Interoperability in Electronic Health Records through ...

Interoperability in Electronic Health Records through the Mediation of Ubiquitous User Model Ma. Lourdes Martínez-Villaseñor1 ,Luis Miralles-Pechuan 1 ,Miguel GonzálezMendoza2 1

Universidad Panamericana Campus M éxico, Augusto Rodin 498, Col. Insurgentes-M ixcoac, M éxico, D.F., M éxico 2 Tecnológico de M onterrey, Campus Estado de M éxico, Edo. M éxico, M éxico {lmartine,lmiralles}@up.edu.mx,[email protected]

Abstract The paradigm of healthcare systems has change from isolated proprietary health records to patient-centric solutions in which government, hospitals and clinics, general practitioners and other stakeholders must cooperate in order to provide improved health services. Enabling interoperability to share heterogeneous medical and administrative information in a secure environment is an issue addressed worldwide. Standards help though are not enough to provide the right information at the right time and place. In this paper we proposed to leverage the interoperability between standards through the mediation of a ubiquitous user model and an automatic process of concept alignment. Keywords: Personal Health Records Interoperability, Electronic, Health Records Interoperability, Ubiquitous User M odeling

1

Introduction

Healthcare systems have been changing in the last decades to a more patient -centric model. It is urgent that information about patient profile, medical history, problems progress, tests, treatments and other related records are no longer gathered in separated silos. The focus of healthcare must be the patient needs and values; patient has to be in control of his/her information. It is important that all parties cooperate and share information in order to provide right information, to the right pers on, at the right time and place [1]. Healthcare is moving towards “solutions which support a continuous medical process and (i) include multiple healthcare professionals and institutions, (ii) utilize ubiquitous computing healthcare environments and (iii) embrace technological advances, typical of the domain of today’s pervasive software applications.” [2] It is incredible that in the digital age, handwritten documents remain prevalent throughout most of the health sector, in spite of all efforts to automate and share electronic health information among general practitioners, patients and other stakeholders. Many efforts have been made to automate medical information developing Electronic Healthcare Records (EHRs) and standards to enable s haring and reusing information amongst healthcare software systems. Nevertheless, health care information is

very complex and therefore interoperability between health systems is hard. Hospital medical records are extremely varied. Millions of different formats and forms are used to capture patient’s personal, social, family and medical history. Patient’s care involves documents of different nature: treatments, tests, progress notes, referrals, imaging, medical charts, and nursing notes just to mention a few. Communication is also many to many given that the doctor communicates with different specialists, patients go to multitude of doctors, treatment services, administrative parties and agents. Healthcare systems and EHRs have been developed worldwide. Health care information is stored in different proprietary formats, and this information is managed in multiple types of hardware and software solutions for diverse business processes. Medical information is stored in structured formats including databases, and unstructured documents. This differences result in a severe interoperability p roblems [3]. Interoperability between EHR is important to deliver more effective and efficient patient care assisting the retrieval and processing of clinical information about patients from and to different health systems. Duplicate testing and prescribing can be reduced [3]. Standards can contribute to exchange medical information in a safe, secure and reliable manner. Several international standards development organizations have been working towards enabling healthcare interoperability. Multiple efforts to address EHR interoperability problems are focus in developing international standard s [1,3]. One of the main problems is that there are too many standards and the adoption of each one of them entails enormous efforts. Standards are also not static; they are constantly evolving in content and structure [2]. There are just a few success cases like MedCom [4] As in other domains, there are syntactic and semantic heterogeneities added to heterogeneities in hardware and software and purpose. Gibbons et al. [5] define interoperability in three categories technical, semantic, and process interoperability in healthcare systems. Despite of all efforts done by governments, standards organizations and different stakeholders involved in healthcare sector, interoperability between EHRs is still an open issue. In this paper, we present a small effort to enable interoperability between EHRs through the mediation of our ubiquitous user model [6,7]. Our framework could help as a mediator between different health systems. Information from different applications could be shared and reused with our user profile. The rest of the paper is organized as follows. State of the art efforts towards EHR interoperability are presented in section 2. An overview of our framework for ubiquitous user model interoperability is described in section 3. In section 4, we present an application scenario of interoperability in EHR through the mediation of ubiquitous user model. Experiments and results are shown in section 5. We conclude and outline future work in section 6.

2

Electronic Health Records Interoperability

In this section we reviewed the most prominent standards, which address the interoperability problems in order to share information across healthcare systems. The major International Organizations for Standardization providing standard solutions for EHR interoperability are: International Organization for Standardization [8], European Committee for Standardization [9], Health Level Seven [10] accredited by American National Standards Institute [11], and Digital Imaging and Communications Medicine [12]. The International Organization for Standardization (ISO) established ISO’s Technical Committee ISO/TC 215, which deals with health informatics. This committee introduces several standards. One of them, ISO/TR 20514 defines the structure and context of an EHR offering a basic-generic EHR. ISO 13606-reference model was published to enhance EHR communication [8]. Health Informatics of the European Committee for Standardization (CEN/TC 251) presented a reference model, an archetype interchange specification, reference archetypes and term lists, security features, and exchange models. CEN/ISO 13606 is a European norm approved also by ISO to achieve semantic interoperability in EHR communication. It defines a Dual Model architecture: Reference Model for EHR representation, and archetypes to provide semantic meaning to the Reference Model structure [9]. Health Level Seven (HL7) term is used as a name for the organization, and as a set of messaging standards. HL7 are successful messaging standards that support two messaging protocols: HL7 Version 2 and HL7 Version 3 [10]. HL7 Version 3 proposed the Clinical Document Architecture (CDA) for exchanging documents across heathcare systems. There are areas of harmonization between standards: “ HL7CDA and CEN 13606 Reference Models and CEN/openEHR archetypes with HL7 Templates. The OpenEHR Reference Model uses the CEN13606 Reference Model, which in turn is used in HL7CDA” [2]. Regarding DIACOM is known as the de facto standard for medical image communication [2,3] . DIACOM standards introduced data structures and services to enable the exchange of medical images and related information across vendors. They presented two EHR standards: Web Access to DICOM Persistent Objects (WADO), and DICOM Structured Reporting. An industry initiative Integrating the Healthcare Enterprise (IHE) encourages the coordinate use of existing standards like DICOM and HL7. They propose storing health documents in a XML repository to facilitate sharing and reusing of EHR information. Dossia proposed a personal health record (PHR) management systems in 2006 [13]. Major technology players, Google and Microsoft have also tried to contribute with personal health record (PHR) management systems [14]. Google Health offered the users a Web-based system to manage health information in 2008 but they retired in 2012 for lack of adoption. Microsoft HealthVault [15] was introduced in October of 2007 and is still active as a personal health record management system. These initiatives didn’t achieve the expected user adoption [14].

In this section, we only write a summary of the major standards, but in real-world there are many more. These standards are dynamic and it is difficult to cope with complex descriptions and documentation. There is a large number of conflicts between them. It is clear that given the great amount of standards, there is no healthcare standard winner. None of them provides a sufficient plug -and-play standard`’n’0`. There is no one-size fit all patient record. Another less explored element to address EHR interoperability is using mapping and alignment tools for exposing, sharing and reusing healthcare information of different EHR repositories and data providers [16]. We propose a solution using mapping and schema alignments to enable EHR interoperability in the following sections.

3

Overview of a Framework for Ubiquitous User Model Interoperability

In a multi-application, multi-device ubiquitous environment, user profile information is scattered in distributed user models. When trying to integrate all this valuable user information, it is important to take into account that highly autonomous profile suppliers and consumers are participating in the interoperability process. This means, first of all, that providers are free to decide what data to store, how to describe the data, set of constrains on the data, and associate an interpretation [14]. Providers also decide what data to share, policies and means of how to share it. Consumers of profile information want to decide when join and leave the system as well. Consumers have also their ways to describe and interpret data. Therefore mechanisms of interoperability must be provided that require the least intervention and effort of the ubiquitous user modeling stakeholders in order to enable interoperability respecting the providers’ and consumers’ autonomy. These conditions stand in learning environments. The user or learner information representation must be machine-readable, and flexible to allow the integration of information of new providers. In previous works [6,7] we presented a framework for ubiquitous user interoperability. The proposed framework enables the interoperability between profile suppliers and consumers with a mixed approach that consists in central ubiquitous user model ontology to provide formal representation of the user profile and a process of concept alignment to automatically discover the semantic mappings between the user models. The central ubiquitous user model interoperability ontology (U2MIO) is a flexible representation of a ubiquitous user model to cope with the dynamicity of a distributed multi-application environment that provides mediation between profile suppliers and consumers. U2MIO can evolve over time to adapt the representation to the changing multi-application environment. The dynamic user profile structure ontology is based in Simple Knowledge Organization for the Web (SKOS) [17] The process of concept alignment is briefly described below. This process automatically discovers the semantic mapping between the concepts of profile su ppliers and consumers and the U2MIO ontology in order to interpret the information from heterogeneous sources, and integrate them into a ubiquitous user model. This process is crucial for the construction and maintenance of the ubiquitous user model; it en ables

interoperability and allows the evolution in time of the U2MIO ontology. We proposed a two-tier matching strategy for the process of concept alignment in a hybrid integration system to provide mediation between heterogeneous sources. This architecture and the process of concept alignment facilitate the participation of new stakeholders in the interoperability process. 3.1 Ubiquitous user model interoperability ontology. The Ubiquitous User Modeling Interoperability Ontology (U2MIO) represents a flexible user model profile that evolves during time according to the recommendations of the concept alignment. The ontology reuses SKOS ontology designing a central concept scheme for the ubiquitous user model and one concept scheme for each profile supplier or consumer. Semantic mapping relations between each stakeholder’s concept scheme concepts and the central user model concept scheme concepts are determined by the process of concept alignment. Semantic relations are set with SKOS properties. This representation supports interoperability overcoming semantic differences and enables the participation of new stakeholders in the interoperability process without effort of the profile information provider or consumer. Figure 1 shows the interrelations between profile stakeholders and the ubiquitous user model concept.

Fig. 1. Interrelations between profile stakeholders and ubiquitous user model.

3.2 Process of concept alignment. Our ubiquitous user modeling framework deals with the profile suppliers’ transfer mechanisms and recollects source documents (sd) in XML, JSON or RDF. If the source is new to the system, a corresponding skos:ConceptScheme (X) is designed and added to U2MIO. The process of concept alignment is based in a two-tier matching strategy (figure 2). First an element level matching step finds a set of concept candidates for alignment for each concept in the source concept scheme. This task is

performed combining three types of similarity measures: a) String similarity based in Dice [18]b) A simple distance of longest substring c) semantic similarity based on WordNet [19].From this step in which we analyze the word similarity between each concept in the source with all concepts in ubiquitous user model concept schema (u2m), we find a set of best suited concepts for alignment (or one best suited concept) in the target (u2m). Next, the method looks for structure similarity. The goal in the structure level matching step is to disambiguate the meaning of the word analyzing its context, this means analyzing the structure and meaning of the neighbor concepts in the same source document. In this step, the similarity between the neighbor concepts in the source and the neighbors of the best suited concept(s) in the target are calculated. After this step, a set of IF THEN rules are applied to determine one-to-one semantic mappings and recommendations of concept and collection additions. The process of concept alignment shown in figure 2 roughly describes the inputs and outputs of each phase. A concept scheme is defined as (C, HC, VC) where C is a set of concepts arranged in a subsumption hierarchy HC. VC is the set of corresponding concept values. Cs is the set of concept labels extracted from the source document. CT is the set of concept labels extracted from the target (ubiquitous user model scheme), and CbT is the set of concepts that are best suited for alignment. R0 (Cs , CTb ) are the highest relations obtained from the element level matching phase and R(Cs , CT) are the final semantic mapping relations found between the concepts of the source document and the ubiquitous user model (target).

Fig. 2. Two-tier matching stategy of the process of concept alignment

4

Application scenario of Interoperability in Electronic Health Records through the Mediation of Ubiquitous User Model

There are many standards of EHRs content, structures and mechanisms for the exchange, Nevertheless as we described in section 2, many health systems model the ir own personal health record or adopt one standard that is not interoperable with other standards used by other health agents, providers or consumers of EHR information. Some personal health record (PHR) management systems have own APIs and web services to enable populating or extracting profile information. Standards and transfer mechanism can change anytime. Prior agreement between stakeholders implies adherence and adoption or/and manual or semiautomatic mapping to enable interoperability.

In this section we propose two experiments in order to establish a proof of concept of two use cases: 1) Interoperability between EHRs through the mediation of ubiquitous user model 2) EHR enrichment reusing Facebook information 4.1

Evaluati on metrics

In order to measure the efficiency and effectiveness of the matching/mapping systems, different metrics have been proposed in the literature [22]. In this work, the evaluation of the process of concept alignment is focused in: • The human effort required by the mapping designer to verify the correctness of the mappings, which is quantified with the metric and partially measures the efficiency of the process. • The quality of the generated mappings quantifying the proximity of the results generated by the process of concept alignment to those expected with four known metrics: precision, recall, f-measure and fall-out .With these metrics a partial measure of the effectiveness of the process is performed. These metrics are based on the notions of true positives (TP), false positives (FP), true negatives (TN) and false negatives (FN). A human expert provided a list of expected matches for the proof of concept example and evaluated the outcomes, deciding if the semantic mapping relations found were correct and recommendations make sense. Exact match relations correctly found by the process, and good recommendations for concept or collection addition were considered as TP. Wrong exact matches were listed as FP. When a relevant exact match was not found by the process, a concept was improperly discarded or a wrong recommendation was made, it was registered as FN. Properly discarded concepts were recorded as TN. 4.2

Experiments

Case 1: Interoperability between EHRs through the mediation of ubiquitous user model. We used FHIR (Fast Healthcare Interoperability Resources) Specification , Patientexample.xml [20] to enhance our Ubiquitous user model interoperability ontology. Next we suppose that the patient wants to use Microsoft HealthVault personal health record (PHR) management system, so in order to enable interoperability between both standards, we performed a process of concept alignment with the Basic Demographic Information [21 ]. The true errors (FP and FN) were generated due to the ambiguity of used label in both sides. Code, text, value, name are frequently used in a schema hindering the semantic matching. The results of the process of concept alignment between HL7 schema and Basic HealthVault PHR are presented in the confusion matrix of table. Even though the Basic profile only has 8 concepts, the semantic meaning of the concepts are difficult to interpret given the level of granularity, especially of country concept.

Table 1. Resulting confusion matrix of the matching process between HL7 and Basic M HVault

Case 1 HL7 vs Basic MHV

Process of concept alignment outcome

Expected matches positive negative

positive

TP=4

FP=2

negative

FN=1

TN=1

Birth year is detailed alone for example. In HL7 gender is associated with “patient” and in Basic HealthVault PHR with “basic”. These schema associations change the outcome of the system adding the concept to the most related collection instead of defining an exact match. The decision is reasonable given that in other cases the alignment is correct, for example spouse or contact gender. The only exact match found was incorrect. Labels where the same (code_value) but the meaning was completely different. Case2: EHR enrichment reusing Facebook information. The ubiquitous user model interoperability ontology contained already information about Facebook User profile. We considered the case of EHR enrichment reusing Facebook information to populate Basic Demographic Information. Table 2. Resulting confusion matrix of the matching process between Facebook and Basic M HVault

Case 2 Facebook vs Basic MHV

Process of concept alignment outcome

positive

Expected matches positive negative TP=6 FP=0

negative

FN=0

TN=2

Table 2 presents the confusion matrix resulting of aligning Basic HealthVault PHR and the ubiquitous user model enhanced with Facebook profile. The results are perfect when you only consider if the semantic relations found were correct , and the system’s recommendation make sense to a human evaluator. Although the system made good decisions, no exact match could be found. Two concepts were discarded correctly, this is no matching was possible,

4.3

Results

The efficiency and effectiveness of the process of concept alignment for case scenario 1 and case scenario 2, are shown in table3. The results of effectiveness show that recall results exceed our needs (medium recall), but our requirements for high precision were not fulfilled in case 1. The same weight was given to precision and recall in order to calculate F-measure and case 1 results are only fair. The greater the overall, less human effort is needed to correct the automatic mapping. In case 1, the automatic process of schema matching was not worth the effort to correct it. Table 3. Efficiency and effectiveness measuring results

Case 1 Case 2 Metric HL7 vs Basic MHV Facebook vs Basic MHV Precision 67% 100% Quality of generated Recall 80% 100% 67% 0% mappings (Effectiveness) Fall-out 73% 100% F-measure 40% 100% Human Effort (Efficiency) Overall Measure

5

Conclusions and Future Work

The paradigm of healthcare systems has change from isolated proprietary health records to patient-centric solutions in which government, hospitals and clinics, general practitioners and other stakeholders must cooperate in order to provide improved health services. Enabling interoperability to share heterogeneous medical and administrative information in a secure environment is an issue addressed worldwide. Standards help though are not enough to provide the right information at the right time and place. In this paper we proposed to leverage the interoperability between standards through the mediation of a ubiquitous user model and an automatic process of concept alignment. Although results prove that the process is making sense in the schema matching decisions, in many cases semantic interoperability is hard and the human effort to mend ambiguities is too much. Nevertheless, automatic schema mapping can be improved and in some cases eases the load of finding consensus between EHRs. For future work, we are trying to implement fuzzy logic to deal with semantic uncertainties. References 1. Benson, T. Principles of health interoperability HL7 and SNOMED. Springer Science & Business M edia, 2012. 2. Begoyan, A. An overview of interoperability standards for electronic health records. USA: society for design and process science, 2007.

3. Eichelberg, M ., Aden, T., Riesmeier, J., Dogac, A., & Laleci, G. B. A survey and analysis of electronic healthcare record standards. Acm Computing Surveys (Csur), 2005, vol. 37, no 4, p. 277-315. 4. Saranummi, N. Regional health economies and ICT services: the picnic experience. IOS Press, 2005. 5. Gibbons P, et al. Coming to terms: Scoping interoperability for health care. 2007. 6. M artinez-Villaseñor, M . D. L., Gonzalez-M endoza, M ., & Hernandez-Gress, N., Neil. Towards a ubiquitous user model for profile sharing and reuse. Sensors, 2012, vol. 12, no 10, p. 13249-13283. 7. M artinez-Villaseñor, M . D. L. Design and Implementation of a Framework for Ubiquitous User Model Interoperability. 2013. Tesis Doctoral. PhD Thesis, Instituto Tecnológico y de Estudios Superiores de M onterrey, Estado de M éxico, M éxico. 8. ISO International Standards, available in www.iso.org, last update 2016. 9. The EN 13606 Association, The CEN/ISO EN 13606 standard, available in http://www.en13606.org/the-ceniso-en13606-standard, last updated 2015, consulted June 14, 2016. 10. HL7, 2004, EHR System Functional M odel: “A M ajor Development Towards Consensus on Electronic Health Record System Functionality”, A White Paper. Health Level Seven, Inc. 11. America National Standards Institute (ANSI), available at http://www.ansi.org /, last update 2016, consulted June 15, 2016. 12. Digital Imaging and Communications M edicine (DICOM ), available at http://dicom.nema.org/ , last update 2016, consulted June 15, 2016. 13. Dossia consortium, Available in http://www.dossia.org, last update 2014, consulted June 14, 2016. 14. Spil, T., & Klein, R., Richard. The personal health future. Health policy and technology, 2015, vol. 4, no 2, p. 131-136. 15. M icrosoft, M icrosoft HealthVault, available in www.healthvault.com, last update 2016, consulted June 14, 2016. 16. Perakis, K., Bouras, T., Ntalaperas, D., Hasapis, P., Georgousopoulos, C., Sahay, R., & Usurelu, D. Advancing patient record safety and EHR semantic interoperability. En Systems, Man, and Cybernetics (SMC), 2013 IEEE International Conference on. IEEE, 2013. p. 3251-3257. 17. M iles, A., M atthews, B., Wilson, M ., & Brickley, D. SKOS core: simple knowledge organisation for the web. En International Conference on Dublin Core and Metadata Applications. 2005. p. pp. 3-10. 18. Dice LR. M easures of the amount of ecologic association between species. Ecology, 1945, vol. 26, no 3, p. 297-302. 19. Fellbaum C. WordNet and wordnets. 2005. 20. FHIR (Fast Healthcare Interoperability Resources) Specification, Patient-example.xml, available in https://www.hl7.org/fhir/patient-example.xml.html, last updated: Sat, Oct 24, 2015 07:43+1100, consulted: June, 15th, 2016. 21. M icrosoft, Basic demographic information example, available in https://developer.healthvault.com/DataTypes/Overview?TypeId=3b3e6b16-eb69-483c8d7e-dfe116ae6092, last updated: 2014, consulted: June, 15th, 2016. 22. Bellahsene, Z., Bonifati, A., Duchateau, F., & Velegrakis, Y. On evaluating schema matching and mapping. En Schema matching and mapping. Springer Berlin Heidelberg,