physicians and medical staff, known as “insiders” of the situation, can be assumed to contribute relatively less brilliant insights than do the scholars, who are regarded as “outsiders” or “observers” of the entire picture. The scholars specializing in big data systems for many years have conducted several relevant research studies and accumulated practical work experience in large big data organizations. These scholars have a solid foundation of technical and theoretical expertise, which enables them to provide constructive suggestions in terms of barriers to medical big data. As revealed in the interviews, all the scholars have experience in cooperating with many universities and hospitals in Taiwan and Canada. Therefore, they perceive the barriers not only from the “outsider” perspective but also from that of a specialist or researcher fully involved in the development and adoption process of big data systems in hospitals and clinics. Hence, the scholars are believed to have a clear
Table 6 Resistance factor weight and rank of medical big data barriers. Dimension
Factor
Weight
Rank
Expertise Barrier
Poor cooperation of staff Data analysis ability Data application ability Multidisciplinary communication Data integration mechanism Poor cooperation of patients Data approachability Unwillingness to share Basic complementary measures Data creditability High initial import costs Heavy staff workloads Limitation of regulation Limitation of data access Limitation of data utilization Responsibility for wrong diagnosis Differences of divisions Protection of patient privacy Limitation of value-added application Lack of incentives
0.047 0.046 0.050 0.049 0.054 0.046 0.054 0.052 0.049 0.049 0.051 0.053 0.045 0.048 0.050 0.046 0.052 0.050 0.058 0.050
9 10 6 7 2 10 2 4 7 7 5 3 11 8 6 10 4 6 1 6
Operation Barrier
Regulation Barrier
Resource Barrier
Market Access Barrier
Table 7 Weighted score. Dimension
Weight
Physician
Medical Staff
Scholar
f i*
fi
Expertise Barrier (EB) Operation Barrier (OB) Resource Barrier (RB) Regulatory Barrier (LB) Market Access Barrier (MAB)
0.194 0.205 0.201 0.191 0.210
7.472 7.181 7.625 7.306 7.569
8.000 8.034 7.914 8.198 7.948
8.667 8.292 8.646 8.729 8.500
10 10 10 10 10
0 0 0 0 0
Table 8 Weighted score. Dimension
Weight
Physician
Medical Staff
Scholar
Expertise Barrier (EB) Operation Barrier (OB) Resource Barrier (RB) Regulatory Barrier (LB) Market Access Barrier (MAB) Svk Qvk
0.194 0.205 0.201 0.191 0.210
0.253 0.282 0.238 0.269 0.243 0.257 0.282
0.200 0.197 0.209 0.180 0.205 0.198 0.209
0.133 0.171 0.135 0.127 0.150 0.144 0.171
Table 5 Supermatrix of component weights. Dimensions
Principal components
Weights
Expertise Barrier (EB) Operation Barrier (OB) Resource Barrier (RB) Regulation Barrier (LB) Market Access Barrier (MAB)
Lack of communication and cross-domain analysis (EBP1) Difficult to obtain patient information (OBP1) Insufficient supporting measures and heavy work (RBP1) Limited use of data and access (LBP1) Limited bonus and lack of incentives (MABP1) Total
0.192 0.206 0.202 0.189 0.210 1.000
8
International Journal of Information Management xxx (xxxx) xxxx
P.-T. Chen, et al.
By analyzing the communication of participants, we outline a list of 20 resistance factors, which are equally divided into the five main groups of barriers. Several relative studies also propose similar results in their research on the limitation of data utilization and protection of patient privacy (Rehman et al., 2016). Other related studies point out the challenge of data integration mechanism, data approachability, and information sharing willingness (Yang et al., 2018). We investigate more factors that are important to the adoption challenges of big data in healthcare. Furthermore, the ANP and VIKOR methods are applied to define the ideal and negative ideal solutions, from which the best strategy to overcome the outlined barriers is suggested. With a thorough understanding of the sequence of importance of resistance factors, managers can plan adequate strategies to solve problems with appropriate priorities. Hence, such a sequence indicates the concrete suggestions toward the pathway that managers can follow to remove development barriers when adopting medical big data systems. Therefore, the most remarkable and urgent barriers that must be eliminated are the limitation of value-added application, poor capability for approaching data, improper mechanism of data integration, and heavy and stressful workloads that medical staff must suffer from the adoption of big data systems. If managers concentrate on first withdrawing such resistance factors and subsequently handling other barriers in a reasonable order, then they can save a considerable amount of time and cost and gradually satisfactorily remove all barriers, leading to the potentially productive development of medical big data systems.
Table 9 Value of Rvk under different v values. v
Physician
Medical Staff
Scholar
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
0.282 0.279 0.277 0.274 0.272 0.269 0.267 0.264 0.262 0.259 0.257
0.209 0.208 0.207 0.206 0.205 0.203 0.202 0.201 0.2 0.199 0.198
0.171 0.168 0.165 0.163 0.16 0.157 0.155 0.152 0.149 0.146 0.144
Table 10 Sequence of strategies. v = 0.5
Physician
Medical staff
Scholar
Rvk 1 Rvk Sequence
0.269 0.731 3
0.203 0.797 2
0.157 0.843 1
vision of the overall scenario and appropriately prevent the developers of such systems from yielding to the barriers. 5.1. Theoretical implications
5.3. Research limitations and future research
“Medical big data” is not a new terminology in the healthcare context and has brought medical institutions numerous applications in surveillance, public health, and research. Our research essentially contributes to the enrichment of research literature. Although most previous studies center on the success or resistance factors of big data systems in various contexts, such as companies, governments, or educational institutions (Ziora, 2015), the literature discussing the adoption of big data in the field of healthcare service is lacking. Big data has not been widely adopted in medical institutions owing to its complexity and demand for massive efforts (Jee & Kim, 2013a). However, if big data is applied in this potential context, then a considerable amount of work can be effectively and adequately handled, especially the reduction of manual work when storing and processing a considerable amount of patient data. By conducting a range of expert interviews, we construct a framework of the main barrier dimensions and resistance factors shared by important stakeholders in the industry. We provide a clear direction for the development of medical big data and give subsequent researchers a solid foundation for further establishing the application and development of big data in healthcare. In addition, our findings provide theoretical insights related to the strategic pathways for the adopters or managers in hospitals and clinics. They can refer to this research as a constructive piece of advice they can consider before developing medical big data systems and can hardly find in previous literature.
Despite the contributive implications introduced by this study, researchers can consider several limitations for future investigations. First, the research setting is restricted because the majority of data is mainly collected inside Taiwan. Although the healthcare industry of Taiwan is acknowledged as high quality and advanced, additional research must be conducted in a wide context to guarantee the credibility and applicability of data into a general healthcare scenario. Second, given the difficulty in approaching healthcare professionals, this research lacks balance among the groups of participants. In particular, although a total of 23 physicians comprise most of the participants, the numbers of medical staff and scholars are small at only 6 and 3 individuals, respectively. Therefore, if the proportion of different groups of participants had been equal, then the comparisons between the contents of each group revealed would have been meaningful and adequate. Finally, only three groups of experts, namely, physicians, medical staff, and scholars, are involved in the research process. We strongly believe that other important stakeholders, such as engineers, governments, and even patients, can also provide crucial information and insights related to the development of medical big data. Future research can lay the groundwork for understanding the viewpoints of such stakeholders and construct new valuable barrier frameworks. With understanding on the resistance factors from a variety of stakeholders engaged in big data, additional opportunities for the successful adoption and implementation of big data systems in the healthcare context can be obtained.
5.2. Managerial implications
6. Concluding remarks
This study aims to examine how barriers to big data development in medical institutions are perceived by three groups of important stakeholders, namely, physicians, medical staff, and scholars. To acquire the most credible and critical data, we initially execute open-ended interviews with these experts, who are expected to share the most valuable knowledge and information regarding barriers to big data in the healthcare scenario. The outcome of the interviews reveals a collection of five main groups of barriers, namely, expertise, operation, resource, regulation, and market access, which are also discussed in innovation resistance theory by Sheth and Ram (1987).
This research applies innovation resistance theory by Sheth and Ram (1987) to uncover adopter resistance toward innovations in the healthcare domain and clarify the barriers to the successful adoption of big data systems in medical institutions. The theory applied is a practical alternative for many other widely used theories, such as TAM, TAM2, TRA, and UTAUT, which mainly focus on consumer acceptance behavior. In this research, we initially determine the resistance factors toward big data development in medical institutions through interviews with experts in the industry. Next, the ANP approach is applied to 9
International Journal of Information Management xxx (xxxx) xxxx
P.-T. Chen, et al.
facilitate the modeling of the factors as a network of factors and alternatives categorized into clusters. This method can help in making the best decisions by evaluating the interrelationships among all identified factors. VIKOR is subsequently used to define the ideal solution (positive ideal solution) and negative ideal solution. The so-called ideal solution refers to the best of all evaluation factors in the alternatives, whereas the negative ideal solution is the worst alternative in the evaluation factors. Numerous barriers are perceived by a range of physicians, medical staff, and scholars, who communicated why many medical institutions have encountered failure in adopting medical big data systems. All barriers are closely related to human expertise, resource allocation, operational procedure, laws and regulations, and market access capability. Unless medical institutions withdraw these barriers, then the adoption of big data systems is impossible. The research findings also show that managers should provide conscious attention to the importance-driven sequence of the identified barriers to follow the most explicit and expedient pathway when solving development problems.
medicine: official journal of the American College of Medical Genetics, 15(10), 802–809. Dorussen, H., Lenz, H., & Blavoukos, S. (2005). Assessing the reliability and validity of expert interviews. European Union Politics, 6(3), 315–337. Farahani, B., Firouzi, F., Chang, V., Badaroglu, M., Constant, N., & Mankodiya, K. (2018). Towards fog-driven IoT eHealth: Promises and challenges of IoT in medicine and healthcare. Future Generation Computer Systems, 78, 659–676. Froschauer, U., & Lueger, M. (2009). Expert interviews in interpretive organizational research. In A. Bogner, B. Littig, & W. Menz (Eds.). Interviewing experts (pp. 217–234). London: Palgrave Macmillan UK. Gagnon, M.-P., Simonyan, D., Ghandour, E. K., Godin, G., Labrecque, M., Ouimet, M., et al. (2016). Factors influencing electronic health record adoption by physicians: A multilevel analysis. International Journal of Information Management, 36(3), 258–270. Gupta, S., Kar, A. K., Baabdullah, A., & Al-Khowaiter, W. A. A. (2018). Big data with cognitive computing: A review for the future. International Journal of Information Management, 42, 78–89. Halamka, J. D. (2014). Early experiences with big data At an academic medical center. Health Affairs, 33(7), 1132–1138. Hännikäinen, J. (2017). When does the yield curve contain predictive power? Evidence from a data-rich environment. International Journal of Forecasting, 33(4), 1044–1064. Haux, R., Koch, S., Lovell, N., Marschollek, M., Nakashima, N., & Wolf, K.-H. (2016). Health-enabling and ambient assistive technologies: Past, present. Future. Hazen, B. T., Skipper, J. B., Ezell, J. D., & Boone, C. A. (2016). Big data and predictive analytics for supply chain sustainability: A theory-driven research agenda. Computers & Industrial Engineering, 101, 592–598. Holden, R. J., & Karsh, B.-T. (2010). The Technology Acceptance Model: Its past and its future in health care. Journal of Biomedical Informatics, 43(1), 159–172. Holmes, T. J., McGrattan, E. R., & Prescott, E. C. (2015). Quid pro quo: Technology capital transfers for market access in China. The Review of Economic Studies, 82(3), 1154–1193. Holzinger, A. (2016). Interactive machine learning for health informatics: when do we need the human-in-the-loop? Brain Informatics, 3(2), 119–131. Holzinger, A., Kieseberg, P., Weippl, E., & Tjoa, A. M. (2018). Current advances, trends and challenges of machine learning and knowledge extraction: From machine learning to explainable AI. Paper Presented at the Machine Learning and Knowledge Extraction, Cham. Holzinger, A., Plass, M., Kickmeier-Rust, M., Holzinger, K., Crişan, G. C., Pintea, C.-M., et al. (2018). Interactive machine learning: Experimental evidence for the human in the algorithmic loop. Applied Intelligence. Hung, Y. W., Hsu, S.-C., Su, Z.-Y., & Huang, H.-H. (2014). Countering user risk in information system development projects. International Journal of Information Management, 34(4), 533–545. Huq, Z. (2005). Managing change: A barrier to TQM implementation in service industries. Managing Service Quality: An International Journal, 15(5), 452–469. Jee, K., & Kim, G.-H. (2013a). Potentiality of big data in the medical sector: Focus on how to reshape the healthcare system. Jee, K., & Kim, G.-H. (2013b). Potentiality of big data in the medical sector: Focus on how to reshape the healthcare system. Healthcare Informatics Research, 19(2), 79–85. Jonathon Northover, J. D. (2014). Big Data or Big promises? : How hospitals leverage what is available today. Executive Insight. Kong, X., Feng, M., & Wang, R. (2015). The current status and challenges of establishment and utilization of medical big data in China. Kordzadeh, N., Warren, J., & Seifi, A. (2016). Antecedents of privacy calculus components in virtual health communities. International Journal of Information Management, 36(5), 724–734. Lee, C.-C., Liang, T.-P., & OuYang, Y.-C. (2002). E-healthcare in Taiwan. International Journal of Healthcare Technology and Management, 4(1-2), 1–14. Liang, X., Sun, X., Shu, G., Sun, K., Wang, X., & Wang, X. (2013). Using the analytic network process (ANP) to determine method of waste energy recovery from engine. Energy Conversion and Management, 66, 304–311. Lin, C.-L. (2015). A novel hybrid decision-making model for determining product position under consideration of dependence and feedback. Applied Mathematical Modelling, 39(8), 2194–2216. Lin, C.-L., & Kuo, C.-L. (2018). A service position model of package tour services based on the hybrid MCDM approach. Current issues in tourism1–33. Lin, C.-L., Shih, Y.-H., Tzeng, G.-H., & Yu, H.-C. (2016). A service selection model for digital music service platforms using a hybrid MCDM approach. Applied Soft Computing, 48, 385–403. Liu, C.-F., Hwang, H.-G., & Chang, H.-C. (2011). E-healthcare maturity in Taiwan. Liu, C. H., Tzeng, G. H., & Lee, M. H. (2012). Improving tourism policy implementation the use of hybrid MCDM models. Tourism Management, 33(2), 413–426. Lorence, D. P., & Spink, A. (2004). Healthcare information systems outsourcing. International Journal of Information Management, 24(2), 131–145. Martin-Sanchez, F., & Verspoor, K. (2014). Big data in medicine is driving big changes. Niemira, M. P., & Saaty, T. L. (2004). An analytic network process model for financialcrisis forecasting. International Journal of Forecasting, 20(4), 573–587. Nilashi, M., Ahmadi, H., Ahani, A., Ravangard, R., & Ibrahim, O. B. (2016). Determining the importance of Hospital Information System adoption factors using fuzzy analytic network process (ANP). Technological Forecasting and Social Change, 111, 244–264. North, D. C. (1990). Institutions, institutional change and economic performance. Cambridge: Cambridge University Press. Opricovic, S., & Tzeng, G.-H. (2004). Compromise solution by MCDM methods: A comparative analysis of VIKOR and TOPSIS. European Journal of Operational Research, 156(2), 445–455. Opricovic, S., & Tzeng, G.-H. (2007). Extended VIKOR method in comparison with outranking methods. European Journal of Operational Research, 178(2), 514–529. Opricovic, S., & Tzeng, G. H. (2002). Multicriteria planning of post-earthquake
Acknowledgements This research was made possible by the support and assistance of a number of people whom I would like to thank. I am very grateful to all the respondents for their valuable opinions. I would like to thank my research assistant Ms. Ya-Ci Ke, Mr. Kuan-Chen Li, Mr. Nguyuen Quoc Duy, Mr. Wei-Zhi Lu, Mr. Kuan-Chung Wang and Miss I-Ching Tsai’s help in literatures and questionnaires collection and organization. This research was supported by the Ministry of Technology and Science under grant number MOST 105-2221-E-006 -260-MY3 and MOST 1082221-E-006-063 and the Medical Device Innovation Center (MDIC), National Cheng Kung University(NCKU) from the Featured Areas Research Center Program within the framework of the Higher Education Sprout Project by the Ministry of Education (MoE) in Taiwan. References Abdel-Basset, M., Mohamed, M., Smarandache, F., & Chang, V. (2018). Neutrosophic association rule mining algorithm for big data analysis. Symmetry, 10(4), 1–19. Abdel-Basset, M., Zhou, Y.-Q., Mohamed, M., & Chang, V. (2018). A group decision making framework based on neutrosophic VIKOR approach for e-government website evaluation. Ali, O., Shrestha, A., Soar, J., & Wamba, S. F. (2018). Cloud computing-enabled healthcare opportunities, issues, and applications: A systematic review. International Journal of Information Management, 43, 146–158. Al-Jarrah, O. Y., Yoo, P. D., Muhaidat, S., Karagiannidis, G. K., & Taha, K. (2015). Efficient machine learning for big data: A review. Big Data Research, 2(3), 87–93. Alston, L. J., Eggertsson, T., & North, D. C. (1996). Empirical studies in institutional change. Cambridge: Cambridge University Press. Archenaa, J., & Anita, E. A. M. (2015). A survey of big data analytics in healthcare and government. Procedia Computer Science, 50, 408–413. Becker, D. K. (2017). Predicting outcomes for big Data Projects: Big Data Project Dynamics (BDPD): Research in progress. Paper Presented at the 2017 IEEE International Conference on Big Data (Big Data). Bellamy, J. L., Bledsoe, S. E., & Traube, D. E. (2006). The current state of evidence-based practice in social work: A review of the literature and qualitative analysis of expert interviews. Journal of Evidence-based Social Work, 3(1), 23–48. Boyd, D., & Crawford, K. (2012). Critical questions for big data: Provocations for a cultural, technological, and scholarly phenomenon. Chang, V. (2017). Computational intelligence for medical imaging simulations. Journal of Medical Systems, 42(1), 10. Chang, V. (2018). Data analytics and visualization for inspecting cancers and genes. Multimedia Tools and Applications, 77(14), 17693–17707. Chang, I. C., Chang, C.-H., Wu, J.-W., & Huang, T. C.-K. (2015). Assessing the performance of long-term care information systems and the continued use intention of users. Telematics and Informatics, 32(2), 273–281. Chang, V., Abdel-Basset, M., & Ramachandran, M. (2019). Towards a reuse strategic decision pattern framework – From theories to practices. Information Systems Frontiers, 21(1), 27–44. Chen, P.-T. (2018). Medical big data applications: Intertwined effects and effective resource allocation strategies identified through IRA-NRM analysis. Technological Forecasting and Social Change, 130, 150–164. Chen, P.-T., & Kuo, S.-C. (2017). Innovation resistance and strategic implications of enterprise social media websites in Taiwan through knowledge sharing perspective. Technological Forecasting and Social Change, 118, 55–69. Chute, C. G., Ullman-Cullere, M., Wood, G. M., Lin, S. M., He, M., & Pathak, J. (2013). Some experiences and opportunities for big data in translational research. Genetics in
10
International Journal of Information Management xxx (xxxx) xxxx
P.-T. Chen, et al. sustainable reconstruction. Computer-Aided Civil and Infrastructure Engineering, 17(3), 211–220. Opricovic, S., & Tzeng, G. H. (2003). Defuzzification within a multicriteria decision model. International Journal of Uncertainty Fuzziness and Knowledge-Based Systems, 11(5), 635–652. Perera, G., Broadbent, M., Callard, F., Chang, C.-K., Downs, J., Dutta, R., et al. (2016). Cohort profile of the South London and Maudsley NHS Foundation Trust Biomedical Research Centre (SLaM BRC) Case Register: current status and recent enhancement of an Electronic Mental Health Record-derived data resource. BMJ Open, 6(3), e008721. Raghupathi, W., & Raghupathi, V. (2014). Big data analytics in healthcare: Promise and potential. Health Information Science and Systems, 2(1), 3. Rehman, M. H. U., Chang, V., Batool, A., & Wah, T. Y. (2016). Big data reduction framework for value creation in sustainable enterprises. International Journal of Information Management, 36(6, Part A), 917–928. Roski, J., Bo-Linn, G. W., & Andrews, T. A. (2014). Creating value in health care through big data: Opportunities and policy implications. Health Affairs (Project Hope). Saaty, T. L. (2006). Rank from comparisons and from ratings in the analytic hierarchy/ network processes. European Journal of Operational Research, 168(2), 557–570. Saaty, T. L. (2007). Time dependent decision-making; dynamic priorities in the AHP/ ANP: Generalizing from points to functions and from real to complex variables. Mathematical and Computer Modelling, 46(7), 860–891. Saaty, T. L., & Begicevic, N. (2010). The scope of human values and human activities in decision making. Applied Soft Computing, 10(4), 963–974. Saaty, T. L., & Bennett, J. P. (1977). Theory of analytical hierarchies applied to political candidacy. Behavioral Science, 22(4), 237–245. Saaty, T. L., & Vargas, L. G. (1980). Hierarchical analysis of behavior in competition prediction in Chess. Behavioral Science, 25(3), 180–191. Schaeffer, C., Booton, L., Halleck, J., Studeny, J., & Coustasse, A. (2017). Big data management in US hospitals: Benefits and barriers. The Health Care Manager, 36(1), 87–95. Scott, W. R. (1992). Organizations: Rational, natural, and open systems. Englewood Cliffs, N.J: Prentice Hall. Scott, W. R. (2001). Institutions and organizations. Thousand Oaks, Calif: London Sage. Shapiro, J. S., Mostashari, F., Hripcsak, G., Soulakis, N., & Kuperman, G. (2011a). Using health information exchange to improve public health. American Journal of Public Health, 101(4), 616–623. Shapiro, J. S., Mostashari, F., Hripcsak, G., Soulakis, N., & Kuperman, G. (2011b). Using health information exchange to improve public health. American Journal of Public Health, 101(4), 616–623. Sheth, J. N., & Ram, S. (1987). Bringing innovation to market: How to break corporate and customer barriers. Wiley.
Sultan, N. (2014). Making use of cloud computing for healthcare provision: Opportunities and challenges. International Journal of Information Management, 34(2), 177–184. Sultan, N. (2015). Reflective thoughts on the potential and challenges of wearable technology for healthcare provision and medical education. International Journal of Information Management, 35(5), 521–526. Sun, S., Cegielski, C. G., Jia, L., & Hall, D. J. (2018). Understanding the factors affecting the organizational adoption of big data. Journal of Computer Information Systems, 58(3), 193–203. Tang, C., Jiang, J., You, S., & Cheng, W. (2017). Healthcare costs associated with cervical Cancer, Precancerous lesions, and genital warts treatment in Taiwan. Value in Health : the Journal of the International Society for Pharmacoeconomics and Outcomes Research, 20(9), A424–A425. Taylor, S., & Todd, P. A. (1995). Understanding information technology usage: A test of competing models. Information Systems Research, 6(2), 144–176. Tzeng, G.-H., Lin, C.-W., & Opricovic, S. (2005). Multi-criteria analysis of alternative-fuel buses for public transportation. Energy Policy, 33(11), 1373–1383. Tzeng, G.-H., Teng, M.-H., Chen, J.-J., & Opricovic, S. (2002). Multicriteria selection for a restaurant location in Taipei. International Journal of Hospitality Management, 21(2), 171–187. Tzeng, G.-H., Tsaur, S.-H., Laiw, Y.-D., & Opricovic, S. (2002). Multicriteria analysis of environmental quality in Taipei: Public preferences and improvement strategies. Journal of Environmental Management, 65(2), 109–120. Venkatesh, V., Morris, M. G., Davis, G. B., & Davis, F. (2003). User acceptance of information technology: A unified model. Wang, P., Liu, B., & Hong, T. (2016). Electric load forecasting with recency effect: A big data approach. International Journal of Forecasting, 32(3), 585–597. Wills, M. J. (2014). Decisions through data: Analytics in healthcare. Journal of Healthcare Management/American College of Healthcare Executives, 59(4), 254–262. Wu, T.-Y., Majeed, A., & Kuo, N. K. (2010). An overview of healthcare system in Taiwan. Yang, Y., Zheng, X., Guo, W., Liu, X., & Chang, V. (2018). Privacy-preserving fusion of IoT and big data for e-health. Future Generation Computer Systems, 86, 1437–1455. Yarbrough, A. K., & Smith, T. B. (2008). Technology acceptance among physicians: A new take on TAM. Medical Care Research and Review : MCRR. Yu, P. (1973). A class of solutions for group decision problems. Yucel, G., Cebi, S., Hoege, B., & Ozok, A. F. (2012). A fuzzy risk assessment model for hospital information system implementation. Expert Systems With Applications, 39(1), 1211–1218. Zeleny, M. (1982). Multiple criteria decision making. New York: McGraw-Hill. Ziora, A. C. L. (2015). The role of big data solutions in the management of organizations. Review of selected practical examples. Procedia Computer Science, 65, 1006–1012.
11