Duke University Standards and Medical Informatics W. Ed Hammond, Ph.D. President, AMIA Vice-chair, HL7 Technical Steering Committee Chair, Data Standards.

Duke University

Standards and Medical Informatics

W. Ed Hammond, Ph.D.

President, AMIA

Vice-chair, HL7 Technical Steering Committee

Chair, Data Standards Working Group, Connecting for Health

Convenor, ISO TC 215 WG2

Professor-emeritus, School of Medicine

Professor-emeritus, Pratt School of Engineering

Adjunct Professor, Fuqua School of Business

Duke University

Duke University

A scenario …

Recently, at my exercise club, my blood glucose measured 112 mg/dl. This elevated value was sent to my composite record then to my PCP and to me. When I logged onto my computer, a flag indicated I had a message in my personal mail at my PCP’s web site. The message ask me to schedule an appointment soon because of the elevated glucose, as well as it was time for my annual physical exam.

I accessed the clinic’s web site and scheduled an appointment with my PCP for the next week. The system identified some additional testing for me, and scheduled me 30 minutes before seeing my PCP for the tests.

I also looked at my on record and noticed that my glucose had been climbing over the past 12 years to its current level.

Duke University

I arrived at the clinic, entered my health card in to a kiosk registering my arrival. My eligibility was automatically checked and my health plan verified. I was directed to the lab for the blood drawing. I was also assigned a number which provided the linkage for me on this visit. Within 2 minutes of my scheduled time, a white board identifying me by number directed me to Exam Room 10. Here the provider performed the annual physical examination, sharing a terminal between us, and discussing how she proposed to deal with the elevated glucose with exercise and weight reduction. Since my cholesterol was also elevated, she decided to start me on Zocor. My dentist had recently started me on an antibiotic that intensifies the action of the cholesterol-reducing drug. My PCP suggested that I complete the antibiotic before I start the Zocor. She also scheduled me to return in 3 weeks to test my liver function because of the drug.

Duke University

This information was put into my personal web page for download into my personal health record.

The exercise program was fed directly into my exercise machine, and my daily progress was monitored and recorded into my personal record. I also gave permission for the data to be uploaded to my PCP, since I thought the added pressure of another eye watching me would increase the incentive for my following the program.

I was also given, interactively, a personal diet to help control my weight. I kept an on-line log in my personal health record.

I also accessed information about the medication I had been given to reduce my cholesterol. I read about side effects and some of the controversy. I knew about the side effects; however, I decided to continue the drug at least for the next month.

Duke University

The Holy Grail of Medical The Holy Grail of Medical Informatics …Informatics …

The Electronic Health The Electronic Health RecordRecord

aka …aka …

Duke University

A changing world of health care

• Our world is expanding– The tremendous expansion of diagnostic tests

available,– The almost individualization of treatment, particularly

drugs– a vastly expanding field of knowledge

• Solution demands use of information technology in health – to contain costs– to reduce medical errors– and to increase quality

• Consumers are becoming more educated and want to be involved

• Integrated health systems are the trend

Duke University

A changing world of health care

• From a private, independent world to a combined and integrated community

• From unconnected, disparate heterogeneous systems to seamlessly connected interoperable systems

• From technologically constrained to technologically rich

• From hospital dominated to person-focused systems: health vs illness

• From billing records to clinically enriched databases

• From concealing data to sharing data

Duke University

Patients – the raison d’etreraison d’etre• Patients are seen asynchronously in a variety of

settings; thus data must give a single, integrated view of the patient.

• Need complete, appropriate data for decision making, to reduce errors and improve care.

• The spectrum of patient care -- home, outpatient, inpatient, intensive care, emergency, nursing homes & specialties.

• Patients are mobile -- data must be accessible internationally

• Patients move -- patient records follow and need to be understandable and useable in the new settings.

Duke University

Why standards in health care?

• There is an assumed and inherent need to share data in the health care setting. The data are of many types and form and will be used for multiple purposes.

• We must share both data and knowledge for both improved health care and for economic reasons.

• Sharing becomes economically possible only if interoperability exists.

• Interoperability occurs only if a full set of standards in health care exist.

Duke University

Standards are an everyday thing!

• VCRs, audio tapes, CDs, DVDs• Bread size - to fit toasters• ATM machines• Air controllers use English language

• Distance between rails for trains• 60 cycle, 110 volt electricity• Shoe sizes, clothes, gloves• Side of road we drive on• Size of paper

Duke University

Too many standards ….

Duke University

Steps to making a standard

• Awareness of need for standard• Critical mass of technical expertise to create

standard– Must insure fairness and not competitive advantage

to any single vendor– Expertise must be both technical and domain

• MUST involve vendors, providers, consultants, government

• Global acceptance important in today’s market• Vendor implementation usually driven by

consumer pressure to implement• Visible reduction in cost and effort of interfaces

using standard necessary for buy-in

Duke University

Different kinds of standards

• Company– DOS

– Windows

• Consortium/Open Source– Unix

– Linux

– JAVA

– M/Mumps

• Industry– DICOM

• Government– NIST

– CMS

– HIPAA/NCVHS

• Voluntary Consensus– ASC X12

– HL7

– NCPDP

– ASTM

– IEEE

Duke University

Consensus Standards

• Volunteer-driven• Not full-time commitment• Uneven levels of participation• Uneven levels of understanding• Required resolutions of negatives• Prone to compromise – leads to ambiguity• Funding constraints• Meet only a few times per year• Specialized balloting process (ANSI: requires

90% approval)

Duke University

How to get there from here …why standards in health care?

• There is an assumed and inherent need to share data in the health care setting. The data are of many types and form and will be used for multiple purposes. Traditionally, these uses have been addressed independently and redundantly.

• We must share both data and knowledge for both improved health care and for economic reasons.

• Sharing becomes economically possible only if interoperability exists.

• Interoperability occurs only if a full set of standards in health care exist.

Duke University

Why haven’t we done it?

• No accepted long term vision of what IT is.• No proven value to those of make purchasing

and financial decisions.• No widespread stakeholder buy-in.• Not considered a core component of health

care.• Resistance to change.• Unwillingness to make decisions and take

action on controversial issues.

Duke University

What are the building blocks?

• Data– Patient-centered– Comprehensive– Aggregated– Organized– High data integrity– Timely– Structured, semantically understandable– Sharable– Accountable– Secure and private

Duke University

How might we use it?

• Information for …– Patient care– Prevention of medical errors– Improved quality of care– Consistency in care– Cost effective care– Shared understanding of health and health care

among patient and provider– Health surveillance and biodefense– Workflow management– Research– Epidemiology– Billing

Duke University

What and how can we learn?

• Knowledge– Clinical trials– Decision support– Disease demographics– Outcomes– Quality indicators– Evidence based medicine

Duke University

What do we get?

• Wisdom– New models for health and health care– More cost effective care– Better understanding of disease and disease

processes– Better relationship among stakeholders– A happier, healthier world

Duke University

Why data standards? (1)

• Patient-centric EHR– Complete, aggregate data about patient

• Patient summary problem list• Current medications list• Allergies• Base demographics• Selected clinical elements• Reimbursement data

– Insurance

– Health Plan

Duke University

Why data standards? (2)

• Population Health Record– Outcomes data– Utilization data– Disease tracking– Detection of disease outbreaks– Detection of bioterrorism events– General health surveillance– Immunization

Duke University

Why data standards? (3)

• Reimbursement– Reimbursement rules– HIPAA transactions requirements– Automation of process– Easier audits for clinical justification– Reduction of use of human resources in

reimbursement process– Analysis of treatment by multivariate parameters

Duke University

Why data standards? (4)

• Research– Clinical Trials

– Drug Trials

– What diseases are prevalent

• By region

• By occupation

• By category

– Variation in outcomes

• Method of treatment

• Provider

• Region

Duke University

Classes of Standards

• External standards not unique to health care– Examples include communication standards, Internet

standards, LAN standards, XML/HTML standards, security standards, etc.

• Application level health data standards absolutely necessary for aggregating and sharing data

• Enhancement health-related standards that improve the process and extend the use of IT. This group includes clinical content and clinical knowledge standards.

Duke University

Classes of Standards - 1

• Basic communication standards that are not specific to health– Communication standards

• Internet standards

• LAN standards

• Web Protocols

– XML

– Security standards

– Authentication standards

– Biometric standards

– Encryption standards

– Digital signature

• Groups producing or influencing these standards

– W3C, IETF, OMG, OASIS, others

Duke University

Classes of Standards - 2

• Standards that relate to the definition, style, and naming of the data itself– Reference Information Model (RIM)– Data types– Terminology– Clinical Documents– Clinical Templates– Data element master set– Business Rules that identify what data elements are

collected: how, when and by whom [implementation manuals, conformance documents, metadata dictionaries}

Duke University

Classes of Standards - 3

• Process standard for message development framework

• Standards associated with data interchange– HL7 V2.4, V2.4 (XML) and in ballot V2.5– HL7 Version 3– DICOM – imaging domain– IEEE/CEN/ISO – medical devices– Others

Duke University

Classes of Standards - 4

• Standards associated with the Electronic Health Record– Architecture, content, format and form, purpose– Privacy and confidentiality– Access– Persistence– Control

Duke University

Standards Related to EHR - 5

• Decision Support Rules– Arden Syntax, GLIF, GEM, Prodigy– Clinical algorithms– CPOE– ePrescribing– Reimbursement Rules

Duke University

Interoperability Standards (1)

• Personal data absolutely MUST be identified when it is sent from the source to the aggregating data base

• That is best (essentially error free) accomplished when there is a single, unique personal identifier

• Because of privacy concerns we have not yet accepted this solution

Duke University

Interoperability Standards (2)

• Reference Information Model– Object Model that provides framework for the

exchange and sharing of health data. EHR model must be based on this model

– HL7 has created such a model, accepted internationally, that is now becoming stable

– HL7 model is high level requiring subsequent refined models for communications and storage of data.

Duke University

Reference Information Model

• An information model needs to underpin all architecture and terminology developments to ensure consistency of approaches and a shared understanding.

Liaw and Grain in a government report

Duke University

0..*

1 0..*

1

HL 7 RIM Core Classes

EntityEntity ParticipationParticipation ActAct

RelationshipRelationshipLinkLink

0..* 0..*

1 1

ActActRelationshipRelationship

1 1

0..* 0..*

ReferralTransportationSupplyProcedureCondition NodeConsentObservationMedicationAct complexFinancial act

OrganizationLiving SubjectMaterialPlaceHealth Chart

PatientEmployeePractitionerAssigned PractitionerSpecimen

RoleRole1

0..*

Duke University

Data Element Definition Set

• Defines every data element that will be collected including when, how and in what form

• Data must be structured• Links data elements to vocabulary sets as well

as RIM• Some work being done in this area by Health

Informatics Standards Board (ANSI) and Australia

Duke University

Data Types

• Simple data types– Numeric, strings, dates, currency, etc.

• Complex data types– Addresses, names, coded data elements

• Tightly coupled with the RIM• Must be consistent with terminology• Must be used (stored) in the EHR as defined by

data type

Duke University

Terminology

• Every data element that will be shared must be defined and coded in a terminology set (text modifiers may be permitted)

• Problem is the existence of too many terminologies, none of which is perfect

• Terminologies may be mapped but costs more money, creates errors and results in the loss of information

• Terminologies required for use must be free, controlled and maintained

• We must have a single, domain-model-based, constantly maintained, and freely distributed world-wide. terminology

Duke University

Drug Terminologies

• Significant progress has been made recently in creating a drug terminology standard. Effort includes starting with VA drug terminology set, adopted by FDA and assigned NDC codes, and mapped into UMLS. HL7 route, form and application device sets are included.

• NLM and SNOMED have apparently reached an agreement that will make SNOMED freely available for use in the U.S.

Duke University

Clinical Document Architecture

• XML-based definition of clinical documents such as discharge summaries, op notes, progress notes, radiology reports, etc.

• HL7 has ANSI approved standards. Work is based on 3 levels: (1) header; (2) header plus body structure and section headings; (3) element content specification and identification

Duke University

Conveying complex concepts• Clinical Data Model or Clinical

Templates

– Defines detail clinical object structures

– Permits constraints on objects

– Examples• Clinical lab battery• Heart Murmur• Blood pressure measurement• Physical exam for chest pain• Protocol for sore throat

• Require registry

Duke University

Decision Support

• For defining knowledge and decision support algorithms

• HL7 brings together several existing efforts in this area– Arden Syntax– Prodigy (UK)– Guideline Interchange Format (GLIF)– GEM

Duke University

Implementation/Conformance

• Most frequently, ambiguity and options remain in standards at all levels. Total interoperability requires a precise definition of what will be sent to whom under what circumstances.

• One example of this approach is the Emergency Department implementation manual called DEEDS.

• The Centers for Disease Control has created a reporting system for health surveillance known as NEDSS will also provide this level of specification.

Duke University

Electronic Health Record

• Requires defining exactly what standards are required

• Issue is where does standard stop and vendor proprietary interests start.

• Includes some architecture and probably categorization of data elements stored.

• Several efforts underway including Good Electronic Health Record (Australia and Europe), HL7 and AMIA

Duke University

Reusable Components

• HL7 Clinical Components Object Working Group (CCOW)

• Defining standards for reusable component software

Duke University

Imaging Standards

• DICOM is international standard for images and pictures and similar media

• JTC1 defines standards for JPEG and MPEG• DICOM also does structured reports similar to

HL7 clinical documents but for radiology and imaging reports. Efforts are being coordinated.

Duke University

Medical Devices

• IEEE provides leadership in this area.• Includes bed-side devices and covers primitive

layer of interface up to application.• Standards include cable, wireless, infrared

connectivity• Standards become international through ISO

Duke University

Security Standards

• At communications level, mostly developed outside health industry but with influence. IETF playing major role.

• Digital Signature and PKI standards are being influenced by health-related participation.

Duke University

Other Standards

• Waveforms• Data Integrity Standards• Presentation Standards• Icon Standards• Functionality standards

Duke University

What is an EHR? … my definition

• It is not a clinical repository.• It’s purpose is to enhance the health and enable

the care of the individual. It’s contents are solely justified for that purpose. When data ceases to contribute, it is removed.

• Much of the data in the inpatient setting has limited persistence - usually the more intense the care, the shorter the persistence.

• There are other repositories – a data warehouse that does contain and retain everything.

• The EHR documents maintenance of care, diagnostic and treatment processes, health status.

Duke University

Population record

• A summary record from all sites and sources of care

• Linkage of data for new sites as care as well as population surveillance, research, quality, analysis

• Data arrives as identified data, available as disidentified

Duke University

The Personal Record

• Model to meet consumer needs and understanding

• Focus on functionality and work management, not clinical repository

• Information display should be driven by intelligent query and understanding of needs

• Couple with appropriate education• Home entry of data - direct or sensors• Person-controlled release• Customizable

Duke University

Business Linkages

Hospital

Intensive Care

EmergencyDepartment

NursingHome

Ambulatory CareClinic

Patient role in control?

Duke University

EHR Interoperability Diagram

PatientEncounter

Provider EHR

Database

PersonalEHR

LongitudinalEHR

Profile

Profile

ProfileResearchDatabase

EnterpriseData

WarehousePersonal

EHR

ProfileDisease Registry

Profile Billing/Claims

Profiles contain business rules

Institution EHR

Database

Duke University

Population or Composite Summary EHR

PatientEncounter

Provider EHR

Database

PopulationEHR

PopulationProfile

Provider EHR

Database

PatientEncounter

Provider EHR

Database

PopulationProfile

Provider EHR

Database

Duke University

Download Process

HL7 Message

ID

DoubleEncryption

SiliconEncoder

Sensitive Demographic

Data

Encrypted IDIdentifying Data,

name, address, etc.

Identifying Data, Translated (e.g. Zip). Aggregated

SummaryLongitudinal

EHR

Summary Data

ID

Duke University

Summary Longitudinal Record

Access list permit

by provider

groupclinicother

Patient controlled access

SummaryLongitudinal

Record

Access log

Feeds PH surveillance, patient safety,epidemiology

Duke University

The future

• I am always a person – a complete entity to the provider I am seeing.

• I don’t have to worry that my known allergies will be missed.

• I have faith that all decisions will be made by someone who knows all about me, my preferences, and my health.

• My data will be interchangeable and understandable.

• My data will be secure and appropriately protected.