During the last decade there has been great interest in systems used for health monitoring and safety [1–5]. One of the key aspects in such systems is to store medical data efficiently in order to capture health status across time. Towards this direction, a number of electronic health records (EHR), or electronic medical records (EMR), have been developed. Some works extend the idea of EHR and focus on developing a personalized virtual patient model (VPM) composed of person’s medical records [6–8].
A factor of high importance for components or systems which require to use data and information exchanged between them [9–11] is the interoperability [10]. Towards this direction, the European Commission’s eHealth Action Plan 2012-2020 promotes semantic interoperability of EHR systems as a crucial challenge in eHealth solutions [12–16] in an effort to provide a roadmap to empower patients and healthcare workers.
Formal modeling of clinical content that can be made available internationally is one of the most promising pathways to semantic interoperability of health information [17]. Over the last two decades many attempts have been made to solve the major issues of health data systems that include semantic interoperability across systems as well as between components of a system, and decision support based on intelligent data analysis. Moreover, computational challenges in health systems due to which standard Information and Communication Technology (ICT) systems are hard to keep up, include the high variety and complexity of data, and high rate of data change ranging from clinical processes to protocols.
In this section we first provide a few details on the Electronic Health Record standardization, followed by a brief overview of DBMS solutions for storing standardized EHRs. Then we focus on the monitoring systems for older adults and how these can be used for assessing frailty.
Electronic health record standardization
An EHR is an electronic version of a patient’s medical history, that is maintained by a healthcare provider over time. It usually includes all the key administrative clinical data relevant to that persons’ care under the particular provider, including demographics, progress notes, problems, medications, vital signs, past medical history, immunizations, laboratory data and radiology reports [18].
Different international organizations have worked on the definition of an EHR architecture. Health Level 7 (HL7) [19] is a set of international standards for clinical data that focus on the application layer, which is "layer 7" in the OSI model. These standards are produced by an international standards organization (HL7 International), and are widely adopted by other bodies such as American National Standards Institute and International Organization for Standardization. The HL7 standards are basically a set of rules that allow clinical information to be shared and processed in a uniform and consistent manner.
The Health Informatics Technical Committee (TC251) of the European Committee for Standardization (CEN/ TC251) [20] has completed a European Standard for the communication of the EHR, called CEN EN13606 whose reference model became an ISO standard in February 2008 under the name ISO 13606. Exploiting this ISO, the openEHR consortium [21] maintains an architecture designed to support the constructions of distributed, patient-centered, life-long, shared care health records. OpenEHR is an open standard specification that is used to facilitate storage, management, retrieval and exchange of health data between different healthcare providers or other interest groups. In openEHR, all health data for a person are stored in a lifelong, person-centered EHR. The openEHR specifications include an EHR Extract specification [22] but in contrast to other standards such us EN 13606 and HL7, the exchange of data between EHR-systems is not their main primary concern. The openEHR framework will be discussed in detail in the “Methods” section.
Since the purpose of this work is to exploit openEHR archetypes and to adapt the openEHR framework for the representation of frailly in older person models, similar works which investigate or adapt the openEHR model for the representation of the required entities are presented next. In one of the earliest works [23], the modeling of a prototype neonatology Electronic Patient Record using openEHR archetypes was introduced, while more recently [24] the suitability of openEHR with its reference model, archetypes and templates, for the digital representation of demographic information and obstetrics related clinical data was investigated. Also the openEHR was elaborated as a tool for modeling Hospital Information Systems on a regional level based on a national logical infrastructure. Other noticeable works include the archetypes reuse for modeling the EHR of children affected by cerebral palsy [25], the adoption of the openEHR standard for the development of an EHR for methadone treatment recording and decision support [26], the validation of openEHR archetypes for the Multiple Sclerosis Functional Composite [27] and the modeling of healthcare authorization and claim submissions using the openEHR dual-model approach [28]. Even though openEHR has been exploited by several researchers and developers, none of the existing solutions has investigated its use for the comprehensive formal modeling of older people’s health, making this an important research issue.
DBMS solutions for storing standardized EHRs
It is generally agreed that the storage and management of medical data is a task of high difficulty, as clinical data are dynamic, sporadic, heterogeneous and complex in nature [29, 30]. As mentioned in [31], despite the fact that medical data share some of the characteristics of the typical data managed by conventional Database Management Systems (DBMS), special attention is required in the design of the corresponding database schemas, due to the unique features that the medical data possess. The selection of database and its corresponding schema design are factors that affect the effective management of clinical data (query performance, scalability, flexibility and extensibility) which usually emerges, especially in systems developed for real-time usage.
The most widely used approach in object-oriented systems is the Object-Relational Mapping (ORM) [32] which is essentially a relational model that allows users to integrate object-oriented features into it. Regarding its application in databases used for storing standarized EHRs, it aims at an exhaustive mapping between the structure of the EHR and the relational database. However, the complex nature and the variability of the information represented by the openEHR model makes the application of ORM in archetype-based systems a complicated and inefficient option, as complicated queries are required even for the retrieval of simple information, leading to low performance [33–35].
A remarkable persistence solution, also based on the relational paradigm, has been proposed by openEHR. This method is called "Node and Path" [36] and uses the relational model to store BLOBs (binary large objects that store serialized subtrees of the archetype XML file extract) into relational tables based on semantic paths defined by the archetypes structure. This method is adopted by numerous archetype-based systems, as mentioned in a survey [37] concerning openEHR storage implementations worldwide until 2013. The wide adoption of a relational persistence solution may be a result of the domination of the relational paradigm in the DBMS field due to its simplicity and hence, the developers’ long experience in it. Nevertheless, its simplicity results in complex queries, which convert data retrieval into an inefficient task [34, 36].
Another notable solution proposed by Wang et al. [38] is Archetype Relational Mapping (ARM), which builds a relational database schema driven by mappings between the openEHR archetypes and relational tables, in contrast with Node + Path which aims at an archetype-independent data storage structure. The comparison conducted in this work between archetype-relational mapping and the BLOB approach showed that the performance of the two methods was similar for single-patient queries, but the performance of ARM was much more efficient for population-based queries.
In other archetype-based implementations, as also mentioned by Frade et al. [37], the adopted storage method was an XML database, which provides flexibility and compatibility advantages. However, XML databases are not capable of providing satisfactory response time when population-based queries are submitted to large datasets, in comparison to their efficiency in single-patient queries [39].
Investigating solutions to overcome the drawbacks coming of the aforementioned relational and XML persistence methods, the NoSQL approach has been examined or adopted in few openEHR-based implementations [31, 33, 34, 40]. NoSQL databases provide a mechanism for storage and retrieval of data which are modeled by means other than the tabular relations used in relational databases. Some of their basic characteristics are that they are schema-free, distributed, open-source, low-cost, horizontally scalable and can store a huge amount of data. All the related works reach a common observation: the NoSQL databases outperform the relational and the XML models in terms of characteristics. More specifically, the results in [31] show that a NoSQL database (produced by MS-SQL) is the best choice for query speed compared to an XML database, while in another comprehensive work [33] the disadvantages of XML and relational databases are pointed out, as experiments showed that the implemented NoSQL database (Couchbase) had better response times than the relational database and both were faster than the XML one. Therefore, the NoSQL database seems to be a promising solution for retrieving results from population-based queries in openEHR-based systems. In another work that examines database persistence of ISO/EN 13606 standardized electronic health record extracts, the comparison of relational and NoSQL approaches, also led the researchers into the conclusion that NoSQL databases perform better in concurrency, using the MongoDB NoSQL database [41].
The NoSQL databases (Couchbase, MongoDB) used in all the aforementioned works are document-based, which pair each key with a complex data structure known as a document. These documents can contain many different key-value pairs, or key-array pairs, or even nested documents. On the other hand, in wide-column or column-family databases, which are a different type of NoSQL databases, data are managed differently. The data are stored in cells identified by a rowkey and a column name, and each column belongs to a column family. While column families need to be defined on the creation of the table, the columns can be created dynamically at runtime. This results in a very flexible schema that can host a virtually unlimited number of columns with the advantage of very fast access/search and data aggregation. The most famous databases of this category are Google’s BigTable, as well as HBase and Cassandra that were inspired from BigTable.
Monitoring systems for elderly
Recently, special attention has been given in electronic technologies that support ageing of older adults in their home environment [42]. There is a significant number of surveys focusing on monitoring systems for older adults, including wearable ECG monitoring systems [3], and tools and technologies in ambient-assisted living [43]. A review on the acceptance of technology for ageing is provided in [44]. The emphasis in monitoring older adults, comes as a result of the constant increase of the ageing population which affects the planning and delivery of health and social care. The age related decline is characterized by reduced physical, physiological and cognitive function, and results to frailty. Frailty is a common clinical syndrome in ageing population that comes as a result of cumulative declines across multiple physiologic systems, and carries an increased risk for adverse health outcomes including falls, hospitalization, disability, and mortality [45]. In order to prevent these adverse outcomes there is the need of monitoring the physiological clinical state of older people using both embedded and behavior monitoring sensors. By integrating such sensor data in an standardized electronic health system, one could achieve efficient interconnection with other heterogeneous systems.
Proposed framework for monitoring frailty
This paper aims to create a modeling approach for the health of the ageing population, using existing or newly developed openEHR archetypes. This is an innovative application since none of the previous works has attempted to describe the required entities for older people’s frailty status using the openEHR model. Additionally, we describe a methodology for a mapping between openEHR archetypes and a wide-column NoSQL database. Such a mapping has not been presented in depth before, although there are some works [46] which use HBase to represent electronic health records.
Our work is part of the FrailSafe project [47] which aims to develop a safe and unobtrusive real life sensing (physical, cognitive, psychological, social, etc) and intervention platform for the ageing population using advanced data analysis tools [48–50], considering that frailty has major health care implications and all persons older than 70 years should be screened for frailty [45].