DOI 10.1515/bmt-2011-0118
Biomed Tech 2012; 57(5):403–411
Nabil A. Alrajeh*, Badr Elmir, Bouchaïb Bounabat and Norelislam El Hami
Interoperability optimization in healthcare collaboration networks Abstract: Interoperability is one of the most challenging concerns that face healthcare information system (HIS) actors. Interoperability implementation in this context may be a data exchange interfacing, a service oriented interaction or even a composition of new composite healthcare processes. In fact, optimizing efforts of interoperability achievement is a key requirement to effectively setup, develop and evolve intra- and interorganizational collaboration. To ensure interoperability project effectiveness, this paper proposes a modeling representation of health processes interoperability evolution. Interoperability degrees of involved automated processes are assessed using a ratio metric, taking into account all significant aspects, such as potentiality, compatibility and operational performance. Then, a particle swarm optimization algorithm (PSO) is used as a heuristic optimization method to find the best distribution of effort needed to establish an efficient healthcare collaborative network. Keywords: healthcare collaborative network (HCN); healthcare information system (HIS); interoperability optimization; particle swarm optimization.
*Corresponding author: Dr. Nabil A. Alrajeh, Biomedical Technology Department, College of Applied Medical Sciences, King Saud University, P.O. Box 91138, Riyadh 11633, Saudi Arabia Phone: +966505268838 E-mail: [email protected] Badr Elmir: Al Qualsadi Research Team, ENSIAS - National Higher School for Computer Science and System Analysis, Université Mohammed V – Souissi, BP 713 Agdal -Rabat, Morocco Bouchaïb Bounabat: Al Qualsadi Research Team, ENSIAS-National Higher School for Computer Science and System Analysis, Université Mohammed V – Souissi, BP 713 Agdal - Rabat, Morocco Norelislam El Hami: Engineers Mohammadia Institute, Ibn Sina Avenue, BP-765, Agdal-Rabt, Morocco
partners’ existing applications. Interoperability implementation project, within HCN and across independent partners, can be seen as a multi-project environment that targets a unified strategic objective of collaboration. Such a situation may involve different teams to interconnect independent information systems. In such an environment, the challenges to effectively establishing interoperability are resource allocation and effort dispatching. This situation involves the interconnection of several automated healthcare processes located within a single organization or across a group of partners in collaboration. This paper studies different approaches to establishing interoperability in HCNs. It represents process interoperability evolution using an innovative model that presents possibilities of several forms of usage and enables the manipulation of interoperability on a large scale. With this perspective, this work uses a RatIop approach [7] as an interoperability assessment method. This work uses also heuristic mechanisms to optimize efforts required to improve the overall interoperability level of a healthcare collaboration situation. The heuristic method used here is particle swarm optimization (PSO) [5], which is known as an efficient approach with high performance in solving optimization problems in many research fields. In this paper, the second section is devoted to interoperability initiatives and approaches in HCNs. The third section describes optimization approaches. It focuses mainly on RatIop as an interoperability assessment approach and on a PSO algorithm as a means to control and to reach optimum effort repartition in multi-integration projects across a HCN. The fourth section discusses the optimization results and analyzes how to use them.
Interoperability in healthcare collaboration networks
Introduction Across a healthcare collaboration network (HCN), interoperability refers to the establishment of new composite process-oriented services and the integration of
HCNs are established to coordinate the delivery of information and services in healthcare ecosystems. Information technology is central to the establishment of an e-health infrastructure that enables the cooperation of
Brought to you by | New York University Bobst Library Technical Services Authenticated Download Date | 5/26/15 10:00 PM
404
N.A. Alrajeh et al.: Interoperability optimization in healthcare collaboration networks
healthcare institutions and allows the sharing of clinical data across disparate applications and systems. Interoperability is a challenge in the IT domain. The issue gets more challenging in HCN as clinical information is so complex. The achievement of interoperability among healthcare organizations has both technical and organizational aspects.
Patients
Researchers
Healthcare collaboration networks HCNs enable the exchange of healthcare information and services between and within organizations. HCNs enable electronic collaboration among entities exchanging data by less reliable or less timely means and among entities who wish to establish the exchange of clinical healthcare information [13]. The reasons for exchanging data and invoking the services of authorized partners are many and varied, including: (i) informing patients of care decisions; (ii) following up on quality of care; (iii) determining if treatments were necessary and reasonable for the purposes of making payments; (iv) responding to healthcare emergencies, such as public health threats; (v) performing studies of population health; and (vi) conducting research into the effectiveness of existing and emerging treatment mechanisms [15] (see Figure 1). Patient-centered healthcare systems involve advanced interactions between: (i) patients; (ii) healthcare providers (hospitals, clinics, physicians, public health providers, specialists, etc.); (iii) independent laboratories; (iv) community pharmacies; (v) public health agencies (local, regional and national); (vi) pharmaceutical and medical device manufacturers; (vii) researchers (academic, government and independent); and (viii) payers (government or private insurers) [15].
Healthcare information system interoperability approaches There are many approaches to effectively establish interoperability across a HCN. This section presents the most used approaches.
Healthcare providers
Payers
Health collaboration network
Medical manufacturers
Independent laboratories
Pharmacies Public health agencies
Figure 1 HCN member classification.
been developed by various organizations, covering every aspect of the healthcare domain [3]. Although, standardization is important in this domain, the need for standardization is not sufficient to establish an enduring interoperability as these standards change over time and there are some entities that may not accept these standards in their information processing.
Adoption of unique, fully integrated systems This integrated approach concentrates on using fully integrated information systems. In these systems, all the necessary structural parts and subsystems are provided by a single vendor, which assures the desired interoperability between them [11]. In terms of integration, this approach remains the best way to achieve a high degree of interoperability. However, high specialization and changing needs of the healthcare domain require using different systems from several vendors.
Custom interfacing for data exchange Healthcare information standards This unified approach suggests the compliance of healthcare information systems (HISs) with a set of commonly accepted data representation and communication standards. As a result, a considerable number of standards have
This federated approach consists of the development of point-to-point custom interfaces between information subsystems [1]. In spite of negative aspects of this approach such as perpetual changes in involved systems and maintenance costs
Brought to you by | New York University Bobst Library Technical Services Authenticated Download Date | 5/26/15 10:00 PM
N.A. Alrajeh et al.: Interoperability optimization in healthcare collaboration networks
405
of interfaces, mainly in large HCNs, it is, in many cases, a unique way to exchange data and establish interoperability.
exchange, a service oriented interaction or even a composition of new composite healthcare processes.
Service oriented interaction
Multi-project context to establish interoperability in HCNs
In addition to the information exchange issues, there is a need for service integration and application reuse. This refers to the concept that HISs should be able not only to access and use services provided by others but also to reuse their functionality. In this manner, it is theoretically possible to build complex information services from the composition of existing ones [1]. This federated approach is used to establish new composite process-oriented services across HCNs. It reuses existing services within entities to provide added value patient-centered services. Process composition Process composition is concerned with automated means for constructing business processes in a HCN. Collaboration among healthcare processes can be supported by linking the underlying sub-supporting systems that are responsible for executing the corresponding subprocesses within each HCN member [2]. This federated approach is used to establish new composite process-oriented services across HCNs. An integrated healthcare process is operationalized in a workflow that can be supported by workflow management technology. Figure 2 summarizes the main cited approaches of interoperability in HCNs.
Optimizing HIS interoperability This section describes the proposed model for optimization of interoperability in a HCN. Interoperability implementation in this case may be a custom interfacing of data
An interoperability implementation project within a collaboration network and across independent organizations can be seen as a multi-projects environment that targets a unified objective of collaboration but involves different teams to interconnect independent information systems [12]. In such an environment, the challenge is resource allocation end-effort dispatching to effectively establish interoperability on a projected level. This section aims to obtain the optimum distribution of effort to establish a specific organizational collaboration situation. The optimal allocation of effort in a multi-project environment refers to an optimization problem for which the objective is to optimize the overall effort and better distribute it in a multi-project implementation of interoperability. In this vein, this work proposes to use RatIop [7] as a measurement method.
RatIop approach Interoperability, like other quality factors, should be measured, monitored and improved throughout the lifetime of an information system. Its assessment requires mechanisms and metrics to measure it. Many works propose interoperability assessment methods and approaches that can be applied to HISs. RatIop is an enterprise-architecture-based assessment approach that aims to quantify on a scalar form the interoperability degree of an information system within its ecosystem [9, 16]. It is designed to assess the interoperability of an integrated e-service delivery in a public administration context and in a business collaboration networks context, as well as to citizen-oriented e-health services
HCN interoperability approaches
Healthcare standards adoption
Fully integrated systems
Customized inferfacing data exchange
Service oriented interaction
Healthcare process composition
Figure 2 HCN interoperability main approaches.
Brought to you by | New York University Bobst Library Technical Services Authenticated Download Date | 5/26/15 10:00 PM
406
N.A. Alrajeh et al.: Interoperability optimization in healthcare collaboration networks
[6–8]. RatIop stipulates that interoperability improvement of an information system is obtained by implementing changes in its supporting system and also by improving on the external interfaces of the interlinked systems. RatIop is a five-step interoperability assessment approach (see Figure 3). It is a parametric method that takes into account three main aspects: – The interoperability maturity level of the environment where the studied HISs are located. It is proposed to use the NEHTA (National E-Health and Transition Authority [14]) Interoperability maturity model to assess the potentiality part of interoperability in a healthcare context. – The degree of compatibility of the external interfaces of the information systems between them. This aspect takes into consideration incompatibilities in different enterprise architecture layers such as process, service, data and infrastructure, as well as the operational performance of the IT infrastructure that supports these systems. – The coupling of the RatIop method with modeling tends to characterize the evolution of the overall interoperability degree of a set of interconnected information systems. This coupling allows monitoring of the efforts needed to improve the interoperability degree of the collaboration network.
Interoperability evolution modeling Let us take a set of “n” HISs (S1, S2,…,Sn). – Each HIS ensures the activity of one healthcare establishment/institution. – Each HIS has the ability to interact with all the others systems.
(1) Delineating the scope
– –
Each HIS is supported by exactly one integrated IT infrastructure. The IT systems have the capacity to interoperate in a homogeneous way with the ecosystem. There are no specific barriers obstructing existing or potential interactions.
Each system Si is associated a ratio ai = RatIop(Si), which represents the interoperability degree within its environment. We aim to monitor the evolution of this indicator in a macroscopic way. The vector I(a1,..,an) evolves in accordance with the effort made to adapt the interconnected system from the current state to the target state. By denoting I = (ai), which represents the current interoperability vector, and I′ = (a′i), which represents the target interoperability vector, we have for each a system Si where a′i = ∑ Eij aj.
(1)
Eij represents the effort necessary on the Si system to improve the Sj system. E = (Eij) represents the matrix effort necessary to reach the target interoperability. I′ = E I
(2)
or ⎛ E11 ⎜E ⎜ 21 ⎜ ... ⎜ ... ⎜ ⎜⎝ E n1
E12 E22 ... ... En 2
... E1n ⎞ ⎛ a1 ⎞ ⎛ a ′1 ⎞ ... E2 n ⎟ ⎜ a2 ⎟ ⎜ a ′ 2 ⎟ ⎟ ⎜ ⎟ ⎜ ⎟ ... ... ⎟ ⎜ ... ⎟ = ⎜ ... ⎟ . ⎟ ⎜ ⎟ ⎜ ⎟ ... ... ⎟ ⎜ ... ⎟ ⎜ ... ⎟ ... Enn ⎟⎠ ⎜⎝ an ⎟⎠ ⎜⎝ a ′ n ⎟⎠
If all the systems Si are compatible with each other and there is no explicit barrier that impedes interaction, Eij is equivalent to the ratio of workload Nij (workload allocated to the improvement of the external interfaces of Si to facilitate the Sj RatIop) over the overall workload allocated to interoperability enhancement. Eij = Nij/Noverall
(3)
(4) Pe rfor
y
nce
tialit
ma
oten
(3) Compatibility
(2) P
In this case, the goal is to find the optimal effort to reach the targeted interoperability vector. So, the objective function to minimize is a ′i − ∑∑ Eij ⋅ai ≤ 0. i
(4)
j
The constraints for each j are given by
(5) RatIop
Figure 3 Five steps of interoperability measurement [7].
(2)
∑E
ij
≤100%
i
Brought to you by | New York University Bobst Library Technical Services Authenticated Download Date | 5/26/15 10:00 PM
(5)
N.A. Alrajeh et al.: Interoperability optimization in healthcare collaboration networks
Eij is to be multiplied by Ni/Noverall with Ni = ∑Nij. To optimize the objective function (4) with respect to the constraints in (5), it is possible to use deterministic techniques such as the gradient function. But with problems with large dimensions, these techniques remain inefficient in terms of performance. Heuristic algorithms such as PSO [5, 10] are a promising discipline to explore in this area.
PSO of interoperability establishment in HCN The PSO, developed by Kennedy and Eberharts in 1995, is an approximation algorithm method proposed for the optimization problem of finding the global minimum [10]. Since then, PSO has been improved by many researchers. The principal of this algorithm is based on the movement of birds searching for food in a flock. This animal behavior is simulated to optimization research. This method generates a group of particles, each one searching for the minimum of fitness by their own knowledge and movement and influenced by the search of his neighbor. If a particle finds a good site, all the others become aware of it more or less directly to take advantage of it. The PSO algorithm has been compared with other evolutionary-based optimization techniques [4]. The authors of [5] compared it with simulated annealing (SA), which is a classical technique that has been successfully used to solve a wide range of optimization problems. The SA algorithm, like most of the meta-heuristics family, is based on gradual local improvement. The SA algorithm can guarantee the optimum solution when the size of the problem is small, but it will take a long time when the problem size is larger [5]. The PSO algorithm can solve a larger problem in an acceptable time, which demonstrates the powerful explorability of the PSO algorithm [5]. In a PSO algorithm, each particle “i” is treated as a point in a space with dimension D, a position Xi, a velocity Vi and a personal best position Xbesti. The personal best position associated with a particle i is the best position that the particle has visited. The best position of all particles in the swarm is represented by the vector Xgbest. Xi = (xi1, xi2,..,xid) is the position of the particle, Vi = (vi1, vi2,..,vid) is the velocity of the particle, Xbesti = (pi1, pi2,..,pid) is the best personnel position, Xgbest = (pg1, pg2,..,pgd) is the best global position of the swarm, 1 ≤ i ≤ n: n is the dimension of the problem representing the position Xi and 1 ≤ d ≤ D: D is the space dimension of the swarm (number of particles). Vid(t+1) = χ (Vid(t)+ρ1[Xbesti(t)-Xi(t)]+ρ2[Xgbest(t)-Xi(t)])
(6)
Xid(t+1) = Xid(t)+Vid(t+1)
407
(7)
where ρ1 = c1r1 and ρ2 = c2r2, with c1 and c2 representing positive acceleration components called social parameters and r1 and r2 representing independent random numbers in the range [0, 1] and χ representing the constriction coefficient. The initialization of the swarm and velocities is usually performed randomly in the search space, following a uniform distribution. The best positions are initially set equal to the initial swarm. After the first time increment, they are moved by the velocity Vi in Equation (6). Then, the algorithm searches for optima by updating generations. The acceleration constants c1 and c2 in Equation (6) represent the weighting of the stochastic acceleration terms that pull each particle towards the Xbesti and Xgbest positions (see Figure 4). c1 represents the confidence that the particle has in itself, and c2 represents the confidence that the particle has in the swarm. In most cases, the acceleration parameters c1 and c2 are dependent on the problems to resolve, and we can make the appropriate choices for these parameters to modify the velocity and to promote convergence [5]. The search procedure of a population-based algorithm such as the PSO algorithm consists of the concept of collaboration; the information regarding the best position of the swarm is communicated to the rest of the particles through Equation (7). Initialization Xi ←Generate the initial particles of the swarm Vi←Generate the initial velocity of the particles Xbesti← Xi Set the best positions to a randomized particle position. Xgbest←Xi Set the best positions of the swarm to a randomized particle position.
Figure 4 Particle movement in a PSO algorithm.
Brought to you by | New York University Bobst Library Technical Services Authenticated Download Date | 5/26/15 10:00 PM
408
N.A. Alrajeh et al.: Interoperability optimization in healthcare collaboration networks
Repeat For i = 1: N (all particles in the swarm) Fitnessi(t)←Evaluate Fitness(Xi) if Fitnessi(t)
Biomed Tech 2012; 57(5):403–411
Nabil A. Alrajeh*, Badr Elmir, Bouchaïb Bounabat and Norelislam El Hami
Interoperability optimization in healthcare collaboration networks Abstract: Interoperability is one of the most challenging concerns that face healthcare information system (HIS) actors. Interoperability implementation in this context may be a data exchange interfacing, a service oriented interaction or even a composition of new composite healthcare processes. In fact, optimizing efforts of interoperability achievement is a key requirement to effectively setup, develop and evolve intra- and interorganizational collaboration. To ensure interoperability project effectiveness, this paper proposes a modeling representation of health processes interoperability evolution. Interoperability degrees of involved automated processes are assessed using a ratio metric, taking into account all significant aspects, such as potentiality, compatibility and operational performance. Then, a particle swarm optimization algorithm (PSO) is used as a heuristic optimization method to find the best distribution of effort needed to establish an efficient healthcare collaborative network. Keywords: healthcare collaborative network (HCN); healthcare information system (HIS); interoperability optimization; particle swarm optimization.
*Corresponding author: Dr. Nabil A. Alrajeh, Biomedical Technology Department, College of Applied Medical Sciences, King Saud University, P.O. Box 91138, Riyadh 11633, Saudi Arabia Phone: +966505268838 E-mail: [email protected] Badr Elmir: Al Qualsadi Research Team, ENSIAS - National Higher School for Computer Science and System Analysis, Université Mohammed V – Souissi, BP 713 Agdal -Rabat, Morocco Bouchaïb Bounabat: Al Qualsadi Research Team, ENSIAS-National Higher School for Computer Science and System Analysis, Université Mohammed V – Souissi, BP 713 Agdal - Rabat, Morocco Norelislam El Hami: Engineers Mohammadia Institute, Ibn Sina Avenue, BP-765, Agdal-Rabt, Morocco
partners’ existing applications. Interoperability implementation project, within HCN and across independent partners, can be seen as a multi-project environment that targets a unified strategic objective of collaboration. Such a situation may involve different teams to interconnect independent information systems. In such an environment, the challenges to effectively establishing interoperability are resource allocation and effort dispatching. This situation involves the interconnection of several automated healthcare processes located within a single organization or across a group of partners in collaboration. This paper studies different approaches to establishing interoperability in HCNs. It represents process interoperability evolution using an innovative model that presents possibilities of several forms of usage and enables the manipulation of interoperability on a large scale. With this perspective, this work uses a RatIop approach [7] as an interoperability assessment method. This work uses also heuristic mechanisms to optimize efforts required to improve the overall interoperability level of a healthcare collaboration situation. The heuristic method used here is particle swarm optimization (PSO) [5], which is known as an efficient approach with high performance in solving optimization problems in many research fields. In this paper, the second section is devoted to interoperability initiatives and approaches in HCNs. The third section describes optimization approaches. It focuses mainly on RatIop as an interoperability assessment approach and on a PSO algorithm as a means to control and to reach optimum effort repartition in multi-integration projects across a HCN. The fourth section discusses the optimization results and analyzes how to use them.
Interoperability in healthcare collaboration networks
Introduction Across a healthcare collaboration network (HCN), interoperability refers to the establishment of new composite process-oriented services and the integration of
HCNs are established to coordinate the delivery of information and services in healthcare ecosystems. Information technology is central to the establishment of an e-health infrastructure that enables the cooperation of
Brought to you by | New York University Bobst Library Technical Services Authenticated Download Date | 5/26/15 10:00 PM
404
N.A. Alrajeh et al.: Interoperability optimization in healthcare collaboration networks
healthcare institutions and allows the sharing of clinical data across disparate applications and systems. Interoperability is a challenge in the IT domain. The issue gets more challenging in HCN as clinical information is so complex. The achievement of interoperability among healthcare organizations has both technical and organizational aspects.
Patients
Researchers
Healthcare collaboration networks HCNs enable the exchange of healthcare information and services between and within organizations. HCNs enable electronic collaboration among entities exchanging data by less reliable or less timely means and among entities who wish to establish the exchange of clinical healthcare information [13]. The reasons for exchanging data and invoking the services of authorized partners are many and varied, including: (i) informing patients of care decisions; (ii) following up on quality of care; (iii) determining if treatments were necessary and reasonable for the purposes of making payments; (iv) responding to healthcare emergencies, such as public health threats; (v) performing studies of population health; and (vi) conducting research into the effectiveness of existing and emerging treatment mechanisms [15] (see Figure 1). Patient-centered healthcare systems involve advanced interactions between: (i) patients; (ii) healthcare providers (hospitals, clinics, physicians, public health providers, specialists, etc.); (iii) independent laboratories; (iv) community pharmacies; (v) public health agencies (local, regional and national); (vi) pharmaceutical and medical device manufacturers; (vii) researchers (academic, government and independent); and (viii) payers (government or private insurers) [15].
Healthcare information system interoperability approaches There are many approaches to effectively establish interoperability across a HCN. This section presents the most used approaches.
Healthcare providers
Payers
Health collaboration network
Medical manufacturers
Independent laboratories
Pharmacies Public health agencies
Figure 1 HCN member classification.
been developed by various organizations, covering every aspect of the healthcare domain [3]. Although, standardization is important in this domain, the need for standardization is not sufficient to establish an enduring interoperability as these standards change over time and there are some entities that may not accept these standards in their information processing.
Adoption of unique, fully integrated systems This integrated approach concentrates on using fully integrated information systems. In these systems, all the necessary structural parts and subsystems are provided by a single vendor, which assures the desired interoperability between them [11]. In terms of integration, this approach remains the best way to achieve a high degree of interoperability. However, high specialization and changing needs of the healthcare domain require using different systems from several vendors.
Custom interfacing for data exchange Healthcare information standards This unified approach suggests the compliance of healthcare information systems (HISs) with a set of commonly accepted data representation and communication standards. As a result, a considerable number of standards have
This federated approach consists of the development of point-to-point custom interfaces between information subsystems [1]. In spite of negative aspects of this approach such as perpetual changes in involved systems and maintenance costs
Brought to you by | New York University Bobst Library Technical Services Authenticated Download Date | 5/26/15 10:00 PM
N.A. Alrajeh et al.: Interoperability optimization in healthcare collaboration networks
405
of interfaces, mainly in large HCNs, it is, in many cases, a unique way to exchange data and establish interoperability.
exchange, a service oriented interaction or even a composition of new composite healthcare processes.
Service oriented interaction
Multi-project context to establish interoperability in HCNs
In addition to the information exchange issues, there is a need for service integration and application reuse. This refers to the concept that HISs should be able not only to access and use services provided by others but also to reuse their functionality. In this manner, it is theoretically possible to build complex information services from the composition of existing ones [1]. This federated approach is used to establish new composite process-oriented services across HCNs. It reuses existing services within entities to provide added value patient-centered services. Process composition Process composition is concerned with automated means for constructing business processes in a HCN. Collaboration among healthcare processes can be supported by linking the underlying sub-supporting systems that are responsible for executing the corresponding subprocesses within each HCN member [2]. This federated approach is used to establish new composite process-oriented services across HCNs. An integrated healthcare process is operationalized in a workflow that can be supported by workflow management technology. Figure 2 summarizes the main cited approaches of interoperability in HCNs.
Optimizing HIS interoperability This section describes the proposed model for optimization of interoperability in a HCN. Interoperability implementation in this case may be a custom interfacing of data
An interoperability implementation project within a collaboration network and across independent organizations can be seen as a multi-projects environment that targets a unified objective of collaboration but involves different teams to interconnect independent information systems [12]. In such an environment, the challenge is resource allocation end-effort dispatching to effectively establish interoperability on a projected level. This section aims to obtain the optimum distribution of effort to establish a specific organizational collaboration situation. The optimal allocation of effort in a multi-project environment refers to an optimization problem for which the objective is to optimize the overall effort and better distribute it in a multi-project implementation of interoperability. In this vein, this work proposes to use RatIop [7] as a measurement method.
RatIop approach Interoperability, like other quality factors, should be measured, monitored and improved throughout the lifetime of an information system. Its assessment requires mechanisms and metrics to measure it. Many works propose interoperability assessment methods and approaches that can be applied to HISs. RatIop is an enterprise-architecture-based assessment approach that aims to quantify on a scalar form the interoperability degree of an information system within its ecosystem [9, 16]. It is designed to assess the interoperability of an integrated e-service delivery in a public administration context and in a business collaboration networks context, as well as to citizen-oriented e-health services
HCN interoperability approaches
Healthcare standards adoption
Fully integrated systems
Customized inferfacing data exchange
Service oriented interaction
Healthcare process composition
Figure 2 HCN interoperability main approaches.
Brought to you by | New York University Bobst Library Technical Services Authenticated Download Date | 5/26/15 10:00 PM
406
N.A. Alrajeh et al.: Interoperability optimization in healthcare collaboration networks
[6–8]. RatIop stipulates that interoperability improvement of an information system is obtained by implementing changes in its supporting system and also by improving on the external interfaces of the interlinked systems. RatIop is a five-step interoperability assessment approach (see Figure 3). It is a parametric method that takes into account three main aspects: – The interoperability maturity level of the environment where the studied HISs are located. It is proposed to use the NEHTA (National E-Health and Transition Authority [14]) Interoperability maturity model to assess the potentiality part of interoperability in a healthcare context. – The degree of compatibility of the external interfaces of the information systems between them. This aspect takes into consideration incompatibilities in different enterprise architecture layers such as process, service, data and infrastructure, as well as the operational performance of the IT infrastructure that supports these systems. – The coupling of the RatIop method with modeling tends to characterize the evolution of the overall interoperability degree of a set of interconnected information systems. This coupling allows monitoring of the efforts needed to improve the interoperability degree of the collaboration network.
Interoperability evolution modeling Let us take a set of “n” HISs (S1, S2,…,Sn). – Each HIS ensures the activity of one healthcare establishment/institution. – Each HIS has the ability to interact with all the others systems.
(1) Delineating the scope
– –
Each HIS is supported by exactly one integrated IT infrastructure. The IT systems have the capacity to interoperate in a homogeneous way with the ecosystem. There are no specific barriers obstructing existing or potential interactions.
Each system Si is associated a ratio ai = RatIop(Si), which represents the interoperability degree within its environment. We aim to monitor the evolution of this indicator in a macroscopic way. The vector I(a1,..,an) evolves in accordance with the effort made to adapt the interconnected system from the current state to the target state. By denoting I = (ai), which represents the current interoperability vector, and I′ = (a′i), which represents the target interoperability vector, we have for each a system Si where a′i = ∑ Eij aj.
(1)
Eij represents the effort necessary on the Si system to improve the Sj system. E = (Eij) represents the matrix effort necessary to reach the target interoperability. I′ = E I
(2)
or ⎛ E11 ⎜E ⎜ 21 ⎜ ... ⎜ ... ⎜ ⎜⎝ E n1
E12 E22 ... ... En 2
... E1n ⎞ ⎛ a1 ⎞ ⎛ a ′1 ⎞ ... E2 n ⎟ ⎜ a2 ⎟ ⎜ a ′ 2 ⎟ ⎟ ⎜ ⎟ ⎜ ⎟ ... ... ⎟ ⎜ ... ⎟ = ⎜ ... ⎟ . ⎟ ⎜ ⎟ ⎜ ⎟ ... ... ⎟ ⎜ ... ⎟ ⎜ ... ⎟ ... Enn ⎟⎠ ⎜⎝ an ⎟⎠ ⎜⎝ a ′ n ⎟⎠
If all the systems Si are compatible with each other and there is no explicit barrier that impedes interaction, Eij is equivalent to the ratio of workload Nij (workload allocated to the improvement of the external interfaces of Si to facilitate the Sj RatIop) over the overall workload allocated to interoperability enhancement. Eij = Nij/Noverall
(3)
(4) Pe rfor
y
nce
tialit
ma
oten
(3) Compatibility
(2) P
In this case, the goal is to find the optimal effort to reach the targeted interoperability vector. So, the objective function to minimize is a ′i − ∑∑ Eij ⋅ai ≤ 0. i
(4)
j
The constraints for each j are given by
(5) RatIop
Figure 3 Five steps of interoperability measurement [7].
(2)
∑E
ij
≤100%
i
Brought to you by | New York University Bobst Library Technical Services Authenticated Download Date | 5/26/15 10:00 PM
(5)
N.A. Alrajeh et al.: Interoperability optimization in healthcare collaboration networks
Eij is to be multiplied by Ni/Noverall with Ni = ∑Nij. To optimize the objective function (4) with respect to the constraints in (5), it is possible to use deterministic techniques such as the gradient function. But with problems with large dimensions, these techniques remain inefficient in terms of performance. Heuristic algorithms such as PSO [5, 10] are a promising discipline to explore in this area.
PSO of interoperability establishment in HCN The PSO, developed by Kennedy and Eberharts in 1995, is an approximation algorithm method proposed for the optimization problem of finding the global minimum [10]. Since then, PSO has been improved by many researchers. The principal of this algorithm is based on the movement of birds searching for food in a flock. This animal behavior is simulated to optimization research. This method generates a group of particles, each one searching for the minimum of fitness by their own knowledge and movement and influenced by the search of his neighbor. If a particle finds a good site, all the others become aware of it more or less directly to take advantage of it. The PSO algorithm has been compared with other evolutionary-based optimization techniques [4]. The authors of [5] compared it with simulated annealing (SA), which is a classical technique that has been successfully used to solve a wide range of optimization problems. The SA algorithm, like most of the meta-heuristics family, is based on gradual local improvement. The SA algorithm can guarantee the optimum solution when the size of the problem is small, but it will take a long time when the problem size is larger [5]. The PSO algorithm can solve a larger problem in an acceptable time, which demonstrates the powerful explorability of the PSO algorithm [5]. In a PSO algorithm, each particle “i” is treated as a point in a space with dimension D, a position Xi, a velocity Vi and a personal best position Xbesti. The personal best position associated with a particle i is the best position that the particle has visited. The best position of all particles in the swarm is represented by the vector Xgbest. Xi = (xi1, xi2,..,xid) is the position of the particle, Vi = (vi1, vi2,..,vid) is the velocity of the particle, Xbesti = (pi1, pi2,..,pid) is the best personnel position, Xgbest = (pg1, pg2,..,pgd) is the best global position of the swarm, 1 ≤ i ≤ n: n is the dimension of the problem representing the position Xi and 1 ≤ d ≤ D: D is the space dimension of the swarm (number of particles). Vid(t+1) = χ (Vid(t)+ρ1[Xbesti(t)-Xi(t)]+ρ2[Xgbest(t)-Xi(t)])
(6)
Xid(t+1) = Xid(t)+Vid(t+1)
407
(7)
where ρ1 = c1r1 and ρ2 = c2r2, with c1 and c2 representing positive acceleration components called social parameters and r1 and r2 representing independent random numbers in the range [0, 1] and χ representing the constriction coefficient. The initialization of the swarm and velocities is usually performed randomly in the search space, following a uniform distribution. The best positions are initially set equal to the initial swarm. After the first time increment, they are moved by the velocity Vi in Equation (6). Then, the algorithm searches for optima by updating generations. The acceleration constants c1 and c2 in Equation (6) represent the weighting of the stochastic acceleration terms that pull each particle towards the Xbesti and Xgbest positions (see Figure 4). c1 represents the confidence that the particle has in itself, and c2 represents the confidence that the particle has in the swarm. In most cases, the acceleration parameters c1 and c2 are dependent on the problems to resolve, and we can make the appropriate choices for these parameters to modify the velocity and to promote convergence [5]. The search procedure of a population-based algorithm such as the PSO algorithm consists of the concept of collaboration; the information regarding the best position of the swarm is communicated to the rest of the particles through Equation (7). Initialization Xi ←Generate the initial particles of the swarm Vi←Generate the initial velocity of the particles Xbesti← Xi Set the best positions to a randomized particle position. Xgbest←Xi Set the best positions of the swarm to a randomized particle position.
Figure 4 Particle movement in a PSO algorithm.
Brought to you by | New York University Bobst Library Technical Services Authenticated Download Date | 5/26/15 10:00 PM
408
N.A. Alrajeh et al.: Interoperability optimization in healthcare collaboration networks
Repeat For i = 1: N (all particles in the swarm) Fitnessi(t)←Evaluate Fitness(Xi) if Fitnessi(t)
