OpenInfobutton
Guilherme Del Fiol, MD, PhD
Department of Biomedical Informatics
University of Utah
Background
1 question out of every
1-2 patients seen (Del Fiol, 2012)
Resources: answer
>90% of questions
“What is the dose of Pediazole for this child?”
Context
“Where is the hospital pain management protocol”
“I need a diabetes patient education handout”
i
“What is this patient’s most recent HbA1c?”
EHR
> 60% of questions
left unanswered
Top 10 question Types
What is the drug of choice for condition x?
80/20
What is the cause of symptom
x?rule
How should I treat condition x?
What test is indicated in situation x?
What is the dose of drug x?
What is the cause of physical finding x?
Can drug x cause (adverse) finding y?
What is the cause of test finding x?
30% Iabout
drugcondition x?
How should
manage
Could this treatment
patient have condition x?
Source: Ely et al. , BMJ 2000; 321:429-32.
11%
9%
8%
7%
6%
6%
5%
4%
3%
2%
Topics
Resources
Context
•
•
•
•
Patient
Provider
Setting
EHR task
Impact of Infobuttons
• Answers to over 85% of questions
• Decision enhancement or learning in over 62%
of infobutton sessions
• Median session time: 35 seconds
• High user satisfaction
• Usage uptake in medications and lab results
•
1 million sessions at Partners Healthcare in 2011
Maviglia et al. J Am Med Inf Assoc, 2006.
Cimino JJ. J Am Med Inf Assoc. 2009.
Del Fiol et al. J Am Med Inf Assoc, 2008.
Architecture
Azithromycin
Female
81 years
User: MD
Order entry
EHR
i
Resource 1
HL7
HL7
HL7
Resource 2
Infobutton
Manager
HL7
HTML
Resource 3
OpenInfobutton
• VA Innovations Project (2010)
– Portland VAMC
– David Douglas, Clayton Curtis (PIs)
• Open source infobutton manager
–
–
–
–
VA system wide infobutton manager
HL7 reference implementation
Foster adoption & innovation
Integrated with 34 knowledge resources
(8 HL7 compliant)
http://www.OpenInfobutton.org
HL7 Context Dimensions
Patient
User
•
•
•
•
•
•
Gender
Age group
Concept of interest
Secondary observations
Patient vs. provider
Discipline / specialty
Goal
Organizational




Task (e.g., order entry, problem list
entry, lab results review)
Care setting
Service delivery location
Location of interest
•
•
•
EHR
Discharge
handout
•
•
•
•
•
•
VA [representedOrganization.id.root=1.3.6.1.4.1.3768]
Problem list review [taskContext.c.c=PROBLISTREV]
Diabetes Mellitus Type 2 [mainSearchCriteria.v.c=250.00&
mainSearchCriteria.v.cs=2.16.840.1.113883.6.103&
mainSearchCriteria.v.ot=Diabetes+Mellitus+type+2]
Female [patientPerson.administrativeGenderCode.c=F]
65 years old [age.v.v=65&age.v.u=a]
ED [encounter.c.c=EMER]
Health care provider [performer=PROV]
Medline Plus
Patient recipient [informationRecipient=PAT]
Patient Spanish speaker [informationRecipient.languageCode.c=es]
Open
Infobutton
Mayo
VA content
Medline Plus
Discharge
handout
Open
Infobutton
<feed>
<title type="text">MedlinePlus</title>
EHR
<subtitle type="text">type 2 Diabetes Mellitus </subtitle>
<updated>2013-03-26T12:13:12.013-06:00</updated>
Mayo
<category scheme="mainSearchCriteria.v.c" term="250.00"/>
<category scheme="mainSearchCriteria.v.cs" term="2.16.840.1.113883.6.103"/>
<category scheme="mainSearchCriteria.v.dn" term="type 2 Diabetes Mellitus "/>
<category scheme="informationRecipient" term="PAT"/>
<entry>
<title>Diabetes Type 2</title>
<link href="http://www.nlm.nih.gov/medlineplus/diabetestype2.html" rel="alternate"/>
VA content
<summary type="html">&lt;p&gt;&lt;a
href="http://www.nlm.nih.gov/medlineplus/diabetes.html"&gt;Diabetes&lt;/a&gt; means your blood glucose,
or blood sugar, is too high. With type 2 diabetes, the more common type, your body does... </summary>
</entry>
</feed>
Discharge
handout
<aggregateKnowledgeResponse>
<feed>
<title type="text">VA patient education</title>
…
</feed>
<feed>
<title type="text">MayoClinic</title>
<subtitle type="text">type 2 Diabetes Mellitus </subtitle>
Medline Plus
<updated>2013-03-26T12:13:12.013-06:00</updated>
<category scheme="taskContext.c.c" term="PROBLISTREV"/>
<category scheme="mainSearchCriteria.v.c" term="250.00"/>
Open
<category scheme="mainSearchCriteria.v.cs"
term="2.16.840.1.113883.6.103"/>
<category scheme="mainSearchCriteria.v.dn" term="type 2 Diabetes Mellitus "/>
Infobutton
<entry>
EHR
<title type="text">Patient education</title>
Mayo
<link
href="http://www.google.com/search?hl=en&amp;btnI=Im+Feeling+Lucky&amp;q=type
2 Diabetes Mellitus +site:www.mayoclinic.com/health/&amp;"/>
</entry>
</feed>
<feed>
<title type="text">MedlinePlus</title>
VA content
…
</feed>
</aggregateKnowledgeResponse>
Software Environment
• Deployment environment
– Any Java application server (tested with
Tomcat)
– Any SQL DB (tested with MySQL)
– UMLS license for terminology access
• Development
– Source code in public GitHub repository
– Eclipse + plug-ins
Ongoing Collaborations
• VA
– Live within HMP (pilot at San Diego VA)
• University of Utah
– Integration with Epic
• Intermountain Healthcare
– Live within problem list
• Duke University
– Medical library search page
• University of Washington
– Pharmacogenomics CDS – feasibility (PhD project)
• New York University
• Nortwestern University
– eMERGE network
Under development
• OpenInfobutton
– Compliance with latest versions of HL7
specification
– Implementations with other terminology
servers (3M Health Data Dictionary)
• Infobutton responder
– Allow local content to be indexed and
searched through HL7 compliant API
Resources
• Project site
– www.openinfobutton.org
• OSEHRA
– http://code.osehra.org/journal/browse/publica
tion/33#
• Source repository
– https://github.com/VHAINNOVATIONS/Innov
ation-182/
Acknowledgements
• Funding source
– VHA CHIO Innovation Project #182
– AHRQ grant 7K01HS018352-03
• HL7 Clinical Decision Support WG
• Portland VAMC
– David Douglas (PI), Clayton Curtis (PI), Saad Aziz
• University of Utah
– Andrew Iskander, Aditya Kalluri
• NIH Clinical Center & NLM
– Jim Cimino, Xia Jing