• Share
  • Email
  • Embed
  • Like
  • Save
  • Private Content
Using a Medical Data Dictionary for Semantic Interoperability

Using a Medical Data Dictionary for Semantic Interoperability



Visit 3Mhis.com/MeaningfulData for more information about semantic interoperability. An integrated, structured medical terminology system can mediate the differences between vocabularies and codes ...

Visit 3Mhis.com/MeaningfulData for more information about semantic interoperability. An integrated, structured medical terminology system can mediate the differences between vocabularies and codes sets within a clinical information system.



Total Views
Views on SlideShare
Embed Views



0 Embeds 0

No embeds


Upload Details

Uploaded via as Adobe PDF

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
Post Comment
Edit your comment

    Using a Medical Data Dictionary for Semantic Interoperability Using a Medical Data Dictionary for Semantic Interoperability Document Transcript

    • Using a Medical Data Dictionary to Comply withVocabulary Standards and Exchange Clinical Data The Interim Final Rule On December 30, 2009, the Department of Health and Human Services (HHS), throughWhat is needed is an its Office of the National Coordinator for Health Information Technology (ONC), issued an interim final rule (IFR) providing details on an initial set of standards,integrated, structured implementation specifications, and certification criteria for electronic health record (EHR) technology. The IFR lists the standards and criteria that EHR systems must meetterminology system in order to be considered “certified.” Under the Health Information Technology for Economic and Clinical Health (HITECH) Act, in order for professionals and hospitalsthat can mediate the to qualify for certain incentive payments, they must demonstrate meaningful use of such certified technology.differences between The certification criteria include functionality that complies with specific standards adopted by HHS. These standards fall into the following four categories:vocabularies and • Vocabulary standardscode sets and easily • Content exchange standards • Transport standardsbe incorporated into • Privacy and security standards Vocabulary standards can be particularly challenging to effectively and efficientlyexisting clinical implement across a healthcare organization or EHR system.information systems. A Practical Approach to Implementing Vocabulary Standards Is there a “right” medical vocabulary that can describe, access, and make sense of all clinical data? Unfortunately, no single healthcare vocabulary or terminology can meet all the needs of all the people who use healthcare information. Each terminology in use today has been designed for different purposes by different healthcare constituencies. ONC recognizes this, which is why it has adopted different vocabulary standards for different certification criteria. ONC has adopted an initial set of vocabulary standards to support the proposed requirements for Stage 1 of meaningful use, which begins in 2011.
    • The vocabulary standards that ONC formats and terminologies from diverse existing resources. However, the resultinghas adopted and against which EHRs systems and sources. When applied product supports standardization acrossmust be tested for certification include in real-time to an organization’s enterprises and enables clinical data tothe following: information system infrastructure, such be exchanged, compared, and aggregated a dictionary can: at local, regional, national, and even • ICD-9-CM international levels. • SNOMED CT® • Describe clinical data in all its possible forms, providing a road ONC’s adoption of the initial set of • Code sets by RxNorm data source map to the content and structure of standards is one of the first steps in its providers the patient database “iterative approach to enhancing the • CPT-4® interoperability, functionality, utility, and • Support encoding of clinical data to • LOINC® remove ambiguities security” of health information technology (HIT) [IFR p. 30]. ONC understands • Vaccines administered (CVX) • Support exchange and comparison that the current state of communicationUnfortunately, ONC has not adopted of data between independent between HIT systems is fraught withindustry-tested implementation computer systems uncertainty and ambiguity, which canspecifications for these vocabulary • Provide structure and content compromise health care quality, efficiency,standards. Eligible professionals and for decision support across and safety.hospitals, as well as systems vendors, care encountersmay be left to their own devices to • Enable users to effectively querysolve the problem of implementing the and report on the databasedifferent vocabulary standards practically, • Support standardization ofefficiently, and effectively. clinical data across enterprises byOver three decades ago, the concept of incorporating industry-standarda medical data dictionary began as a controlled medical vocabulariesmeans of supporting clinical information (CMVs) as its foundationsystems and the computerized patient The key to accomplishing these tasks isrecord. A well-designed, intelligent the incorporation of de facto CMVs (e.g.,medical data dictionary with terminology LOINC®) into the medical data dictionaryservices can enable a healthcare itself. Adding and “mapping” (cross-organization or EHR system to comply referencing) industry-standard CMVs to awith vocabulary standards and exchange dictionary are labor-intensive efforts thatclinical data, received in all of its various many organizations cannot manage withCPT is a registered trademark of the American Medical Association.2 3M Health Information Systems
    • What’s Not Making Sense in Clinical Data Today? Computers easily process enormous “reconcile and reunite” all of these volumes of data, but computers also different test codes and results in the “expect” unambiguous data in a specific patient’s record? How can clinicians form. Using computerized clinical data “see” all of Doe’s glucose results means more than simply storing numbers together for effective comparison and and words. Terms must be clearly defined comprehensive analysis? and placed in a context. How can a healthcare organization make What does the word “cold” mean? well-informed decisions regarding futureThe vocabulary must be • An accident victim brought into the expansion and resource deployment? ER tells the attending physician, “I As part of the strategic planning process,carefully monitored to feel cold.” a healthcare organization’s team of risk managers, care providers, and financialavoid duplication, • A pulmonologist tells a 58-year-old male patient that he is suffering officers wants to examine population from COLD (Chronic Obstructive studies by facility to determine how bestsupport synonyms, and Lung Disease). to deploy an expensive special service like cardiology. How can the team get ancompletely describe • You call your family practitioner for an appointment and tell the accurate report of all cardiology-specific services and treatments delivered across receptionist, “I have a bad coldterms from all areas of that’s not getting better.” the organization? In all of these situations, people aremedicine—these If the word “cold” is recorded on a patient’s chart during any of these three defining the particular term, code, or information query according to a particularcharacteristics lay at the scenarios, how does that word become an accurate and meaningful part of the EHR? context. In the first example, “cold” can be a sensory perception, a pulmonary How can the term be translated so thatheart of a medical data outcomes can be understood and managed diagnosis, or an upper respiratory viral infection. The human mind easily resolves and the medical events themselvesdictionary. accurately described and interpreted? the ambiguity of a word like “cold,” especially when the term is considered in What is the “correct code” for a its context. But how can a computer track glucose result? and correctly interpret “cold,” especially when it can appear in various forms and An organization currently receives many different contexts? laboratory test results from two different external vendors, a government lab, and Bottom line: The vocabulary must be several internal STAT labs. Each external carefully monitored to avoid duplication, system uses its own distinct set of test support synonyms, and completely result codes, which differ from the internal describe terms from all areas of medicine; set of codes used by the organization’s these characteristics lay at the heart of a STAT labs. Patient John Doe has glucose medical data dictionary. test results from one or more of these labs—how can the organization’s order entry and results review applications www.3Mhis.com 3
    • What Is a Medical Data Dictionary?A medical data dictionary is a database that describes the organization and logical structure of the medical data found in a clinicaldatabase. It contains “metadata”—or “data about data”—that describes the content, structure, and relationships between clinical data.In short, a medical data dictionary “translates,” precisely defines, and effectively accesses the contents of the EHR.What Information Does Today’s EHR Contain?A look at the information stored in today’s • Hospital information systems • Controlled medical vocabularytypical EHR uncovers the Tower of Babel • Medical record coding (CMV) formats (a typical patientthat exists in healthcare: EHR contains ICD-9-CM codes, • Billing CPT® codes, LOINC® codes, etc.)Data comes from and resides in many Data from these diverse sources exists in • Proprietary formats from “home-different systems and databases: many different formats: grown” and commercial laboratory, • Laboratory • Coding and classification hospital information, billing, and • Radiology system formats pharmacy systems and databases • PharmacyWho Is Trying to Use the Information in the EHR?The data in the EHR is valuable and needs regardless of data format or terminology A well-designed and intelligent medicalto be available to many different people, (e.g., “MI,” “myocardial infarction,” data dictionary defines the connectionsranging from clinicians and administrators ICD code 410, etc.). By the same token, between “MI,” “myocardial infarction,”to researchers and government regulatory users should be able to retrieve all those and ICD code 410 within an EHR, andagencies. The data should be retrieved and cases—and even the associated patient also makes it possible to retrieve andreturned to these people in the appropriate records—for further study in their chosen return the term most appropriate for aformat and with the correct degree of data format. given audience.granularity for the audience. Data needs tobe accessed in various ways and for manydifferent purposes: patient clinical reports,statistical studies, ad hoc reporting,regulatory requirements, etc.Not only must all of this data in the EHRbe integrated, it must also be “normalized”into a form that can easily be shared byall audiences. For example, if users arecompiling organization-wide reports onacute myocardial infarctions, they wantquick access to all cases in the database,4 3M Health Information Systems
    • Other “Language” in the EHR: Interface Terminologies (Organization- andOrganizations can User-specific Terms)send messages to one A harsh fact of life for today’s typical Organizations can send messages to one healthcare organization is that it has another only to find that because theanother only to find that invested heavily in legacy information content of the message is not standardized, systems. Such information systems they still are not successfully “talking”because the content of transmit data among themselves by means of interfaces, and each interface system to one another. This is the reason that HL7 version 3 is focusing on vocabularythe message is not has its own “language” or terminology. Many legacy systems use their own standardization in its development and also why ONC considers vocabulary proprietary codes to describe clinical data, standards so important.standardized, they and these codes are not “understood” by any other system. The number and Site-specific Terminologiesstill are not successfully types of interfaces and proprietary coding The people within a healthcare systems are unique and specific to each organization will continue using their“talking” to one another. healthcare organization. accustomed “languages” and vocabularies. If they are familiar with a particular legacyThis is the reason that The limits of Health Level 7 (HL7) system and its idiosyncrasies, they will continue to use that system. If site-specific Interfaces do not evaluate the content ofHL7 version 3 is the messages they transmit. The healthcare pharmacy formularies are in place, those will continue to be used in addition to any industry-standard interface protocol, HL7, other globally standard formulary. All offocusing on vocabulary has standardized the message structure these “interface terminologies” and site- (syntax) of healthcare information specific entities are legitimate componentsstandardization in its transmissions. Thus, every organization of a healthcare organization’s “working that is HL7-compliant can send messages vocabulary” and should be a part of andevelopment and also to other HL7-compliant organizations; as long as the message syntax is correct, HL7 intelligent medical data dictionary.why ONC considers considers the transmission successful. However, in HL7 version 2, while the syntax may be correct, the actual contentvocabulary standards of the transmission can be gibberish.so important. www.3Mhis.com 5
    • How Does a Medical Data Dictionary Benefit Healthcare Organizations?The power of the medical data dictionary Rather than replace legacy systems, the inaccurate, or imprecise data to becomeis its ability to encompass and create the organization can use the dictionary with part of the patient’s record. For example,links between industry standard CMVs, terminology services to translate between to facilitate an automated pharmacycoding sets, and organization-specific such systems and normalize the data they ordering system, a medical data dictionaryvocabularies. By integrating both global send into the EHR. can and should include a standardizedand site-specific vocabularies, the medical pharmacy CMV (such as First Data Bank)data dictionary allows data in an EHR Accommodating future adoption that includes an accurate list of Nationalto be cross-referenced to standards of vocabulary standards. A well- Drug Codes (NDCs). When a dictionaryeveryone can understand, rather than designed medical data dictionary can can “enforce” accurate drug codes andlocking the data up into yet another seamlessly integrate classification systems positively identify the drug(s) beingproprietary language that only a such as ICD or CPT® along with other ordered, human errors and mistakes canselect few can speak. When a dictionary standard vocabularies and classification be reduced.can seamlessly translate universal and systems. Whatever HHS’s future adoptedorganization-specific data “behind the standards turn out to be, they can be Facilitating interoperability.scenes,” the healthcare organization enveloped in a medical data dictionary as The challenge here is for an organizationcan save time, money, and resources by additional CMVs and consequently to achieve comparable data with itsavoiding the retooling or replacement of cross-referenced to the other CMVs still business partners and move towardlegacy systems and the re-training of in use within a healthcare organization or standardized code sets. A dictionary thatits people. vendor system. adheres to sound vocabulary principles can identify a concept even when it isProtecting an organization’s Increasing quality. Two of the identified by numerous codes; these codes touchstones of an intelligent medical may even represent both past and presentinvestment. When a dictionary has data dictionary are its insistence on usage. For example, an ICD-9-CM codebeen designed to be both extensible unambiguous data and its use of that is no longer acceptable to a claimsand flexible, it can encompass all of industry-standard CMVs to help reduce payer must still be part of the dictionary tothe diverse “languages” spoken by an the opportunities for misinterpreted, support historical data.organization and its legacy systems.An Industry-tested Implementation ToolThe 3M™ Healthcare Data Dictionary to create a very practical implementation is operational in the “real world,” having(3M HDD) was created over 15 years ago, of informatics principles and industry- been mapped in multiple commercialand it has been continuously expanded and standard vocabularies. Supporting such healthcare enterprises and all of themaintained throughout that time. It is the applications as clinical information hospitals and clinics supported by moreresult of combining industry experience exchange, data warehousing, order than one hundred Military Treatmentand professional and academic expertise entry, and results review, the 3M HDD Facilities (MTFs) worldwide in the CPT is a registered trademark of the American Medical Association.6 3M Health Information Systems
    • Department of Defense’s AHLTA project. The strengths of 3M’s dictionary lie in its:3M’s strategy has been to partner with • Structure and depth/breadth • Ability to map “local extensions”healthcare organizations to help them of content • Clinical foundations/expertisemake optimal use of their patient data,with the goal of improving quality of care, • Architectureoutcomes, costs, and competitiveness.3M Healthcare Data Dictionary’s Structure: Depth and Breadth of ContentThe 3M HDD comprises an information the vocabulary concepts should be used that clinical observations can be attributedmodel, vocabulary, and knowledge base. and how data can be combined to create to the correct patient, clinical observer, meaningful database records that represent and time sequence, and describes theInformation model clinical events. It can be thought of appropriate information domains andThe 3M HDD uses an information model as a set of “grammar rules” that show types of values that should be presentto accurately represent clinical data in how data interacts with other data. The (see Figure 1 below).the 3M™ Clinical Data Repository. information model establishes temporalAn information model describes how and spatial contexts for patient data, so A medication order consists of: Drug What drug is being ordered? Dose What dosage is being ordered? Route How will the drug be given? Frequency How often will the drug be given? Start time When (date/time) will the drug first be given? End time When (date/time) will the drug stop being given? Ordered by Who ordered this drug? Order # What is the order number? A medication order can be represented in the computer as: MedicationOrder ::=SET { drug Drug, dose Decimal, route Route, frequency Frequency, startTime DateTime, endTime DateTime, orderedBy Clinician, orderedNum OrderNumber}Figure 1. How concepts are placed in a 3M Healthcare Data Dictionary information model www.3Mhis.com 7
    • Vocabulary: Industry-standard CMVs and codingand classification systems Concepts and concept IDsThe vocabulary component of the 3M • All Patient DRGs (AP-DRGs) In the 3M HDD, a concept is a unique,HDD identifies and represents the various • All Patient Refined DRGs definable idea or item that has a verymedical concepts found in clinical data, (APR-DRGs) specific, known meaning (e.g., cold,and it is organized to support synonyms temperature, sensation, viral infection, • CPT®and other lexical characteristics. As of infection, diagnosis) or a combination ofJanuary 2010, the 3M HDD contains • HCPCS concepts (“chest x-ray”). In the 3M HDD,over 2.6 million active concepts, over • PTXT (from the 3M™ HELP each concept is defined by both a human-17.9 million representations, and over System) readable text description and an assigned,16 million relationships. The source unique numerical identification, referred • Customer vocabularies (legacyvocabularies include: to as an “NCID” (“Numerical Concept systems, local and organization- Identifier,” which has no intrinsic meaning • SNOMED CT® specific terms) or significance in itself). No redundant • RxNorm In content, the 3M HDD encompasses: concepts are allowed, since they defeat • Unified Medical Language System the purpose of a controlled vocabulary. • Encounter data and demographics (UMLS) To demonstrate the need for concept IDs • Laboratory and a controlled vocabulary, consider how • LOINC® • Microbiology the word “cold” can be used in medical • National Drug Codes (NDCs) from language, as shown below in Figure 2. the First Data Bank Pharmacy • Pharmacy database • Diagnostic and procedural coding • ICD-9-CM • Findings, signs, and symptoms • Diagnostic Related Groups (DRGs) • Problem lists and diagnoses Concept ID Definition Cold, #123 A sensory perception (“patient complains of feeling cold”) Cold, #569 A pulmonary diagnosis (Chronic Obstructive Lung Disease) Cold, #784 An upper respiratory viral infection (“common cold,” “cold,” “flu,” etc.)Figure 2. Sample of 3M Healthcare Data Dictionary’s concepts and NCIDsCPT is a registered trademark of the American Medical Association.8 3M Health Information Systems
    • Synonyms—expanding the • Different representations of the • The terms preferred in specific same concept, either in a natural contexts (for example, “dyspnea”3M HDD vocabulary language or other coded format can be designated as the term forMuch of the richness of the 3M HDD • Common variants of a term, such a cardiologist, while “shortness ofvocabulary comes from its use of as acronyms or even common breath” can be the preferred termsynonyms (see Figure 3, below, for misspellings for a lay person)examples). In its use of synonymy, thevocabulary includes: • Foreign language equivalents • Synonyms, homonyms, and (human languages—French, eponyms (names derived from Spanish, Portuguese, etc.—can people or places) be added) Synonym examples: Acute Sinusitis ACUTE SINUSITIS Acute sinusitis, NOS Sinusitis, acute Acute infection of nasal sinus, NOS Acute inflammation of nasal sinus, NOS C0149512 (UMLS) 621850 (hospital-specific interface ID) Possible contexts for a synonym— Synonym Context Acute Sinusitis Problem list display Acute infection of nasal sinus Explanation C0149512 UMLS code 621850 Interface codeFigure 3. Sample of how the 3M Healthcare Data Dictionary uses synonyms and can specify a context www.3Mhis.com 9
    • The 3M™ Healthcare Data Dictionary’s knowledge baseThe 3M HDD’s knowledge base consists of semantic networks and hierarchies that describe the complex relationships existingbetween concepts in the vocabulary. These relationships can be hierarchical (parent-child or “is-a”) or non-hierarchical (“is-a-component-of”). Figure 4 (below) is an example of how the knowledge base can describe the relationships between the components ofa CHEM 4 laboratory test. Lab Test is a component of is a CHEM 4 Glucose Chloride Potassium Sodium Lab resultFigure 4. Sample of the 3M Healthcare Data Dictionary’s knowledge base as applied to a CHEM 4 lab testFigure 5 (below) shows how the 3M dictionary’s information model, vocabulary, and knowledge base tie together to create ameaningful database record that represents a clinical event. Information model Database record MedicationOrder: :=SET { MedicationOrder { drug Drug, drug Ampicillin, dose Decimal, dose 500, route Route, route Oral, frequency Frequency, frequency Q6H, startTime DateTime, startTime 08/01/09 10:01, endTime DateTime, endTime 08/11/09 23:59, orderedBy Clinician, orderedBy John Doe MD, orderedNum OrderNumber } orderedNum A234567 }Figure 5. Sample of a 3M Healthcare Data Dictionary information model (left) and the resultant database record (right) populated with vocabularyconcepts and organized by the knowledge base10 3M Health Information Systems
    • The 3M Dictionary’s Ability to “Map” Source CMVs and “Local Vocabularies”All industry-standard CMVs can coexist Mapping “local extensions” The advantages of mappingin the 3M HDD because of a process Because of the mapping process, an The ability to create local extensionscalled “mapping,” which cross-references organization’s “local” terminologies means that the healthcare organization’selements in each CMV with a concrete, can also be integrated with the standard users can continue using their own terms,unambiguous concept in the 3M HDD. CMVs, along with such interface while the dictionary seamlessly handles requirements as interface codes, billing the “translation” of such terms behind theWhy mapping is necessary codes, etc. Site-specific mapping is time scenes. Mapping provides today’s resource • There is no single, universally and resource intensive, requiring and time-strapped healthcare enterprise accepted and applied standard highly trained and experienced clinical with several advantages: vocabulary for the healthcare personnel who: • People do not need to be industry. • Understand each CMV’s inherent “retrained” in the language of • Existing coding systems are characteristics (e.g., “molecular” another computer system. They incomplete. combinations) continue using the terms they know • The HL7 version 2 standard • Consider each CMV’s limitations and understand. specifies message structure only; it (e.g., the lack of explicit • Existing information systems do not does not specify the actual data that relationships, reuses codes, etc.) need to be replaced or redesigned. is sent within the structure. • Design a mapping strategy that: • Individual entities within an • Every healthcare organization uses organization can retain specific - Follows vocabulary principles different terms and codes. information structures or sets, such - Meets user needs as preferred formularies. • It is impractical and too expensive to replace all legacy systems. - Is flexible and extensible • A specific form of the concept can be displayed in a specific context. Concept ID Representation Context 123 1234-5 LOINC code 123 NAS Interface code for Site #1 123 Serum Sodium User display for Site #1 123 CL357 Interface code for Site #2 123 S. Sodium User display for Site #2Figure 6. A sample of how the 3M HDD can represent a serum sodium laboratory result in different contexts3M staff expertise and credentialsThe 3M clinicians and informaticists • American Medical Informatics • Healthcare Information andwho support the 3M HDD and provide Association (AMIA) Management Systems Societymapping services belong to the following • HL7 Vocabulary Technical (HIMSS)groups and associations: Committee • American Health Information • Laboratory and clinical committees Management Association (AHIMA) of LOINC® www.3Mhis.com 11
    • 3M™ Healthcare Data Dictionary Architecture and “Platform Independence”The 3M HDD was designed to meet open • HL7 Common Terminology Finally, because the 3M dictionary is anarchitecture standards, allow for platform Services (CTS) independent entity, it can be updated withindependence, and conform to the • ASN.1 information model, new medical knowledge without rewritingfollowing industry standards: translatable to XML the application programs that use it. • Linux platform • Application Programming • HL7 messaging Interfaces (APIs) to access the 3M HDDConclusionTechnology and industry drivers—not • Effectively integrate information These challenges are best met throughthe least of which is HHS and ONC’s systems to increase the healthcare a working medical data dictionary thatinterim final rule—are quickly making facility’s competitive strength in the is an integral part of a clinical dataa medical data dictionary mandatory community and reduce the costs repository. In design and architecture,for any healthcare organization or associated with poor information the 3M Healthcare Data DictionaryEHR vendor that wishes to remain infrastructure. can support existing technologiescompetitive. The potential power of a • Meet the challenges posed by and continued use of legacy systems,medical data dictionary lies in its ability both industry and governmental providing a flexible and extensibleto help healthcare organizations meet mandates for uniform data “vocabulary server” that addressesthese challenges: standards and standardized regulatory pressures and serves the unique clinical vocabularies, coding, needs of both business partners and • Use the power of computers to and coding sets. healthcare consumers. deliver decision support to care providers in diagnostic, therapeutic, • Bridge communication gaps and management arenas. between healthcare providers, • Significantly improve outcomes business partners, and consumers— and contribute to the quality and even as technology and software safety of patient care by helping evolve—and allow expansion into to reduce medical errors resulting Internet-enabled technologies, from misinterpreted, inaccurate, or e-health relationships, and other imprecise patient data. communication vehicles.Health Information Systems575 West Murray Boulevard 3M is a trademark of 3M Company. CPT SNOMED and SNOMED CT are registered Please recycle. Printed in U.S.A.Salt Lake City, UT 84123 is a registered trademark of the American trademarks of the International Health © 3M 2010. All rights reserved.800-367-2447 Medical Association. LOINC is a registered Terminology Standards Development Published 03/10www.3Mhis.com U.S. trademark of Regenstrief Institute, Inc. Organisation. 70-2009-9098-7