Improving RFID adoption in Taiwan's healthcare industry based on a DEMATEL technique with a hybrid MCDM model

(1)

Improving RFID adoption in Taiwan's healthcare industry based on

a DEMATEL technique with a hybrid MCDM model

Ming-Tsang Lu

a

, Shi-Woei Lin

b

, Gwo-Hshiung Tzeng

c,d,

⁎

a

Graduate Institute of Management Science, National Chiao-Tung University, 1001, Ta-Hsueh Road, Hsin-Chu 300, Taiwan

b_{Department of Industrial Management, National Taiwan University of Science and Technology, 43, Keelung Road, Section 4, Taipei 106, Taiwan} c

Graduate Institute of Urban Planning, National Taipei University, 151, University Road, San Shia 237, Taiwan

d

Institute of Project Management, Department of Business and Entrepreneurial Management, Kainan University, Taoyuan 338, Taiwan

a b s t r a c t

a r t i c l e i n f o

Article history: Received 16 June 2011

Received in revised form 11 June 2013 Accepted 11 June 2013

Available online 27 June 2013 Keywords:

Healthcare industry

Radio frequency identiﬁcation (RFID) Decision making trial and evaluation laboratory (DEMATEL)

DANP (DEMATEL-based ANP)

Multiple criteria decision making (MCDM) VIKOR

The use of radio frequency identification (RFID) technology has progressed tremendously in recent years. In the healthcare industry, the decision to adopt RFID technology is a problem requiring a multi-criteria decision analysis that involves both qualitative and quantitative factors. The evaluation of this decision may be based on imprecise information or uncertain data. Furthermore, there can be significant dependence and feedbacks between the different criteria and alternatives. However, most conventional decision models cannot capture these complex interrelationships. As a result, in this study we develop a general evaluation framework for industry evaluation, improvement and adoption of RFID. We use a hybrid Multiple Criteria Decision Making (MCDM) method known as DDANPV that combines DEMATEL (decision making trial and evaluation laboratory), DANP (DEMATEL-based ANP), and VIKOR to evaluate the factors that influence the adoption of RFID. Specifically, we study the adoption of RFID in Taiwan's healthcare industry. Wefind that technology integration is the most influential criterion and the strongest driver in the adoption of RFID of Taiwan's healthcare industry.

1. Introduction

Radio frequency identification (RFID) is a communication technolo-gy that uses radio waves to exchange data. RFID has three components: (1) an antenna for transmitting and receiving signals; (2) a transponder programmed with the identification information; and (3) an RF module (reader) with a decoder or transceiver. RFID has many applications and is an increasingly valuable tool for enabling automatic identification and management. For many industries, RFID is not only a new alternative to existing tracking methods but is also a solution for a range of previously cost-prohibitive innovations in internal control and supply chain coor-dination[34,46].

RFID has existed for decades. This technology was originally used to identify and trackﬂying aircrafts during the Second World War. Until recently, RFID was deemed to be too expensive and limited in functionality for many commercial applications. As the prices of RFID equipment and RFID tags have dropped in recent years, RFID applications have become increasingly prevalent. Cost is no longer a barrier. However, RFID has not been extensively adopted by the healthcare industry. The relatively conservative attitudes of healthcare providers have prevented hospitals from using the latest information

technologies. Furthermore, technology adoption often depends on a critical mass being reached; a manager's decision to adopt a new tech-nology often depends on the techtech-nology's diffusion rate, which, in turn, depends on the decisions made by other managers. Furthermore, even if a hospital decides to evaluate the relative costs and beneﬁts of implementing RFID technology, no comprehensive evaluation and adoption model exists that can be used as a reference for the adoption of RFID in the healthcare industry. Thus, it is inappropriate to focus only on the cost of a new IT technology as the primary factor in its adoption[4,7,9,50].

Most of the conventional multi-criteria decision analysis (MCDA) models cannot handle the analysis of complex relationships among different hierarchical levels of criteria. However, the decision to adopt RFID requires a decision model that performs just that analysis. In this paper, we develop a hybrid MCDM model called DDANPV that com-bines DEMATEL, DANP, and VIKOR. DDANPV overcomes the limitations of existing decision models and can be used to help us analyze the fac-tors that inﬂuence industry adoption of RFID technology. In particular, we use Taiwan's healthcare industry as an example to study the inter-dependence of the factors that inﬂuence the adoption of RFID in the healthcare industry, as well as to evaluate alternative RFID adoption processes to achieve the desired levels of performance from RFID technology.

This paper is organized intoﬁve sections.Section 2reviews the literature on the implementation of RFID in the healthcare industry. ⁎ Corresponding author. Tel.: +886 3 3412456; fax: +886 3 3412430.

E-mail addresses:mingtsang.lu@gmail.com(M.-T. Lu),shiwoei@mail.ntust.edu.tw

http://dx.doi.org/10.1016/j.dss.2013.06.006

Contents lists available atScienceDirect

Decision Support Systems

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / d s s

(2)

We will discuss the advances in evaluating the RFID adoption process, the selection criteria for adopting RFID technology, the decision models currently being used to determine whether RFID technology should be adopted, and the speciﬁc problems related to evaluating the RFID adoption process.Section 3introduces the hybrid MCDM method called DDANPV. InSection 4, we use Taiwan's healthcare industry as an empirical example to illustrate how DDANPV could help select the best RFID adoption method and discuss the results. InSection 5, we draw conclusions.

2. The effects of evaluating the RFID adoption model in the healthcare industry

The purpose of this section is to survey the relevant studies in the RFID adoption process, to investigate and compare various evaluation frameworks, and to identify possible factors that in_{ﬂuence the RFID} adoption process in the healthcare industry. Due to the lack of previous research on the criteria used in evaluating RFID for adoption, this study expands upon a general evaluation framework used in other industries and compiles four primary factors—technology, organization, environ-ment and cost—with the goal of identifying the criteria that are most crucial for the adoption of RFID.

2.1. Related literature on the factors inﬂuencing RFID adoption in the healthcare industry

RFID is one of the most promising technologies with the potential to increase supply chain visibility and improve process efficiency [45]. Once goods have RFID tags attached, their whereabouts can be tracked automatically by radio readers. With applications in transportation pay-ments, asset management, retail sales, and item tracking, RFID technol-ogy provides greater inventory visibility, improves business and control processes, and enhances supply management efficiency[26,47]. Hence, many industries are in various stages of applying RFID to experimental projects to improve operational efficiency and gain competitive advan-tages[5]. RFID has also been receiving considerable attention in the healthcare industry because it addresses the vexing problem of locating people and things in healthcare operations, as demonstrated in the case study examined in this project. RFID applications can be classified into two or more major categories based on different objectives in the healthcare industry. However, we only use two alternatives (“patient tracking management performance (A1)” and “asset tracking manage-ment performance (A2)”) from our project as examples to clearly illus-trate two relatively good uses for RFID applications. Thefirst set of applications is mainly designed for managing the patient-tracking sys-tem. For example, RFID is used in patient-tracking to automate the check-in process and other outbound logistical processes (i.e., activities that outsource the service to the customer in a service environment). In a healthcare setting, outbound logistics involve getting the right patient to the right place at the right time[21]. The second set of applications is also used for tracking purposes, but these applications are used to con-trol assets. RFID offers active tags for tracking various healthcare assets, such as wheelchairs, infusion pumps and crash carts. In the healthcare environment, assets (e.g., equipment and staff) are essential to provid-ing healthcare services to patients[21].

Schmitt et al.[38]reviewed related work and derived 25 adoption factors from the technological, organizational, and environmental dimensions of the RFID process. These researchers extracted theﬁve most important factors affecting the process of RFID adoption and diffusion in the automotive industry. These factors included compat-ibility, costs, complexity, performance, and top management support, as well as most of the more technological characteristics. Schmitt et al.[38]concluded that the RFID adoption and diffusion processes were still in the early stages and that the basic technological issues had to be solvedﬁrst. However, the organizational and environ-mental factors were found to be less important. Similarly, the

inter-organizational factors did not play essential roles because most of the RFID deployments in the automotive industry were intra-organizational applications.

Brown and Russell[6]conducted an exploratory investigation to identify the factors that may inﬂuence RFID adoption in South African retail organizations. A combination of quantitative and qualitative data based on six retailers were collected and analyzed using the Technology, Organization, and Environment (TOE) framework. Brown and Russell[6]

expounded upon the intention to adopt RFID technology using techno-logical factors (i.e., relative advantage, compatibility, complexity, and cost), organizational factors (i.e., top management attitude, information technology expertise, organization size, and organizational readiness), and external factors (i.e., competitive pressure, external support, and the existence of change agents).

In addition to the TOE framework mentioned above, the key barriers to RFID adoption also stem from the high technology expen-ditures, such as the software and hardware costs, required by RFID

[20]. When an organization plans to adopt RFID, both the implemen-tation costs and the maintenance costs need to be evaluated carefully. Lean information technology budgets suggest that new technologies need to demonstrate compelling business reasons for adoption while promising beneﬁts and short payback periods. As a result, most compa-nies are still waiting for RFID technology to drop in price to make it a more affordable investment[12,20,36]. In addition to the cost-beneﬁt analysis mentioned above, many factors contributing to the adoption of RFID are similar to the factors contributing to the recent adoption of e-commerce technology[12].

Previous studies on RFID adoption have not focused on all three TOE dimensions. Many authors have restricted their discussion to only a few key factors. For example, Hoske[13]highlighted the cost factor, while Jones et al.[18]examined private and public policies on RFID. Thus, in this paper, we take the TOE framework as a basis and add cost, resulting in technology, organization, environment, and cost (TOEC) as the four dimensions of our research framework. The factors relevant to the adoption of RFID within each dimension will be discussed below.

2.2. Criteria for evaluating the RFID adoption process

The criteria for evaluating the RFID adoption process are described below.

Technology dimension (D1): Technological factors, also referred to as“innovation characteristics” in several studies on organizational adoption processes[36]. Technology integration, technology com-petence, and security concerns have all been suggested as important to the adoption of RFID technology and are used in our evaluation framework[37,39].

Organization dimension (D2): Characteristics of the organization that is implementing the new technology are shown by Orlikowski

[32]to be highly relevant to the adoption process. Several studies have supported thisfinding with respect to RFID adoption, with fac-tors such as top management support,firm size, and organizational readiness considered to be potential influences[36,37,39]. Environment dimension (D3): Orlikowski [32]highlights the role and influence of the external environment in an organization's decision to adopt new technology. Competitive pressure, partner support, and regulatory support are regarded as among the most important external factors[36,37,39].

Cost dimension (D4): The beneﬁts of any new innovation should exceed the costs of adopting it[36]. Therefore, the costs associated with a new technology have a major bearing on the decision of its adoption. In this respect, RFID technology is no exception[39]. Most companies still have doubts about whether the costs associated

(3)

with RFID can be offset by its promised beneﬁts. The cost of RFID tags have been widely mentioned[1,17]in this discussion, as these costs determine the feasible level of tagging: item-level, case-level, or palette-level[13]. In this study, we investigate the related costs of RFID such as hardware, software, implementation, and maintenance. Our evaluation framework focuses on TOEC as the four dimensions that signiﬁcantly impact RFID adoption in the healthcare industry. Within each dimension, there are also lower-level criteria based on related factors that were considered in previous studies. Our entire evaluation framework, including both the dimensions and the criteria, is presented inTable 1.

3. DDANPV— A hybrid MCDM model for evaluating and improving RFID adoption

DDANPV is comprised of three stages. First, we use the DEMATEL method to uncover the relationship between the criteria and their network structure in the presence of interdependence and feedback among criteria. DEMATEL is more suitable in real-world applications than traditional methods, which assume independence among criteria

[8,14,15,23,24,33,43]. Second, we combine DEMATEL with the ANP

method to form DANP (DEMATEL-based ANP) to obtain influential weights for each dimension and criterion in our evaluation structure. Third, we incorporate these weights into the VIKOR method to rank the performance of the alternatives presented to the decision-maker and identify the gaps that each alternative has to an as yet non-existent aspired alternative; this approach provides us with a roadmap to how we can improve upon each alternative by reducing the perfor-mance gaps of each criterion and dimension relative to their aspired levels through innovation and research in the future. In short, the evaluation framework contains three main stages: (1) use DEMATEL to construct the influential network relation map (INRM) among the di-mensions and criteria; (2) use DANP to calculate the influence weights of each dimension and criterion; and (3) use VIKOR to rank the alterna-tives and improve the performances of the alternaalterna-tives.

3.1. The DEMATEL technique for constructing INRM

The DEMATEL technique has been successfully used to identify critical success factors in the adoption and assessment processes for emergency[48]and knowledge management[14]. This method can conﬁrm the interdependence of variables/criteria and restrict the re-lations that reﬂect the characteristics within an essential systemic and developmental trend. The method can be summarized in the follow-ing steps[14,23,48]:

Step 1: Find the initial average matrixA by assigning scores to each factor. Suppose we have n factors. Respondents (experts or stake-holders) are asked to rate the direct effects that factor i has on factor j using an integer scale ranging from 0 to 4 to repre-sent the range from“absolutely no inﬂuence (0)” to “very high inﬂuence (4)”. We then calculate the mean score among the respondents to arrive at element aijand form the initial average matrixA = [aij]n × n.

Step 2: Normalize the direct inﬂuence matrix D. Using matrix A, the normalized direct-relation matrixD = [dij]n × n is calculated using Eqs.(1) and (2).

D ¼ z A ð1Þ z¼ min 1= maxiX n j¼1 aij; 1= maxjX n i¼1 aij 8 < : 9 = ;; i; j∈ 1; 2; :::; nf g ð2Þ

Step 3: Calculate the total inﬂuence matrix T. The total inﬂuence matrix T can be obtained by summing the direct effects and all of the indirect effects using Eq.(3),

T ¼ D þ D2_{þ D}3_{þ … þ D}h_{¼ D I þ D þ D} 2_{þ … þ D}h−1h_ð_I−D_{Þ I−D}_ð _Þ−1i

¼ D I−D h_ð_I−D_Þ−1_;

whereI is denoted as the identity matrix and (I − D)(I − D)−1₌_{I. Then,}

T ¼ D I−Dð Þ−1; whenh→∞; Dh

¼ 0½ _nn ð3Þ

whereD = [dij]n × n, 0≤ dijb 1, 0 ≤ ∑idij≤ 1,

0≤∑jdij≤ 1, and at least one (but not all) of the columns or rows of the summation is equal to 1 inXn

j¼1

dijand

Xn i¼1

dij, and thus

we can guarantee thatlimh→∞Dh¼ 0½ nn.

We can denote the row and column sums of the total-inﬂuence matrixT as column vectors r and s respectively:

T ¼ tijh i

nn; i; j ¼ 1; 2; …; n; ð4Þ

Table 1

Explanation of criteria.

Dimensions/criteria Descriptions Proposed

in ref. Technology (D1)

Technology integration (C1)

Technology integration reduces incompatibility between legacy systems and enhances the

responsiveness of information systems.

[11,51]

Technology competence (C2)

Network technologies and enterprise systems that provide a platform on which the RFID applications can be built, installed in the organization.

[6,45]

Security concern (C3)

The degree to which the Internet platform is deemed secure for exchanging data and conducting online transactions.

Examples include personal data protection and security in using the RFID technology.

[2,51]

Organization (D2)

Top management support (C4)

Top management can provide vision, support, and a commitment to create a positive effect on the RFID adoption process.

[25]

Firm size (C5) Largeﬁrms typically have the resources

necessary to experiment, pilot, and decide what technology and standards they require.

[6,45]

Organizational readiness (C6)

Organizations must be prepared to make business process changes, and potential sites need to make adjustments for RFID if beneﬁts are to accrue.

[6]

Environment (D3)

Competitive pressure (C7)

By adopting RFID,firms may benefit from better inventory visibility, greater operation efficiency,

and more accurate data collection.

[6]

Partner readiness (C8)

Partner readiness refers to the degree to which aﬁrm's customers and suppliers are willing and ready to conduct business activities using RFID.

[3,49]

Regulatory support (C9)

This concept is similar to government policies that affect IT diffusion.

[50]

Cost (D4)

Hardware costs (C10)

The hardware costs of RFID adoption. [16]

Software costs (C11)

The software costs of RFID adoption. [16]

Implement costs (C12)

The implementation cost of RFID adoption, including work disruption, initial installation, management of associated change, etc.

[16]

Maintenance costs (C13)

The cost of maintaining the operation of the RFID system.

(4)

r ¼ ri½ n1¼ Xn j¼1 tij 2 4 3 5 n1 ; s ¼ sjh i_n1¼ X n i¼1 tij " #_′ 1n ð5Þ

where the superscript′ denotes transpose.

If ridenotes the row sum∑j = 1n tijof the ith row of matrixT, then ridenotes the sum of the direct and indirect effects that factor i has on all of the other factors. If sidenotes the column sum from matrixT, then sidenotes the sum of the direct and indirect effects that factor i has received from all of the other factors. Furthermore, (ri+ si) provides an index of the strength of the influences that are given and received; that is, (ri+ si) shows the degree of the total influences factor i has in this system. Therefore, if (ri− si) is positive, then factor i has a net influence on the other factors, and if (ri− si) is negative, then factor i is, on the whole, being influenced by the other factors

[43].

3.2. Combine the ANP method forﬁnding the inﬂuence weights of the criteria

We define the total influence matrix Tc= [tij]n × nby the criteria and TD= [tijD]m × m by the dimensions; TD can be obtain from Tc. Next, we normalize the total influence matrix Tcby each dimension and normalize the influence matrix TDby the total row sums shown asTcαand TDαrespectively to find the DANP influential weights by dimension. Then, the unweighted supermatrixW can be obtained by transposing the normalized total influence matrix Tcα to bring it into congruence with the definition of an ANP supermatrix, i.e., W = (Tcα)′. We can subsequently obtain the weighted supermatrix Wα₌_T

DαW (i.e., the normalized supermatrix W). Finally, the DANP inﬂuence weights can be obtained by taking the limg→∞Wαg, where g represents any number as a power. The procedures can be described inﬁve steps:

Step 1: The total inﬂuence matrix for criteria Tc= [tij]n × n. The total inﬂuence matrix Tcfor the criteria is shown below:(6)

ð6Þ

Step 2: The normalized total inﬂuence matrix for criteria Tcα. The normalized total inﬂuence matrix Tcαfor the criteria is shown below.(7)

ð7Þ

For example, an explanation for the normalization ofTcα11on dimension 1 based on dimension 1 (α11) is shown by Eqs.(8) and (9). d11ci ¼ Xm1 j¼1 t11ij ; i ¼ 1; 2; …; m1 ð8Þ Tcα11¼ t11c11=d11 c1 ⋯ t 11 c1j=d11 c1 ⋯ t 11 c1m1=d 11 c1 ⋮ ⋮ ⋮ t11ci1=d11 ci ⋯ t 11 cij=d 11 ci ⋯ t 11 cim1=d 11 ci ⋮ ⋮ ⋮ t11cm11=d 11 cm1 ⋯ t 11 cm1j=d 11 cm1 ⋯ t 11 cm1m1=d 11 cm1 2 6 6 6 6 6 4 3 7 7 7 7 7 5 ¼ tα11c11 ⋯ tα11c1j ⋯ tα11c1m1 ⋮ ⋮ ⋮

tα11ci1 ⋯ tα11cij ⋯ tα11cim1

⋮ ⋮ ⋮ tα11cm11 ⋯ t α11 cm1j ⋯ t α11 cm1m1 2 6 6 6 6 6 4 3 7 7 7 7 7 5 ð9Þ

where tcijα11= tcij11/dci11 denotes the element of normalized influence for the element tcij11 (shows that the element of i influences other j (j = 1, 2,..., m1) in which dimension 1 in flu-ences dimension 1 of total influence matrix) divided by the sum d11ci d11ci ¼ Xm1 j¼1 t11ij ; i ¼ 1; 2; …; m1 0 @ 1

A of each row (criterion i inﬂuences all other criteria in dimension 1).

Step 3: Find the unweighted supermatrixW by transposing the normalized total matrixTcα. Because the total influence matrix Tcmatches andfills the interdependence among dimensions and criteria, we can transpose the normalized total influence matrix Tcα by the dimensions based on the basic concept of ANP resulting in the unweighted supermatrix W = (Tcα)′ as shown by Eq.(10).(10)

ð10Þ

Step 4: Find the weighted normalized supermatrixWα_{. To obtain the} weighted supermatrixWα_{from the unweighted supermatrix} W, we can multiply the normalized total influence matrix TDα by the unweighted supermatrixW. The normalized total influ-ence matrixTDαcan be obtained by using to normalize total in-fluence matrix TDin process as shown from Eq.(11)to Eq.(12).

TD¼ t11D ⋮ ti1D ⋮ tn1D ⋯ ⋯ ⋯ t1j_D ⋯ ⋮ tijD ⋯ ⋮ tnjD ⋯ t1nD ⋮ tinD ⋮ tnnD 2 6 6 6 6 6 4 3 7 7 7 7 7 5 ð11Þ

We normalized the total inﬂuence matrix TDof the dimensions (Eq.(11)) and obtained a new normalized total inﬂuence ma-trixTDαof dimensions as shown by Eq.(12)(where tDαij= tDij/di anddi¼ Xn j¼1 tijD). TαD¼ t11D=d1 ⋯ t 1j D=d1 ⋯ t 1n D=d1 ⋮ ⋮ ⋮ ti1D=di ⋯ t ij D=di ⋯ tin D=di ⋮ ⋮ ⋮ tn1D=dn ⋯ t nj D=dn ⋯ tnn D=dn 2 6 6 6 6 4 3 7 7 7 7 5 ¼ tα11D ⋯ tα1jD ⋯ tα1nD ⋮ ⋮ ⋮

tαi1D ⋯ tαijD ⋯ tαinD

⋮ ⋮ ⋮ tαn1D ⋯ tαnjD ⋯ tαnnD 2 6 6 6 6 4 3 7 7 7 7 5 ð12Þ

(5)

Next, we multiplied the normalized total inﬂuence matrix of the dimensions_TDα, with the unweighted supermatrixW to ob-tain the new weighted supermatrixWα_{(i.e., by the normalized} matrix) as shown Eq.(13).

Wα¼ TDαW ¼ tα11D W 11 ⋯ tαi1D W i1 ⋯ tαn1D W n1 ⋮ ⋮ ⋮ tα1jD W 1j ⋯ tαijD W ij ⋯ tαnjD W nj ⋮ ⋮ ⋮ tα1nD W 1n ⋯ tαin D W in ⋯ tαnn D W nn 2 6 6 6 6 4 3 7 7 7 7 5 ð13Þ

Step 5: Find the limit of the weighted supermatrixWα_{by raising it to a} sufficiently large power g (i.e. g → ∞). If we raise the weighted supermatrix Wα _{to a suf}_{ficiently large power g, then the} weighted normalized supermatrixWα_{converges and becomes} a long-term stable supermatrix, i.e., lim_g→∞Wαg, where g represents any number as a power. Consequently, we can obtain what DANP calls the influential weights (i.e., global influential weights).

3.3. The VIKOR method for ranking and improving the alternatives Opricovic[27]proposed the compromise ranking method (VIKOR) as a technique that could be implemented within the MCDM model

[28–31,40–42]. If the feasible alternatives are represented by A1, A2, …, Ak,…, Am, the performance scores of alternative Akin each criteri-on j can be denoted by fkj(k = 1, 2,..., m; j = 1, 2,..., n); wjis the in ﬂu-ential weight (by DANP) of the jth criterion, where j = 1, 2,…, n., and n is the number of criteria. We deﬁne the best fj⁎ values (aspired level) and the worst fj−values (tolerable level) of all of the criterion functions, j = 1, 2,…, n. Next, we began the development of the VIKOR method using the following form of the Lp− metric:

Lp_k¼ X n j¼1 wj f j−fkj = fj−f−j h ip 8 < : 9 = ; 1=p ð14Þ

where 1≤ p ≤ ∞; k = 1, 2, …, m; the weight wjis derived from the DANP (the so-called DDANPV method combines the DEMATEL, ANP, and VIKOR methods). To formulate the ranking and gap measures, L_kp = 1 _{(as S}

k) and L_kp =∞ (as Qk) are used in the VIKOR method

[27,28,30,31,41,42]. Sk¼ L p¼1 k ¼ Xn j¼1 wj f j−fkj = fj−f−j h i ð15Þ Qk¼ L p¼∞ k ¼ max j f j−fkj = fj−f−j j¼ 1; 2; ⋯; n j g n ð16Þ

The compromise solutionmin

k L p

kshows that the synthesized/integrated

gap is the minimum and, as a result, will be selected, as its value is the closest to the aspired level. In addition, the group utility (average gap) is emphasized when p is small (such as p = 1); however, if p is inﬁnite, then the individual maximum regrets/gaps gain more importance in prior improvement (basic concept from Yu[44]and Freimer and Yu

[10]) of each dimension/criterion. Consequently, minkSk stresses the maximum group utility for the majority (in other words, shown for min-imizing average gap); however, minkQkstresses selecting the minimum from the maximum individual regrets/gaps (in other words, shown which maximum gap for prior improvement). Following the above afore-mentioned ideas, weﬁnd that the compromise ranking and improvement algorithm VIKOR has four steps as described below.

Step 1: Obtain an aspired or tolerable level. We calculate the best fj⁎ values (aspired level) and the worst fj− values (tolerable level) of all of the criterion functions, j = 1, 2, …, n. For example, the performance value of each criterion can be obtained by using questionnaires with a scale ranging from 0 point (complete dissatisfaction) to 10 points (the best sat-isfaction). Therefore, we can set the aspired level as fj∗= 10 and the worst value as fj−= 0. As a result, in this research, we are setting fj∗= 10 as the aspired level and setting fj−= 0 as the worst value for normalization, in contrast to the traditional approach, which sets fj ¼ maxkfkj and f−j ¼ minkfkj. We propose this new idea for improvement to avoid the traditional approach of “choosing the best among the inferior choices/options/alternatives (i.e., pick the best apple in a barrel of rotten apples)”. The original per-formance rating matrix can be converted into a normalized gaps-rating matrix [rkj]m × n(where the rating rkjshows the gap of alternative k in j criterion; how can we reduce the gaps of each criterion and dimension based on the inﬂuential network relation map for achieving the aspired level?) using the following equation:

rkj¼ f j−fkj = fj−f−j ð17Þ

Step 2: Calculate the means of group utility and maximal regret. These gap-values can be computed using the rating-weighted Sk= ∑j = 1n wjrkj(i.e., the synthesized/integrated gap for all criteria) and Qk¼ max j rkjjj¼ 1; 2; …; ng

n

(shown which the maximal gap of alternative k for prior improvement in each dimension and overall criteria respectively).

Table 2

The initial inﬂuence matrix A for the criteria.

Criteria C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12 C13 C1 0.000 2.400 2.400 2.400 1.800 2.800 2.000 3.000 1.600 2.600 3.000 2.800 3.000 C2 2.200 0.000 3.200 2.400 1.200 2.200 2.400 2.000 2.800 1.800 2.200 2.200 1.800 C3 2.400 2.400 0.000 3.000 1.200 1.800 2.200 2.200 3.400 2.200 2.200 1.800 1.800 C4 3.400 2.200 2.800 0.000 2.600 3.200 2.200 2.200 2.600 2.400 2.800 2.800 3.000 C5 2.200 2.000 2.200 2.200 0.000 2.800 2.000 1.600 2.000 2.800 2.400 2.200 2.600 C6 2.200 1.800 2.400 2.400 2.400 0.000 2.000 2.600 2.400 1.600 1.600 2.800 2.400 C7 2.000 1.800 2.000 2.400 2.000 2.400 0.000 1.800 1.400 1.600 1.800 2.400 2.200 C8 2.400 2.400 2.400 2.200 1.600 2.400 2.800 0.000 1.400 2.200 2.400 2.600 2.200 C9 3.000 1.600 2.800 2.800 2.200 2.400 2.400 2.400 0.000 2.200 1.600 2.200 2.800 C10 2.600 2.000 1.400 2.200 2.200 2.000 1.800 1.800 1.800 0.000 2.400 2.600 3.000 C11 3.200 1.600 2.600 2.400 2.200 2.200 1.800 1.600 1.400 2.400 0.000 2.800 2.200 C12 2.800 2.200 2.000 2.600 2.400 2.400 2.000 1.800 2.200 2.400 2.600 0.000 3.200 C13 2.800 2.000 1.800 2.400 2.800 2.400 2.600 2.600 1.600 2.400 2.200 2.200 0.000

(6)

Step 3: Calculate the index value. This value can be measured by the following equation: Rk¼ v Sk−S = S −−S þ 1−vð Þ Q k−Q= Q −−Q; v∈ 0; 1½ ð18Þ where S¼ miniSi(traditional approach); or let S* = 0 (no gap, the aspiration level is achieved in our approach); S−¼ maxiSi (traditional approach), or let S−= 1 (the worst situation in our approach); Q¼ miniQi (traditional ap-proach), or let Q* = 0 (no gap, the aspiration level is achieved in our approach); and Q−¼ maxiQi(traditional approach), or let Q−= 1 (the worst situation in our approach). Eq.(18)can be rewritten as Rk= vSk+ (1− v)Qkwhen S* = 0 and Q* = 0 (i.e., all criteria have achieved their corresponding aspiration levels), and S−= 1 and Q−= 1 (i.e., the worst situation). How do decision makers determine the v value? When v = 1, only the average gap (the average regret) is considered in each dimension or overall; when v = 0, only the maximum gap in the improvement is considered a priority for the criterion in each dimension or overall. The value obtained from miniSi represents the maximum group utility (the minimum average gap indicator), and the value obtained from maxiQirepresents the maximum regret (the largest gap shown as priority im-provement). Thus, v represents the weight of the strategy. Generally v = 0.5, which can be adjusted depending on the case under consideration from the view-points of dimensions and overall for improvement priority; v = 1 indicates that only the average gap is considered, and v = 0 indicates that only the maximum gap is prioritized for improvement individually.

The compromise ranking method (VIKOR) is applied to determine the compromise solution by measured gaps. This solution is useful for decision-makers because it offers the maximum group utility for the majority (shown by min S, i.e., shown for minimizing average gap) and the maximum regret (basic concept from Yu[44]) of the minimum number of individuals of the opponent (shown by min Q, i.e., shown which maximum gap for priority improvement) to reduce the gaps for improving the performance values in each criterion and dimension for-ward to achieving the aspired levels in each dimensions and criterion.

4. An empirical case study on RFID adoption in Taiwan's healthcare industry

In this section, we present an empirical study using the proposed DDANPV model to evaluate, select, and improve upon the best alter-native for RFID adoption in Taiwan's healthcare industry.

4.1. Background and problem descriptions

In the environment of a hospital, where service demand might be unpredictable, and the infrastructure is often complex, the speed with which critical medical assets can be located might determine the out-come of a hospital's mission to save lives[19,22,35]. RFID has received considerable attention because it promises to meet the challenge of tracking people and locating items within a large building complex. However, the adoption of new technology involves an analysis of its costs and beneﬁts, and managers in the healthcare industry need an evaluation framework to help them decide whether to adopt RFID technologies, and if so, what type or conﬁguration of RFID applica-tions they should adopt.

Table 3

The normalized direct-inﬂuence matrix D for the criteria.

Criteria C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12 C13 C1 0.000 0.075 0.075 0.075 0.056 0.087 0.062 0.093 0.050 0.081 0.093 0.087 0.093 C2 0.068 0.000 0.099 0.075 0.037 0.068 0.075 0.062 0.087 0.056 0.068 0.068 0.056 C3 0.075 0.075 0.000 0.093 0.037 0.056 0.068 0.068 0.106 0.068 0.068 0.056 0.056 C4 0.106 0.068 0.087 0.000 0.081 0.099 0.068 0.068 0.081 0.075 0.087 0.087 0.093 C5 0.068 0.062 0.068 0.068 0.000 0.087 0.062 0.050 0.062 0.087 0.075 0.068 0.081 C6 0.068 0.056 0.075 0.075 0.075 0.000 0.062 0.081 0.075 0.050 0.050 0.087 0.075 C7 0.062 0.056 0.062 0.075 0.062 0.075 0.000 0.056 0.043 0.050 0.056 0.075 0.068 C8 0.075 0.075 0.075 0.068 0.050 0.075 0.087 0.000 0.043 0.068 0.075 0.081 0.068 C9 0.093 0.050 0.087 0.087 0.068 0.075 0.075 0.075 0.000 0.068 0.050 0.068 0.087 C10 0.081 0.062 0.043 0.068 0.068 0.062 0.056 0.056 0.056 0.000 0.075 0.081 0.093 C11 0.099 0.050 0.081 0.075 0.068 0.068 0.056 0.050 0.043 0.075 0.000 0.087 0.068 C12 0.087 0.068 0.062 0.081 0.075 0.075 0.062 0.056 0.068 0.075 0.081 0.000 0.099 C13 0.087 0.062 0.056 0.075 0.087 0.075 0.081 0.081 0.050 0.075 0.068 0.068 0.000 Table 4

The total inﬂuence matrix Tcfor the criteria.

Criteria C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12 C13 C1 0.471 0.444 0.492 0.513 0.432 0.519 0.457 0.479 0.423 0.480 0.501 0.526 0.543 C2 0.485 0.335 0.470 0.467 0.374 0.456 0.425 0.410 0.418 0.415 0.435 0.462 0.461 C3 0.496 0.408 0.384 0.488 0.379 0.451 0.424 0.420 0.437 0.430 0.439 0.456 0.467 C4 0.604 0.467 0.536 0.479 0.483 0.565 0.493 0.488 0.480 0.507 0.528 0.560 0.579 C5 0.493 0.399 0.449 0.469 0.346 0.481 0.421 0.406 0.401 0.450 0.448 0.470 0.492 C6 0.489 0.391 0.451 0.471 0.411 0.397 0.418 0.430 0.409 0.414 0.423 0.482 0.483 C7 0.441 0.357 0.402 0.430 0.366 0.427 0.323 0.372 0.348 0.377 0.391 0.431 0.436 C8 0.497 0.410 0.454 0.468 0.391 0.469 0.442 0.357 0.384 0.432 0.447 0.480 0.480 C9 0.538 0.407 0.486 0.507 0.428 0.492 0.452 0.448 0.361 0.454 0.447 0.492 0.520 C10 0.488 0.387 0.413 0.453 0.397 0.445 0.402 0.398 0.381 0.356 0.434 0.466 0.488 C11 0.514 0.384 0.453 0.468 0.404 0.459 0.409 0.401 0.380 0.434 0.374 0.480 0.475 C12 0.535 0.425 0.467 0.504 0.436 0.494 0.443 0.433 0.426 0.461 0.476 0.430 0.533 C13 0.521 0.409 0.449 0.485 0.435 0.482 0.448 0.443 0.398 0.449 0.454 0.482 0.429 Note:1 n2 Xn i¼1 Xn j¼1 tp ij−tp−1ij tp ij

100%= 2.439%b 5%, i.e., significant confidence level is 97.561%, where p = 15 denotes the number of experts and tijpis the average influence of criterion ion

criterion j. Here n denotes the number of criteria, with n = 13 and n2

(7)

4.2. Data collection

The data in this study were collected from 15 experts with profes-sional management and decision-making experience in the healthcare industry. Most of these experts had worked in the healthcare industry for more than ten years, and their responses were collected via personal interviews and questionnaires in May 2011. The objects of this ques-tionnaire are the experts and not the users, with the goal of analyzing user behavior. In this respect, it is not the distribution of the sample size that is at issue but rather the consensus of the expert opinions. In other words, we need to test the consensus of the experts. If the number of experts increases, the degree of consensus should increase so that the differences in their responses will decrease. For thefifteen experts, a 97.561% significance confidence level is obtained (see the notes below

Table 4).

4.3. Construct the network relation map using DEMATEL

The DEMATEL technique introduced in Section 3.1 is used to analyze the interrelationships between the 13 criteria summarized

from the literature. First, the direct influence matrix A for the criteria is obtained (seeTable 2). Next, the normalized direct-influence ma-trix_{D for criteria can be calculated by Eq.}(1)(seeTable 3). Third, the total direct influence matrices Tcfor the criteria andTDfor the di-mensions are calculated based on Eq.(3)(seeTables 4 and 5). Finally, the in_{fluence network relation map (INRM) can be constructed using} the vectors r and s in the total direct influence matrix TD(seeTable 6), as shown inFig. 1.

4.4. Using DANP to calculate the influence weights for each criterion The in_{fluence weights (global weights) for the 13 criteria can be} calculated by using DANP, as shown inTables 7–9. The results show that the experts consider technology integration, top management support, and organizational readiness as the most important criteria, with influence weights of 0.094, 0.089, and 0.088, respectively; they are least concerned with software costs and hardware costs, with influence weights of 0.062 and 0.061, respectively. In the tech-nology dimension, the experts consider techtech-nology integration to be the most important criterion. In the organization dimension, the ex-perts think that top management support is the most important criteri-on. In the environment dimension, the experts consider competitive pressure to be the most important criterion. In the cost dimension, the experts consider maintenance costs to be the most important criterion. These_{findings reveal that the experts believe technology integration} should not be overlooked by managers when selecting a method to evaluate the RFID adoption process. Additionally, wefind that the ex-perts are less concerned about technology and environmental dimen-sions, as the means of these dimensions are substantially lower than those of the other dimensions.

4.5. Compromise ranking by using VIKOR

We apply the VIKOR method to determine the compromise rankings after calculating the inﬂuence weights for the criteria using DANP inSection 4.4. The results of our calculations (Table 10) show that the total gaps are the largest in asset tracking management performance (0.408), meaning that the alternative program shouldﬁrst improve Table 5

The total inﬂuence matrix TDfor the dimensions.

Dimensions D1 D2 D3 D4 D1technology 0.443 0.453 0.433 0.468 D2organization 0.475 0.456 0.438 0.486 D3environment 0.444 0.442 0.387 0.449 D4cost 0.454 0.455 0.414 0.451 Table 6

The sum of inﬂuences given and received on the dimensions.

Dimensions ri si ri+ si ri − si D1technology 1.796 1.815 3.612 −0.019 D2organization 1.856 1.806 3.662 0.050 D3environment 1.722 1.672 3.394 0.050 d4cost 1.774 1.854 3.628 −0.081

Technology

Patient tracking

management performance

Technology competence

Security concern

Top management support

Organizational readiness

Firm size

Organization

Environment

Cost

Technology integration

Partner readiness

Regulatory support

Competitive pressure

Hardware cost

Implement cost

Software cost

Maintenance cost

Asset tracking

management performance

(8)

asset tracking management performance and then improve patient-tracking management performance. Therefore, in the optimal RFID adoption application, managers should focus on how to improve asset tracking management performance to achieve the desired level of performance.

4.6. Implications and discussion

There are several important results of our study. First, according to our DANP results, technology integration is the most important crite-rion for evaluating RFID adoption with an inﬂuence weight of 0.094. By reducing the incompatibility between legacy systems and enhancing the responsiveness of information systems, technology integration exerts an important effect on the adoption and diffusion of new tech-nologies in an organization and helps improve performance via a reduc-tion in cycle times, better customer services, and lower procurement

costs[3,11]. Similar to thefindings in other industries studies of new technology adoption, wefind that a firm's ability to convert new tech-nology into core capabilities is essential and that techtech-nology integration is the most significant factor when evaluating RFID adoption in the healthcare industry.

Second, top management support is the second most important criterion, with an inﬂuence weight of 0.089. This ﬁnding also echoes the results obtained in previous studies, where top management support is shown to be a key factor in overcoming resistance to changes caused by new technology adoption and diffusion[13]. Thus, managers in the healthcare industry should regard strong commitment and support within top management as key to successful RFID adoption.

Third, compromise ranking from VIKOR (seeTable 10) shows that, between the choice of the two RFID applications, the system with better patient-tracking management system (total gaps = 0.373) is pre-ferred to the system with better asset-tracking management system Table 7

The unweighted supermatrixW = (Tcα)′.

Criteria C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12 C13 C1 0.335 0.375 0.381 0.374 0.366 0.365 0.365 0.364 0.373 0.374 0.374 0.371 0.373 C2 0.313 0.264 0.315 0.291 0.299 0.296 0.299 0.302 0.288 0.301 0.289 0.299 0.298 C3 0.351 0.361 0.303 0.335 0.336 0.339 0.336 0.335 0.339 0.325 0.337 0.330 0.329 C4 0.367 0.377 0.386 0.333 0.376 0.385 0.368 0.368 0.372 0.367 0.369 0.370 0.365 C5 0.262 0.256 0.256 0.281 0.238 0.285 0.267 0.262 0.265 0.272 0.270 0.270 0.277 C6 0.371 0.366 0.359 0.386 0.386 0.330 0.365 0.370 0.363 0.361 0.362 0.360 0.359 C7 0.338 0.340 0.334 0.338 0.343 0.335 0.313 0.369 0.357 0.341 0.344 0.343 0.348 C8 0.351 0.330 0.330 0.334 0.332 0.341 0.354 0.305 0.353 0.337 0.337 0.332 0.341 C9 0.311 0.330 0.336 0.328 0.325 0.324 0.333 0.326 0.290 0.322 0.320 0.325 0.311 C10 0.231 0.231 0.237 0.231 0.239 0.228 0.229 0.233 0.234 0.205 0.241 0.239 0.242 C11 0.249 0.249 0.248 0.246 0.243 0.242 0.244 0.247 0.239 0.253 0.220 0.255 0.254 C12 0.260 0.262 0.257 0.260 0.256 0.267 0.264 0.262 0.259 0.269 0.273 0.232 0.268 C13 0.260 0.258 0.258 0.263 0.262 0.263 0.263 0.258 0.267 0.273 0.267 0.274 0.236 Table 8

The normalized supermatrixWα₌_T D α_W. Criteria C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12 C13 C1 0.083 0.078 0.086 0.088 0.074 0.089 0.081 0.085 0.075 0.061 0.064 0.067 0.069 C2 0.093 0.064 0.090 0.091 0.073 0.089 0.082 0.079 0.080 0.061 0.064 0.068 0.068 C3 0.095 0.078 0.073 0.093 0.073 0.086 0.080 0.079 0.082 0.063 0.064 0.066 0.068 C4 0.096 0.075 0.085 0.077 0.078 0.091 0.080 0.079 0.078 0.061 0.064 0.068 0.070 C5 0.094 0.076 0.086 0.089 0.066 0.091 0.081 0.078 0.077 0.063 0.063 0.066 0.069 C6 0.094 0.075 0.087 0.090 0.079 0.076 0.079 0.081 0.077 0.060 0.061 0.070 0.070 C7 0.095 0.077 0.086 0.090 0.077 0.090 0.070 0.080 0.075 0.060 0.062 0.069 0.069 C8 0.094 0.078 0.086 0.090 0.076 0.091 0.084 0.068 0.073 0.061 0.063 0.068 0.068 C9 0.097 0.073 0.087 0.091 0.077 0.089 0.081 0.080 0.064 0.062 0.061 0.067 0.071 C10 0.097 0.077 0.082 0.090 0.079 0.088 0.079 0.079 0.075 0.052 0.063 0.068 0.071 C11 0.097 0.073 0.086 0.090 0.078 0.088 0.080 0.079 0.074 0.063 0.054 0.069 0.069 C12 0.096 0.076 0.084 0.090 0.078 0.088 0.079 0.078 0.076 0.062 0.064 0.058 0.071 C13 0.097 0.076 0.083 0.089 0.080 0.088 0.081 0.080 0.072 0.063 0.064 0.068 0.060 Table 9

The stable matrix of DANP when the power limit is g→ ∞, i.e., limg→∞Wαg.

Criteria C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12 C13 C1 0.094 0.075 0.085 0.089 0.076 0.088 0.080 0.079 0.075 0.061 0.062 0.067 0.069 C2 0.094 0.075 0.085 0.089 0.076 0.088 0.080 0.079 0.075 0.061 0.062 0.067 0.069 C3 0.094 0.075 0.085 0.089 0.076 0.088 0.080 0.079 0.075 0.061 0.062 0.067 0.069 C4 0.094 0.075 0.085 0.089 0.076 0.088 0.080 0.079 0.075 0.061 0.062 0.067 0.069 C5 0.094 0.075 0.085 0.089 0.076 0.088 0.080 0.079 0.075 0.061 0.062 0.067 0.069 C6 0.094 0.075 0.085 0.089 0.076 0.088 0.080 0.079 0.075 0.061 0.063 0.067 0.069 C7 0.094 0.075 0.085 0.089 0.076 0.088 0.080 0.079 0.075 0.061 0.063 0.067 0.069 C8 0.094 0.075 0.085 0.089 0.076 0.088 0.080 0.079 0.076 0.061 0.062 0.067 0.069 C9 0.094 0.075 0.085 0.089 0.076 0.088 0.080 0.079 0.076 0.061 0.063 0.067 0.069 C10 0.094 0.075 0.085 0.089 0.076 0.088 0.080 0.079 0.075 0.061 0.062 0.067 0.069 C11 0.094 0.075 0.085 0.089 0.076 0.088 0.080 0.079 0.075 0.061 0.063 0.067 0.069 C12 0.094 0.075 0.085 0.089 0.076 0.088 0.080 0.079 0.075 0.061 0.062 0.067 0.069 C13 0.094 0.075 0.085 0.089 0.076 0.088 0.080 0.079 0.076 0.061 0.062 0.067 0.069

(9)

(total gaps = 0.408). As mentioned previously, a patient-tracking man-agement system that used RFID in the check-in process demonstrated the utility of automated outbound logistical processes. In the healthcare

setting, this means getting the right patient to the right place at the right time. In addition, using RFID effectively could reduce the number of staff required to manage the patient check-in process, which results in an overall improvement in patient-tracking management performance. RFID can also be used to track healthcare assets, such as wheelchairs, infusion pumps, and crash carts. In the healthcare environment, assets (both equipment and staff) are essential to providing healthcare services to a patient.

Fourth, according to DEMATEL, we could look at the interrelation-ship among dimensions and criteria based on the influence network relation map (INRM) to help improve each dimension and criterion (seeFig. 2). The INRM shows that the environment dimension (D3) and the organization dimension (D2) are the highest priority for im-provement. Thisfinding means that managers should first improve these two dimensions because they are the most important relative to the other dimensions. Thus, the environment and the organization dimensions can be regarded as the critical dimension for evaluating and improving the RFID adoption process in the healthcare industry. In addition, with respect to the technology dimension (D1): technol-ogy competence (C2) is the most influential criterion and should be improved uponfirst, followed by technology integration (C1) and se-curity concerns (C3) (seeFig. 2for more details). In addition, with re-spect to the organization dimension (D2), top management support (C4) is the most influential criterion and should be improved upon first, followed by firm size (C5) and organizational readiness. With re-spect to the environment dimension (D3), regulatory support (C9) is the most influential criterion and should be improved upon first, followed by partner readiness (C8) and competitive pressure. With respect to the cost dimension (D4), hardware costs (C10) is the most influential criterion and should be improved upon first, followed by implementation costs (C12), software costs (C11), and maintenance costs (C13). Each of the evaluation dimensions and criteria identify the necessary behaviors for inducing RFID adoption in the healthcare industry. Therefore, managers should evaluate all of the dimensions and criteria for the RFID adoption process in accordance withFig. 2. While this evaluation method could in principle be used by most of the healthcare industries in the world, differences do exist, and the relative importance of the 13 criteria may vary according to the particulars of each healthcare industry. Managers should compare the evaluation methods for each RFID adoption model before deciding upon the best RFID application to suit their needs.

5. Conclusions

The dimensions and criteria outlined in this study serve as bridg-ing mechanisms for the evaluation of RFID adoption processes. Prior literature has identified the dimensions and criteria that influence the evaluation of adopting RFID. The main contributions of this study are twofold. First, the evaluation of technology adoption is a decision-making problem that is composed of complex dependences and interactions. In this paper we used previous studies to develop a TOEC framework to evaluate RFID adoption in the healthcare industry. Second, we combine the DEMATEL, DANP and VIKOR methods to de-velop an evaluation method known as DDANPV to prioritize the rela-tive influence-weights of the TOEC dimensions and criteria. DDANPV could handle the complex interactions and interdependences among dimensions and criteria and produce results that allow us to build a vi-sual cause-and-effect diagram for evaluating the various adoption processes. Additionally, we demonstrate how the results could provide guidance to managers by identifying the key criteria for decision-making andfinding the best way to improve existing RFID adoption processes.

This DDANPV method provides a general evaluation framework for industry evaluation and adoption of RFID and a guide for future managers in the healthcare industry even if they do not completely understand how to evaluate the details of the various RFID adoption Table 10

The inﬂuence weights for the criteria used in evaluating the alternatives and improving total performance by VIKOR.

Dimensions/ criteria Local weight

Global weight (by DANP) Patient tracking management performance (A1) Asset tracking management performance (A2) Technology (D1) 0.254 0.181 0.210 Technology integration (c1) 0.371 0.094 (1) 0.140 0.180 Technology competence (c2) 0.295 0.075 (9) 0.280 0.260 Security concern (c3) 0.334 0.085 (4) 0.140 0.200 Organization (D2) 0.253 0.390 0.342

Top management support (c4) 0.353 0.089 (2) 0.400 0.320

Firm size (c5) 0.299 0.076 (7) 0.460 0.440 Organizational readiness (c6) 0.348 0.088 (3) 0.320 0.280 Environment (D3) 0.234 0.316 0.374 Competitive pressure (c7) 0.340 0.080 (5) 0.500 0.480 Partner readiness (c8) 0.337 0.079 (6) 0.260 0.300 Regulatory support (c9) 0.323 0.075 (8) 0.180 0.340 Cost (D4) 0.259 0.289 0.327 Hardware cost (c10) 0.235 0.061 (13) 0.340 0.260 Software cost (c11) 0.241 0.062 (12) 0.300 0.320 Implement cost (c12) 0.259 0.067 (11) 0.220 0.360 Maintenance cost (c13) 0.265 0.069 (10) 0.300 0.360 Ak − 1.00 Total gaps 0.373 0.408 Notes:

1. Example for local weights calculations from global weights:

– The local weight for D1(sum global weights from criteria (c1, c2, c3)) was calculated

as follows.

0:094 þ 0:075 þ 0:085 ¼ 0:254; weight c1equals 0:094 0:254

¼ 0:371; …; then 0:371 þ 0:295 þ 0:334 ¼ 1:

– The local weight for D2(sum global weights from criteria (c4, c5, c6)),was calculated

as follows.

0:089 þ 0:076 þ 0:088 ¼ 0:253 and weight c4equals 0:089 0:253

¼ 0:353; …; then 0:353 þ 0:299 þ 0:348 ¼ 1:

– The local weight for D3(sum global weights from criteria (c7, c8, c9)), was calculated

as follows.

0:080 þ 0:079 þ 0:075 ¼ 0:234 and weight c7equals 0:080 0:234

¼ 0:341; :::; then 0:340 þ 0:337 þ 0:323 ¼ 1:

– The local weight for D4(sum global weights from criteria (c10, c11, c12, c13)),was

calculated as follows.

0:061 þ 0:062 þ 0:067 þ 0:069 ¼ 0:259 and weight equal 0:061 0:259 ¼ 0:235; :::; then 0:235 þ 0:241 þ 0:259 þ 0:265 ¼ 1:

– The local weight for overall dimensions was calculated as follows: 0.254 + 0.253 + 0.234 + 0.259 = 1.

2. Example for gaps performance for patient tracking management performance: – Calculating total performance by global weights:

0:094 0:141 þ 0:075 0:280 þ 0:085 0:140 þ 0:089 0:400 þ … þ 0:069 0:300

¼ 0:373:

– Calculating total performance by local weights:

0:254 0:181 þ 0:254 0:390 þ 0:234 0:316 þ 0:254 0:289 ¼ 0:373: – Integrating performance from criteria (c1, c2, c3) to dimension (D2) by local

weights (c1, c2, c3):

(10)

models. Moreover, the INRM diagram helps decision makers under-stand how to improve their evaluations of RFID adoption processes. Future research could expand the DDANPV method into a general evaluation framework for industry adoption of new technologies.

There are several limitations to this study that require further ex-amination. First, this study was conducted by surveying a relatively limited number of experts. A larger sample would have allowed for a more sophisticated analysis of evaluation procedures, which would have generalized the results of this study. Second, this study uses crisp numbers as opposed to fuzzy numbers. Future studies could incorporate fuzzy numbers to estimate the relative in fluence-weights of each influence on the evaluation method. Third, the TOEC evaluation criteria are selected from a review of prior literature on TOE and cost evaluation, which excluded some possible influences on the RFID evaluation process. Future studies could use different methods, such as longitudinal studies and interviews, to identify other criteria. Finally, to provide more objective information on the applicability of the proposed TOEC evaluation model, future studies could use case studies of particular performance evaluations and thus prove the practicality of the general evaluation framework for the industry evaluation and adoption of RFID proposed in this study.

References

[1] Z. Asif, M. Mandviwalla, Integrating the supply chain with RFID: a technical and business analysis, Communications of the Association for Information Systems 15 (24) (2005) 393–427.

[2] J. Ayoade, Security implications in RFID and authentication processing framework, Computer and Security 25 (3) (2006) 207–212.

[3] A. Barua, P. Konana, A.B. Whinston, F. Yin, An empirical investigation of Net-enabled business value: an exploratory investigation, MIS Quarterly 28 (4) (2004) 585–620.

[4] E. Bendoly, A. Citurs, B. Konsynski, Internal infrastructural impacts of RFID per-ceptions and commitment: knowledge, operational procedures, and information-processing standards, Decision Sciences 38 (3) (2007) 423–449.

[5] B. Bilge, I. Ozkarahan, Strategic tactical and operational production-distribution models: a review, International Journal of Technology Management 28 (2) (2004) 151–171.

[6] I. Brown, J. Russell, Radio frequency identiﬁcation technology: an exploratory study on adoption in the South African retail sector, International Journal of Information Management 27 (4) (2007) 250–265.

[7] N.G. Carr, IT doesn't matter, Harvard Business Review 81 (5) (2003) 41–49.

[8] F.H. Chen, T.S. Hsu, G.H. Tzeng, A balanced scorecard approach to establish a performance evaluation and relationship model for hot spring hotels based on a hybrid MCDM model combining DEMATEL and ANP, International Journal of Hospitality Management 30 (4) (2011) 908–932.

[9] C.C. Chao, J.M. Yang, W.Y. Jen, Determining technology trends and forecasts of RFID by historical review and bibliometric analysis from 1991 to 2005, Technovation 27 (5) (2007) 268–279.

[10]M. Freimer, P.L. Yu, Some new results on compromise solutions for group decision problems, Management Science 22 (6) (1976) 688–693.

(11)

[11] D.L. Goodhue, M.D. Wybo, L.J. Kirsch, The impact of data integration on the costs and beneﬁts of information systems, MIS Quarterly 16 (3) (1992) 293–311.

[12] W. Hong, K. Zhu, Migrating to internet-based e-commerce: factors affecting e-commerce adoption and migration at theﬁrm level, Information Management 43 (2) (2006) 204–221.

[13] M.T. Hoske, RFID: adoption increases despite costs, Control Engineering 51 (7) (2004) 46–47.

[14] Y.H. Hung, S.C. Chou, G.H. Tzeng, Knowledge management adoption and assess-ment for SMEs by a novel MCDM approach, Decision Support Systems 51 (2) (2011) 270–291.

[15] C.Y. Huang, J.Z. Shyu, G.H. Tzeng, Reconﬁguring the innovation policy portfolios for Taiwan's SIP mall industry, Technovation 27 (12) (2007) 744–765.

[16] M.C. Jones, R.C. Beatty, Towards the development of measures of perceived beneﬁts and compatibility of EDI: a comparative assessment of competing ﬁrst order factor models, European Journal of Information Systems 7 (3) (1998) 210–220.

[17] P. Jones, C. Clarke-Hill, D. Hillier, D. Comfort, The beneﬁts, challenges and impacts of radio frequency identiﬁcation (RFID) for retailers in the UK, Marketing Intelligence and Planning 23 (4) (2005) 395–402.

[18] P. Jones, C. Clarke-Hill, P. Shears, D. Hillier, D. Comfort, Radio frequency identiﬁcation in retailing and privacy and public policy issues, Management Research News 27 (8/9) (2004) 46–56.

[19] B. Kaplan, The medical computing“lag”: perceptions of barriers to the application of computers to medicine, International Journal of Technology Assessment in Health Care 3 (1) (1987) 123–126.

[20] B. Kinsella, The Wal-mart factor, Industrial Engineer 35 (11) (2003) 32–36.

[21] L.S. Lee, K.D. Fiedler, J.S. Smith, Radio frequency identiﬁcation (RFID) implementa-tion in the service sector: a customer-facing diffusion model, Internaimplementa-tional Journal of Production Economics 112 (2) (2008) 587–600.

[22] S.H. Lee, A.W. Ng, K. Zhang, The quest to improve Chinese healthcare: some fundamental issues, International Journal of Health Care Quality Assurance 20 (5) (2007) 416–428.

[23] J.J.H. Liou, G.H. Tzeng, H.C. Chang, Airline safety measurement using a hybrid model, Journal of Air Transport Management 13 (4) (2007) 243–249.

[24] C.H. Liu, G.H. Tzeng, M.H. Ming-Huei Lee, Improving tourism policy implementation — the use of hybrid MCDM models, Tourism Management 33 (2) (2012) 239–488.

[25] C. Loebbecke, J. Palmer, RFID in the fashion industry: Kaufhof Department Stores AG and Gerry Weber International AG, fashion manufacturer, MIS Quarterly Executive 5 (2) (2006) 15–25.

[26] E.W.T. Ngai, T.C.E. Cheng, S. Au, K. Lai, Mobile commerce integrated with RFID technology in a contained depot, Decision Support Systems 43 (1) (2007) 62–76.

[27] S. Opricovic, Multicriteria Optimization of Civil Engineering Systems, Faculty of Civil Engineering, Belgrade, 1998. , (in Serbian).

[28] S. Opricovic, G.H. Tzeng, Multicriteria planning of post-earthquake sustainable reconstruction, Computer-Aided Civil and Infrastructure Engineering 17 (3) (2002) 211–220.

[29] S. Opricovic, G.H. Tzeng, Fuzzy multicriteria model for post-earthquake land-use planning, Natural Hazards Review 4 (2) (2003) 59–64.

[30] S. Opricovic, G.H. Tzeng, Compromise solution by MCDM methods: a comparative analysis of VIKOR and TOPSIS, European Journal of Operational Research 156 (2) (2004) 445–455.

[31] S. Opricovic, G.H. Tzeng, Extended VIKOR method in comparison with outranking methods, European Journal of Operational Research 178 (2) (2007) 514–529.

[32] W. Orlikowski, CASE tools as organisational change: investigating incremental and radical changes in systems development, MIS Quarterly 17 (3) (1993) 309–340.

[33] Y.P. Ou Yang, H.M. Shieh, J.D. Leu, G.H. Tzeng, A novel hybrid MCDM model combined with DEMATEL and ANP with applications, International Journal of Oper-ations Research 5 (3) (2008) 160–168.

[34] B. Öztaysi, S. Baysan, F. Akpinar, Radio frequency identiﬁcation (RFID) in hospitality, Technovation 29 (9) (2009) 618–624.

[35] A. Oztekin, F.M. Pajouh, D. Delen, L.K. Swim, An RFID network design methodology for asset tracking in healthcare, Decision Support Systems 49 (1) (2010) 100–109.

[36] G. Premkumar, M. Roberts, Adoption of new information technologies in rural small businesses, International Journal of Management Science 27 (4) (1999) 467–484.

[37] C. Ranganathan, S. Jha, Adoption of RFID technology: an exploratory examination from supplier's perspective, Proceedings of the Eleventh American Conference on Information Systems, 11–14 August, Omaha, USA, 2005, pp. 2195–2199.

[38] P. Schmitt, F. Thiesse, E. Fleisch, Adoption and diffusion of RFID technology in the automotive industry, in: H. Österle, J. Schelp, R. Winter (Eds.), Proceedings of the 15th European Conference on Information Systems, St Gallen, Switzerland, 2007.

[39] A. Sharma, A. Citurs, Radio frequency identiﬁcation (RFID) adoption drivers: a radical innovation adoption perspective, Proceedings of the Eleventh American Conference on Information Systems, 11–14 August, Omaha, USA, 2005, pp. 1213–1218.

[40] G.H. Tzeng, S.H. Tsaur, Y.D. Laiw, S. Opricovic, Multicriteria analysis of environ-mental quality in Taipei: public preferences and improvement strategies, Journal of Environmental Management 65 (2) (2002) 109–120.

[41] G.H. Tzeng, M.H. Teng, J.J. Chen, S. Opricovic, Multicriteria selection for a restaurant location in Taipei, International Journal of Hospitality Management 21 (2) (2002) 171–187.

[42] G.H. Tzeng, C.W. Lin, S. Opricovic, Multi-criteria analysis of alternative-fuel buses for public transportation, Energy Policy 33 (11) (2005) 1373–1383.

[43]G.H. Tzeng, C.H. Chiang, C.W. Li, Evaluating intertwined effects in e-learning programs: a novel hybrid MCDM model based on factor analysis and DEMATEL, Expert Systems with Applications 32 (4) (2007) 1028–1044.

[44]P.L. Yu, A class of solutions for group decision problems, Management Science 19 (8) (1973) 936–946.

[45]Y.M. Wang, Y.S. Wang, Y.F. Yang, Understanding the determinants of RFID adoption in the manufacturing industry, Technological Forecasting and Social Change 77 (5) (2010) 803–815.

[46] J. Woods, K. Peterson, C. Hirst, Maturing open RFID applications will reshape SCM. , Available at http://www4.gartner.com/DisplayDocument?doc_cd=112865S2003, (Accessed 12 September 2005).

[47]N.C. Wu, M.A. Nystrom, H.A. Lin, H.C. Yu, Challenges to global RFID adoption, Technovation 26 (12) (2006) 1317–1323.

[48]Q. Zhou, W. Huang, Y. Zhang, Identifying critical success factors in emergency management using a fuzzy DEMATEL method, Safety Science 49 (2) (2011) 243–252.

[49]K. Zhu, S. Dong, S.X. Xu, K.L. Kraemer, Innovation diffusion in global dimensions: determinants of post-adoption digital transformation of European companies, European Journal of Information Systems 15 (6) (2006) 601–616.

[50]K. Zhu, K.L. Kraemer, Post-adoption variations in usage and value of E-business by organizations: cross-country evidence from the retail industry, Information Systems Research 16 (1) (2005) 61–84.

[51]K. Zhu, K.L. Kraemer, S. Xu, The process of innovation assimilation byﬁrms in different countries: a technology diffusion perspective on e-business, Management Science 52 (10) (2006) 1557–1576.

Ming-Tsang Lu is a Ph.D. student in the graduate institute of management science, National Chiao Tung University, Taiwan. His research interests include Multiple Criteria Decision Making, application of fuzzy theory to information systems, and information technology.

Shi-Woei Lin is a risk and decision analyst. His current research interests include the application of game theory to risk-informed regulation and the use of mathematical models to aggregate experts' uncertainty judgments. Other interests include decision biases and methods for effective risk communication, both to decision makers and to the general public.

Gwo-Hshiung Tzeng was born in 1943 in Taiwan. In 1967, he received a bachelor's degree in business management from the Tatung Institute of Technology (now Tatung University), Taiwan. In 1971, he received a master's degree in urban planning from Chung Hsing University (Now Taipei University), Taiwan. In 1977, he received a Ph.D. in management science from Osaka University, Osaka, Japan.

Gwo-Hshiung Tzeng was an Associate Professor at Chiao Tung University, Taiwan, from 1977 to 1981, a Research Associate at Argonne National Laboratory from July 1981 to January 1982, a Visiting Professor in the Department of Civil Engineering at the University of Maryland, College Park, MD, from August 1989 to August 1990, a Visiting Professor in the Department of Engineering and Economic System, Energy Modeling Forum at Stanford University, from August 1997 to August 1998, a professor at Chaio Tung Univer-sity from 1981 to 2003, and a Chair Professor at Chiao Tung UniverUniver-sity. He was named a National Distinguished Chair Professor (Highest Honor offered by the Ministry of Educa-tion Affairs, Taiwan) and Distinguished Research Fellow (Highest Honor Offered by NSC, Taiwan) in 2000. His current research interests include statistics, multivariate analysis, networks, routing and scheduling, multiple criteria decision making, fuzzy theory, applica-tion of hierarchical structure analysis to technology management, energy, the environment, transportation systems, transportation investment, logistics, locations, urban planning, tourism, technology management, electronic commerce, global supply chain, etc. He was awarded a Highly Cited Paper (March 13, 2009) ESI“Compromise solution by MCDM methods: A comparative analysis of VIKOR and TOPSIS” as published in the “EUROPEAN JOURNAL OF OPERATIONAL RESEARCH” on July 16th, 156(2), 445–455, 2004, which has recently been identiﬁed by Thomson Reuters' Essential Science Indicators SM as one of the most cited papers in theﬁeld of Economics and Business.

He received the MCDM Edgeworth-Pareto Award from the International Society on Multiple Criteria Decision Making (June 2009), the world Pinnacle of Achievement Award in 2005, and the National Distinguished Chair Professor Award (highest honor offered) of the Ministry of Education Affairs of Taiwan; additionally, he is a three time recipient of a dis-tinguished research award and was twice named a disdis-tinguished research fellow (highest honor offered) of the National Science Council of Taiwan. He is also a Fellow IEEE Member (since September 30, 2002). He organized a Taiwan afﬁliate chapter of the International As-sociation of Energy Economics in 1984 and he was the Chairman of the Tenth International Conference on Multiple Criteria Decision Making, July 19–24, 1992, in Taipei, the Co-Chairman of the 36th International Conference on Computers and Industrial Engineering, June 20-23, 2006, Taipei, Taiwan, and the Chairman of the International Summer School on Multiple Criteria Decision Making 2006, July 2–14, Kainan University, Taiwan. He is a member of IEEE, IAEE, ISMCDM, World Transport, the Operations Research Society of Japan, the Society of Instrument and Control Engineers Society of Japan, the City Planning Institute of Japan, the Behavior Metric Society of Japan, and the Japan Society for Fuzzy Theory and Systems and participates in many societies of Taiwan. He is an editor-in-chief of the International Journal of Information Systems for Logistics and Management.