OpenMRS Concept Management Tutorial presented on 9 Dec 2015 at the OpenMRS Worldwide Summit in Singapore. Presented by Andy Kanter and Ellen Ball. 4 hour presentation.
1. Andrew S. Kanter, MD MPH FACMIa,b,c
a Intelligent Medical Objects, Inc., Chicago, USA
b Department of Biomedical Informatics, Columbia University, New York, NY, USA
c Department of Epidemiology, Mailman School of Public Health, Columbia University,
New York, NY, USA
Ellen Ball
Partners In Health, Boston, USA
OpenMRS Concept Management
OpenMRS Worldwide Summit
9 December 2015
Singapore
2. Topics
• Terminology 101
• OpenMRS data model and concepts
• Controlled terminology and reference mappings
• Management of concept dictionary
• Usage on forms and reports
• Future
3. Introduction and Disclosure
• Andy (ask2164@cumc.columbia.edu)
• OpenMRS Leadership (Terminology and Meta Data Lead)
• Direct Columbia International eHealth Lab
• Department of Biomedical Informatics
• Department of Epidemiology/MSPH
• Board Member/Director of Clinical Integration for
Intelligent Medical Objects (IMO)
•Ellen (eball@pih.org)
•Implemented OpenMRS at Partners In Health Haiti,
Rwanda, Lesotho, Malawi, Peru, Liberia, and Sierra Leone
5. Why vocabulary matters…
● Clinical users of EHRs resist the constraints of structured
documentation
● Users and administrators underestimate the complexity and
difficulty of data mining
● Data is dirty, misplaced, and/or incomplete
● Humans think conceptually, systems store data literally
● Everything we want to do depends on how meaning is
recorded in the information system. Clinical intent is
paramount and you get one chance to capture it correctly!
9. Terminology about Terminology (cont)
● Domains
● Granularity—broader vs. more specific
● Pre-coordination
● Post-coordination
10. Concepts
● The actual meaning is a phrase or even a paragraph.
● Developed at the right level for the user
● Severe right knee pain
● Liver dysfunction
● Can have many different descriptions but all have the same
meaning
● Assigned a non-sensical numeric identifier
● Meaning often developed through relationships to other concepts
● One description often flagged as the default name
11. Concepts
● Goal: default description (fully specified name) sufficient
to understand the concept
● Unambiguously defined
● Have one domain
● Can provide more semantics around concept than default
description
● Fully specified name includes appended domain, e.g., cough
(finding) vs. cough (symptom)
12. Descriptions
● A collection of text strings or terms
● perennial allergic rhinitis
● seasonal allergies (hay fever)
● allergic rhinitis, seasonal
● hay fever
● perennial rhinitis
● perennial allergies
● …
13. Descriptions
● May need context for full understanding
● Fever
● Patient reported they felt feverish
● Patient reported they took their temperature with
thermometer
● Healthcare provider took temperature and was elevated to…
● Acronym - Careful— ARV = “Anti-rabies vaccination” or
“Antiretroviral”?
● Pragmatics
● Brain tumor
● malignant neoplasm of brain/Neoplasm of brain/Brain mass
● Breast CA
● Breast cancer / Breast carcinoma
14. Description attributes
● Unique code
● Audience
● MD, ancillary health, patients
● Length (cell phone, etc.)
● Search friendly (word order important)
● Display to user vs. recognize as mapped to concept
● Locale, language, country, etc.
15. Case style
● Right case**
● CHF (congestive heart failure)
● Sentence case*
● Spine fracture
● Title case
● Spine Fracture
● Upper case
● SPINE FRACTURE
16. Words
● Definition
● Not obvious
● Alphanumerics separated by non-alphanumerics
● What about apostrophes like Alzheimer’s or peau d’orange?
● Words ensure consistency with searching
● Not every concept will have a description with all
misspellings or word variants
● Hepatic failure vs. liver failure
17. Relationships and Mappings
● One of the defining features of an ontology, i.e.,
relationships between concepts
● Drawing the lines between concepts or between
concepts and codes
● Relationship types
● Can be more complex than parent-child (Is-A)
● “Severe anemia” is narrower-than Anemia
● Other examples, has-location, has-severity, has-laterality
18. User interface terminology
(descriptions)
AMI (alternate term)
Myocardial infarction, acute (entry
term)
Acute MI (alternate term)
Acute myocardial infarction (preferred)
Reference terminology
Acute ischemic heart
disease
Ischemic heart disease
Structural disorder of the heart
myocardial disease
heart disease
disease of cardiovascular system
Myocardial infarction
Mycardial necrosis
Concepts
Acute myocardial infarction
Words
Heart, cardiac, myocardium, myocardial, infarction, CV, attack, AMI, acute, …
19. Mappings (type of relationship)
● One or more external codes mapped to each
concept
● ICD10 code B54.9
● SNOMED code 2423424211
● UMLS code C0018621
● Need relationship type
● Is it broader than, narrower than, same as…?
● Important for inference
20. Mappings and Inference
● Malaria
● Same as SNOMED CT 61462000 (Malaria)
● Same as ICD-10 B54 (Unspecified malaria)
● Severe malaria
● Narrower than SNOMED CT 61462000 (Malaria)
● In both eyes
● Narrower than SNOMED CT 54485002 (ophthalmic
use)
22. Administrative terminology
● Used primarily for classification
● Major examples include:
● ICD (ICD-10-WHO, ICD-10-CM, etc.)
● CPT®
● Not particularly good for capturing clinical data
● Often used for billing and reimbursement and some
reporting
23. Administrative terminology
● ICD-10-CM is now mandated for use in the US as of 10/15
● Differences between ICD-9-CM, ICD-10 and ICD-10-CM
● 13,000 ICD-9-CM to 68,000+ for ICD-10-CM
● 3-5 digits for ICD-9 compared to 3-7 for ICD-10
● ICD-9 had only a few alpha codes, all ICD-10 codes start with
a letter
● Combination codes for conditions and common symptoms
or manifestations and for poisonings and external causes
● Added laterality
26. Interface terminology
● List of terms or phrases
● Supports clinician entry into electronic systems
● Multiple descriptions may mean the same “concept”
● May have unique identifiers
● Major examples include:
● IMO Problem (IT), Procedure (IT)
● Vanderbilt Terminology
27. Groups
● Used for providing a list for user selection
● Used for providing Allergen class-ingredients
● Can be published value set for quality reporting
● Extensional value sets used for meaningful use
● Asthma, active diagnosis with set list of ICD or SNOMED CT codes
● Can be programmatic for decision support
● Intensional value set based on logic such as
● All children of SNOMED code xxxxx
● Includes with children A, B, C but excludes D
28. Pre-coordination
● More user friendly
● Examples
● Acinar cell carcinoma of the pancreas
● Severe right knee pain
● Recurrent intravascular papillary endothelial hyperplasia of
the right middle finger
● Recurrent intravascular papillary endothelial hyperplasia of
the right ring finger…….
● Combinatorial explosion
30. Examples
Pre-coordination Post-coordination
Acinar cell carcinoma of the
pancreas
carcinoma of pancreas + acinar
cell carcinoma
Severe right knee pain knee pain + right + severe
Recurrent intravascular papillary
endothelial hyperplasia of the
right middle finger
intravascular papillary endothelial
hyperplasia + middle finger
structure + right
31. Terminology Process
1. Core terminology content development including mapping to standards
(code mapping)
2. Specialized domain content development (including subsetting of
content, expansion of content, etc.)
3. Mapping of user requirements to specific concepts (field mapping)
4. Deployment of content within the software platform (including
searching within forms, data capture tools, etc.)
5. Meta-data modeling and information modeling including schema design
6. Ontolological work including building of aggregate indicators and
measures (including maps to standard quality measures, etc.)
7. Reporting/Analysis using common algorithms, formulae and concepts
8. Transactional translation or tagging for on-the-fly encoding of concepts
including natural language processing
34. OpenMRS concept dictionary
•A collection of concepts
•CIEL, PIH, Kenya, etc.
•Forks, subsets, and supersets
•Local or central management
35. Concept creation workflow
Paper form,
list of data
fields, or
indicators
Concept
analysis in
existing forms
Propose new
concept in CIEL
or use existing
concept
Add language,
description,
synonyms, and
mappings
51. Why not just use ICD-10 or SNOMED?
• Admin/Reference terms change which
require changing reports and forms
• Clinicians don’t use terms like
• Other disease of blood & blood-forming organ
• SNOMED is post-coordinated
• Hard to say fracture of RIGHT arm
52. So why should OpenMRS share concepts?
• Interoperability of data between
applications and between organizations
• Ability to share forms, data collection
tools
• Ability to share reports
• Ability to share decision support rules
54. Leveraging Maps for Reporting
• There are multiple CIEL concepts
mapped to the same ICD or SNOMED
code
• Use Reference_Reference_Map to build
subsumption queries
• CIEL/OCL to add map for particular
value sets
57. Concept management scenarios
Standalone
All concepts
managed locally
PIH Malawi
Master/Slave
Concepts maintained
on central server
CIEL with subscription
PIH Haiti with mds
PIH Rwanda with sync
Central Curation
Open Concept Lab
(OCL)
58. CIEL Concept Dictionary
• Contains most diseases, procedures and
medications (>49,000 concepts)
• Mapped to SNOMED CT, ICD-10, 3BT,
RxNorm, LOINC and CVX codes.
• Several Languages:
SNOMED CT 49,514
ICD-10-WHO 40,015
RxNORM 5,599
LOINC 390
3BT 7,703
68,275 en 4001 vi 62 bn 30 rw
32,630 es 2,737 fr 51 ru 29 ht
11,760 nl 242 sw 51 ti 13 am 7 om
77. Open Concept Lab- Jonathan Payne
• Beta customer is Kenya EMR
• Working with Kenyan Community and
ITECH
• 9 months behind schedule
• Focusing on API then UI
• Initial Beta testing complete
78. Open
MRS
OpenMRSSubscription
Subscription Process
• Create OCL user to get an OCL API token
• Install OCL Subscription Module in your OpenMRS instance and
configure to subscribe to a specific source
• On first synchronization, pulls entire dictionary
• On subsequent synchronization, pulls latest changes only (e.g. new
concepts, updates, deletes, retires)
• Does NOT overwrite local concepts or concept metadata (based on
concept and concept metadata UUIDs)
Open
Concept
Lab
OCL API
OCL Subscription
Module
83. OpenHIE and
Terminology
Management
Terminology
Management
Service
2
1
2
1
• OCL as source of content
for the TS.
• Requires local TS.
• Appropriate for high-
volume, real-time
transactions (e.g. code
validation, lookups,
transformations, etc.).
• OCL provides canonical
source(s) to HIE,
subscription service, &
collaborative
management tool.
• NOT for real-time, high-
volume transactions.
• Alleviates need for local
service.
84. Terminology Sustainability
• Looking for additional community
leadership (Judy, Jonathan, etc.)
• Basic support and funding from Columbia
is running out
• Looking for sustaining support ($150K/y)
• Partnering with OCL/IMO
85. Proposed OCL Sustainability Model
FREE BASIC PREMIUM ENTERPRISE
Target • Existing CIEL User-base • Researchers,
harmonization,
terminology geeks
• Dictionary managers,
e.g. AMPATH, PIH, CIEL
• Governments or
institutions managing
terminology as a core
service; require
guaranteed level of
service
Features • Access to all OCL
functionality for CIEL
dictionary only
• Limits on the number of
subsets you can
create/manage
• OpenMRS Subscription
to CIEL dictionary
• Includes access to CIEL
community content
• Limited API access
• Access to major
terminology sources in
addition to CIEL (ICD-10,
LOINC, SNOMED, etc.)
• No limit on collections
• Ability to propose
content for curation in
one of the “managed”
dictionaries (i.e. CIEL)
• Create your own
sources
• Full API Access
• Guaranteed level of
service for terminology
curation
• Assistance importing
local/proprietary
terminology sources
• Configuration of
organizational workspace
• Additional training and
services available
Initial User
Base
• OpenMRS + CIEL
Subscriptions: >100
• MCL: 16k
lookups/searchers; 2k
unique visitors in last
year
• THRIVE/WHO • Partners In Health • Kenya Ministry of Health
86. OCL Roadmap
2015 Q3
• OCL Launched with Kenya MOH!
• Basic functionality complete:
–Full-text search
–Create users and organizations
–Build your own sources and create/edit concepts and
mappings
–Export of sources using AWS
• CIEL dictionary imported
• All functionality implemented through APIs
• OpenMRS subscription to a single source (e.g. CIEL
dictionary)
2015 Q4
• Begin implementing sustainability model and signing up paid
clients
• Optimized search (e.g. better weighting of search terms to
improve likelihood of finding the correct result)
• Full support for creating and managing collections (i.e.
references to concepts from other sources)
• Import WHO ICD-10 source
• CIEL transition to managing dictionary on OCL instead of in
OpenMRS
• Secured access to OCL website and API (e.g. https encryption)
• Stability and performance improvements (esp. imports, exports)
Potential Future Features
• FHIR API compatibility
• Import additional sources, including SNOMED CT, LOINC
• RSS feeds of changes to sources, collections, and concepts
• Social functionality
• Improved organization management - better control of access to content for members of an organization
• Ability for users to "star" sources, collections, and concepts
• Collection/source comparisons
• Ability for users to "follow" organizations or other users