Lessons Learned from Pilot Project using SNOMED-CT for Clinical Observations

Lessons Learned from Pilot Project using SNOMED-CT for Clinical Observations John B. Hughes, M.D., C.M., Associate professor, McGill University Michel A. Lortie, ing. Fall 2010 Infoway Partnership Conference Vancouver, 16 Nov 2010

Health Informatics Research Institute of St-Mary s Hospital Center NPO for research into health informatics and clinician adoption Core funding from St-Mary s Hospital Foundation Resultant of a review of the state of the art of EMRs done in preparation for McGill accreditation by the College of Family Physicians of Canada

HiRES Health Informatics for Research, Education and Service; a project to Research the computable patient record needed for teaching, research and service in the academic environment Design a clinician defined clinical content specification mechanism for Primary Care Teaching Facilities Create a curated Pan Canadian clinical content standards library through endorsement by the College of Family Practice of Canada and the Royal College of Physicians and Surgeons of Canada.

Pilot 2008 Description In 2008, the principals at HIRI undertook a 6 month pilot project to investigate the use of SNOMED-CT as a coding standard for the computable medical record Physicians in a poly-clinic in Westmount and in a retirement home in East-end Montréal were provided with a software tool that captured SNOMED-CT coded findings, observations and diagnoses

Pilot 2008 Description (cont) Intent was to link findings, observations and diagnoses to a differential diagnoses engine (i.e. DxPlain) to create a computable target for investigation

Pilot 2008 - Description (cont) Selected coding of Obs, Findings and Diagnoses Observations, Findings Diagnoses Candidate coding of Obs, Findings and Diagnoses Candidate Differentials Candidate coding of Obs, Findings and Diagnoses Selected coding of Obs, Findings and Diagnoses EMR User I/F Current obs, Finding and Diagnoses SNOMED-CT Coding Bowser Candidate coding of Obs, Findings and Diagnoses Query based on Obs, findings and Diagnoses Selected coding of Obs, Findings and Diagnoses Candidate differentials SNOMED coded Patient Medical records Prior obs, finding and Diagnoses SNOMED / DxPlain transcoder SNOMED-CT New and historic Obs, Findings and Diagnoses DxPlain coded Candidate Differentials DxPlain Advisor Query based on Obs, findings and Diagnoses Candidate differentials DxPlain Differentials database

Pilot 2008 Description (cont) The acquisition / coding tool was modeled after the Clue browser and operated in process to an EMR system in use at both sites The paradigm used for the acquisition was one of composing findings, observations and diagnoses into a prose SOAP note by providing hints to the browser then selecting the appropriate concept and qualifiers as presented by the browser

Pilot 2008 Description (cont)

Pilot 2008 Outcomes Most participating physicians opted out of project Poly-clinic users abandoned the system ahead of the retirement home physicians Bridging of SNOMED-CT coding to DxPlain coding proved highly problematic The use of a structured prose approach to data acquisition was deemed to be sub-optimal

Pilot 2008 Findings Initial search speed of browser was considered inadequate for production environments. In spite of subsequent significant enhancements to the search speed; project never recovered from initial poor perception Vocabulary of the search was felt to be too formal and restrictive; the resulting prose note not acceptable.

Pilot 2008 Findings (cont) Coding of diagnoses and extant bridge to ICD 9 was considered very valuable with respect to billing Difficulties with the coding of findings and observations coupled to the bridging to DxPlain made assessment of a computable feature impossible

Pilot 2008 Preferred direction Physicians proposed that a simpler browser aimed at collecting coded diagnoses would fit better within their practice Physicians felt that pre-coded acquisition templates, structured according to subject (e.g. diabetes, COPD, etc), would be an improvement over the stilted prose the browser produced

Post-Pilot 2008 direction After investigation of experiences in the UK and, in recognizing the problem of physician uptake with respect to any EHR, let alone a computable record, it was felt that establishing a pan-canadian effort to collect the practice knowledge of health care professionals would better serve the physician community Archetypes and Templates were selected to serve as this new collection tool

Canadian Archetype Initiative Derived from work done elsewhere, notably in the UK. Unique characteristics include: Mandatory coding, compliant with Infoway guidelines Use of a richer constraint language Enhanced user interface Secure network linking contributors to the repository managed by St-Mary s for the CFPC

Outline IT standards have had a major positive impact on many facets of health data capture, communication and analysis. Certainly, this led to better research with faster results. The question is whether current e-health standardization efforts around semantic interoperability are equally beneficial. We believe, as is now recognized internationally, it is because of too much focus on business aspects and not enough meaningful involvement of different the types of clinicians, educators and researchers

Some dissident voices Do we want industry writing the nation s health IT standards? CCHIT s standards are, in fact, mainly approvals of whatever industry is doing. Dana Blankenhorn. Who will control the coming health IT standards? ZDNet. February 27th, 2009 (http://healthcare.zdnet.com/?p=1905) Kush and others had asked the AHIC earlier this year to form a federally sponsored workgroup to help integrate clinical research data needs into EHRs, but the panel did not grant the request. It was dominated by representatives of the health care delivery and payer communities. Nancy Ferris. Post-AHIC health IT standards process gets under way. http://govhealthit.com/articles/2008/11/postahic-health-it-standards-process-getsunder-way.aspx

A widespread misconception A variety of clinical terminology standards exist in the health care system, giving meaning to raw data and allowing for semantic interoperability. The Systemized Nomenclature of Medicine (SNOMED) + Clinical Terms (CT) is one example of a clinical reference terminology that provides for semantic interoperability.

The sad consequences Misunderstanding concepts, and excessive focus on information models, data models, etc. Leads to the many mistakes in concept-based terminologies; and many ambiguities in electronic patient records in which such terminologies are used.

Using Information codes in does the Service not of prevent Health ambiguities as to what is The same type of location code used described: how many disorders in relation to three different events might or are listed? might not refer to the same PtID Date ObsCode location. Narrative 5572 04/07/1990 26442006 Three references of hypertension closed fracture of shaft for of the femur same 5572 04/07/1990 81134009 patient denote three Fracture, times closed, the same spiral disease. 5572 12/07/1990 26442006 closed fracture of shaft of femur 5572 12/07/1990 9001224 Accident in public building (supermarket) 5572 04/07/1990 79001 Essential hypertension If the same fracture code is used for the same patient If two different on fracture different codes dates, then are used in relation to If two different tumor these codes might or might The same codes fracture are used observations code used made in on relation not refer the to same in relation to observations the same to two different made on day patients different for the same can not patient, refer fracture. they to dates for the same the patient, same fracure. might refer to the same fracture they may still refer to the same tumor. 0939 24/12/1991 255174002 benign polyp of biliary tract 2309 21/03/1992 26442006 closed fracture of shaft of femur 2309 21/03/1992 9001224 Accident in public building (supermarket) 47804 03/04/1993 58298795 Other lesion on other specified region 5572 17/05/1993 79001 Essential hypertension 298 22/08/1993 2909872 Closed fracture of radial head 298 22/08/1993 9001224 Accident in public building (supermarket) 5572 01/04/1997 26442006 closed fracture of shaft of femur 5572 01/04/1997 79001 Essential hypertension 0939 20/12/1998 255087006 malignant polyp of biliary tract

A Little Philosophy Realism vs. Concept

Desiderata for Controlled Medical Vocabularies in the Twenty-First Century James J. Cimino 1998 Department of Medical Informatics, Columbia University, New York, USA Builders of medical informatics applications need controlled medical vocabularies to support their applications Vocabulary content, concept orientation, concept permanence, Non-semantic concept identifiers, Poly-hierarchy, formal definitions, Rejection of "not elsewhere classified" terms, multiple granularities, multiple consistent views, context representation, Graceful evolution, and recognized redundancy.

From Concepts to Clinical Reality: An Essay on the Benchmarking of Biomedical Terminologies Barry Smith 2004 Department of Philosophy and National Center for Biomedical Ontology, University at Buffalo, Buffalo, NY 14260, USA Institute for Formal Ontology and Medical Information Science, Saarland The orthodox approach rests on the view that the fixation of meanings is best brought about through the alignment of terminologies on what are called concepts. concept orientation was in some respects an important step forward in terminology development,

Four loose families of concepts can be distinguished, Linguistic view, concepts are general terms whose meanings have been somehow regimented (or, as in some variants of the view, they are these meanings themselves). Psychological view, concepts are mental entities, analogous to ideas or beliefs. Epistemological view, concepts are units of knowledge (as the latter term is used in phrases such as knowledge representation, knowledge modeling, knowledge-based systems, and the like Ontological view, concepts are universals, kinds, attributes or properties (i.e. they are something like general invariant patterns) on the side of entities in the world.

Each one of these views might, in and of itself, be in a position to sustain a coherent methodology for the fixation of meanings in terminologies. However, elements of all four views are to be found mixed together in different combinations in the standard literature, in ways which provide strong evidence for the thesis that no single reading of the term concept can sustain all of the expectations which have become associated with its use

When I use a word, Humpty Dumpty said in a rather scornful tone, it means just what I choose it to mean neither more nor less. The question is, said Alice, whether you can make words mean so many different things. The question is, said Humpty Dumpty, which is to be master that is all. Lewis Carroll

MODELS Man tries to make for himself in the way that suits him best a simplified and intelligible picture of the world and thus to over come (sig. understand) the world of experience, for which he tries to some extent to substitute this cosmos (sig. picture) of his. This is what the painter, the poet, the speculative philosopher and the natural scientist do, each in his own fashion... one might suppose that there are any number of possible systems... all with an equal amount to be said for them; and this opinion is no doubt correct, theoretically. But evolution has shown that at any given moment out of all conceivable constructions one has always proved itself absolutely superior to all the rest. Einstein, A. The World as I See It (1931)

The Ogden and Richards (1923) semiotic triangle

Three levels of reality From Werner CEUSTERShttp://www.org.buffalo.edu/RTU 1. The world exists as it is prior to a cognitive agent s perception thereof;

Reality exist before any observation; From Werner CEUSTERShttp://www.org.buffalo.edu/RTU R MEdbASE Werner Ceusters Research

Reality exist before any observation From Werner CEUSTERShttp://www.org.buffalo.edu/RTU Humans had a brain well before they knew they had one. Trees were green before humans started to use the word green. R And also most structures in reality are there in advance.

Three levels of reality From Werner CEUSTERShttp://www.org.buffalo.edu/RTU 1. The world exists as it is prior to a cognitive agent s perception thereof; 2. Cognitive agents build up in their minds cognitive representations of the world;

From Werner CEUSTERS http://www.org.buffalo.edu/rtu B The ontology author acknowledges the existence of some Portion Of Reality (POR) R

From Werner CEUSTERShttp://www.org.buffalo.edu/RTU B Some portions of reality escape his attention. R

Three levels of reality From Werner CEUSTERShttp://www.org.buffalo.edu/RTU 1. The world exists as it is prior to a cognitive agent s perception thereof; 2. Cognitive agents build up in their minds cognitive representations of the world; 3. To make these representations publicly accessible in some enduring fashion, they create representational artifacts that are fixed in some medium.

He represents only what he considers relevant; From Werner CEUSTERShttp://www.org.buffalo.edu/RTU RU O1 1 RU B1 1 B Both RU 1 B1 and RU 1 O1 are representational units referring to #1; RU 1 O1 is NOT a representation of RU 1 B1 ; RU 1 O1 is created through concretization of RU 1 B1 in some medium. O R #1

THUS These concretizations are NOT supposed to be the representations of the cognitive representations; We should not be in the business of concept representation Or should we?

Proposed Solution: Referent Tracking! From Werner CEUSTERShttp://www.org.buffalo.edu/RTU Purpose: explicit reference to the concrete individual entities relevant to the accurate description of each patient s condition, therapies, outcomes,...

Numbers instead of words From Werner CEUSTERShttp://www.org.buffalo.edu/RTU Method: Introduce an Instance Unique Identifier (IUI) for each relevant particular (individual) entity 78

The principle of Referent Tracking From Werner CEUSTERShttp://www.org.buffalo.edu/RTU John Doe s #1 instance-of at t 1 person inst-of at t 2 #10 John Smith s liver #2 instance-of at t 1 liver inst-of at t 2 #20 liver tumor #3 instance-of at t 1 tumor inst-of at t 2 #30 tumor was treated with RPCI s #4 irradiation device instance-of at t 1 #5 instance-of at t 1 #6 treating clinic device inst-of at t 2 inst-of at t 2 #5 #6 #40 RPCI s was treated with irradiation device

In defense of the Desiderata James J. Cimino * 2005 Departments of Biomedical Informatics and Medicine, Columbia University, New York, NY 10032, USA I suggest a path that acknowledges the importance of representing reality, as best we can know it, but accepts the need for concepts to help us, among other things, reason under uncertainty. I consider this the realistic path.

COMPUTATIONAL TECHNOLOGY FOR EFFECTIVE HEALTH CARE IMMEDIATE STEPS AND STRATEGIC DIRECTIONS William W. Stead and Herbert S. Lin, Editors Committee on Engaging the Computer Science Research Community in Health Care Informatics Computer Science and Telecommunications Board Division on Engineering and Physical Sciences NATIONAL RESEARCH COUNCIL OF THE NATIONAL ACADEMIES THE NATIONAL ACADEMIES PRESS Washington, D.C. www.nap.edu 2009

Principle 1: Focus on Improvements in Care Technology Is Secondary The most important principle for guiding evolutionary change in health care is to focus efforts on achieving the desired improvements in health care rather than on the adoption of health care IT as a goal in itself.

Principle 2: Seek Incremental Gain from Incremental Effort If programs can be structured so that small investments yield visible success, stakeholders and the relevant decision makers are more likely to be persuaded to continue along such a path.

Principle 3: Record Available Data So That They Can Be Used for Care, Process Improvement, and Research Systematic improvement of health care is datadriven. providers should aggregate as much data as feasible about people, processes, and outcomes from all sources, acknowledging the never-ending challenge of maintaining reasonable degrees of patient confidentiality in such a data collection effort.

Principle 4: Design for Human and Organization Factors design systems to support people in doing the right thing provide incentives for and eliminate barriers to doing those things. barriers and incentives can be sociological, psychological, emotional, cultural, legal, economic, or organizational. Human-centered design pays attention to all of these factors as they relate to technical function and form.

Principle 5: Support the Cognitive Functions of All Caregivers, Including Health Professionals, Patients, and Their Families cognitive support needs to center on high-level decision making (e.g., diagnosis) for populations, patients, or situations, cognitive support needs to span transactional tasks such as test ordering or prescribing. cognitive support is not well served by the taskspecific automation systems that make up the majority of today s health care IT.

Principle 6: Architect Information and Workflow Systems to Accommodate Disruptive Change architect health care IT for flexibility to support disruptive change rather than to optimize today s ideas about health care. it is axiomatic that health care will change dramatically into the future.

Principle 7: Archive Data for Subsequent Re-interpretation health care IT should provide the capability of recording any data collected in their measured, uninterpreted, original form, archiving them as long as possible to enable subsequent retrospective views and analyses of those data. Advances in biomedical science and practice will change today s interpretation of data.

Principle 8: Seek and Develop Technologies That Identify and Eliminate Ineffective Work Processes develop technologies that allow identification and elimination of ineffective work processes and implementation of new approaches to achieving their purpose. automation of work processes developed in an era when paper was the medium for communicating and archiving is fraught with cost and unintended consequences.

Principle 9: Seek and Develop Technologies That Clarify the Context of Data develop technologies that present new information in the context of other information available about the patient and relevant biomedical knowledge. the combination of new biomedical technologies, together with increased access to data through health care IT, is increasingly overwhelming health professionals ability to make sense of individual findings e.g. Alert fatigue New approaches are needed to present information in context so that patterns and choices stand out.

Summary of EMR generations First generation: The Collector - simple systems that provide a sitespecific solution for the need to access clinical data which is imported through scanning or other forms of aggregation Second generation: The Documenter - basic systems that clinicians use at the point of care to adequately document rather than merely access clinical data Third generation: The Helper - Systems that include episodic and encounter data and use decision support tools to assist clinicians, functional in at the minimum both ambulatory and inpatient settings Fourth generation: The Partner - Advanced systems that provide more decision support capabilities and that are operational and accessible across the continuum of care, and providing sufficient credibility as to become the patient's legal medical record Fifth generation: The Mentor - Complex and fully integrated systems that include all previous capabilities and that are a main source of decision support in guiding patient care for both clinicians and consumers

13th International Congress on Medical Informatics September 2010 Cape Town, South Africa