My name is H. C. Mullins. I am a Family Physician with 19 years' experience as a primary care practitioner in a rural community, followed by 20 years as a clinician, educator/administrator in an academic setting. I have been involved in primary care informatics for the last nine years, consulting, teaching, conducting research, and actively participating in national and international primary care informatics organizations as Chair of the FP/PC Working Group of AMIA and US representative to Working Group V of IMIA. I greatly appreciate the opportunity to appear before this committee today to represent primary care practitioners, particularly the Family Practice/Primary Care Working Group of AMIA. My testimony will be based on: the recommendations of a consensus conference on standards for clinical vocabularies cosponsored by the FP/PC Working Group of AMIA and Working Group V of IMIA; an analysis of published studies; a recent informal survey of primary care informaticians and professional organizations; and my own experience.
There is an increasing realization that clinical data, held in ambulatory primary care medical records , is essential to provide complete and accurate information needed for both clinical and administrative services at multiple levels in an effective health information system.
As illustrated in the diagram from the classic paper on " The Ecology of Medical Care" by Kerr White, the capture of primary care health data can provide detailed information about a large majority of the health problems of over 95% of a given population.
From this detailed information, clinical datasets can be derived to support patient care, clinical audit, decision support, and practice-based research; management datasets will support activities such as billing, accounting, planning, etc.
This data flow can be accomplished in a progression of several distinct processes:
Each requires a different structuring language, mapped to each other: a clinical vocabulary for terming clinical information; a coding scheme for encoding and classifying information; and casemix coding for various groupings.
Below is a schematic of the data flow progression:
Process |
Purpose |
Volume |
|
TERMING ß |
Clinical records, guidelines, audit, decision support |
Tens of thousands of clinical terms |
|
CLASSIFYING and ENCODING ß |
Administrative and clinical data sets for billing; clinical and health services research; statistical analyses, epidemiology |
Thousands of categories |
|
GROUPING |
Resource management, costing, health systems research and planning |
Hundreds of groups |
The key building block in such an arrangement is a fully functional, structured, ambulatory primary care electronic medical record, from which flows the data needed by clinicians, educators, researchers, payers, managed care organizations, federal agencies, and others. This data can be captured only in the Primary Care setting .
On Nov. 1-2, 1995, a conference sponsored by the Family Practice/Primary Care Working Group of AMIA, entitled "Moving Toward International Standards in Primary Care Informatics: Clinical Vocabulary" was held in New Orleans. It was an extraordinary collaborative process which succeeded in gathering 79 participants from 9 countries who achieved an unlikely result: an agreed, shared-vision consensus on clinical vocabularies for primary care. Full details of the meeting, including the report of consensus and recommendations, are reported in the conference summary provided in the "Attachment" section of my testimony.
Excerpts appropriate to this hearing are the:
Available published studies, admittedly -- in most cases, -- , do not answer questions about the efficacy of the various available clinical vocabularies in coding clinical information for use in electronic clinical information systems. These include my own study (included in the "Attachments" section of my testimony) as well as those of some members of this Committee. Additional studies are recommended. For studies of primary care clinical vocabularies to be meaningful and useful:
1. What medical/clinical codes and classifications do you use in administrative transactions now? What do you perceive as the main strengths and weaknesses of current methods for coding and classification of encounter and/or enrollment data ?
In the US, ICD-9CM and CPT are generally the medical/clinical codes and classifications used. Strengths and weaknesses include:
| Strengths: | Weaknesses: |
|
|
2. What medical/clinical codes and classifications do you recommend as initial standards for administrative transactions, given the time frames in the HIPPA ? What specific suggestions would you like to see implemented regarding coding and classification ?
ICD-(n) or ICD(n) /CPT are recommended as initial standards for coding and classification for administrative transactions.
For clinical vocabularies, it is recommended that ICPC, Read, and SNOMED be cross-mapped, and aggressively evaluated in actual use at the point of care by primary care practitioners, using standardized methodologies . It would be premature, misleading, and destructive to recommend one vocabulary over another. (Conference Summary Report of "Moving Toward International Standards in Primary Care Informatics: Clinical Vocabulary, page 17).
It is recommended that a coding and classification strategy be implemented that incorporates the data flow previously described which features a progression of several distinct information handling processes:
Each requires a different structuring language, mapped to each other: a clinical vocabulary for terming clinical information; a coding scheme for encoding and classifying information; and casemix coding for various groupings.
Clinical data, held in primary care electronic medical records, is essential to provide complete and accurate information needed for both clinical and administrative services at multiple levels in an effective health information system. Capture of primary care health data can provide detailed information about a large majority of the health problems of over 95% of a given population.
From this detailed information, clinical datasets can be derived to support patient care, clinical audit, decision support, and practice-based research, and management datasets can support administrative activities such as billing, accounting, planning, etc.
3) If some version of ICD is to be used for administrative transactions, do you think it should be ICD-9-CM or ICD-10-CM and ICD-10-PCS, assuming that field evaluations are generally positive ?
Until performance evaluation is determined, there is no basis for an opinion about which versions of codes and classifications should be used for administrative transactions. The tool that best fits the job should be determined by performance evaluation.
4) Recognizing that the goal of P. L. 104-191 is administrative simplification, how, from your perspective, would you deal with the current coding environment to improve simplification and reduce administrative burden, while obtaining medically meaningful information ?
The most efficient and effective way to get information that is meaningful to clinicians, is for the clinicians themselves to enter the data at the time and point of care, in a manner that works for them, using a clinical vocabulary that supports the level of detail that they wish to record and retrieve. With increasingly sophisticated automatic mapping mechanisms -- from vocabularies to classification systems -- simplification and reduction of administrative burden is expected.
5) How should the ongoing maintenance of medical/clinical code sets and the responsibility, intellectual input, and funding for maintenance be addressed for the classification systems included in theses standards? What are the arguments for having these systems in the public domain versus the private sector, with or without copyright?
The maintenance of medical/clinical code sets should involve official professional organizational input both to ensure 1) clinical accuracy and 2) continuing professional endorsement. Central responsibility and funding are necessary to assure long term stability. Copyright is essential to prevent unauthorized use and corruption. If the code sets are available, stable, and reasonably priced, they could reside in either the public or private sector, or in some combination of both.
6) What would be the resource implications of changing from the coding and classification systems that you currently are using in administrative transactions to other systems? How do you weigh the costs and benefits of making such changes ?
I have no data on the resource implications of changing from current coding and classification systems used in administrative transactions to other systems. However, a conclusion may be drawn that administrative coding would replace manual systems, thus decreasing personnel. Software and training would require moderate cost outlay. Effective clinical systems, allowing for paperless offices would decrease personnel needed for transcribing and coding. Moderate costs would be required for upgrading existing hardware systems to include wireless and portable interfaces at the point of care. If the cost of the clinical system is perceived by the primary care provider as --first, and foremost -- aiding that provider in patient care -- it would be enthusiastically embraced and funded. The use of recorded clinical data for administrative and epidemiological use would then be relatively cost free at the user level.. Benefits for administrative simplification and improved patient care far outweigh the costs. The exception would be facilities with little to no computerization, which would require large outlay of funds for hardware. Today this is a rarity.
7. A Coding and Classification Implementation Team has been established within the Department of Health and Human Services to address the requirements of P. L. 104-191; the teams charge is enclosed. Does your organization have any concerns about the process being undertaken by the Department to carry out the requirements of the law in regard to coding and classification issues ? If so what are those concerns and what suggestions do you have for improvements ?
There is a concern that there will not be adequate primary care input into the process. In the past Primary Care has been unsuccessful in shaping the current and evolving electronic information systems in the U.S.. The reasons are multiple and arise from many quarters within and outside the discipline. Even though there are scattered pockets of individual and institutional excellence in primary care informatics and electronic information systems, they are minimally organized and without an adequate infrastructure. This makes it difficult to identify appropriate people and activities for input into this process. It is therefore suggested that the Department exhibit unusual aggressiveness in seeking the input and involvement of primary care on committees and in hearings. It is also suggested that the extensive experience with primary care electronic information systems in other countries be obtained by testimony, conferences, and other methods.
NATIONAL COMMITTEE ON VITAL AND HEALTH STATISTICS, SUBCOMMITTEE ON HEALTH DATA NEEDS, STANDARDS AND SECURITY
H. C. Mullins, M.D.
Professor, Family Practice
University of South
Alabama
Mobile, Alabama
Crozer-Keystone Center for Family Health
Springfield,
Pennsylvania
Past Chair, Family Practice/Primary Care Working Group, AMIA
Additional Questions for NCVHS Hearing on Coding and Classification Systems
1.Can one system serve most, if not all, purposes, e.g., clinical care, surveillance, quality assessment/improvement, clinical research, billing/management? If not, which systems can serve most of theses functions ?
One coding and classification system cannot serve all purposes. There are no valid studies available on which vocabularies serve which, if any, functions in an electronic medical record adequately or best. Recommendations are the initial approval of SNOMED, Read and ICPC as vocabularies for terming, with rigorous evaluation of actual use in electronic medical records and information systems. ICD-(X) is recommended as a classification system for encoding. ICPC is especially suited to record "episodes," which is essential for any meaningful collection and analysis of illness data. The recording of episodes is not solely a coding scheme function, but is an issue for the structure of the electronic medical record, as well.
2. Is it administratively simpler to use the same disease classification for administrative transactions and for statistical reporting ?
While it may be more simple to use the same disease classification for administrative transactions and statistical reporting, that disease classification may not be the best tool for the job in terms of data quality. To obtain the highest quality data at a particular level of detail or aggregation for a particular purpose, a tool designed for that purpose is required. Until systems have been thoroughly tested against clear criteria devised for specific purposes, administrative transactions and statistical reporting will require a suite of information tools, each devised for its specific purpose, whether it be terming, encoding or grouping.
3. To what extent do you feel that your discipline and practice setting are well represented by current systems for coding health conditions, diagnoses, services and procedures in administrative and financial transactions ?
The nature of primary care is such that it is not well represented by current coding systems in the US. A patient's visit to a primary care provider for a problem, frequently does not result in making a traditional diagnosis, but the reason for the encounter is recorded in terms of symptoms, findings, or problems. Since patients are seen at multiple visits over time for a single episode, to ensure accuracy it is absolutely essential to capture data so that it can be structured in the form of episodes. There is general agreement among primary care providers that for the above reasons, ICD based classifications "do not work" for primary care. Many feel that:
· ICPC is the best Primary Care Classification. It focuses on the experience in terms of the population served and is excellent for epidemiologic purposes. ICPC, and its extended versions are being used around the word to successfully capture "reason for encounter" and "episodes," but must be linked to other systems for administrative purposes.
· Read code, which is used extensively in the UK, is focused on individual patient care, was devised for use in the electronic medical record, and is able to capture concepts and terms about real patients at the point of care (terming). It is mapped to ICD-(x) for administrative and analytic purposes.
· The combined use of Read with ICPC (coupling) in a primary care ambulatory electronic medical record has been found to be immeasurably beneficial by allowing freedom of clinical expression along with a simple way of recording population trends.
· No reports on the use of SNOMED to populate working electronic medical record systems have been reviewed.
4. What problems do you encounter linking data coded with different classification systems and trying to crosswalk between (or among) classification systems ?
Linking data coded with different classification systems and crosswalking between or among them is doable but complex. This capability is being simplified by developing and/or emerging engines. There is incompatibility between ICD-9 and ICD-10. There is, of course, a loss of detail mapping from a more detailed system to a less detailed system, often accompanied by a change in clinical meaning.
5. What is the impact on and implications for current (and emerging) medical/clinical classification systems as we migrate towards computer-based patient records (cpr)? To what extent can the major classification systems currently in use serve -- in part, as vocabulary for -- the cpr? And if another system is recommended as the vocabulary for cpr, how can we assure that it crosswalks relatively easily to the classification systems currently used in administrative and financial transactions ?
Classifications are not vocabularies, do not capture the level of detail required for clinical care, are not designed for populating an electronic medical record, and should not be used -- even in part -- for that purpose. Crosswalks are essential but, as mentioned previously, are not simple between vocabularies and classifications. Mapping is available for many existing systems and can be developed to emerging standards. Sophisticated, automated mapping processes are emerging, which should simplify the process.
6. For presenters recommending a particular coding or classification system, what is the market acceptance for the system and current scope of its use ? What are the mechanisms for low cost distribution ?
No particular coding or classification system is recommended. Instead, a suite of vocabularies (ICPC, Read, and SNOMED) is recommended for use and testing. "The overwhelming consensus of the participants was that, at this time, it would be premature, misleading, and destructive to recommend any one vocabulary over another" (Conference Summary Report of "Moving Toward International Standards in Primary Care Informatics: Clinical Vocabulary, page 17).
The number of countries using various coding and vocabulary systems is shown in the following table (Conference Summary Report, Page 14).
Information on costs and mechanisms for distribution is incomplete. ICPC is available worldwide and is distributed by Oxford University Press. Costs not available. SNOMED is available for worldwide use from the College of American Pathologists. Costs not available. Read code is available the U. S. from Computer Automated Medical Systems at $200/yr per practitioner and $2000/yr per hospital.
7. How might NCVHS work with the Department to assure that the United States coordinates development of an international medical dictionary, classification/coding system, etc., including the terms, a process to keep continuously such terminology updated, and a server to deliver the content to whoever needs access ?
· Clinical/medical coding must be an international activity. In the continually shrinking global community, health concerns and health statistics, especially Public Health, are no longer limited to national boundaries. This should be approached in a multifaceted manner. We should:
· Support and build upon Conferences such as the "Moving Toward International Standards in Primary Care Informatics: Clinical Vocabulary" reported here.
· Establish a National Center for Coding and Classification adequately funded and charged with the responsibility of maintenance, etc. And,
· Promote and assist in the establishment of an International Coordinating Body to identify, develop, adopt and promote international standards.
Etc, etc.
8. Is it practical to move to a single procedure classification on the timetable required for initial implementation of administrative standards or should the standards continue the current practice of requiring different procedure coding systems for the ambulatory and inpatient sectors ?
I have no data or experience to comment on that question.
9. If a medical/clinical code set or classification system is selected as a standard, should providers be able to use all the available codes within the set or system or should those requiring the information (payers) be allowed to restrict reporting of certain codes ?
Both 1) providers should be able to use all the available codes within the set or system and 2) payers should be allowed to restrict the reporting of certain codes. This should not be viewed as a matter of restrictions, but of ensuring -- through collaborative processes and a common understanding of the information needs of the system as a whole -- that the required data is recorded and retrieved. The provider should have the freedom and system capability to record in the electronic medical record whatever information is desired in order to care for that particular patient, plus any other data that he/she may wish to collect for his/her own audit, research, and/or other purposes. That information must be structured using a clinical vocabulary, or "terming." The termed information can then be encoded by mapping to a classification system such as ICD-(X) , or "encoding." The payer (or other user of data) can then derive the information required for defined data sets restricted to their particular purposes, otherwise known as "grouping." In order for theses data sets to be complete, it may be necessary to prompt the provider for additional or specific information to be collected at the time of service.