praveen r. rao , stanley a. edlavitch , jeffrey l. hackman ,...
TRANSCRIPT
TEMPLATE DESIGN © 2008
www.PosterPresentations.com
Our Vision
Potential Funding Sources
Architectural Overview
References
Introduction
Collaborators
Motivations
1. Centers for Medicare and Medicaid Service, Office of Actuary, 2009,
http://www.cms.hhs.gov/NationalHealthExpendData/downloads/proj2009.pdf.
2. American Cancer Society, Surveillance Research, 2009,
http://www.cancer.org/downloads/PRO/2009_cases_deaths_by_age.pdf.
3. Fenstermacher D, Street C, McSherry T, Nayak V, Overby C, Feldman M: The Cancer Biomedical
Information Grid (caBIG). Conference Proceedings IEEE Engineering in Medicine and Biology
Society, 2005.
4. Weber GM, Murphy SN, McMurry AJ, Macfadden D, Nigrin DJ, Churchill S, Kohane IS: The Shared
Health Research Information Network (SHRINE): A Prototype Federated Query Tool for Clinical
Data Repositories, Journal American Medical Informatics Association, Sep 2009.
5. Stead WW, Lin HS: Computation Technology for Effective Health Care: Immediate Steps and
Strategic Directions. Washington D.C.: The National Academies Press, 2009.
6. Hristidis V: Information Discovery on Electronic Health Records, CRC - Taylor & Francis,
December 2009.
7. Rao PR, Moon B: Locating XML Documents in a Peer-to-Peer Network Using Distributed Hash
Tables. IEEE Transactions on Knowledge and Data Engineering, Volume 21, No 12, pp 1737-1752,
December 2009.
8. Rao PR, Moon B: An Internet-Scale Service for Publishing and Locating XML Documents.
Proceedings of 25th IEEE International Conference on Data Engineering (ICDE ’09), Shanghai,
China, March 2009.
9. Scott Boag, Don Chamberlin, Mary Fernandez, Daniela Florescu, Jonathan Robie, Simon J:
XQuery 1.0: An XML Query Language. World Wide Web Consortium Recommendation 23, January
2007.
10. Nebraska Health Information Initiative, http://www.nehii.org/.
National Institutes of Health (NIH)
National Science Foundation (NSF)
Agency for Healthcare Quality and Research (AHRQ)
Missouri Life Sciences Research Board (MLSRB)
4.48
0.00
0.50
1.00
1.50
2.00
2.50
3.00
3.50
4.00
4.50
5.00
Tri
llio
ns
National Health Expenditure
Based on data provided by US Department of Health and Human Services [1]
Source: Centers for Medicare and Medicaid Services,
Office of Actuary, National Health Statistics Group
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
All Ages Under 45 45 and Over Under 65 65 and Over
Mill
ion
s
Cancer cases (2009)
Men Women Both Genders
Based on data provided by
American Cancer Society [2]
Key Challenges
Sample Queries
Incidence: What is the incidence of small cell lung cancer in a non-
smoker male between 2007 through 2010?
Regimen: What is the best treatment regimen for melanoma? What
are the alternative regimens?
Staging: How many patients were diagnosed with prostate cancer
stage II-B?
Radiation side-effects: What kind of cardiac side-effects were
observed in patients receiving radiation to left breast?
Chemotherapy side-effects: How safe is R-CHOP regimen for my
condition?
Survival rate: What is the 5-year survival rate for a patient with
stage II-B colon cancer?
Remission: What are the chances of complete remission for
cervical cancer patient treated with adjuvant chemoradiotherapy?
Praveen R. Rao1, Stanley A. Edlavitch2, Jeffrey L. Hackman3, Timothy P. Hickman2, Douglas S. McNair4, and Deepthi S. Rao5
1 School of Computing and Engineering, University of Missouri-Kansas City, Kansas City 2 School of Medicine, University of Missouri-Kansas City, Kansas City 3 Truman Medical Centers, Kansas City 4 Cerner Corporation, Kansas City 5 Argentine Family Health, Kansas City
Today the nation faces one of the toughest
challenges in health care due to high
operating costs.
$2.3 Trillion in 2008
National Health Expenditures, 2008
Cancer is the second most common cause of deaths in the US.
Health care costs are rising steadily. It is estimated that in the year 2019, US
will spend $4.48 trillion for health care.
Vast amounts of information (e.g., electronic health records, drug data, data
from clinical diagnosis) remain largely untapped due to the lack of suitable IT
solutions. Data sources evolve over time and are heterogeneous.
Data sharing and collaboration, and large-scale management of health care
data have been identified as the key IT challenges to advance the nation’s
health care system [3].
The Institute of Medicine (IOM) envisions the development of a learning
healthcare system.
IOM’s quality aims: safe, timely, effective, efficient, equitable, patient-centered.
Can we provide cost-effective, more efficient, and higher quality care to patients by
sharing health care data?
How can we share health care data on a very large-scale (e.g., petabytes of data)?
How can we manage non-standardized data sources that evolve with time?
How can we pose a single query to execute across multiple data sources?
Can we protect the privacy of patients and ownership of patient data?
Can we ensure HIPAA compliance?
<?xml version="1.0" ?>
<ClinicalDocument>
<id extension="49912" root="2.16.840.1.113883.3.933"/>
<patient>
<name>
<given>John</given>
<family>Doe</family>
</name>
<genderCode code="M" codeSystem="2.16.840.1.5.1"/>
<birthTime value="20020924"/>
</patient>
<component>
<StructuredBody>
<component>
<section>
<code code="10160-0" codeSystem="2.16.840.1.113883.6.1" codeSystemName="LOINC" />
<title>Medications</title>
<entry>
<Observation> <code code="84100007" codeSystem="2.16.840.1.113883.6.96" codeSystemName="SNOMED CT" displayName=" medication
history"/>
<value xsi:type="CD" code="195967001" codeSystem="2.16.840.1.113883.6.96" codeSystemName="SNOMED CT" displayName="Emphysema">
<value xsi:type="CD" code="91143003" codeSystem="2.16.840.1.113883.6.96" codeSystemName="SNOMED CT" displayName="Albuterol" />
<originalText>
<reference value="m1"/>
</originalText>
</value>
</Observation>
</entry>
<entry>
<Observation> <code code="84100007" codeSystem="2.16.840.1.113883.6.96" codeSystemName="SNOMED CT" displayName="medication
history"/>
<value xsi:type="CD" code="32398004" codeSystem="2.16.840.1.113883.6.96" codeSystemName="SNOMED CT" displayName="Squamous cell
lung carcinoma">
</value>
</Observation>
</entry>
<entry>
……………………………………………………
</entry>
</section>
</component>
<component>
<section>
<code code="10164-2" codeSystem="2.16.840.1.113883.6.1" codeSystemName="LOINC" />
<title>History of Presen t Illness</title>
<text>65 year old gentleman with a history of Emphysema presented to our hospital
with cough since two months. He experienced severe bouts of cough since two
days and an episode of Hemoptysis yesterday evening. A CT scan image showed a
3x2.5 cm mass in the apex of right lung and was confirmed through biopsy
about the possibility of squamous cell lung carcinoma.
</text>
</section>
</component>
</StructuredBody>
</component>
</ClinicalDocument>
An Example XML Document [6]
HL7 Clinical Document Architecture
Potential Impact
Individual hospital level
Allow a more rapid review of events by quality improvement staff
Significantly decrease the time spent gathering information for nationally
reportable databases such as Core Measures and PQRI (physicians quality
reporting initiative)
National and international level
Monitor side-effects of new medicines
Assess best practices
Conduct comparative effectiveness research
Leverage knowledge from clinical trials
Superior decision support and data mining tools
Local, state, and national level health information exchange (HIE) initiatives
Limitations of current data integration systems
E.g.,National Cancer Institute’s caBIG [3], SHRINE [4], NeHii [10]
Creating a mediated schema and semantic mappings is very cumbersome in a
federated database model as the number of data sources increases.
Requires sufficient domain knowledge for each data source.
Web services based architecture requires explicit specification of data location in
the query – coarse grained selection of data sources and lacks scalability.
Collaborative Data Network
Missouri Regional Life Sciences Summit (2010), Kansas City , MO
Design Principles
A CDN benefits from the marriage of two successful technologies
Peer-to-peer computing (P2P)
Scalability, fault tolerance, load balancing, decentralized design
XML/JSON standards
Model heterogeneity of data sources, non-standardization
Rich query expressiveness (e.g., XQuery [10])
We propose a new design for sharing Electronic Health Records (EHRs) called a
Collaborative Data Network (CDN).
Data resides with the data provider – it can implement local access control policies
and protect the privacy of patients.
Location oblivious queries - a user can pose a single query across multiple data
sources.
Natural language
parser
XQuery
Generator
Clinical
terminology/
thesaurus
XQuery processor
XML/P2P based Collaborative Data
Network
Query
Publisher
Schema
Recommender
Data
XML signature
Location oblivious
queries
Query results
User feedback
Form-based query
web serviceweb service
Feedback Analyzer
Gossip
ClinicalDocument
id patient component
Observation
value value
Emphysema AlbuterolSNOMED CT
Observation
value
Squamous cell lung
carcinoma
component
section
text
65 year old gentleman with a history of Emphysema
presented to our hospital ...... about the possibility of
squamous cell lung carcinoma.
codeSystem
2.16.840.1.113883.6.1
XML/P2P based CDN
(1) Single XQuery query
(2) Find matching XML documents
and their publishers (e.g., using psiX [7,8])
(3) Return matching
documents/publishers
(4) Data or query
shipping
Chemotherapy data
(1) Single XQuery query
(3) Query results
(2) Join across multiple data sources
Radiotherapy data
HIPAA Compliance
(5) Query results
XML Document Tree
Location Oblivious Queries
By design, the actual data resides with the publisher (a.k.a. data provider).
Data protection and local access control policies can be implemented similar
to a federated design.
Data ownership can be achieved in a CDN.
Our CDN can protect the privacy of patients.
Only the publisher has the permission to modify its data.
By design, a user posing a query cannot access an XML document of a data
provider if he/she is not authorized by the data provider.
Data can be encrypted while being transferred through the network.
Cloud ComputingCloud computing can reduce the infrastructure setup and maintenance costs for
health care providers
Our proposed collaborative platform can be adapted to run in the cloud.
Data provider stores actual data locally and accesses the cloud via the
Internet.
We have a working prototype of psiX [7,8] at http://vortex.sce.umkc.edu/psix
that demonstrates proof-of-concept.
XML/P2P based CDN