Semantic Web Application for PCAST Health IT Recommendation: Streamlining EHR Metadata Entry

noiseboliviaSecurity

Nov 5, 2013 (4 years and 8 days ago)

65 views



1

Semantic Web Application for PCAST Health IT Recommendation
:
Streamlining EHR Metadata Entry

Eric Rozell
*

Tetherless World Constellation, Rensselaer Polytechnic Institute, Troy, New
York
,
USA

AST_2011_Rozell_Eric_A2


ABSTRACT

The purpose of this assignment is
to
propose an application
of Semantic Web technology to one of the recommendations
of the PCAST Health IT recommendation.

1

INTRODUCTION

The President’s Council of Advisors on Science and Tec
h-
nology
(PCAST)
recently assembl
ed a report containing a
set of recommendations for moving forward on an infr
a-
structure for health information technology (Health IT) and
its potential to improve healthcare

[1]
. The report discussed
potential benefits of such an infrastructure, including

more
informed, and therefore effective, physicians and popul
a-
tion
-
scale healthcare research. They also provided reco
m-
mendations on the technology that are critical to the Health
IT infrastructure. Their focus was on the application of an
extensible mark
up language, such as XML, and a metadata
vocabulary standard for the interchange of electronic health
records (EHR
s
),
and ways of implementing the necessary
security infrastructure to protect patients.



In the second section of the PCAST report, the cou
ncil
presents the current status of Health IT, and the limitations
of the present
-
day impl
ementations of EHRs. On page 26
,
the first limitation listed is Workflow Disruption and Do
c-
umentation Burden. The main idea behind this limitation is
that physician
s often complain about the added burden of
EHRs, as they must spend some of their valuable time ty
p-
ing in orders and notes from patient visits. Although
staff
members can complete data entry, this increased burden for
the staff in small practices may not
be economically feas
i-
ble. The council states that streamlined data entry through
checklists and templates is an area of opportunity in sof
t-
ware development.


By combining the recommendation that an extensible
markup language is used for the exchange of

EHRs, and the
functional requirement that metadata entry be streamlined
and enabled through checklists and templates, we have a
clear application area for Semantic Web technologies. In
many cases, the metadata that must be attached to an EHR


*

To whom correspondence should be addressed.

based on
reg
ular office visit, or for a common diagnosis,
such as a cold or strep throat (streptococcal pharyngitis) is
the same for each patient. If the doctor used a semantic
knowledge base based on the Resource Description Fram
e-
work (RDF) [2] for patient data, we
could apply a descri
p-
tion logic framework, such as provided by the Web Ontol
o-
gy Language (OWL) [3], to infer much of the metadata that
must be attached to an EHR.


OWL is a vocabulary for the Semantic Web that supports
the description of classes of enti
ties inherent in Web r
e-
sources and the relationships between t
hem. OWL DL is a
flavor of OWL that places restrictions on the language, e
n-
suring the computability of inference entailments by reaso
n-
ing engines. OWL DL has reasoning properties that would
sa
tisfy the functional requirement of streamlined metadata
entry. In particular, the language supports the definition of
classes that have specific property requirements
, called va
l-
ue restrictions
. For instance, in OWL DL, a class called
RedBarn

can be cre
ated with
two subclass constraints:

1)

an instance of
RedBarn

is an instance of
Barn

2)

an instance of
RedBarn

has triple

hasColor red

Using these two constraints, we can make two kinds of i
n-
ferences using existing open source tools (see Pellet, [4]). If
we hav
e an instance of
Barn

that has the triple
hasColor
red
, then we can infer that the instance is in the class
RedBarn
. Also, if we have an instance of
RedBarn
, we can
make two inferences, that the instance is in the class
Barn
,
and we should have the triple

hasColor RedBarn
.


A doctor could establish classes that group diagnoses, or
information entry requirements, with repetitive EHR
metadata attachments. In the strep throat diagnosis exa
m-
ple, there could be a class
StrepThroatDiagnosis
with the
followin
g subclass constraints:

1)

an instance of
StrepThroatDiagnosis

has triple
has
Infection

StreptococcalPharyngitis

2)

an instance of StrepThroatDiagnosis has triple
prescribed Amoxicillin

So now, the doctor simply needs to check the box for the
StrepThroatDiagnosis when writing up a patient visit, rather
than inputting the individual metadata elements. One of the
benefits of OWL DL is that the ontology is easily extended
Rozell, E.


AST_2011_
Rozell_Eric_A2




2

and shared because

it is based on RDF, so communities of
doctors can create and share the common diagnoses they
have established.


In summary, the PCAST
report has recommended an e
x-
tensible markup language for the exchange of EHRs. If the
eventual infrastructure is ba
sed on RDF, then the constru
c-
tion of systems that utilize OWL DL inferencing will be
possible. With OWL DL inferencing, we can easily meet
the functional requirement proposed in the report to stera
m-
line metadata entry. Even if the infrastructure is based

on
XML, client systems that transform the XML metadata into
RDF and visa versa can be created to support reasoning with
Semantic Web technologies.

REFERENCES

[1] President’s

Council of Advisors on Science and Technology,

Report to
the President Realizing the Full Potential of Health Information
T
ec
h-
nology to

I
mprove

H
ealthcare for

A
mericans: The Path

F
orward”.

[2]
Graham Klyne, Jeremy Caroll, Brian McBride, “Resource Description
Framework (RDF): Concepts and Abstract Synta
x,” W3C Recomme
n-
dation, 10 February 2004.

[3] Michael Smith, Chris Welty, Deborah McGuinness, “OWL Web Onto
l-
ogy Language Guide,” W3C Recommendation, 10 February 2004.

[4] “Pellet OWL Reasoner,”
http://www.mindswap.org/2003/pellet/
.