Sunteți pe pagina 1din 9

2018 17th IEEE International Conference On Trust, Security And Privacy In Computing And Communications/ 12th

IEEE International Conference On Big Data Science And Engineering

PHeDHA: Protecting Healthcare Data in Health


Information Exchanges with Active Data Bundles
Wesam H. Fadheel Raed M. Salih Leszek T. Lilien
Department of Computer Science
Western Michigan University
Kalamazoo, MI 49008, USA
{wesam.fadheel; raedmahdi.salih; leszek.lilien}@wmich.edu

Abstract— Health Information Exchanges (HIEs) collect and confidentiality, integrity and security of individuals’ Electronic
disseminate electronic patient healthcare data (EHRs/EMRs) Protected Health Information (ePHI) [1, 2]. HIPAA Security
among different healthcare providers to improve the quality and Rule requires implementation of three levels of safeguards
reduce the cost of healthcare services. However, the dissemination including physical, administrative and technical [2]. It should be
of patient data raises privacy and security concerns due to ease of
noted that the Security Rule does not cover PHI that is
copying and unauthorized dissemination of electronic data. This
paper proposes a HIE system called PHeDHA (Protecting transmitted or stored on paper or provided orally, while the
Healthcare Data in HIEs with Active Data Bundles), which provides Privacy rule covers the confidentiality of PHI in all formats
privacy and security protection for patient data during their including electronic, paper and oral.
transmission via an HIE among different healthcare providers. If healthcare providers adopt and adhere to HIPAA Privacy
PHeDHA uses as its basis the scheme named Active Data Bundles and Security Rules, the exchange of individuals’ health data can
with Trusted Third Party (ADB-TTP). As the name suggests, ADB- be facilitated and enhanced.
TTP is based on an integration of a trusted third party (TTP) with
Active Data Bundles (ADBs). An ADB is a software object that A. Health Information Exchange (HIE)
keeps patient healthcare data as sensitive data; includes metadata
describing these sensitive data and prescribing their use (via data
Healthcare providers include hospitals, clinics, labs and
access and privacy policies specified within metadata); and healthcare research institutes, insurance companies and
encompasses a policy enforcement engine (called a virtual machine government agencies. Typically, healthcare information is
or VM), which controls and manages how the ADB behaves. In stored across multiple healthcare providers, who cooperate to
particular, the VM assures ADB’s data integrity and enforces provide technology, governance and support for so called Health
its policies specified as a part of metadata. We describe and discuss Information Exchanges.
the conceptual model for PHeDHA, based on ADB-TTP. We are A Health Information Exchange (HIE) enables cooperating
currently evaluating PHeDHA via simulation experiments. healthcare providers and patients to access and securely transfer
patients’ healthcare data electronically, in this way improving
Keywords—active data bundles; health information exchange;
the quality, safety and cost of patient care [3]. Each HIE must
privacy; security; trusted third party.
adopt all HIPAA regulations, including Privacy and Security
I. INTRODUCTION Rules.
Patient healthcare data include Electronic Health/Medical
The Health Insurance Portability and Accountability Act
Records (EHRs/EMRs), which are records of patients’ long-term
(HIPAA) [1] is a legal framework for protecting individuals by
health history, containing doctor visit summaries, diagnoses,
assuring privacy and security of healthcare data. The U.S.
prognoses, laboratory and radiology tests, treatment
Department of Health and Human Services (HHS) issued
descriptions, etc. [4].
HIPAA Privacy Rule [1] to implement HIPAA requirements and
There is a difference between EMRs and EHRs. EMRs are
regulations. The Rule promotes high-quality healthcare services
legal records for patients that are created in a single healthcare
for individuals in order to improve the public health and well-
provider facility (like a hospital or a clinic). In contrast, EHRs
being. This can be achieved by providing appropriate safeguards
are the summaries of EMRs for a given patient that are collected
and restriction conditions (or rules) to protect the privacy of
by authorized healthcare providers from more than one
Protected Health Information (PHI) as well as dictate its use and
healthcare provider visited by the patient [5].
disclosure among healthcare providers1 [1, 2].
HIE manages multi-directional flows of electronic patients’
Since most of the PHI is created, stored, used and
healthcare information among different healthcare providers as
disseminated electronically using computer systems, the HHS
shown in Figure 1.
included a Security Rule in HIPPA [1] to ensure the
1
Examples of healthcare providers include physicians, nurses, dentists,
pharmacists, healthcare researchers as well as organizations employing them
(hospitals, clinics, medical labs, medical research institutes, etc.)

2324-9013/18/31.00 ©2018 IEEE 1187


DOI 10.1109/TrustCom/BigDataSE.2018.00164
Under HIPAA requirements, healthcare providers must
anonymize data by removing all “identifiers” from healthcare
records before sharing them with other healthcare providers or
other authorized parties [7, 8].
C. Privacy and Confidentiality in HIE
We define “security as the right not to have one’s activities
adversely affected via tampering with one’s objects, and
privacy as the right to have information about oneself left
alone” [9].
User privacy is a user’s right to protect and control her data.
As a special case, patient privacy refers to the patient’ right to
protect and control her healthcare-related data [23, 24]. Patients
should be able to determine when, how and which portions of
their health information are disclosed or disseminated [11].
User confidentiality is the right of an individual or user to
Figure 1. Example of Health Information Exchange (HIE).
keep her personal and medical information private unless she
gives her permission to disclose a part or the whole of this
There are three exchange procedures used by HIEs to share
information to another party [10]. As a special case, patient
patient information among multiple healthcare providers [3]:
confidentiality is the right of a patient to keep her healthcare or
1) Directed exchange – enabling healthcare providers to medical information private unless she gives her permission to
coordinate secure electronic transmissions (sending and disclose a part or the whole of this information to another party
receiving) of patient information. [11, 12].
2) Query-based exchange – enabling healthcare providers to
find and/or request information on a patient from other D. Paper Contribution and Organization
healthcare providers. This paper proposes a secure health information exchange
3) Consumer-mediated exchange – empowering patients to by a HIE using a scheme known as Active Data Bundles with
aggregate and control the use of their health information by Trusted Third Party (ADB-TTP). ADB-TTP is based on the use
healthcare providers. of a trusted third party (TTP) and Active Data Bundles (ADBs).
ADBs protect carried patient healthcare data by encapsulating
B. Protected Health Information (PHI) them as their sensitive data; by including metadata, which
describe the sensitive data and prescribe their use (via policies,
Table I shows eighteen PHI/ePHI identifiers which are
including access and privacy policies); and by encompassing an
created, stored, and maintained by or on behalf of healthcare
execution engine (called a virtual machine or VM). VM controls
providers.
and manages how the ADB behaves; among others, how it
Table I. HIPAA Protected Health Information (PHI/ePHI) [2]. enforces privacy policies.
The paper is organized as follows. Section II discusses
No. PHI/ePHI Identifiers related work. Section III presents motivation and the problem
1 Names (patient, healthcare provider) of protecting healthcare data disseminated among diverse
2 Addresses authorized healthcare providers. Section IV describes structure
3 Geographic subdivisions and operations of ADB-TTP. Section V presents the baseline
4 All elements of dates related to the individual (dates of birth,
marriage, death, etc.) HIE and the proposed HIE system named Protecting Health
5 Telephone numbers Data in HIE with ADBs (PHeDHA—pronounced FEH-dah,
6 Facsimile numbers which rhymes with “Jedah”). Section VI presents Conclusions,
7 Driver’s license numbers and Section VII presents future works.
8 Electronic mail addresses
9 Social security numbers
10 Medical record numbers
11 Health plan beneficiary numbers II. RELATED WORK
12 Account numbers, certificate/license numbers Zhou et al. [13] propose Hospital Information eXchange
13 Vehicle identifiers and serial numbers
14 Device identifiers and serial numbers (HIX) for hospital healthcare business-services system. The
15 Web Universal Resource Locators (URLs) HIX is a cloud-based computing support for business functions,
16 Internet Protocol (IP) address numbers including integrated mobile medical information services and
17 Biometric identifiers mobile health information services. The HIX uses the cloud-
18 Full face photographic images and any comparable
based technology and business model of the Monetary
Authority of Singapore (MAS). It provides total solutions for
hospital healthcare business services system, including efficient

1188
hospital information management enabling shorter waits for healthcare providers exchanging patient healthcare data via
registration, payments, and doctor visits. It also facilitates HIE. Any request for patient data is mediated by HIE. For
doctor-patient communication and reduces costs of hospital instance, when Clinic requests a patient’s healthcare data
healthcare services. Security in HIX relies on security solutions through HIE, HIE distributes the request to healthcare providers
built into the MAS technology. that the patient visited. Each healthcare provider that has the
Afazl et al. [14] propose a HIE framework providing requested data, sends them to HIE. Then, HIE responds to the
a secure access to patient data and adhering to the Clinic’s request by sending to it all data collected from different
interoperability rules. HIE framework includes six healthcare providers.
components: Health Life Horizon: Personal Health Record Clearly, dissemination of patients’ healthcare data among
(HLH PHR), which provides infrastructure for individual users many diverse providers increases the risk of disclosing (or
to manage their shareable data; Keys Manager (KM), which leaking) patients’ private information to unauthorized parties.
manages keys for encryption of PHR records; Authentication For protecting privacy of patients’ sensitive data
Manager (AM), which authenticates various categories of users; disseminated within HIE, we propose the system named
Transaction Manager (TM), which releases/accepts resource PHeDHA (Protecting Health Data in HIE with ADBs), which is
contents to be published on HLH PHR for authenticated users; based on the ADB-TTP scheme.
PHR Repository (PHR-R), which stores contents published on
HLH PHR; and External Interfaces, which provide patient data
in a secure manner, based on proper authentication and IV. STRUCTURE AND BASIC OPERATIONS OF ACTIVE DATA
authorization. The proposed framework implements the L- BUNDLES WITH TRUSTED THIRD PARTY (ADB-TTP)
diversity algorithm to ensure individual privacy when A. ADB Structure
publishing data for statistical analysis. An ADB (a.k.a. AB or Active Bundle) [19] is a software
Lijun and Jianchao [15] propose a patient identification
construct bundling together three components (cf. Figure 2):
service for the Chinese healthcare administration, which adopts
a centralized HIE architecture to protect patient healthcare 1) Sensitive data representing a digital content that needs
information. The proposed identification service is based on to be protected from unauthorized access, including privacy
Patient Identifier Cross-Reference (PIX) to process shared violations, data leaks, unauthorized dissemination, etc.
information; the Master Patient Identifier (MPI) service to 2) Metadata containing information describing sensitive
construct or select a suitable unique identifier for the HIE data and prescribing their use, such as access and privacy
domain; and Match Engine (the core of HIE) to provide an policies for the sensitive data.
identification service using a cross-reference algorithm (which 3) Virtual machine (VM) controlling and managing how
helps to assure uniqueness of patient identifications). HIE the ADB encompassing it behaves. VM is the component that
depends on a unique patient identity (such as a national ID card makes its ADB active because it is implemented via executable
number) to match and integrate all patient healthcare mobile code and responsible for performing all ADB
information provided by various healthcare institutions. operations. The essential tasks of the VM include assurance of
Ma et al. [16] propose a hybrid cloud prototype to support ADB data integrity, and enforcement of the privacy policies and
digital home-based self-care. It composes OpenStack and data dissemination rules specified by ADB metadata for ADB
Amazon Web Services to exchange health information among data [19, 20].
patients and multiple stakeholders. The prototype exchanges
health data uses the HL7 CCD format [17]. The authors claim
that using open-source cloud computing technologies for
sharing health information is secure, scalable, cost-effective,
and partially solves interoperability and privacy problems.
Yu et al. [18] propose a city-wide HIE system for Shanghai,
the largest HIE system in China. It includes six districts
Sensitive data
hospitals and 40 community health centers, covering 39 million
patients and saving at least 48 million RMB. The system is
based on SOA (Service-Oriented Architecture) and complies Metadata
with industry standards. Healthcare data from different
providers are integrated and stored in a central data repository, Virtual machine (VM)
and linked by a city-wide general patient index (using unique
patient healthcare identifiers). The authors indicate that data
security and privacy remain the major challenges. Figure 2. The basic structure of an Active Data Bundle (ADB) [20].

The use of the term “virtual machine (VM)” in the ADB


III. MOTIVATION AND PROBLEM STATEMENT
context is different than in the operating systems and related
Protecting exchanged patients’ sensitive data is commonly
considered the main problem in HIEs. Figure 1 shows diverse

1189
areas.2 This term is used here since ADB’s VM still exhibits origination host either automatically or interactively with the
general characteristics of a VM—esp. enabling mobility, help of a user-friendly ADB Creator (ADBC) software. With an
flexibility, and compatibility [21]. assistance from ADBC, ADB is going through a sequence of
actions to assure confidentiality and integrity of the ADB’s
B. ADB Hosts and ADB Lifecycle sensitive data during the planned ADB dissemination.
Any ADB can be transmitted from its origination host to its ADBC uses data encryption and hashing techniques
destination host via a sequence of intermediate hosts multiple times to protect confidentiality and integrity of ADB’s
(optionally using a secure transmission for even better privacy components (sensitive data, metadata, and VM). To protect
and security). The intermediate and destination hosts are privacy of ADB’s sensitive data, ADBC encloses the
collectively known as visited hosts (VHs). origination host’s privacy policies for these data in the ADB’s
ADB lifecycle includes three phases: ADB creation, ADB metadata.
dissemination, and ADB enabling [19]. ADBC bundles ADB’s components, completing ADB
Any ADB lifecycle phase may require ADB reduction construction. Now, ADB-TTP sends to the TTP the enabling
and/or ADB termination [19]. ADB reduction involves a partial record, which includes decryption keys and hash values
ADB destruction, which occurs if only a part of sensitive ADB (generated by ADBC) required by the destinations for accessing
data are threatened. ADB data. The ADB is ready for sending to the destination
ADB termination always involves a complete ADB hosts.
destruction. We distinguish: (a) a triggered ADB termination, 2) ADB Dissemination
which occurs before a given ADB completes its data-delivery Using a secure transmission, ADB-TTP sends the newly
task, in response to a threat that all sensitive data in this ADB created ADB to one or more destination hosts, possibly via
will be disclosed by the attack materializing this threat; and (b) a sequence of intermediate forwarding hosts [19].
a pre-planned ADB termination, which occurs at the end of the
ADB enabling phase performed on the last host visited by the 3) ADB Enabling
given ADB while executing it multi-destination data Enabling of an ADB starts when it reaches a destination host
dissemination job.3 (DH). ADB enabling comprises four groups of activities (cf.
Figure 3): ADB Authentication, ADB Authorization, ADB
C. Partial or Complete ADB Destruction via Evaporation Integrity Verification, and ADB Policy Enforcement [6, 24].
and Apoptosis
A partial ADB destruction occurs via the mechanism of
evaporation (an analogy to a phenomenon in physics). An
evaporating ADB destroys only a part of its data, namely the
hottest data. The hottest data are identified by having data-
sensitivity level exceeding the clearance level that an entity
trying to access it possesses.
The complete ADB destruction occurs via apoptosis (an
analogy to a cell-biology phenomenon). The apoptosizing ADB
self-destroys completely (destroying its data, metadata, and
VM), irretrievably, and in a privacy-preserving and secure
manner.
As an example, suppose that a given ADB has sensitive data
with Confidential, Secret, and Top Secret data-sensitivity
levels. If there is a threat posed by an entity (a person, an
application, etc.) with the Confidential clearance, then all Secret
and Top Secret data are the “hottest” data (”hotter” than
Confidential data), and they evaporate. If there is another threat,
this time posed by an entity with the Restricted clearance
(below the Confidential level), then ADB apoptosizes.
D. More Details of ADB Lifecycle and Operations
The three ADB lifecycle phases are as follows.
1) ADB Creation Figure 3. ADB enabling steps performed by VM at a destination host [6].
In this phase, a given ADB is created for a user at the

2
A system VM (a.k.a. a hardware VM) enables access to a complete system implementations, it delivers replication, emulation, and facilitates
platform—including its operating system, the underlying hardware, performance optimization [21, 22].
networking resources, input/output devices, and GUI [21]. A process VM 3
We allow for multi-destination ADB dissemination, in which a single ADB
(a.k.a. an application VM) provides user applications with a high-level visits and delivers data to multiple destination hosts.
abstraction through a high-level programming language. In its various

1190
Figure 4. The baseline scenario for HIE: Components and process flow for Clinic’s request for P-EMR from HIE.

The enabling record (sent by ADB-TTP to the TTP during 3) Control Unit (CU): used to process any request verified by
ADB Creation) is requested from TTP and used for Steps a - c RV.
of ADB Enabling. 4) Database (DB): a local database holding a subset of P-
The flow of control during ADB Enabling, shown in Figure EMRs.
3, includes four steps: 5) Query Interface (QI): used to send a query (a request): in
a) If authentication fails, then VM apoptosizes the entire ADB case of a P-EMR requester (e.g., Clinic), to send a request
and ADB enabling ends. Otherwise (when authentication to HIE; in case of HIE, to forward a request to healthcare
succeeds), authorization starts. providers.
b) If authorization fails, then VM apoptosizes the entire ADB
and ADB enabling ends. Otherwise, integrity verification A. The Baseline Scenario for HIE
starts. In our baseline scenario for HIE, shown in Figure 4, Clinic
c) If integrity verification fails, then VM apoptosizes the requests HIE for an up-to-date P-EMR.
entire ADB and ADB enabling ends. Otherwise, policy The actors in this scenario are: (a) Clinic acting as an
enforcement starts. originating host with a user (a doctor) who is health data
d) During policy enforcement, VM enforces the privacy requester for an up-to-date P-EMR; (b) HIE acting as
policies (stored in the ADB’s metadata) and delivers to DH a mediator between the healthcare data requester and other
only the data that the policy permits DH to access. healthcare providers; and (c) Hospital acting as a healthcare
provider in possession of the requested up-to-date P-EMR.
The flow of requesting up-to-date P-EMR (cf. Figure 4) is
V. THE BASELINE HIE AND THE PHEDHA SYSTEM as follow:
A HIE system consists of five components (cf. Figure 4): Doctor at Clinic requests for an up-to-date P-EMR
1) Interface Application (App): used as an interface which 1) A Doctor at Clinic uses App to request for P-EMR; App
receives a user’s request for Patient Electronic Medical forwards the request to Clinic RV.
Record (P-EMR). 2) After a positive verification, Clinic RV passes the request
2) Request Verifier (RV): used to verify any request received to Clinic CU.
either from App or from other providers. 3) Clinic CU queries its local Clinic DB for the up-to-date P-
EMR.

1191
Figure 5. Basic scenario for PHeDHA: ADB-TTP components and process flow for Clinic’s request for P-EMR from HIE.

4) Local Clinic DB indicates that its copy of P-EMR is HIE responds to Clinic request
outdated. for the up-to-date P-EMR
5) Clinic CU forwards the P-EMR request to Clinic QI. 16) After a positive verification of the up-to-date P-EMR
6) Clinic QI sends the request to HIE RV. received from Hospital, HIE RV sends it to HIE CU.
HIE forwards Clinic request for the up-to-date 17) HIE CU sends the up-to-date P-EMR to the local DB so
P-EMR to Hospital4 that HIE DB can update its copy of the P-EMR.
7) After a positive verification of the Clinic request, HIE RV 18) HIE DB sends an ACK (confirmation) to HIE CU.
passes it to HIE CU. 19) HIE CU sends the response with the up-to-date P-EMR to
8) HIE CU queries its local DB for the up-to-date P-EMR. Clinic RV.
9) HIE DB indicates that its copy of P-EMR is outdated. Clinic responds to Doctor at Clinic
10) HIE CU forwards the Clinic P-EMR request to HIE QI. 20) After a positive verification of the up-to-date P-EMR
11) HIE QI sends the request to Hospital RV. received from HIE, Clinic RV sends it to Clinic CU.
Hospital responds to the HIE request 21) Clinic CU sends the up-to-date P-EMR to its local DB so
for the up-to-date P-EMR that the local DB can update its copy of the P-EMR.
12) After a positive verification of the HIE request, Hospital 22) Clinic DB sends an ACK to Clinic CU.
RV passes it to Hospital CU. 23) Clinic CU sends the up-to-date P-EMR to its App, which
13) Hospital CU queries its local DB for the up-to-date P- delivers the up-to-date P-EMR to Doctor who requested it.
EMR. B. The PHeDHA Scenario for HIE
14) Hospital DB finds the requested up-to-date P-EMR and
The analogous request-and-response scenario for PHeDHA,
sends the response with P-EMR to Hospital CU as the
which uses the ADB-TTP scheme, is significantly different than
response to the query.
the above baseline HIE scenario.
15) Hospital CU sends the up-to-date P-EMR to HIE RV. Figure 5 illustrates the data and control flow for the
4
In general, HIE can forward a party’s request for P-EMR to not just one but
many providers.

1192
scenario, involving Clinic, HIE and Hospital. Note that the step.
PHeDHA scenario includes not only HIE but also some details 9) H-ADB’s VM executes ADB Policy Enforcement (on HIE
of the modules used in Clinic and Hospital to implement the CU), including privacy enforcement, to release the up-to-
ADB-TTP scheme. In addition to being a mediator (as in the date P-EMR to HIE.
baseline HIE scenario), HIE in PHeDHA also acts as a TTP 10) H-ADB’s VM releases the up-to-date P-EMR (till now
(providing the enabling record to a given ADB when ADB is encapsulated within H-ADB) to HIE CU.
processed at HIE and Clinic). Please note that HIE’s TTP
11) HIE CU sends the released up-to-date P-EMR to the local
activities are not shown in Figure 5.
HIE DB so that HIE DB can update its copy of the P-EMR.
In PHeDHA, P-EMR is protected by being encapsulated by
Note that Steps 5-11 can be repeated N times when N ADBs
an ADB. Hence, this scenario includes ADB creation, ADB
(with N copies of the P-EMR) are received by HIE CU
dissemination, and ADB enabling (till termination by
from healthcare providers (such as Hospital, Laboratory,
apoptosis).
etc., shown in Figure 1).5
The following steps are not shown in Figure 5: 12) HIE DB merges all released up-to-date P-EMRs (received
1) Doctor at Clinic requests for an up-to-date P-EMR from HIE CU) into a single P-EMR and sends it to HIE
(a sequence of steps identical to Steps 1-6 in the baseline CU.
HIE scenario). 13) HIE CU attaches metadata to the merged P-EMR. These
2) HIE forwards Clinic request for the up-to-date P-EMR to metadata integrate (merge) metadata of all individual P-
Hospital (a sequence of steps identical to Steps 7-11 in the EMRs released to HIE-CU (in one or more repetitions of
baseline HIE scenario). Step 10).
3) After a positive verification of the HIE request, Hospital 14) HIE CU encapsulates the result of Step 13 within VM, thus
RV passes it to Hospital CU (identical to Step 12 in the completing construction of the HIE-ADB.
baseline HIE scenario). 15) HIE CU sends HIE-ADB to Clinic RV (for verification of
4) Hospital CU queries its local DB for the up-to-date P-EMR HIE-ADB).
(the step identical to Step 13 in the baseline HIE scenario).
Clinic responds to Doctor at Clinic
The steps following Step 4 (given directly above) are shown
16) After a positive verification, Clinic RV forwards HIE-ADB
in Figure 5; they are renumbered below starting with 1.
to Clinic CU for enabling.
Continued: Hospital responds to the HIE request 17) HIE-ADB’s VM executes ADB Authentication (on Clinic
for the up-to-date P-EMR using ADB CU). If this authentication fails, HIE-ADB is apoptosized
1) Hospital DB finds the requested up-to-date P-EMR and by VM. Otherwise, VM proceeds to the next step.
sends the response with P-EMR to Hospital CU as the 18) HIE-ADB’s VM executes ADB Authorization (on Clinic
response to the query. CU). If this authorization fails, HIE-ADB is apoptosized
2) Hospital CU attaches metadata (“MD”) to P-EMR. by VM. Otherwise, VM proceeds to the next step.
3) Hospital CU encapsulates the result of Step 2 within VM, 19) HIE-ADB’s VM executes ADB Integrity Check (on Clinic
thus completing creation of the Hospital ADB (H-ADB). CU), verifying P-EMR. If this check fails, HIE-ADB is
4) Hospital CU sends H-ADB to HIE RV (for verification of apoptosized by VM. Otherwise, VM proceeds to the next
H-ADB). step.
20) HIE-ADB’s VM executes ADB Policy Enforcement (on
HIE responds to Clinic request Clinic CU), including privacy enforcement, to release the
for the up-to-date P-EMR up-to-date P-EMR to Clinic.
5) After a positive verification, HIE RV forwards H-ADB to 21) HIE-ADB’s VM releases the up-to-date P-EMR (till now
HIE CU. encapsulated within HIE-ADB) to Clinic CU.
6) H-ADB’s VM executes ADB Authentication (on HIE CU). 22) Clinic CU sends the up-to-date P-EMR to the Clinic DB so
If this authentication fails, H-ADB is apoptosized. that Clinic DB can update its copy of the P-EMR.
Otherwise, VM proceeds to the next step.
23) Clinic DB sends ACK to Clinic CU. Note that there is no
7) H-ADB’s VM executes ADB Authorization (on HIE CU). merging of P-EMRs by the Clinic DB here, so Clinic CU
If this authorization fails, H-ADB is apoptosized by VM. has this P-EMR copy already.
Otherwise, VM proceeds to the next step.
24) Clinic CU sends the up-to-date P-EMR to App, which
8) H-ADB’s VM executes ADB Integrity Check (on HIE delivers the up-to-date P-EMR to Doctor who requested it.
CU), verifying P-EMR. If this check fails, H-ADB is
apoptosized by VM. Otherwise, VM proceeds to the next

5
Note that in case that multiple up-to-date E-PMR are released to HIE-CU, Laboratory) might bring to HIE-CU a partial up-to-date E-PMR with the
they are “partial up-to-date E-PMRs.” For example, H-ADB (from Hospital) most recent lab results for the patient.
might bring to HIE-CU a partial up-to-date E-PMR describing the most
recent hospital visits of the patient in question on, while L-ADB (from

1193
VI. CONCLUSIONS There are four granularity types:
Health information exchange shares healthcare data among 1) Granularity by data type: a patient expresses her desire to
healthcare providers to improve the quality, safety, and cost of block or not a specific data record/element during
healthcare services. However, sharing patient healthcare data electronic exchanges.
among many diverse providers increases the risk of disclosing 2) Granularity by provider: a patient expresses her desire to
patients’ sensitive data to unauthorized parties. allow specific healthcare providers or staff categories (e.g.,
This paper presents the basic structure and operations of the nurses) to access her healthcare data during electronic
PHeDHA system for protecting patient healthcare data exchanges.
exchanged among different healthcare providers in a health 3) Granularity by time range: a patient expresses her desire to
information exchange (HIE). PHeDHA aims to alleviate allow access to pre-defined subsets of her data within
privacy and security concerns related to these data exchanges. a specific time range during electronic exchanges.
We are currently designing proof-of-concept simulation 4) Granularity by purpose: a patient allows her data to be
experiments using J-Sim [25] to validate effectiveness and accessed according to a specific purpose during electronic
efficiency of PHeDHA. We plan to simulate both the baseline exchanges.
HIE scenario, which shows interactions among healthcare
providers through HIE, and the analogous PHeDHA scenario,
which differs from the baseline scenario only by using in it REFERENCES
Active Data Bundles and Trusted Third Parties. The results [1] “HIPAA Survival Guide,” Lions Publishing. Accessed in February 2018
obtained in both simulations will provide quantitative data for from: http://www.hipaasurvivalguide.com.
a comprehensive evaluation of PHeDHA versus the baseline [2] “HIE Privacy and Security Overview and Resource List,” National Rural
system. Health Resource Center. Accessed in February 2018 from:
https://www.ruralcenter.org/resource-library/hie-privacy-and-security-
overview-and-resource-list.
VII. FUTURE WORK [3] “Health Information Exchange (HIE),” HealthIT.gov. Accessed in
February 2018 from: https://www.healthit.gov/providers-
We plan two extensions to the presented work. First, we professionals/health-information-exchange/what-hie#query-
plan to use consent models [26] by enhancing ADB privacy based_exchange
policies [19, 6] with consent rules. The consent models are [4] “Electronic Health Records Overview,” National Center for Research
intended to maintain and control patients’ data disclosures in Resources, NIH, Bedford, MA, April 2006. Accessed in June 2011 from:
http://www.ncrr.nih.gov/publications/informatics/ehr.pdf.
electronic exchanges in HIE by enforcing patients’ selected
[5] D. Garets and M. Davis, “Electronic Medical Records vs. Electronic
preferences and conditions [26].
Health Records: Yes, There Is a Difference,” White Paper, HIMSS
There are five consent models: Analytics, Chicago, IL, January 2006.
1) No-consent: patient does not participate in electronic [6] R.M. Salih, “Using Agent-based Implementation of Active Data Bundles
exchanges (no data are exchanged). for Protecting Privacy in Healthcare Information Systems,” Ph.D. Thesis,
Department of Computer Science, Western Michigan University,
2) Opt-in: patient expresses her desire to participate in Kalamazoo, MI, April 2018.
electronic exchanging of all her data or some pre-defined [7] S. Martínez, D. Sánchez, and A. Valls, “A Semantic Framework to Protect
subset of her data. the Privacy of Electronic Health Records with Non-numerical Attributes,”
3) Opt-out: patient expresses her desire to automatically J. of Biomedical Informatics, vol. 46(2), New York, April 2013, pp. 294-
303.
participate in electronic exchanging of pre-defined subsets
of her data. [8] T. Mabotuwana, C.S. Hall, R. Van Ommering, R. Tellis and M. Sevenster,
“An HL7 Data Pseudonymization Pipeline,” IEEE Intl. Conf. on
4) Opt-out with exceptions: patient expresses her desire to Healthcare Informatics (ICHI), Dallas, TX, October 2015, pp. 303-309.
participate in electronic exchanging (with specific [9] A. Al-Gburi, A. Al-Hasnawi and L. Lilien, “Differentiating Security from
exceptions) of all her data or some pre-defined subsets of Privacy in Internet of Things—A Survey of Selected Threats and
her data. Controls,” Chapter 9 in: K. Daimi et al., Computer and Network Security
Essentials, Springer International Publishing, Cham, Switzerland, 2018,
5) Opt-out with restrictions: patient expresses her desire to pp. 153-172.
participate in electronic exchanging (with restrictions) of [10] R. Shirey, “RFC 2828 - Internet Security Glossary,” Internet FAQ
all her data or some pre-defined subsets of her data. Archives, June 2012. Available at: http://www.faqs.org/rfcs/
ADBs will enforce consent rules with their privacy policies rfc2828.html
during ADB dissemination among healthcare providers. [11] R.C. Barrows and P.D. Clayton, “Privacy, Confidentiality, and Electronic
Medical Records,” J. of the American Medical Informatics Association,
As the second extension, we plan to add granularity rules vol. 3(2), March/April 1996, Avenel, NJ, pp. 139-148.
[26] to ADB privacy policies. [12] “Patient confidentiality, “Encyclopedia of Surgery. June 2011.
Granularity refers to a patient’s concern or fear of how her Accessed in February 2018 from: http://www.surgeryencyclopedia.com/
personal information is perceived or used during the electronic Pa-St/Patient-Confidentiality.html
exchange. This concern and fear can be eliminated by enforcing [13] F. Zhou, F. Cheng, L. Wei and Z. Fang, “Cloud Service Platform-
granularity rules that grant the desired level of control over Hospital Information Exchange (HIX),” IEEE 8th Intl. Conf. on e-
specific sensitive information included in the patient’s Business Engineering, Beijing, China, October 2011, pp. 380-385. DOI:
10.1109/ICEBE.2011.35.
healthcare data.
[14] M. Afzal, M. Hussain, M. Ahmad, and Z. Anwar, “Trusted Framework

1194
for Health Information Exchange,” Frontiers of Information Technology [21] “Virtual Machine,” Wikipedia. Accessed in February 2018 from:
(FIT), Islamabad, Pakistan, December 2011, pp. 308-313. https://en.wikipedia.org/wiki/Virtual_machine.
[15] W. Lijun and C. Jianchao, “The Identification Service in Health [22] J. Smith and R. Nair, Virtual Machines: Versatile Platforms for Systems
Information Exchange,” IEEE 16th Intl. Conf. on Computational Science and Processes, Morgan Kaufmann Publishers, San Francisco, CA, 2005.
and Engineering, Sydney, Australia, December 2013, pp. 1161-1166. [23] R.M. Salih, L.T. Lilien, and L. Ben Othmane, “Protecting Patients’
[16] J. Ma, C. Peng and Q. Chen, “Health Information Exchange for Home- Electronic Health Records Using Enhanced Active Bundles,” 6th Intl.
Based Chronic Disease Self-Management—A Hybrid Cloud Approach,” ICST Conf. on Pervasive Computing Technologies for Healthcare
5th Intl. Conf. on Digital Home, Guangzhou, China, November 2014, (PervasiveHealth 2012), San Diego, CA, May 2012, 4 pages, DOI:
pp.246-251. DOI: 10.1109/ICDH.2014.54 10.4108/icst.pervasivehealth.2012.248719
[17] “HL7/ASTM Implementation Guide for CDA® R2 -Continuity of Care [24] R.M. Salih, L. Ben Othmane and L. Lilien, “Privacy Protection in
Document (CCD®) Release 1,” (n.d.). Accessed in February 2018 from: Pervasive Healthcare Monitoring Systems with Active Bundles,” Intl.
http://www.hl7.org/implement/standards/product_brief.cfm?product_id= Conf. on Advanced Software Engineering (ICASE), in conjunction with
6 ISPA 2011, Busan, Korea, May 2011, pp. 311-315. DOI:
[18] Y. Guangjun, C. Wenbin, Z. Li, B. D. W, G. Jianlei and L. Hui, 10.1109/ISPAW.2011.60.
“Implementation of a city-wide Health Information Exchange solution in [25] J-Sim Official, (n.d.). Accessed in February 2018 from:
the largest metropolitan region in China,” IEEE Intl. Conf. on https://sites.google.com/ site/ jsimofficial
Bioinformatics and Biomedicine (BIBM), Shenzhen, China, December [26] M. M. Goldstein, A. L. Rein, P. P. Hughes, J. K. Lappas, S. A. Weinstein
2016, pp. 795-798. and B. Williams, “Consumer Consent Options for Electronic Health
[19] L. Ben Othmane and L. Lilien, “Protecting Privacy in Sensitive Data Information Exchange: Policy Considerations and Analysis,” White
Dissemination with Active Bundles,” Seventh Annual Conf. on Privacy, Paper, Health Information Technology, Washington, DC, March 2010,
Security and Trust (PST 2009), Saint John, New Brunswick, Canada, https://www.healthit.gov/sites/default/files/choicemodelfinal032610.pdf
August 2009, pp. 202-213.
[20] L. Ben Othmane, “Active Bundles for Protecting Confidentiality of
Sensitive Data Throughout Their Lifecycle,” Ph.D. Thesis, Department of
Computer Science, Western Michigan University, Kalamazoo, MI,
December 2010.

1195

S-ar putea să vă placă și