Strategies to Access Patient Clinical Data from Distributed Databases

ao Rafael Almeida

1,2

, Olga Fajarda

, Arnaldo Pereira

and Jos

e Lu

ıs Oliveira

Institute of Electronics and Informatics Engineering of Aveiro (IEETA), University of Aveiro, Aveiro, Portugal

Department of Computation, Computer Science Faculty, University of A Coru

na, A Coru

na, Spain

Keywords:

Clinical Research, Electronic Health Records, Observational Studies, Common Data Model, Semantic Web.

Abstract:

Over the last twenty years, the use of electronic health record systems has become widespread worldwide,

leading to the creation of an extensive collection of health databases. These databases can be used to speed

up and reduce the cost of health research studies, which are essential for the advance of health science and

the improvement of health services. However, despite the recognised gain of data sharing, database owners

remain reluctant to grant access to the contents of their databases because of privacy and security issues, and

because of the lack of a common strategy for data sharing. Two main approaches have been used to perform

distributed queries while maintaining all data control in the hands of the data custodians: applying a common

data model, or using Semantic Web principles. This paper presents a comparison of these two approaches by

evaluating them according to parameters relevant to data integration, such as cost, data quality, interoperability,

extendibility, consistency, and efﬁciency.

1 INTRODUCTION

Health research studies are determinant for the ad-

vance of health science and the improvement of health

services. Pharmaceutical and public health surveil-

lance, the development of new treatments, the ex-

pansion of knowledge about diseases and monitor-

ing health crises are essentially done using health re-

search studies (Nass et al., 2009). This type of work

involves several time-consuming and expensive steps,

namely, the identiﬁcation and recruitment of consent-

ing subjects, and the gathering of the data, which in

some cases means following the recruited subjects

over a long period. However, health research studies

can be speed up and much cheaper, if they are done

using data collected for other purposes, like data from

health-related registry systems or data collected from

previous studies (Cheng and Phillips, 2014).

Nowadays, due to the worldwide generalisation of

electronic health record (EHR) systems and the digi-

tisation of health-related information, a vast number

of electronic health databases, containing diversiﬁed

clinical digital data, exists (Geissbuhler et al., 2013).

Besides turning the research more efﬁcient, by saving

time and money, the use of these databases for health

research studies has the advantage of increasing the

quality of the research, especially when combining

data from several databases (Piwowar and Chapman,

2010). Furthermore, the use of existing databases pre-

vents the collection of duplicate data and gives the

researcher access to a larger, more diverse popula-

tion, as well as to certain groups of people, which,

for example, do not participate in clinical trials, such

as children and older people (Schneeweiss and Avorn,

2005). Moreover, every clinical trial puts the research

subjects through some risk and, therefore, the sub-

stitution of a clinical trial by the secondary use of

clinical digital data prevents unnecessary risk (Doolan

et al., 2017). Even when clinical trials are necessary,

e.g. for the development of new therapies, existing

health care data can be used to identify clinical trial

participants, and, consequently, accelerate this com-

plex process (Ohmann and Kuchinke, 2007; Pakho-

mov et al., 2007). Drug safety surveillance is, essen-

tially, done using EHRs, because some adverse drug

events are only observed after the release of the drug

to a larger, diversiﬁed population (Triﬁr

o et al., 2014).

Retrospective cohort studies and case-control stud-

ies are other kinds of health research studies that can

be done using existing health databases (Ganz et al.,

2014; Reisner et al., 2015).

However, despite the recognition of the ines-

timable value of the secondary use of existing digi-

tal clinical data, and the importance of the open data

movement and the FAIR Data principles (Wilkinson

et al., 2016), health database owners remain reluctant

466

Almeida, J., Fajarda, O., Pereira, A. and Oliveira, J.

Strategies to Access Patient Clinical Data from Distributed Databases.

DOI: 10.5220/0007576104660473

In Proceedings of the 12th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2019), pages 466-473

ISBN: 978-989-758-353-7

in sharing the content of their databases (Pisani and

AbouZahr, 2010). Even the data obtained through

public research funding projects are not shared with

the research community (Lopes et al., 2015). The re-

luctance of the health database owners to share their

data is due to several reasons. The main reasons con-

cerns data ownership, intellectual property rights and

the lack of a common strategy for data sharing.

Two main approaches are used to enable the ac-

cess to clinical data from distributed databases, with-

out losing patient data privacy: (i) applying a common

data model or (ii) using Semantic Web (SW) princi-

ples. In this paper, we compare these two approaches

according to parameters relevant to data integration,

such as cost, data quality, interoperability, extendibil-

ity, consistency, and efﬁciency.

The rest of the paper is organized as follows: in

Section 2 we present an overview of existing solu-

tions, in Section 3 we describe the CDM, and the SW

approaches, in Section 4 we discuss and compare both

approaches, and ﬁnally in Section 5 we conclude the

paper.

2 RELATED WORK

Several solutions have been developed for the se-

cure sharing of patient clinical data from distributed

databases. CALIBER

, for instance, is a research

platform consisting of a combination of highly trained

staff, tools and data resources, “research ready”

variables extracted from linked electronic health

records coming from England’s hospital records, pri-

mary care, social deprivation information, and cause-

speciﬁc mortality data. The resources available con-

sists of data up to 2016 including more than 10 mil-

lion people with approximately 400 million person-

years of follow-up. The main purpose of CALIBER is

to promote an open community developing methods

and tools to accelerate replicable science across all

clinical and scientiﬁc disciplines spanning the trans-

lational cycle (from drug discovery through to public

health). However, the process to gain access to CAL-

IBER resources is very slow and bureaucratic.

PopMedNet

is a scalable and extensible open-

source platform to simplify the implementation

and operation of distributed health data query-

ing networks. This platform was developed by

HMORN (Health Maintenance Organisation Re-

search Network), a consortium of 19 U.S. regional

healthcare delivery organisations. Through a set of

http://www.ucl.ac.uk/health-informatics/caliber

http://www.popmednet.org/

web-based services and tools, PopMedNet enables

the creation and use of distributed data networks. It

supports both menus driven queries and distributed

analyses using complex, single-use or multi-use pro-

grams and returns aggregated counts of eligible study

cohorts (Brown et al., 2012).

OHDSI (Observational Health Data Sciences and

Informatics)

is an international, interdisciplinary and

multi-stakeholder project with the aim to develop ap-

plications to access and analyse large-scale observa-

tional health data. This collaborative was initiated at

the end of the Observational Medical Outcomes Part-

nership (OMOP) project, in order to continue the re-

search started. The OMOP was a public-private US

project with the objective to develop solutions to per-

form medical product safety surveillance using obser-

vational healthcare databases (Hripcsak et al., 2015).

The main outcome of the OMOP consortium was the

creation of the OMOP Common Data Model (CDM),

which standardises the content, structure and conven-

tion of healthcare databases (Overhage et al., 2011).

The OMOP CDM is considered to be the most com-

plete an efﬁcient common data model available (Kahn

et al., 2012; Ogunyemi et al., 2013; Ross et al., 2014).

Over the last ﬁve years, besides continuing to improve

the OMOP CDM, the OHDSI community developed

several analytic tools, namely, Achilles, HERMES,

HERACLES, and CIRCE. In 2016, this community

released a web-based platform, called ATLAS

, that

integrates features form the previously mentioned ap-

plications. This web-based platform provides tools to

browse standardised vocabularies, explore databases,

deﬁne cohorts, and make a population-level analysis

of observational data converted to the OMOP CDM.

The European Medical Information Frame-

work (EMIF)

is a European project, launched in

2013, with the purpose of improving the access of

researchers to patient-level data from distinct health

data repositories across Europe. The EMIF Platform

is an integrated system where researchers can browse

three different levels of information: metadata, aggre-

gated data, and raw data. Every Data Custodian con-

trols to whom and the level of information that can be

shared (Trifan et al., 2018). Several solutions have

been developed to simplifying the access to health

data, in order to meet the needs of the Data Custodi-

ans involved in the project. EMIF has adopted OMOP

CDM for EHR data harmonisation, as also the use

of solutions to infer knowledge through query federa-

tion.

Applying the idea of having a common data

http://www.ohdsi.org/

http://www.ohdsi.org/web/atlas/

http://www.emif.eu

Strategies to Access Patient Clinical Data from Distributed Databases

467

model, there are some methodologies and tools with

a similar goal, such as the Semantic Web frame-

works (Berners-Lee et al., 2012). The principles of

SW and Linked Data (LD) (Speicher et al., 2015) can

be used to solve data integration and interoperability

problems. One of the pillars for the realisation of the

SW is the way data is represented. The Resource De-

scription Framework (RDF) covers this important is-

sue, with the data model proposed by the World Wide

Web Consortium (W3C) in a suite of normative spec-

iﬁcations (Schreiber and Raimond, 2015).

Nowadays, semantic technologies are at the core

of many systems that support data-intensive research

areas, as is the case with system biology, inte-

grative neuroscience, bio-pharmaceutics and transla-

tional medicine, just to mention a few cases (Chen

et al., 2013). In addition, numerous repositories are

using the SW data model that can be accessed over the

Internet (Zaveri and Ertaylan, 2017), due to the exis-

tence of stable standards and best practice guidelines.

Bringing together people and machines, the semantic

technologies offer the ability to describe data better

and to map and link distributed datasets. In this way,

an information network is created that can be used by

searching the information from a single entry point.

The literature reports the use of various SW so-

lutions to integrate data from EHR systems. To in-

crease the usability of EHR systems, (Lasierra et al.,

2017) described a method to model patient-centered

clinical EHR workﬂows. To allow interoperable shar-

ing of patient data between healthcare organisations,

(Alamri, 2018) proposes a semantic-mediation archi-

tecture to support semantic interoperability. By us-

ing this intermediate layer, the clinical information is

exploited using richer ontological representations to

create a “model of meaning” for enabling semantic

mediation. For the facilitation of RDF data manage-

ment and query federation across several repositories,

(Sernadela et al., 2017) developed SCALEUS

, a se-

mantic web migration tool that can be deployed on

top of traditional systems to bring knowledge, infer-

ence rules, and query federation to the existent data.

In a single package, it includes a triplestore support-

ing multiple independent datasets, simpliﬁed API and

services for data integration and management, and

a SPARQL query engine, supporting real-time infer-

ence mechanisms and optimised text searches over the

knowledge base. This platform was used to facilitate

RDF data management and query federation across

the several tools of the RD-Connect initiative, an EU

FP7 project which aimed to create an integrated plat-

form connecting databases, registries, biobanks and

clinical bioinformatics for rare disease research.

http://bioinformatics-ua.github.io/scaleus/

3 QUERYING METHODOLOGIES

3.1 Querying Pipeline

Common technical and governance solutions must be

developed to simplify the access to health data. An

approach to do so is using the methodology presented

by (Fajarda et al., 2018), where a pipeline is used to

achieve the querying process, as shown in Figure 1.

An implementation of this kind of solution was done

in the EMIF project. The pipeline considers three

main roles:

• the Researcher, someone who needs to query sev-

eral databases, to which he has no direct access,

to conduct research;

• Data Custodians, individuals responsible for ad-

ministering their databases;

• the Study Manager, the person responsible for

managing the research study and act as an inter-

mediary between the Researcher and Data Custo-

dians.

The study starts with a researcher who wants to

query some databases. This person creates a study

request, writing his/her question in the EMIF Cat-

alogue

, a platform that allows researchers to ﬁnd

databases which fulﬁl their particular research study

requirements. This platform, also, allows the research

to select the desired databases that s/he would like to

query (Silva et al., 2018).

After receiving the study request, the Study Man-

ager starts a workﬂow using TASKA

, a work man-

agement system (Almeida et al., 2018) that will sup-

port the whole study orchestration.

The ﬁrst task of the workﬂow consists of the co-

hort/query deﬁnition, which results in a script. Us-

ing TASKA, this script can be sent to all the selected

Data Custodians at once. Every Data Custodian ex-

ecutes the script in their database, and the results of

the querying are, then, exported to the Study Man-

ager. After receiving all the results, the Study Man-

ager compiles them to answer the Research’s request.

Finally, the Research receives the response to his/her

request, and the pipeline ends.

The Study Manager has an important role in this

pipeline since s/he has direct access to the Data Custo-

dians and must have the knowledge necessary to work

with the query deﬁnition tools. Consequently, a per-

son not familiar with these tools can easily query sev-

eral databases of her/his choice and to which s/he has

no direct access.

https://emif-catalogue.eu/

https://bioinformatics.ua.pt/taska

HEALTHINF 2019 - 12th International Conference on Health Informatics

468

Figure 1: Workﬂow of the querying process (Fajarda et al., 2018).

Concerning the Data Custodian stage of the

pipeline, two main approaches can be followed. The

ﬁrst one using a common data model, and the second

using the SW principles.

3.2 Common Data Model Approach

The Command Data Model approach requires that

Data Custodians’ databases use a shared schema to

all. Therefore, a model must be delineated, which is

currently not a problem, since there are already some

common data models deﬁned for observational stud-

ies, e.g. the OMOP CDM. This common data model

is, already used in several countries (Hripcsak et al.,

2015). However, to be used the Data Custodians need

to convert their database into the OMOP CDM, using

Extract, Transform and Load (ETL) methodologies.

The OHDSI provides documentation of best practices

to perform the transformation, including several tools

to support the data migrations. In the early stages,

the OMOP CDM migration process was very com-

plex, however, currently, this procedure is optimised,

and OHDSI created several tools to guide the different

specialised entities involved. These specialised enti-

ties are:

• Local data experts and CDM experts, which to-

gether design the ETL transformation, without

creating the migration script;

• People with medical knowledge, which deﬁne the

code mappings;

• A technical person, which creates and implements

the ETL scripts following the speciﬁcations de-

ﬁned previously.

In the ﬁnal stage of the migration, all the entities in-

volved need to ensure the quality control of the imple-

mentation, this validates the process and ensures that

the data is consistent. However, despite all of these

tools and protocols design to help these entities, it is

still impossible to fully automate this process. An-

other disadvantage of this process is the need for peo-

ple with medical knowledge, which can be an expen-

sive resource.

Assuming that this procedure was done in all the

available databases, the Study Manager and the Data

Custodians can use some tools to extract and analyse

the data, e.g. ATLAS. With a local installation of AT-

LAS, the Study Manager can deﬁne a cohort and send

the resulting extraction script to all the Data Custodi-

ans involved in the study. The Data Custodians can,

then, execute the script received, in their local AT-

LAS installation, which provides a result, that can be

analysed and ﬁltered, before being sent to the Study

Manager. This procedure ensures that Data Custodi-

ans have full control over their data and keeps non-

authorised users away from patients data, preserving

data privacy.

3.3 Semantic Web Approach

An alternative to using a common data model is the

use of Semantic Web technologies. This approach re-

Strategies to Access Patient Clinical Data from Distributed Databases

469

quires that Data Custodians use a common ontology

to specify the knowledge of the domain. The Web

Ontology Language (OWL) (Hitzler et al., 2012) is

the W3C semantic language to describe the entities

of domains, providing classes, properties, individuals,

and data values. Ontologies have been used by several

communities to structure knowledge domains. Just to

give an example, the Gene Ontology (GO)

deﬁnes

concepts to describe gene function along three differ-

ent aspects: molecular function, cellular component,

and biological process. Many more biomedical on-

tologies and terminologies can be found on the NCBO

BioPortal

(Whetzel et al., 2011).

The Semantic Web approach relies on the con-

version of the original data into RDF format. This

conversion must be performed for each distributed

database by using a convenient solution, such as the

SCALEUS tool. After the creation of the semantic

model for the domain of interest and the data con-

version into RDF, the SPARQL query language is

the convenient tool for extracting knowledge from the

created semantic database.

Figure 2 presents a snippet of the query pipeline.

In this scenario, the Study Manager uses SPARQL to

create the desired query and send it to the Data Cus-

todians. Then, the Data Custodians sends back the

patients’ data to the Study Manager, for interpretation

and compilation of results.

Figure 2: Semantic Web approach.

The SCALEUS solution provides a set of data

connectors and interfaces that help in the translation

process to a user pre-deﬁned model (ontology) to be

negotiated between the Data Custodians. Although

migration tasks are simpliﬁed, the need for consensus

among Data Custodians may make it difﬁcult to use

this option.

http://www.geneontology.org/

https://bioportal.bioontology.org/

4 DISCUSSION

Choosing the best strategy for using data from het-

erogeneous and distributed repositories is a task that

impacts across the course of an entire project. As no

universal formulas are covering all kinds of possibil-

ities, it is desirable for decision-makers to be aware

of the strengths and weaknesses of the most widely

used and documented available options. Some selec-

tion criteria can be pointed out from the authors’ ex-

perience based on their collaboration in the EMIF and

RD-Connect projects:

• Cost - Sum of costs of implementation and train-

ing of users;

• Data Quality - Ability to serve the purpose of

users;

• Interoperability - Interoperability relates with

machine-readability and machine-actionability,

describing the extent to which solutions can au-

tomatically exchange and interpret data;

• Extendibility - Possibility to extend and add fur-

ther information a-posteriori;

• Consistency - Refers to the data structure co-

herence between the different Data Custodians’

databases;

• Efﬁciency - Efﬁciency in producing an answer to

a research question.

Table 1 presents a summary of the evaluation of

both methodologies according to the considered se-

lection criteria, indicating the most favourable ap-

proach for each criterion.

Table 1: Assessment of the methodologies.

Common

Data Model

Semantic

Web

Cost + -

Data quality +/- +/-

Interoperability - +

Extendibility - +

Consistency + -

Efﬁciency + -

Legend. + : better; +/- : tied; - : worst

Data migration requires signiﬁcant investments

in infrastructures, software solutions and human re-

sources. Considering that for both approaches the

infrastructure is already in place and that solutions

are open and free (e.g. OHDSI, SCALEUS), hu-

man effort become more relevant in the cost equation.

HEALTHINF 2019 - 12th International Conference on Health Informatics

470

Both approaches require medical knowledge. How-

ever, since the OMOP CDM migration is more pop-

ular, this pipeline is already optimised, reducing the

costs. Furthermore, the Semantic Web approach has

less adhesion in this scenario, causing more costs to

support this transition.

This data transition demands a great understand-

ing of the institutional data and its structure, which is

a requirement to produce a solid migration. During

this process, the data owners need to specify how to

deal with poor quality data. This is done in an ini-

tial stage of the data migration, where Data Custodi-

ans have the responsibility to ensure the data quality

of their databases, thus optimising the analyst’s work,

which avoids errors during the migration.

The ability of systems and applications to collab-

orate at a machine-machine level is a requirement for

automating the extraction of knowledge from hetero-

geneous and distributed data repositories. For this

collaboration to be possible between the different sys-

tems, they must communicate using a set of standards

that enable the intelligible communication of infor-

mation using, preferably, the Internet. For the ﬁrst

approach, machine-machine interoperability is more

difﬁcult. The data model used in this approach is not

as appropriate as that deﬁned for SW solutions. On

the other hand, the existence of a series of standards

ensures that the second approach meets those require-

ments in an easier way to implement.

After putting a system in production, its data

model may need changes to reﬂect the changes in

the reality of interest. This need can, in the limit,

lead to the whole system having to be changed in

depth. The ﬁrst approach is based on the use of entity-

relationship data models that scale poorly comparing

to SW solutions. In fact, changes to semantic data

models do not signiﬁcantly change the systems in pro-

duction, ensuring a good extendibility.

Regardless of the approach chosen, Data Custodi-

ans will have a shared data structure. The Common

Data Model approach has already a well-deﬁned data

structure. Additionally, the different data representa-

tion formats are normalised in the OMOP CDM ap-

proach, keeping the same conventions consistent in

the data model. Another aspect is the vocabulary def-

inition and mapping due to the existence of several

clinical terms. Those have been mapped onto OMOP

Vocabularies, improving the ability to analyse and

search the databases. Furthermore, the vocabulary

deﬁnition helps researchers ﬁnd relevant drug codes.

For instance, if a researcher wants to ﬁnd a drug by

its National Drug Code (NDC), s/he can do it easily

searching for it in ATLAS or ATHENA, which is also

a standardised process, due to the consistency in the

cohort deﬁnition, analysis design and results report-

ing. Succinctly, using OHDSI tools in OMOP CDM

databases, allows the observational research to be per-

formed by institutional groups, generating systematic

scientiﬁc practices, where research guidelines can just

be followed. In contrast, in the SW approach, vocab-

ulary and relationships are not standardised, needing

to be negotiated in advance.

The efﬁciency in observational studies is mainly

based on the Data Custodians’ response delay. This

lack of response’s speed can be turned in months or

even years of waiting to get a ﬁnal answer which may

be one of the biggest challenges that researchers need

to deal, due to data accessing permissions restriction.

The pipeline presented, intends to reduce this delay,

mainly due to all the technology involved, which fa-

cilitates the querying process. Furthermore, the role

division reduces some boundaries that existed due to

the lack of agreements and rules. A Data Custodian

can quickly and easily query his/her database, anal-

yse the result, make the necessary adjustments, by ﬁl-

tering some sensitive data, and sent it to the Study

Manager. This is possible mainly due to data classiﬁ-

cation, and tools prepared to work with it. We could

also analyse the efﬁciency of both approaches individ-

ually. However, the most signiﬁcant delay is the coor-

dination of people, which is enriched in the pipeline.

5 CONCLUSIONS

Observational data research offers the opportunity to

chart empirically-demonstrated scientiﬁc work and si-

multaneously produces an empirical evaluation of the

quality of the evidence generated, useful for mean-

ingfully informing decision-making processes. In or-

der to support such studies, while ensuring data pri-

vacy and security, a strategy for querying different

databases in a mediated way is needed. In this paper,

we analysed two different strategies to perform dis-

tributed queries in health databases. Both approaches

use open-source solutions and can offer alternative

pipelines to help researchers answer their questions

without the need for direct access to data. The ﬁrst

approach is based on the use of a common data model

and the second on the application of Semantic Web

principles.

The approaches were evaluated based on a set

of selection criteria created from the authors’ expe-

rience in the application of each of the approaches.

Both approaches are similar when we consider the

data quality criteria. The Common Data Model so-

lution is more performant for the cost, consistency,

and efﬁciency. When considering interoperability and

Strategies to Access Patient Clinical Data from Distributed Databases

471

extendibility, the Semantic Web approach is more

favourable.

ACKNOWLEDGEMENTS

This work was partially funded by the NETDI-

AMOND project, grant number POCI-01-0145-

FEDER-016385. AP is supported by FCT, grant

PD/BD/142877/2018.

REFERENCES

Alamri, A. (2018). Semantic health mediation and access

control manager for interoperability among health-

care systems. Journal of Information Technology Re-

search, 11:87–98.

Almeida, J., Ribeiro, R., and Oliveira, J. L. (2018). A mod-

ular workﬂow management framework. In Proceed-

ings of the 11th International Conference on Health

Informatics (HealthInf 2018).

Berners-Lee, T., Hendler, J., and Lassila, O. (2012). The

semantic web. Scientiﬁc American, 284:34–43.

Brown, J., Balaconis, E., Mazza, M., Syat, B., Rosen, R.,

Kelly, S., Swan, B., and Platt, R. (2012). Ps1-46:

Hmornnet: shared infrastructure for distributed query-

ing by hmorn collaboratives. Clinical medicine & re-

search, 10(3):163–164.

Chen, H., Yu, T., and Chen, J. Y. (2013). Semantic web

meets integrative biology: a survey. Brieﬁngs in

Bioinformatics, 14:109–125.

Cheng, H. G. and Phillips, M. R. (2014). Secondary analy-

sis of existing data: opportunities and implementation.

Shanghai archives of psychiatry, 26(6):371.

Doolan, D. M., Winters, J., and Nouredini, S. (2017). An-

swering research questions using an existing data set.

Medical Research Archives, 5(9).

Fajarda, O., Silva, L. A. B., Rijnbeek, P. R., Van Spey-

broeck, M., and Oliveira, J. L. (2018). A methodology

to perform semi-automatic distributed ehr database

queries. In HEALTHINF, pages 127–134.

Ganz, M. L., Wintfeld, N., Li, Q., Alas, V., Langer, J., and

Hammer, M. (2014). The association of body mass

index with the risk of type 2 diabetes: a case–control

study nested in an electronic health records system in

the united states. Diabetology & metabolic syndrome,

6(1):50.

Geissbuhler, A., Safran, C., Buchan, I., Bellazzi, R.,

Labkoff, S., Eilenberg, K., Leese, A., Richardson, C.,

Mantas, J., Murray, P., et al. (2013). Trustworthy reuse

of health data: a transnational perspective. Interna-

tional journal of medical informatics, 82(1):1–9.

Hitzler, P., Kr

otzsch, M., Parsia, B., Patel-Schneider, P. F.,

and Rudolph, S. (2012). Owl 2 web ontology language

primer (second edition). w3c recommendation.

Hripcsak, G., Duke, J. D., Shah, N. H., Reich, C. G., Huser,

V., Schuemie, M. J., Suchard, M. A., Park, R. W.,

Wong, I. C. K., Rijnbeek, P. R., et al. (2015). Obser-

vational health data sciences and informatics (ohdsi):

opportunities for observational researchers. Studies in

health technology and informatics, 216:574.

Kahn, M. G., Batson, D., and Schilling, L. M. (2012).

Data model considerations for clinical effectiveness

researchers. Medical care, 50.

Lasierra, N., Schweitzer, M., Gorfer, T., Toma, I., and

Hoerbst, A. (2017). Building a semantic model

to enhance the user’s perceived functionality of the

ehr. Studies in Health Technology and Informatics,

228:137–141.

Lopes, P., Silva, L. B., and Oliveira, J. L. (2015). Chal-

lenges and opportunities for exploring patient-level

data. BioMed research international, 2015.

Nass, S. J., Levit, L. A., Gostin, L. O., et al. (2009). The

value, importance, and oversight of health research.

Ogunyemi, O. I., Meeker, D., Kim, H.-E., Ashish, N.,

Farzaneh, S., and Boxwala, A. (2013). Identifying

appropriate reference data models for comparative ef-

fectiveness research (cer) studies based on data from

clinical information systems. Medical care, 51:S45–

S52.

Ohmann, C. and Kuchinke, W. (2007). Meeting the chal-

lenges of patient recruitment. International Journal

of Pharmaceutical Medicine, 21(4):263–270.

Overhage, J. M., Ryan, P. B., Reich, C. G., Hartzema, A. G.,

and Stang, P. E. (2011). Validation of a common data

model for active safety surveillance research. Jour-

nal of the American Medical Informatics Association,

19(1):54–60.

Pakhomov, S., Weston, S. A., Jacobsen, S. J., Chute, C. G.,

Meverden, R., Roger, V. L., et al. (2007). Electronic

medical records for clinical research: application to

the identiﬁcation of heart failure. Am J Manag Care,

13(6 Part 1):281–288.

Pisani, E. and AbouZahr, C. (2010). Sharing health data:

good intentions are not enough. Bulletin of the World

Health Organization, 88(6):462–466.

Piwowar, H. A. and Chapman, W. W. (2010). Public shar-

ing of research datasets: a pilot study of associations.

Journal of informetrics, 4(2):148–156.

Reisner, S. L., Vetters, R., Leclerc, M., Zaslow, S., Wol-

frum, S., Shumer, D., and Mimiaga, M. J. (2015).

Mental health of transgender youth in care at an ado-

lescent urban community health center: a matched ret-

rospective cohort study. Journal of Adolescent Health,

56(3):274–279.

Ross, T. R., Ng, D., Brown, J. S., Pardee, R., Hornbrook,

M. C., Hart, G., and Steiner, J. F. (2014). The hmo

research network virtual data warehouse: a public data

model to support collaboration. EGEMS, 2(1).

Schneeweiss, S. and Avorn, J. (2005). A review of uses

of health care utilization databases for epidemiologic

research on therapeutics. Journal of clinical epidemi-

ology, 58(4):323–337.

Schreiber, G. and Raimond, Y. (2015). Rdf 1.1 primer. w3c

working group note.

Sernadela, P., Gonz

alez-Castro, L., and Oliveira, J. (2017).

Scaleus: Semantic web services integration for

HEALTHINF 2019 - 12th International Conference on Health Informatics

472

biomedical applications. Journal of Medical Systems,

41:1–11.

Silva, L. B., Trifan, A., and Oliveira, J. L. (2018). Montra:

An agile architecture for data publishing and discov-

ery. Computer methods and programs in biomedicine,

160:33–42.

Speicher, S., Arwe, J., and Malhotra, A. (2015). Linked

data platform 1.0. w3c recommendation.

Trifan, A., D

ıaz, C., Oliveira, J., et al. (2018). A method-

ology for ﬁne-grained access control in exposing

biomedical data. Studies in health technology and in-

formatics, 247:561–565.

Triﬁr

o, G., Coloma, P., Rijnbeek, P., Romio, S., Mosseveld,

B., Weibel, D., Bonhoeffer, J., Schuemie, M., Lei, J.,

and Sturkenboom, M. (2014). Combining multiple

healthcare databases for postmarketing drug and vac-

cine safety surveillance: why and how? Journal of

internal medicine, 275(6):551–561.

Whetzel, P. L., Noy, N. F., Shah, N. H., Alexander, P. R.,

Nyulas, C., Tudorache, T., and Musen (2011). Biopor-

tal: enhanced functionality via new web services from

the national center for biomedical ontology to access

and use ontologies in software applications. Nucleic

Acids Research, 39:W541–W545.

Wilkinson, M., Dumontier, M., Aalbersberg, I., Appleton,

G., Axton, M., Baak, A., Blomberg, N., Boiten, J., da

Silva Santos, L., Bourne, P., Bouwman, J., Brookes,

A., Clark, T., Crosas, M., Dillo, I., Dumon, O., Ed-

munds, S., Evelo, C., Finkers, R., Gonzalez-Beltran,

A., Gray, A., Groth, P., Goble, C., Grethe, J., Heringa,

J., ’t Hoen, P., Hooft, R., Kuhn, T., Kok, R., Kok,

J., Lusher, S., Martone, M., Mons, A., Packer, A.,

Persson, B., Rocca-Serra, P., Roos, M., van Schaik,

R., Sansone, S., Schultes, E., Sengstag, T., Slater, T.,

Strawn, G., Swertz, M., Thompson, M., Van Der Lei,

J., Van Mulligen, E., Velterop, J., Waagmeester, A.,

Wittenburg, P., Wolstencroft, K., Zhao, J., and Mons,

B. (2016). The fair guiding principles for scientiﬁc

data management and stewardship. Scientiﬁc Data, 3.

Zaveri, A. and Ertaylan, G. (2017). Linked data for life

sciences. Algorithms, 10:126.

Strategies to Access Patient Clinical Data from Distributed Databases

473