The Workgroup on Computer-based Patient Records of the National Committee on Vital and Health Statistics held public hearings on September 16-17, 1999, at the Hubert H. Humphrey Building, Washington, D.C. Present:
The NCVHS Workgroup on the Computer-based Patient Record (CPR) held hearings on September 16-17, 1999, in an ongoing process to study the issues and advise the Department of Health and Human Services on adoption of uniform data standards for patient medical record information (PMRI) and the electronic exchange of such information. The Workgroup heard presentations from presenters on four panels. In addition to the hearings, the Workgroup discussed a briefing letter to the Secretary of Health and Human Services and its future informational needs.
William Jessee, M.D., Medical Group Management Association (MGMA)
Dr. Jessee discussed key components of patient records, key characteristics of data and data linkages, and some limitations of the paper record environment. He described clinical, administrative, and research functions as prime candidates for standardization, and suggested the work of David Eddy on clinical logic trees as a possible approach to standardize clinical data. Capturing “low-hanging fruit” is critical in the standardization process, and linkage of access to AHCPR clinical guidelines would be an essential feature in achieving a return on investment.
Among MGMA’s concerns about CPR development are the following: (1) The government has a leading role to play in the development of a standard data dictionary; (2) universal provider, employer, payer, and patient identifiers should be required; (3) the government should lead in developing standardized interchange protocols from external databases to be able to pool data from multiple sources; (4) flexibility in CPR systems must allow for variations in service delivery within organizations; (5) health information must integrate with data from external knowledges bases to apply rules-based, logic-driven clinical-decision support; (6) privacy and confidentiality must be protected; (7) open technologies and architectures must be part of the solution; (8) incentives would help to advance the standardization process; (9) the government should promote inclusiveness in the standardization development effort; (10) proposed standards must be field reviewed and pilot tested before widespread adoption; and the government should focus on data exchange, data vocabulary standards, certification procedures, security standards, and public education on new standards.
MGMA’s specific recommendations to the Workgroup include the following: (1) The federal government should lead in the development of standards for CPR content and format; (2) standards to ensure functionality of CPR systems should include data definitions and formats, data security, processing data in a uniform manner, and transmission; (3) the highest level of record confidentiality should be ensured without imposing onerous administrative burdens on practices, and no administrative simplification provisions should be enacted until enactment of comprehensive standards for health data privacy and security; (4) standardization should be prospective with a lengthy transition period; (5) the government should continue working with standards-development groups; (6) mechanisms to assure wide input to the standards development process from medical record users should be identified; and (7) education of the health care industry and public is critical on the need for standardization, implementation process, and security measures.
Alfred Buck, JCAHO
Dr. Buck proposed a definition of a patient medical record and noted that, in addition to recording the management of an individual’s health care, CPR can assist in organizational management, process improvement and assessment, and research and education. Among issues to be resolved are confidentiality legislation, universal patient and provider identifiers, standardized chart format, standardized data element fields and definitions, and core data sets. Dr. Buck also itemized a series of data quality issues that must be resolved, including data element definitions and categories and domains of measurement.
Dr. Buck urged that narrative entries into the patient record be addressed through standard vocabulary and terms. Applicable standards and other criteria must apply to both data elements and narrative.
The government and the private sector together should facilitate identification of significant activities and consensus development, and identifying and evaluating expert opinion, and should serve as consolidator and regulator; Dr. Buck concurred with Dr. Jessee’s view of the government’s role.
Dr. Buck emphasized the role of quality and value in the standards process. He proposed key components of a definition of health care quality, including compliance with science, consistency with wisdom-based processes, outcomes, goals, measures, customer satisfaction and perception, and oversight and accountability, including accreditation. He stated that patient medical record initiatives should facilitate definition of health care benefits, structure of access, allocation of resources, and alignment of research and education. Such electronic contributions can help meet health care challenges and to refine policy.
Stephen Lamb, National Committee on Quality Assurance
Mr. Lamb described the shortcomings of the limited information system infrastructure, which has hampered the use of HEDIS measures to assess the quality of health care delivery. He also discussed several financial costs and scientific and health-care limitations present in today’s environment, which are likely to be ameliorated by improved information systems and the CPR. Mr. Lamb pointed out that HEDIS measures often drive the nation’s public health agenda, and if it is not measured, it does not become a priority.
Drs. Buck and Jessee underscored the importance of resolving the confidentiality and privacy issues with the input of consumers to avoid a situation with excessive confidentiality provisions. Mr. Lamb highlighted the importance of individual identifiers. In a discussion of quality, Dr. Jessee contrasted the litigious medical environment for provider error and adverse effects with the risk-free environment for reporting airline pilot error. He described a forthcoming initiative on patient safety in office-based practices that includes research, educational, communications, and advocacy components. Pressures on productivity have led to diminished data capture and increased likelihood of error. Technology can alleviate administrative burdens. Dr. Buck attributed delays in standards development to lack of consensus rather than available methodology, in addition to lack of experience and data.
The participants discussed difficulties in making the business case for CPR. It is not reasonable to expect return on investment in the short term, but some larger practices are seeing benefits for having made the investment in electronic records. If value were increased by standardization, noted Dr. Jessee, outcome comparisons could be made between practices. Dr. Buck pointed to early evidence that shows that when quality equates with greater value, payback is possible; but cost is still a prime issue for providers. Mr. Lamb stated that the AMA and NCQA are discussing CPT codes. He noted that without mastering the financial drivers, widespread implementation will be difficult to achieve.
Herman Jenich, IPRO
Mr. Jenich noted the need for comparable PMRI. He commented that validation at 5 percent is an awesome task and a great expense. He identified five categories of system limitations—clinical, membership, provider, and vendor data, and data integrity—and gave examples of each. He also described difficulties in getting data for HEDIS measures and problems in performance measurement when vendors do not collect data, the health plan does not collect the data from the vendor, or the formats of the two entities are too difficult to match. Using different systems at this point is a hindrance for health plans in consolidating data from various sources.
Mr. Jenich recommended that the Workgroup address limitations in existing code schemes and the need for unique patient and provider identifiers, because automation may perpetuate and accentuate existing shortcomings. Any patient medical record system must consider a broad spectrum of potential users; the success of a CPR standard depends on the commitment of individuals who enter the data. Clinicians and administrative staff will be most compliant with a system that provides timely information on their performance, with comparisons to peer groups they believe are meaningful. The standards development process should stress training in coding and data entry, and guidance (i.e., national minimum standards) in editing and auditing electronic PMRI data. PMRI standards should be flexible, and CPR records should ensure data security and patient confidentiality.
Mr. Jenich asserted that PMRI standards will help both purchasers and providers of health care to measure health plan performance, encourage the development of fully automated record keeping, and eliminate the need for costly abstraction from paper records to support performance measurement. Both government and the private sector must be active participants in the process of standards development and implementation. Mr. Blair noted that the government’s role is to study PMRI and its exchange, not to develop standards.
Stanley Griffith, M.D., Indian Health Service (IHS)
Dr. Griffith presented an overview of the IHS’s health information systems and then discussed the importance of data quality. He enumerated types of data quality problems, such as erroneously recorded data, missing data, and data residing outside the system. Quality varies among the elements in patient records: Medications and vital signs data are generally excellent, while data on patient education, smoking history, and cause and place of injury are not good. The most problematic data are information that clinicians do not enter into the electronic system.
Dr. Griffith listed conditions under which data quality will improve, including when individuals who use the data enter the data, when all staff depend on the electronic record rather than the hard copy, and when jobs are designed to make it easier for staff to do their jobs well and data is collected as a byproduct. Vested interests seem to drive the industry toward better data. Dr. Griffith asserted that achieving the fully functional CPR can be the greatest step in improving data quality, which relates directly to improved quality of health care for individuals, improved public health care, more effective service delivery, improved financial well-being, and improved legal risk management. He urged the federal government to identify which aspects of the necessary infrastructure are not adequately addressed by current development efforts: information models, coding systems, data dictionaries, communications, secure aspects, and fill in the blanks with public resources.
Mr. Jenich emphasized the importance of capturing all the information possible, noting that behavioral health is the most difficult data to capture because of confidentiality issues. Confidentiality measures increasingly are hindering data collection efforts.
Dr. Griffith discussed the need for intelligent services that will code data at the point of service. Dr. Kolodner noted that one goal is to collect clinical data-level specificity relating to patient care, with the system elevating that information to a higher level of aggregation.
Dr. Griffith offered additional information about the size of the IHS health system.
Gary J. Arvary, M.D., Skylands Medical Group
Dr. Arvary described the decrease in physicians’ disposable income to purchase technological tools; he asserted that physicians are apathetic towards CPR because they cannot afford it. He cited survey results that show decreasing use by physicians of the CPR from 1997 to spring 1999. He postulated that if all providers used CPR and entered structured data, the insurance carriers, academic institutions, and the government would all want that structured data for accurate analyses—as would the providers. Dr. Arvary suggested that commonly used, but imperfect, coding systems be maintained in the future to avoid losing data. He asserted that trying to make data dictionaries all inclusive is probably unworkable; he suggested making a small portion of it codable with the rest unstructured. Mr. Arvary put forth the case for pen-based computerized systems.
In summary, Dr. Arvary recommended the following: (1) Developers of software should be subsidized to standardize the record, using such techniques as a standardized (not all-inclusive), paginated medical vocabulary that is easy to use with pen-based computers; (2) parts of the record should be fragmented and transferred over analog phone lines, the entire note can be transferred over high band-width lines; (3) physicians should be subsidized to purchase the systems to begin the health information infrastructure; (4) the structure of the health information infrastructure should be regional to reflect the regional character of cultures in general, referrals, insurers, and Medicare; (5) in the private sector, network administrators should be existing small groups and IPAs; and (6) subsidies should be given to the use of communication lines. Dr. Arvary predicted that within 10 years, such a network will be self-sufficient and part of a national health information infrastructure with both for-profit and not-for-profit segments.
Jeffrey Rose, M.D., Kaiser Permanente, Colorado Region
Dr. Rose described how his organization created a CPR system and the features of that system designed to solve the following problems: inadequate patient information for clinical decisions, inadequate evidence to guide diagnostic and treatment choices, and an overload of inadequate or incorrect information that leads to incorrect conclusions.
Using the metaphor of Ariadne’s thread, Dr. Rose explained that the Kaiser system in Colorado is more complex than just a CPR: It facilitates an entire process of efficient, effective health care. The thread connects clinical processes, documentation, guidelines, and reminders; it manages the coded vocabularies SNOMED and First Data Bank NDC. The system adheres to HL7 and TCP/IP standards to collect clinically accurate, comparable data, and permits an evolutionary path as medicine changes. The system is scalable, open in architecture, compatible with emerging Web technology, patient centered, and flexible across provider specialties and patient populations.
Kaiser’s paperless clinical information system includes an ambulatory CPR in 25 facilities, plus workstations in hospitals for continuity of care; three thousand users in a team care mode; 3,400 PC workstations in every exam room and office and 175 printers. All work is conducted online; medical records have been archived. More than 1 million medical records are held by the system, with 350,000 active charts. On one server, the system handles all ordering, results review, and desktop medicine. The system is managed remotely, with a single point-of-contact help desk.
The most difficult aspect of this system is that the process must fit into the care delivery service on a moment-to-moment basis. The goal is to improve the care process and care information for the provider, and to collect as much comparable data as possible to improve evidence-based medicine. The clinical system is based on controlled medical terminology and centered on the patient medical record. It offers simultaneous, multiclinician access.
In response to the Workgroup’s prepared questions, Dr. Rose asserted that the current state of data quality, integrity, comparability, and accuracy impedes analysis of valid outcomes, quality, performance, and research. Specific limitations include vocabulary, code sets that relate to vocabulary, billing requirements that do not document clinically relevant care, and cultural limitations.
He recommended that the government support a clinically accurate standard vocabulary, with SNOMED as the reference library; mandate representation in a Web-enabled form; and offer incentives for clinical accuracy. Inadequate PMRI poses serious barriers, dangers, and costs to the medical community and the public. The government should play an important role in developing standards and infrastructure, devising adequate and appropriate privacy legislation, promoting unique identifiers, and rewarding value rather than cost.
In addition to physicians’ declining disposable income, Dr. Arvary cited such barriers to adoption of CPR as lack of adequate expression and lack of typing skills. He further explained the benefits of a pen-based electronic record system. By contrast, Dr. Rose stated that the doctors in his group do not favor “mobile computing”; each exam room is equipped with a full PC workstation, and half the physicians complete the charting before they leave the exam room. Dr. Rose explained some advantages of SNOMED.
Dr. Rose asserted that billing systems have generated perversion of codes; a remedy is to enter clinical data once accurately and then use that information for such purposes as billing. He noted that CPR search engines currently are not strong enough, but that when the applications become less limited, physicians will use codified information without being aware they are doing so.
Dr. Ferrans noted that hidden communications costs exist in electronic systems. Dr. Rose discussed incentives for physicians to invest in CPR: tangible advantages for the physicians that are intrinsically rewarding. The success of his organization’s adoption is due to good training and good change management. He stated that the application service provider model will come into play, probably as subscriptions to a service and medical records maintained in a secure repository.
Dr. Arvary cited examples of value to a physician provided by CPR: when the physician encounters an unfamiliar disease and an Internet search provides information quickly, when fewer charts piling up reduce office stress, and when decisions with significant impact must be made in light of otherwise insufficient information.
Dr. Rose weighed the advantages and disadvantages of various vocabularies, including SNOMED. Drs. Arvary and Rose concurred that adequate knowledge and technology exist to maintain adequate security for medical records, and they discussed government versus private responsibility for security. Dr. Rose asserted that the government should be involved in selecting a controlled, common medical terminology, standard representation of data that is amenable to Web browsing, and a standard way to communicate securely. He asserted that when the infrastructure is in place, the medical industry will adopt CPR.
Kaiser Permanente’s successful implementation of CPR has been due to understanding the technical issue, the business, and its capabilities. The system has intrinsic value to the physicians, and the change was managed carefully, with adequate attention to adequate training and appropriate workload adjustments in the transition period. Financial incentives were offered to navigate the changes. Dr. Rose asserted that the eventual payoff for adopting the new technology is in the billions of dollars. Dr. Arvary noted that the environments of smaller practices must be considered in standards development.
Members revised a draft version of their interim letter to the Secretary that describes the Workgroup’s progress toward fulfilling its congressional mandate to produce a report on standards for PMRI by August 2000.
Betsy Humphreys, National Library of Medicine (NLM)
Ms. Humphreys described the Uniform Medical Language System (UMLS), which aims to overcome problems in information retrieval caused by disparities in language used by different users and in different types of information systems. She also described NLM’s meta-thesaurus. UMLS knowledge sources are free under a licensing agreement with NLM, but, Ms. Humpreys noted, UMLS neither compounds nor resolves licensing problems associated with different code sets. Regarding HIPAA, UMLS provides a uniform distribution mechanism for all standard codes (except NDC) and for all major vocabulary candidates. A 1996-97 test found that existing vocabulary systems incorporate most of the terminology needed to describe patient conditions and that a combination of systems is better than any single system.
Since 1997, Ms. Humphreys stated, progress has been made in eliminating duplication of effort, for example, in the collaborations between LOINC and SNOMED, and between CDC and LOINC. NLM has collaborated with the VA and Department of Defense to develop mechanisms to support ongoing maintenance. In the GCPR, LOINC is about to be named as a HIPAA standard in the claims attachment standard. A new meta-thesaurus contract will make funds available to federal agencies for small, related projects. Also since 1997, several tight feedback loops have been established between system builders and vocabulary developers, including LOINC and the National Cancer Institute.
Ms. Humphreys discussed the importance of supporting the development and maintenance of terminologies to permit access at relatively low cost, in support of the federal government’s need for good standard, comparable data.
Ms. Humphreys agreed with Dr. Cohn about the near-term importance of mapping to ensure the success of the entire effort, but pointed to difficulties that arise when new versions of vocabularies are introduced. She noted the opportunity afforded by GCPR to make decisions that affect many federal agencies, and for those decisions to be tested and used. She also lauded the international cooperation involved in the FDA’s standards work with ACRI on a unified device nomenclature system, and commented that the VA will soon serve as a laboratory for the integration of UMLS as a provider of source terminology.
Ms. Humphreys stated that clinical vocabularies should drive information systems. She observed that the level of greatest payoff should be determined by the degree to which controlled vocabulary is used in data elements in a record. She suggested as an R&D project the wedding of voice data entry with reference terminologies, plus an intelligence that sorts for personal vocabulary idiosyncracies. The participants discussed the vocabulary maintenance process, including the need for education on systems and provision of tools to enhance best use. NLM is focusing some efforts on natural language processing.
Dr. Kolodner discussed the incremental nature of implementation, the possible necessity to mandate taking certain small steps, and the possibility of pooling government resources to make the investment. Ms. Humphreys noted NLM’s interest in doing the mapping to help the private sector recover its costs, but a good model and a pool of users are lacking.
Asked to make specific recommendations, Ms. Humphreys discussed the role of government vis-a-vis next versions. She cited Kent Spackman’s work on a methodology to deliver change sets over time. She recommended moving to ICD-10-CM and stated that she could not envision a solution that did not include SNOMED. She once again recommended that the government provide support to the process.
Members of the Workgroup discussed managing their work plan to permit adequate time to develop their recommendations. Members suggested additional areas for information gathering: government-to-government efforts on medical device terminology standards, the GCPR project, an update on nursing terminologies, First Data Bank, business case for standards, business case for CPR, build-versus-buy decisions based on return on investment, hospital CIOs and other end users, CPRI’s survey of CHIM on business case issues, and CHIME or the Clinic Club or the Advisory Board Company for the CIO viewpoint.
Mr. Blair convened these hearings of the Workgroup on Computer-Based Patient Records of the National Committee on Vital and Health Statistics on September 16, 1999. He explained that the purpose of the group, pursuant to the mandate of the Health Insurance Portability and Accountability Act (HIPAA), is to study the issues related to the adoption and exchange of uniform data standards for patient medical record information (PMRI). The Workgroup, whose efforts are currently in the data-gathering stage, will report its recommendations to the Secretary of Health and Human Services (DHHS) by August 2000. Workgroup members and other participants introduced themselves.
Primarily for the benefit of the live Internet audience, Mr. Blair described a brief history of the Workgroup. In December 1998, focus areas were established: message format standards, medical terminologies, data quality, business-case issues, differences in state laws, and relationships with the emerging national health information infrastructure. The Workgroup held hearings in March, May, and June, and plans additional hearings at the present meeting and again in October. The first two panels of the day focused on areas of data quality and how quality affects uniform standards for PMRI.
William Jessee, M.D., President and Chief Executive Officer, Medical Group Management Association (MGMA)
Dr. Jessee explained that MGMA, the nation’s oldest and largest medical group practice organization, represents 211,000 physicians in 8,300 physician group practices. The association includes large practices, such as Ford, Mayo, and others; multispecialty clinics; small single-specialty practices; hospital-based clinics; academic practice plans; management services organizations; and physician practice management companies. In his presentation, Dr. Jessee outlined the issues, concerns, and expectations of medical group practices for CPR.
The key components of patient records are health status, health care, and administrative data. Data originates from multiple sources for different uses by many different users; data may take many forms, including text, voice, and image; and data cover a range of time spans. A comprehensive record should include general information including demographics, personal identification numbers, medical and genetic history, social and family history, and legal information; referral information, including physician information, diagnoses, insurance authorization information, appointment information, and follow-up notes; and diagnosis and treatment information, including history and physician findings, lab reports, medications, allergies and immunizations, treatment procedures, and follow-up. In order to enable assessment of effectiveness and efficiency, a CPR should link to patient financial information, including insurance data, charges procedure codes, procedure charges, authorized visits or procedures, number of visits left in a treatment protocol, and insurance expiration data.
Dr. Jessee described some limitations of paper records, including inaccessibility, illegibility, and nonportability. A barrier to the use of electronic records has been lack of standardization. Advantages to data standardization include the facilitation of uniform, and more accurate analysis, performance assessment, and full integration of clinical and administrative information.
From MGMA’s perspective, the area of clinical functions (chief complaint, social and family history, systems, physical examination) is a prime candidate for standardization, especially because of the great variability in the content of each component. Dr. Jessee cited the work of David Eddy on clinical logic as a possible approach to standardization. For the most common complaints or illnesses, algorithms can be constructed that depict the logical flow of the clinician’s evaluation and decision-making process. Upon construction of a clinical logic tree, a series of standardized queries can produce a standard record, while still accommodating variability in patient-specific differences. Logic trees can be converted to menu-driven tools to collect standardized data on most patients much of the time.
Dr. Jessee urged focusing on common conditions for which good evidence exists on appropriate management; an initial focus on this low-hanging fruit is appropriate before broadening the scope of CPR. CPR also permits incorporating real-time recommended treatment options, highlighting potential treatment conflicts, generating reminders for patient follow-up and prescription renewals, etc. Next steps include integration of access to AHCPR clinical guidelines through the National Guideline Clearinghouse. In electronically documenting the history of a patient with congenital heart failure, a clinician could hyperlink to the most current guidelines on care. Dr. Jessee considers this capability to be an essential feature in the return on investment in guidelines and standards.
A second prime area for standardization is administrative functions. The patient record drives the referral process; uniform records and standardized administrative data sets would facilitate that process and access to both referring physician and specialist. Linking patient records to resource use facilitates analysis of clinical processes and patient outcomes.
Standardization should permit consumers of medical services to be direct sources of input into their own records, the value of which has been documented in Shelley Greenfield’s self-report studies of diabetic patients. User-friendly interfaces and monitoring devices enable patients to maintain electronic log books, thus reducing staff workload and improving data quality. Patients should also be able to verify the accuracy of the data and to update administrative information. Dr. Jessee described the Internet-based ENM coding tool that produces a visit note; patients can access their own information on the Internet and hyperlink to related care information on managing the patient’s condition.
Research functions represent a third area in which standardization is essential: for individual medical practices seeking to benchmark performance, for larger research and public health analyses, and for management and administrative research purposes.
Dr. Jessee enumerated MGMA’s concerns about CPR development. Development of a standard data dictionary is a critical priority in which the government must play a leading role. Acknowledging its controversial nature and noting MGMA’s work in the area, Dr. Jessee urged the introduction of universal provider, employer, payer, and patient identifiers. He also urged the government to lead in developing standardized interchange protocols from external databases in order to pool data from multiple sources. CPR systems must be sufficiently flexible to allow for diverse practices and variations in service delivery among elements within organizations, and must integrate health information with data from external knowledge bases to apply rules-based, logic-driven clinical-decision support. Protection of the privacy and confidentiality of patient health data is a prime concern that systems can address by standards to authenticate user identity and authorization. Open technologies and architectures are essential to maximize access by authorized users. The cost of CPR is an important concern to Dr. Jessee’s constituency, and to many it is a major impediment with little or no perceived return on investment. A long, phased-in implementation approach may be indicated, and uniform data standards leading to industry interoperability may prove to be an incentive.
The federal government should serve as a focal point in standardization and in inclusion in the process of users of CPR and the authors of patient records. Proposed standards must be rigorously field reviewed and pilot tested before widespread adoption. The government should extend its influence to data exchange and data vocabulary standards, certification procedures and security standards, and to public education efforts on new standards. Federal funding of demonstration projects, incentives, and research grants would be useful ways to promote CPR adoption.
MGMA’s specific recommendations to the Workgroup include the following: (1) The federal government should lead in the development of standards for CPR content and format; (2) standards to ensure functionality of CPR systems should include data definitions and formats, data security, processing data in a uniform manner, and transmission; (3) the highest level of record confidentiality should be ensured without imposing onerous administrative burdens on practices, and no administrative simplification provisions should be enacted until enactment of comprehensive standards for health data privacy and security; (4) standardization should be prospective with a lengthy transition period; (5) the government should continue working with standards-development groups; (6) mechanisms to assure wide input to the standards development process from medical record users should be identified; and (7) education of the health care industry and public is critical on the need for standardization, implementation process, and security measures.
Alfred Buck, Executive Vice President, JCAHO
The Joint Commission is a national accreditor of health care entities. Dr. Buck highlighted points from his prepared testimony, including a proposed definition of a patient medical record: data or information documented and retrievable easily from one or more sources, hard copy or electronic, pertinent to an individual, used to describe and measure their health care status and health care interventions over time and across settings. In addition to the management of an individual’s health care, CPR can be used for organization management (with safeguards), process improvement and assessment, and research and education.
Dr. Buck listed issues that remain to be resolved, including confidentiality legislation; universal patient and provider identifiers; standardized chart format; standardization of data element fields and definitions; core data sets; refinements in the front sheet; universal billing form; aggregation rules; coding, coding audit methodologies, and methodologies to follow and aggregate reimbursements. Data quality issues to be resolved include data element definitions; data field determinations; technical standards; categories and domains of measurement, including reliability, validity, interpretability, collection, and calculation methods; determination of proprietary or nonproprietary status (the latter is preferred); defining categories and episodes of care coding and coding accuracy; core measurement sets; risk adjustment methodologies; and audit methodology, aggregation rules, cohort generation, edit standards, and management of missing data.
Dr. Buck urged that in individual patient records, narrative entries be addressed through standard vocabulary and terms, standard entry categorization, format rules, etc. Applicable standards and criteria for accountabilities, identification and validation of source, completeness, time of entry, and integrity must apply to data elements as well as to the narrative.
The roles of government, in partnership with the private sector, should be as: (1) facilitator in identifying significant activities and consensus development, and identifying and evaluating expert opinion, and (2) consolidator and regulator. Dr. Buck concurred with the previous testimony on the government’s role.
He asserted that the CPR and data quality issues must address and communicate aspects of quality and value at aggregate levels related to health care. Quality, costs, and improvement are key factors that define value. Data and data-based products must depict these aspects in meaningful ways. Data and data management infrastructure must be refined enough to allow the depiction of probabilities related to clinical outcomes resulting from various interventions, including preventive measures. Key components of an interim proxy definition for health care quality would involve compliance with science, consistency with wisdom-based processes, outcomes, goals, measures, customer satisfaction and perception, and oversight and accountability, including accreditation.
In summary, Dr. Buck stated that patient medical record initiatives are fundamental to the achievement of more substantial database resolution for health care challenges and refinements of policy; the initiatives include definition of health care benefits, structure of access, allocation of resources, and alignment of research and education.
Stephen Lamb, Vice President of Public Policy, National Committee for Quality Assurance (NCQA)
Mr. Lamb noted the collaboration between the Joint Commission and NCQA on the confidentiality program, stating his intention to provide the Workgroup with a copy of the report. Mr. Lamb also offered to share NCQA’s experiences on the importance of educating the media and the public on performance issues.
Mr. Lamb explained that NCQA has used Health Plan Employer Data and Information Set (HEDIS) measures to assess the quality of delivery of health care, but a limited information system infrastructure is proving to be a barrier in evaluating certain aspects of care. NCQA published The Information Systems Roadmap to alert the managed care industry to the types of information systems that will be required in the future. Mr. Lamb discussed the financial costs and scientific and health-care limitations encountered in today’s environment: the necessity to hire staff to abstract paper records onsite (as opposed to accessing electronic records), coding inaccuracy rates as high as 50 percent, and insufficient granularity in data to make meaningful decisions. Mr. Lamb stated that improved information systems and the CPR will help to minimize these deficiencies.
Mr. Lamb noted that HEDIS measures often drive the nation’s public health agenda. If an activity is measured, health plans focus resources on improving that activity. Mr. Lamb pointed out that some important measures, such as ACE inhibitor use, are not part of HEDIS. Thus, the information available to purchasers, consumers, providers, regulators, and others to allow them to assess the quality of health care and the service that is provided, is not as robust as it could be. The problems with getting good information are present at all levels in the system.
Dr. Buck underscored the importance of Congress resolving the confidentiality issue before the mandate of the Workgroup can be fulfilled. Dr. Jessee added that security is another important aspect of this issue. Consumer input is critical to avoid creating situations, such as that experienced in Maine, with more confidentiality restrictions than wanted or warranted. Mr. Lamb cautioned that the ability to identify individual member records is essential to close the quality improvement circle for audits and accreditation; if individuals opt out of the system altogether, problems are encountered. Dr. Buck added to the confidentiality discussion concerns about errors and adverse effects and tort reform that may emerge from engaging in performance improvement efforts.
Dr. Jessee contrasted the airline industry’s risk-free environment for reporting pilot errors with medicine’s litigious response to provider error. The movement of risk procedures from hospitals to outpatient facilities has provoked several states (e.g., California, New Jersey, and Florida) to regulate performance of procedures in office-based practices. He hopes to launch an initiative on patient safety in office-based practices that includes research, educational, communications, and advocacy components. With managed care’s new pressures on throughput in office-based practices, data capture has diminished and the likelihood of error increased. The challenge is to find ways for technology to alleviate administrative burdens. Dr. Fitzmaurice commended JCAHO and NCQA on their efforts to reduce errors in medicine.
On the issue of frustration due to lack of standards, Dr. Buck discussed the lack of consensus, rather than available methodology, plus lack of experience and data. Two themes that have emerged from safety analyses are insufficient staff training and insufficient staffing levels. An organization that can produce a rational, economically sound, constructive maintenance framework that supports constructive assignments, economic commitments to achieve appropriate staffing levels will be ahead of the game. Dr. Jessee noted problems due to lack of elapsed-time records in emergency rooms, difficulties in notifying patients taking a newly recalled drug, and disconnects related to specialists’ records included with primary care records.
Dr. Jessee stated that making a business case is a difficult proposition. Some practices (e.g., Intermountain) that have made large investments in information systems now are beginning to quantify return on investment, for example, in terms of reduced antibiotic use and therefore reduced patient risk and injury. It is not reasonable at this time to expect short-term returns on investment. If the value were increased by standardization, such as by enabling keeping better records and the pooling of data, outcome comparisons between practices would be possible. He noted the typical procrastination in computer purchase is the expectation of a faster-better-cheaper product to be put out on the day after the major purchase.
Dr. Fitzmaurice discussed the business-case scenario whereby incentives might be offered to physicians who are less productive if they spend additional time keeping better patient records. Dr. Buck asserted that early, fragmentary evidence shows that quality may pay insofar as it provides value to a variety of users, but cost is still a great consideration. Dr. Buck also urged that the concept of improvement be added to the equation of value equals quality over cost.
Mr. Lamb stated that NCQA is speaking with the American Medical Association about CPT codes. He noted that the employer community is applying pressure in a controversial move to accelerate implementation of a system that considers hybrid submissions to be nonreporting, in an effort to force investment at the health-plan level. He added that practitioners must coordinate their investment decisions with the multiple health plans they contract with. Without understanding the financial drivers, it will be difficult to achieve widespread implementation.
Herman Jenich, Associate Vice President for Managed Care, IPRO
Mr. Jenich stated that IPRO serves as the Medicare peer review organization for New York State, holds Medicaid oversight contracts in several states, and serves a variety of private customers. From the viewpoint of a data user, he discussed IPRO’s work in the managed care area, including HEDIS auditing, encounter data validation as required by states and the Health Care Financing Administration (HCFA), and IPRO’s clinical quality-improvement studies.
Mr. Jenich concurred on the need for comparable medical record information. He noted that in determining the meaningfulness of information, an acceptable amount of bias depends on the data user’s needs and on the validation techniques to detect differences among health care organizations’ data; validating at 5 percent, given that the current data infrastructure is a monumental task and expense, and sometimes it is based on judgment rather than hard data.
Mr. Jenich identified five categories of system limitations: clinical, membership, provider, and vendor data, and data integrity. Clinical data (e.g., claims and encounter data) are limited by proprietary coding schemes and the need for multiple coding schemes; modification of standard codes; proprietary forms that do not capture all relevant data; variation in data editing rigor; multiple, hard-to-integrate data systems; and incomplete data. Problems with membership data, although generally of good quality, include inconsistent patient identifiers that impede longitudinal record linkage and data captured in multiple systems, and inability to link members who transition from one population to another or dependents whose relationships to the primary subscriber change. Provider data is often housed in separate systems, provider identifiers are often different across health plans, and different sites of the same provider often use different provider identifiers. Vendor data refers to both the wide array of ancillary service providers (e.g., pharmacy, behavioral health, home health) and basic health plan functions, such as processing claims and encounters data which health plans may delegate to an outside entity. Problems in performance measurement arise when vendors do not collect data, the health plan does not collect the data from the vendor, or the formats of the two entities are too difficult to match.
Some health plans are unable to consolidate claims, encounter, membership, provider, and vendor data required for performance measurement, Mr. Jenich explained, particularly when they use different systems to track each data type and/or when they have recently undergone a merger, acquisition, or major system upgrade. Mr. Jenich described difficulties in getting data for HEDIS measures and in determining by the codes which diabetic patients should receive retinal exams.
Mr. Jenich recommended to the Workgroup that they address variations and limitations in the existing code schemes and the need for unique patient and provider identifiers for linking multiple databases; automation may perpetuate and accentuate existing shortcomings. He urged that any patient medical record system must consider a broad spectrum of potential users and noted that the success of a CPR standard rests on the commitment of individuals who enter the data. Clinicians and administrative staff will be most compliant with a system that provides timely information on their performance with comparisons to peer groups they believe are meaningful. The standards development process should stress training in coding and data entry, and guidance (i.e., national minimum standards) in editing and auditing electronic PMRI data. PMRI standards should be sufficiently flexible to account for changes in health care delivery and information system technology, including an institutionalized process to review and improve standards. CPR records should also provide for data security and patient confidentiality.
Mr. Jenich asserted that PMRI standards will help both purchasers and providers of health care to measure health plan performance, encourage the development of fully automated record keeping, and eliminate the need for costly abstraction from paper records to support performance measurement. Both government and the private sector must be active participants in the process of standards development and implementation. Mr. Blair noted that the government role remains to study PMRI and its exchange, not to develop standards.
Stanley Griffith, M.D., Indian Health Service (IHS)
Dr. Griffith presented a brief overview of the IHS’s health information systems, which have been used by IHS and major health plans in the private sector for 15 years; he referred to his written testimony for additional details.
Dr. Griffith discussed the importance of data quality in commission or avoidance of medical errors, noting that bad data leads to bad policy, which can have adverse effects on an individual’s health care. The IHS has gathered data quality information informally, through standardized audits, and in research studies on data quality.
Data quality problems take many forms, including erroneously recorded data, missing data, lack of specificity, unstructured data, data residing outside the system, and variations in data quality across sites. Data quality also varies among elements in a patient’s records. Quality in medications and vital signs is generally high; immunizations and Paps are good quality; patient education and smoking history, and cause and place of injury, are not good quality. Dr. Griffith stated that the greatest problem is with data that clinicians do not enter into the electronic system because of lack of consensus on coding systems, data dictionary, lexicons, and the full infrastructure spectrum of issues.
Dr. Griffith stated that data quality will improve when: (1) those who enter data also use the data; (2) all staff depend on the electronic record, rather than the hard copy; (3) data entry is not redundant; (4) data is more fully structured and codified in a consistent manner; (5) more elements of an organization depend on the same information; (6) efficient and accurate data entry is conducted at points of service; (7) robust and accurate voice recognition system permits data entry; and (8) jobs are designed to make it easier for staff to do their jobs well and data is collected as a byproduct. The theme of vested interests appears to drive the move toward better data.
Achieving the fully functional CPR can be the greatest step in improving data quality, which relates directly to improved quality of health care for individuals, improved public health care, more effective service delivery, improved financial well-being, and improved legal risk management. Dr. Griffith urged the federal government to identify which aspects of the necessary infrastructure are not adequately addressed by current development efforts: information models, coding systems, data dictionaries, communications, secure aspects, and fill in the blanks with public resources.
Mr. Jenich asserted the importance of capturing all the information available, particularly secondary diagnoses and procedures. He noted that behavioral health is difficult to capture because of confidentiality issues. Providers, states, and societies are encountering tighter patient confidentiality restrictions, making it increasingly difficult to do HEDIS measures. For example, identifying the correct people in a population group is difficult; sometimes it is hard to determine who is a diabetic.
Mr. Blair discussed the excessive burden on a clinician of improving specificity by searching references. Dr. Griffith asserted the need for intelligent systems that will code at the point of service, citing his own frustration with a 40-minute search for a CPT code for a routine coronary angiography. A fully functional CPR would gather data in codified fashion, do a review of systems, a physical examination, and a diagnosis with a taxonomy that describes level of risk and complication level of diagnosis. Systems, not clinicians, should be able to provide increased specificity.
Dr. Kolodner noted that a goal is to capture clinical data-level specificity relating to patient care; the system would elevate that information to a higher level of aggregation, perhaps with some interactive questions and answers. Mr. Jenich concurred that coding must be done at point of service and must be provider friendly. He asserted that the important issue is to facilitate what the coding system will be.
Dr. Griffith stated that IHS serves 1.5 million users in 500 facilities in 39 states, about half the number of the VA’s ambulatory users.
Gary J. Arvary, M.D., Skylands Medical Group
Dr. Arvary described factors leading to the managed care environment, which has created greater administrative overhead for the private-practice physician. He cited his own example; in 1995, his practice was 50 percent managed care, and he had an adequate amount of disposable income for technology, compared with today, when he is 95 percent managed care, including 60 percent capitated and the rest discounted fee for service. He asserted that physicians are apathetic towards CPR now because they cannot afford it.
AAFP studies showed that the number of primary care physicians who use the CPR has dropped from 4.3 percent in 1997 to 3.2 percent in spring 1999. Dr. Arvary noted that the negative influences are superseding the positive. On the other hand, he postulated, if all providers used CPR and entered structured data, insurance carriers, academic institutions, and the government would all want that structured data for accurate analysesas would providers, who would like structured problem lists and allergy and medication lists to help avoid errors. Better data is entered by physicians, in part because of definitional differences among providers, insurance carriers, academic institutions, and the federal government. At minimum, physicians also would like to have the problem, allergy, and medication lists electronically available to them at remote locations.
Dr. Arvary suggested that developing a new health information structure to substitute for the commonly used ICD and CPT codes results in lost and obsolete diagnoses, and difficulty in interpreting previous records; he suggested that these imperfect codes nevertheless be maintained.
Most physicians record their notes in the exam room during a 12-minute examination, a daunting task to be achieved with a desktop computer, but easier with a pen-based, slate-type computer. Physicians also care about the effect of technology on their staff.
Language is missing from CPR and cannot be duplicated with any structured device. He asserted that trying to make data dictionaries all inclusive is unnecessary and probably unworkable; he suggested making a small portion of it codable, with the rest unstructured. He also recommended a standard structured, coded medical vocabulary that remains the same; consistency will make it easier for physicians to remember the location of modifiers. In addition, removal of the scroll bar will speed up the process.
Dr. Arvary’s findings from a physician survey indicated that 92 percent prefer to use a pen-based computer and 7 percent a desktop. Handwritten medical records accurately reflect the physician interaction with the patient. A pen-based system would permit the addition of modifiers in minimal time. In a centralized information structure health information network, it would be possible to transfer structure and unstructured data through high band-width lines, a technology whose cost is declining.
On the basis of his anecdotal experience and on surveys, Dr. Arvary recommended the following: (1) Developers of software should be subsidized to standardize the record, using such techniques as a standardized (not all inclusive), paginated medical vocabulary easy to use with pen-based computers; (2) parts should be fragmented and transferred over analog phone lines, while the entire note can be transferred over high band-width lines; (3) physicians should be subsidized to purchase the systems to begin the health information infrastructure; and (4) the structure of the health information infrastructure should be regional to reflect the regional character of cultures in general, referrals, insurers, and Medicare. The network administrators, in the private sector, should be existing small groups and IPAs. Subsidies should be given to the use of communication lines.
Dr. Arvary sees such a network self-sufficient in up to 10 years as part of a national health information infrastructure with for-profit and not-for-profit segments. For-profit segments can offer directory services, website services, and Web development for their subscribers. He asserted that although providers are the largest portion of the health information infrastructure, data must be provided to insurers, academics, and the government for it to be of value.
Jeffrey Rose, M.D., Director, Clinical Information Systems, Kaiser Permanente, Colorado Region
Dr. Rose described the main problems facing health care, how Kaiser Permanente solved those problems with its system, how their system works, and how the system is based on standards. He then answered selected written questions previously posed by the Workgroup.
Dr. Rose asserted that between 50 and 66 percent of the potentially controllable rising costs in health care result from increases in the volume and intensity of medical services or the content of care that is determined by clinical decisions not by salaries, administration, research, Y2K, or fraud and abuse. Without information tools, clinicians cannot tame these costs. Dr. Rose quoted Albert Einstein, “We can’t solve problems by using the same kind of thinking we used when we created them.” The main problems are inadequate patient information for clinical decisions, inadequate evidence to guide diagnostic and treatment choices (and to maximize effectiveness), and paradoxical data toxicity, an overload of redundant, inaccurate, uninformative, or confusing facts that lead to incorrect conclusions.
In order to manage care and not just costs, Dr. Rose suggested, information and information technology are necessary. Using the metaphor of Ariadne’s thread (a string that would lead Theseus back through the labyrinth from the Minotaur), the character of this information system is more complex than just a CPR; it must facilitate a process of efficient care that is educational for both provider and patient, helps providers remember, and documents the legal record of care.
Dr. Rose stated that Kaiser Permanente in Colorado has built Ariadne’s thread in about eight years. The thread connects clinical processes, documentation, guidelines, and reminders; it manages the coded vocabularies SNOMED and First Data Bank NDC. The system adheres to HL7 and TCP/IP standards to collect clinically accurate, comparable data, and permits an evolutionary path as medicine changes. The system is scalable, open in architecture, compatible with emerging Web technology, patient centered, and flexible across provider specialties and patient populations.
Kaiser Permanente Rocky Mountain Division’s paperless clinical information system includes an ambulatory CPR in 25 facilities, plus workstations in hospitals for continuity of care; three thousand users in a team care mode; 3,400 PC workstations in every exam room and every office and 175 printers. All work is conducted online; medical records have been archived. More than 1 million medical records are held by the system, with 350,000 active charts. On one server, the system handles all ordering, results review, and desktop medicine. The system is managed remotely, with a single point-of-contact help desk.
The most difficult aspect of this revolutionary system is that the process must fit into the care delivery process on a moment-to-moment basis. The goal is to improve the care process and care information for the provider, and to collect as much comparable data as possible to improve evidence-based medicine.
Dr. Rose described the structure of the clinical system, which is based on controlled medical terminology, centered on the patient medical record, provides simultaneous multiclinician access, and offers the option of lock up and transfer to facilitate ease in sequential providers accessing a selected record. He presented use statistics, including 880,000 clinician accesses to the medical record monthly, 99 percent of visits charted, 84 percent SNOMED coded assessments. Everyone uses this system; there is no alternative. The system supports both structured and unstructured data (including dictation and free text).
At this point, Dr. Rose addressed the Workgroup’s prepared questions. He agreed that the current state of data quality, integrity, comparability, and accuracy impairs the ability to do valid outcome, quality, performance, and research studies. Specific limitations include vocabulary, code sets as they relate to vocabulary, billing requirements that do not document clinically relevant care, and cultural limitations. The private sector is not making satisfactory progress in data representation and messaging in the electronic world.
The government should do the following: (1) support a clinically accurate standard vocabulary; SNOMED should be chosen as the reference library; (2) mandate representation in a Web-enabled form, such as XML; and (3) move away from inadequate codes and toward clinically accurate terminology (pay us for being clinically accurate). The PMRI must be as comparable, precise, complete, and accurate as is reasonably possible. If this kind of PMRI is not achieved, Dr. Rose foresees continuing iatrogenic injuries that cost billions of dollars a year, inability to advance rapidly to evidence-based practice, ongoing and increasing waste and costs, continuing inappropriate use of pharmaceuticals and technology, uncontrollable quality, uncontrollable expense, uninformed decision making, and absence of value and quality-based health management (which should be the goal).
Vocabulary should be merged to a single reference model with common definitions and contextual meaning (SNOMED) for all reporting, reimbursement, research, and government-funded programs. A standard dictionary is essential. ICD and CPT do not permit collecting data that is clinically accurate and complete, but it is reasonable to establish a path and time frame for migration from current clinically inadequate coding sets to the more accurate one.
The federal government has an important role in developing standards and infrastructure, and in devising adequate, but not punitive, privacy legislation. The government must also recognize the importance of unique health record numbers. Government guidance to reward value rather than cost, to reward better rather than doing more, is essential to move to optimal-quality health care. Dr. Rose presented the Clinician’s Bill of Rights and Responsibilities.
Dr. Arvary described factors, in addition to physicians declining disposable income, that impede adoption of CPR: lack of adequate expression and physicians inadequate typing skills. He noted research done at Stanford and other places on pen-based computer systems; it remains for the government, insurers, and academic institutions to determine what information should be structured. Dr. Arvary concurred with Dr. Rose about the need for a standardized, codified medical vocabulary. It is possible now, with a pen-based or desktop computer, to use ICD or CPT or another reference terminology. Unstructured data, generally important only to the physician who enters it, can be added more quickly with a pen-based computer.
Dr. Rose added that doctors generally enter information largely derived from a controlled medical terminology that is natural for them to use. The doctors in his group do not favor mobile computing; each exam room is equipped with a full PC workstation because they prefer functionality and larger screens to portability. In Dr. Rose=s system, clinicians find it as easy to enter a rapid, legible, structured, codified note in our system as it is to write a sloppy, illegible one in the chart. Probably half the physicians finish the charting before they leave the exam room, Dr. Rose observed, but it seems to be a matter of style.
Dr. Rose discussed the need for the private sector to create new interactive techniques that use a reference terminology either directly or through mapping. SNOMED is both an application and a reference terminology. He noted the need for a standard to which an application or billing terminology is mapped. His organization chose SNOMED because it is a manageable, relatively atomic vocabulary with well-structured terms in good taxonomies. By contrast to SNOMED’s atomicity, ICD’s structure of a thousand terms in one statement that has a single code is not useful.
Dr. Cohn noted problems with administrative policy terminologies and with billing and payment policy. Dr. Rose stated that billing systems have generated perversion of codes. In order to remedy the situation, he asserted the need to enter data once accurately and then use that information for purposes such as billing. Dr. Arvary discussed the CPT-ICD situation, which would be 100 percent more accurate than the current situation, which is rife with clerical errors. He noted that CPR search engines currently are not strong enough.
Dr. Ferrans concurred that communication costs are a hidden part of ownership of electronic systems and identified as additional barriers not having a comparable PMRI user interface and lack of standards. He questioned what it would take for clinicians in the center of the CPR adoption bell curve to begin investing in a system. Dr. Rose stated that clinicians must feel there is something in it for them that is directly and immediately tangible, such as an application that people can be taught adequately, that is worthy of rewarding them, and that is itself rewarding. He credited the success of his organization’s success to good training and change management. The next generation of Internet offers solutions to issues related to cost of communication, privacy, and uniform representation of data. Dr. Rose believes the application service provider model will come into play, most likely as subscriptions to a service and medical records kept in a secure repository, controlled by either government or an entrepreneur.
Dr. Arvary cited several examples of the CPR/reference information system as providing real value to a physician; when he or she confronts an unfamiliar disease and an Internet search provides quick information; when fewer charts piling up reduce office stress. He urged that financial incentives be offered early. Dr. Rose added to the mix the stressful situation of making difficult, decisions with significant impact in light of insufficient information.
Dr. Rose urged moving away from ICD and CPT, and toward SNOMED-RT as a standard. He sees no disadvantage to licensing that kind of vocabulary, but does see value lost in time delays in getting to a standard. Dr. Rose characterized First Data Bank’s intention to merge the Medispan and First Data Bank vocabularies as difficult. He also noted that mandating NDC would be a mistake. Because both Multum and First Data Bank have a standardized vocabulary, both are able to send alerts on drug-drug interactions, reminders, and drug allergies.
In a discussion of privacy, Dr. Arvary concurred with Dr. Rose that adequate knowledge and technology exists now to maintain sufficient security for medical records. Breaches of privacy occur with the consent of patients who provide information to employers, insurers, or attorneys. Dr. Arvary stated that medical groups that establish Internet networks should handle a data repository, as should IPAs; the process should be handled by physicians who have something to lose if information is lost. Dr. Rose—noting that if a private company is subject to the same legislation that the government is subject to, the private company may do a better job than the government—asserted that it matters less who controls the repository than what are the rules that govern it. Dr. Ferrans stated that certain for-profit third parties may be trusted.
Dr. Rose explained his experience with SNOMED, which allows gathering of accurate, comparable information expected to be valuable in issuing alerts and reminders, and in performing outcome and meta-analysis of the data. In 1991, Kaiser selected SNOMED—the richest, most expressive, most natural and complete, unambiguous vocabulary available at the time, which was still inadequate (e.g., it did not include the words cancer or fever); over the years, they have improved it. SNOMED permits fast ordering and expression of assessments, complaints, and physical examination and other historical items in a comparable way. They have developed powerful search and display engines. Dr. Rose stated that they are collecting vast amounts of codified data that will help with further refinement.
Dr. Rose discussed the need for the government to be involved in developing a controlled, common medical terminology, a standard way of representing that data that is amenable to Web browsing, and a standard way to communicate securely. He concurred with Dr. Arvary on the need for basic infrastructure to manage change and the culture, providing applications by subscription service or other ways. He pointed to the need for access to the same information technology as patients. He noted the parallel between the creation of standards and then of the Internet, and of standards and the electrical industry, with the medical community; when standards are created, the medical industry will take it forward.
Dr. Rose responded to Mr. Blair’s question about what it took for Kaiser Permanente’s Rocky Mountain Division to implement its current system. He cited the importance of understanding the technical issue and especially the business and its capabilities. With that knowledge and the development of something of immediate and intrinsic value, the task is to manage cultural change. Physicians’ workload was cut in half for a month and they received extensive training. They were paid to participate in the design and then to learn the system. He echoed Dr. Arvary’s statement that not only must there be immediate reward for using the new system, but the clinician who is willing to adopt the new technology must be compensated; the payoff down the line is in the billions of dollars. Malpractice carriers lower premiums if doctors use CPR; perhaps there would be a similar reward for using SNOMED according to new standards.
Dr. Arvary acknowledged Kaiser Permanente’s accomplishments, but noted that in the Northeast, no big employer hires many physicians, and physicians make individual decisions about their practices. It is important, therefore, that whatever standards are agreed to must respond to the environment of the medical office.
Drs. Fitzmaurice and Kolodner coordinated an extensive, detailed discussion of revisions to a draft version of the Workgroup’s interim letter to Secretary Donna Shalala. Dr. Cohn noted that the draft letter agreed to by the Workgroup will be considered, possibly revised, and then adopted by NCVHS committee members. The letter describes the progress of the Workgroup toward fulfilling its Congressional mandate to produce a report on standards for PMRI by August 2000. The Workgroup completed its review of the draft letter on the following day.
The meeting was recessed and reconvened on September 17, 1999. Mr. Blair again called the meeting to order.
Ms. Humphreys discussed the NLM’s perspective on clinical terminology development, beginning in the 1960s with NLM’s pioneering thesaurus used in automated retrieval systems. NLM launched its Unified Medical Language System (UMLS) in 1986 and ran, with the Agency for Health Care Policy and Research, a vocabulary test in 1996-97.
The UMLS project aims to build machine-readable knowledge sources usable by various software products to (1) overcome problems in information retrieval caused by disparities in language used by different users and in different types of information systems, and (2) to determine which of many information sources is relevant to any particular question. NLM built intellectual middleware for system developers, including the meta-thesaurus.
The meta-thesaurus is a vehicle for building and maintaining maps between standard HIPAA codes and the more specific patient record, and for distributing information about which vocabularies are valid for specific segments of transactions. The meta-thesaurus incorporates many different biomedical terminologies, a specialist lexicon and lexical programs that deal with the syntactic aspects of language and variation, and a semantic network. The 1999 version includes 627,000 concepts and 1.1 million terms from around 50 source vocabularies, including all HIPAA code sets except NDC. UMLS knowledge sources, updated and distributed annually, are free under a license agreement with NLM. UMLS neither compounds nor resolves licensing problems associated with different code sets. Regarding HIPAA, UMLS provides a uniform distribution mechanism for all standard codes except NDC and for all major candidates for patient record vocabularies. Ms. Humphreys described the impact of the UMLS project on vocabulary matters.
The 1996-97 vocabulary test found that existing vocabulary systems incorporate most of the terminology needed to describe patient conditions; the combination of systems is better than any single system; and the test was organizationally and technically successful. The test found that the bulk of vocabulary creation for patient conditions had been accomplished; the model of the patient record determines the range of vocabulary needed; a workable strategy was needed to eliminate unnecessary duplication of effort among vocabulary developers to support ongoing maintenance enhancements of key complementary systems; and establishment of a closed feedback loop between clinical system builders and vocabulary developers to match the terminology with the data element layouts and model of patient record is needed.
Since 1997, progress has been made in eliminating duplication. Agreements have been reached between LOINC and SNOMED, CDC and LOINC, and CAP SNOMED and the National Health Service to merge SNOMED and REED. Ms. Humphreys noted that collaboration on the MEDRA would be beneficial.
Ms. Humphreys suggested that semi-automated or automatically assisted mapping between SNOMED and CPT for procedures and ICD-9-CM for diagnoses would be helpful. NLM recruited AMA, CAP, and NCHS, and the group has conducted preliminary work preparatory to a final draft and testing of the mapping and related decision rules. It is not clear when action will resume on this project.
NLM has collaborated with the VA and Department of Defense to develop mechanisms to support ongoing maintenance. This effort will fund the basic expansion, development, and distribution of LOINC for the next three years. LOINC, a chosen target coding system for a major project within the Government Computer-based Patient Record (GCPR) Project, is about to be named as a HIPAA standard in the claims attachment standard. An initial focus will be on defining a standard request submission format and devising a process open to public involvement and scrutiny. Labeled versions of LOINC, accessible on the Web, will help in resolving issues. The contract vehicle will permit federal agencies to seek assistance to move LOINC forward. The new contract for development and support of the meta-thesaurus back-end system also provides funding sufficient to assist federal agencies on small, related projects.
Since 1997, several tight feedback loops have been established between system builders and vocabulary developers, a situation likely to produce a better product for both sides. LOINC and the National Cancer Institute are among them; MEDRA’s pricing mechanism permits core service subscribers to submit requests for additional terminology. SNOMED is a looser example; development of major new versions complicates the process, especially in terms of maintenance and change management.
Ms. Humphreys discussed the importance of supporting the development and maintenance of terminologies to permit access at relatively low cost, particularly because of the federal government’s need for good standard, comparable data.
Ms. Humphreys concurred with Dr. Cohn that mapping in the near term is critical to ensure the success of all else that is happening, but that the movement to major new versions complicates the effort; a systematic process must be hammered out. She suggested that the federal government subsidize use of vocabularies important for government projects, but not before experience demonstrates their benefits and problems. Nevertheless, federal support for terminology development is justifiable based on the needs of just DHHS.
Ms. Humphreys lauded the GCPR’s efforts, especially because of the opportunity it affords to make decisions that affect many government agencies, and for these decisions to be tested and used. She reiterated that multiple vocabularies work better in concert than do single vocabularies. She also lauded the international collaboration inherent in the FDA’s work with ACRI on a unified device nomenclature system. An underlying reference terminology is essential, but different clinician groups may use different approaches. The meta-thesaurus provides the underpinning that permits groups to use their own terminologies. It appears as though the VA will soon serve as a lab for the integration of UMLS as a provider of source terminology. Dr. Griffith commented that GCPR currently focuses on clinical rather than administrative uses of data. Ms. Humphreys concurred with the testimony of Drs. Rose and Cimino that clinical vocabularies should drive information systems, and then administrative data be derived from that data. It should not be necessary to have separate intellectual processes to aggregate or categorize data.
Ms. Humphreys observed that the level of greatest payoff must be determined in terms of the degree to which controlled vocabulary is used in data elements in a record. The GCPR and other projects should identify where the apparent value is highest and gradually implement that level of control. She suggested as an R&D project the marrying of voice data entry with reference terminologies, plus an intelligence that sorts for personal vocabulary idiosyncracies. Dr. Cohn discussed the limitations imposed by the paper environment on the transition to electronic systems. Ms. Humphreys emphasized the need for encouragement and support for the maintenance and development of clinical terminologies that meet the needs of emerging systems, which is independent of the timing of getting to the point of mandating a particular standard.
Ms. Humphreys cited Mark Tuttle’s proposal to provide up-front support for vocabulary development to permit its distribution at lower costs and maintenance in a useful way, plus encouragement for use. Dr. Ferrans suggested requiring participants to contribute to the maintenance process if they were subsidized. Ms. Humphreys stated that this would require an infrastructure capable of responding to the input. She also called attention to the need for education on systems and to provide tools to make it difficult to submit incomplete information or not to find a term already in the vocabulary.
Ms. Humphreys said she will submit references on natural language support and noted that good information on that topic appears on NLM’s website. Researchers at NLM use UMLS knowledge sources plus experimental programs to determine the sense of paragraphs and sentences by determining semantic types using UMLS knowledge as natural language processing. For example, she noted, the suppressible synonym tag in the meta-thesaurus allows language processors to ignore or delete the more esoteric abbreviations or alternative names.
Dr. Kolodner discussed the incremental nature of implementation, the possible necessity to mandate taking certain small steps, and the possibility of pooling government resources to make the investment. Ms. Humphreys again cited the long-term benefit to government operations of subsidies. NLM has been interested in doing the mapping to help people try to recover the costs of what they are doing, but they don’t have a good model and they don’t have users. Mapping requires direct input by the people who most clearly understand the system.
Ms. Humphreys discussed the role of government vis-a-vis next versions. She cited Kent Spackman’s work on a methodology to deliver change sets over time. She recommended moving to ICD-10-CM and stated that she could not envision a solution that did not include SNOMED. She recommended that the government provide support to the process.
Dr. Cohn stated that because of his formal involvement in the CPT-5 process, he recuses himself from related discussions and decision making.
Members of the Workgroup continued revisions to the draft letter to the Secretary. A copy of the revised draft letter is to be circulated to the Workgroup.
Members of the Workgroup discussed managing their work plan to permit adequate time to develop their recommendations. Members suggested additional areas for information gathering: government-to-government efforts on medical device terminology standards, the GCPR project, an update on nursing terminologies, First Data Bank, business case for standards, business case for CPR, build-versus-buy decisions based on return on investment, hospital CIOs and other end users, CPRI’s survey of CHIM on business case issues (Ms. Amataykul will furnish a copy to the members), and CHIME or the Clinic Club or the Advisory Board Company for the CIO viewpoint. Ms. Amataykul will search for testimony on the business case, medical terminology, and other topics.
Mr. Blair requested members to inform him via e-mail of their availability for meetings in January and February. He thanked the members for their participation and adjourned the meeting.
I hereby certify that, to the best of my knowledge, the foregoing summary of minutes is accurate and complete.
/s/Jeffrey S. Blair 5/30/00