Electronic Health Record Meets Digital Library

Abstract

Linking the electronic health record to the digital library is a Web-era reformulation of the long-standing informatics goal of seamless integration of automated clinical data and relevant knowledge-based information to support informed decisions. The spread of the Internet, the development of the World Wide Web, and converging format standards for electronic health data and digital publications make effective linking increasingly feasible. Some existing systems link electronic health data and knowledge-based information in limited settings or limited ways. Yet many challenging informatics research problems remain to be solved before flexible and seamless linking becomes a reality and before systems become capable of delivering the specific piece of information needed at the time and place a decision must be made.

Connecting the electronic health record to the digital library also requires positive resolution of important policy issues, including health data privacy, government envouragement of high-speed communications, electronic intellectual property rights, and standards for health data and for digital libraries. Both the research problems and the policy issues should be important priorities for the field of medical informatics.

The idea of linking automated clinical data and knowledge-based information to support health care, research, and education gained prominence in the early 1980s when the vision of Integrated Advanced Information Management Systems (IAIMS) was first articulated.¹ The National Library of Medicine's Unified Medical Language System (UMLS) project was initiated in 1986 to facilitate the development of IAIMS systems capable of linking and integrating different types of machine-readable biomedical information, including patient records, the biomedical literature, factual databanks, and expert systems.²

As expected by those involved with IAIMS, UMLS, or both, the amount of useful patient data, clinical information, and biomedical knowledge in electronic form has increased dramatically since the 1980s. Also as anticipated, computers have continued to decrease in cost and size while increasing in speed, functionality, and ease of use. Not so generally predicted, but even more critical for the integration of the electronic health record and the digital library, the spread of the Internet and the development of the World Wide Web have simultaneously:

• Reduced the technical complexity of integrating access to disparate legacy systems developed on different technical platforms

• Provided de facto technical standards to which new electronic resources can be designed

• Made it possible for researchers, health professionals, and patients both to retrieve and to create information at home as well as in the library, office, clinic, and hospital

• Increased the perceived benefit of having a computer and an Internet connection

These developments make integrated access to computer-based health records and related knowledge-based information increasingly feasible. The rise of Web technology has also contributed to expansive definitions of both the electronic health record and the digital library. In fact, these two concepts now meet at the edges.

Electronic Health Record Defined

In a 1998 concept paper,³ the National Committee on Vital and Health Statistics (NCVHS) built on previous work by the American Medical informatics Association⁴ to describe three types of computer-based health records. In the NCVHS view, patient, personal, and population health records are needed to facilitate co-ordination, research, and assessment for clinical care and public health and to permit individuals to participate more actively in their own health care. Patient records “record clinical care and are used by delivery systems in which doctors, nurses, and other health professionals provide an array of hospital, primary care, and other ambulatory and institutional health services.”³ Electronic patient records are increasingly using multimedia formats,⁵ and they are being aggregated into clinical data repositories and warehouses by many medical centers.^6,7,8

Personal or consumer-oriented health records are “for individual use, including assessment of health status and linkage with physicians' records.”³ They may also include knowledge-based information, such as health education and disease management advice, that is specifically relevant to the person, or links to such information, as are currently available in several Web-based systems.⁹

Population health records are derived “from the health care system and have been made as non-identifiable as possible for public health and research applications. They may also incorporate survey data.”³ Population health records are used for monitoring public health and the outcomes of care and also for health services research, including quality management. In other words, the population health record envisioned by the NCVHS is a database derived from both health care and health survey data. The population health data sets now accessible via the Web generally do not combine the two types of data, and their interactive search capabilities are limited, in large part by the need to preserve confidentiality.^10,11

Digital Library Defined

The term “digital library” may have originated with the National Science Foundation in 1994, when the first opportunity for federal digital library grants was announced. The concept is not sharply defined, and librarians and computer scientists tend to emphasize different aspects of it. A good working definition was proposed by Clifford Lynch in 1995, that is, an electronic information access system that offers the user a coherent view of an organized, selected, and managed body of information.¹² The digital library focuses on information accessible via the Internet and encompasses the following materials:

• Materials converted to digital format, including retrospective print and non-print materials. NLM's Profiles in Science, which provides access to the papers of eminent biomedical scientists, is an interesting example.¹³

• Materials newly published in digital format but also published in print format. Many current bio-medical journals fall into this category.

• Materials published solely in digital format, including content that would not or could not be published in print format. To some, this category is the core of the digital library. Many solely electronic text documents can be printed without loss of content, but an increasing number incorporate multimedia components and large data sets that are not amenable to print. Other resources, such as the Visible Human¹⁴ and GenBank,¹⁵ are essentially unusable except in electronic format.

A digital library is not a single entity. It requires the technology to link different resources and assumes that the linkages between individual digital collections and information services will be transparent to the end user. Most existing systems have yet to achieve this transparency, but there are robust examples of usefully integrated digital collections, including some in biomedicine. The linked information resources and services of NLM's National Center for Biotechnology Information represent one of the best current instantiations of the digital library concept at a national level.

Given an expansive definition of the digital library, collections of data aggregated from individual health records, like the clinical data warehouse or the population health data set, can be viewed as part of the larger digital library needed to support biomedical research, education, and informed health care decisions. This is a useful view for institutions that develop and provide access to digital health data and to knowledge-based information. Regarding aggregated health data as part of the digital library not only opens up new funding opportunities but may also encourage fruitful multidisciplinary cooperation on problems common to knowledge-based information and aggregated health data, including permanent retention of electronic information or the need to implement variable user access privileges.

Requirements for Connecting the Electronic Health Record and the Digital Library

Although aggregations of health data can be profitably viewed as components of a digital health library, connecting individual electronic health records to other electronic information in the digital library remains a highly desirable goal. Such connections are useful both to find the universe of “patients like this one” and to facilitate access to scientific evidence, accepted treatment practices, and other decision support tools relevant to particular patients. Early in the UMLS project, three requirements for effective linking between disparate sources of machine-readable health information were identified—technical connections, organizational connections, and conceptual connections.¹⁶ Today, it is clear that a fourth requirement for linking the electronic health record and the digital library is public policy that permits the technical, organizational, and conceptual connections to be made.

One policy issue that casts a long shadow over the development of electronic health records and their effective connection to the digital library is health data privacy. In the United States, a lack of consensus about the legitimate uses of health data, and about the appropriate balance between these uses and a patient's right to confidentiality, has thus far stymied efforts to pass federal health data privacy legislation.

The absence of federal legislation increases public unease about the electronic storage and transmission of identifiable health information. The failure to enact U.S. health privacy legislation by Aug 21, 1999, triggered the Health Insurance Portability and Accountability Act's requirement that the Department of Health and Human Services issue privacy regulations that will have the force of law.¹⁷ Proposed regulations were issued in November 1999 and have elicited about 40,000 comments from interested parties, reflecting the degree of controversy surrounding public policy in this area. These comments must be analyzed and addressed before final federal privacy regulations can be issued.

In addition to privacy, policy issues such as government support for telecommunications, intellectual property rights, and standards also affect our ability to forge effective technical, organizational, and conceptual connections between the electronic health record and the digital library. These policy issues are discussed in the context of the particular type of connections to which they apply.

Technical Connections

The technical connections needed to link the electronic health record and the digital library include the computing equipment, telecommunications, platform-independent software, logon procedures, and access controls needed to ensure access to different information sources. These technical matters already have been greatly simplified by the spread of the Internet and the Web. It is safe to assume that general technical advances, not driven specifically by health or digital library applications, will continue to make it easier to connect the electronic health record and the digital library. Technical aspects of format standards for all types of information (including health records and electronic publications) will continue to converge and to increase in flexibility and functionality, as the current move to XML illustrates.^18,19 Technology will drive solutions to now-difficult problems, including efficient user access control (maybe via systems that combine biometric scanning and small portable devices with embedded computers) and clinical data entry (maybe via voice recognition systems that make use of intelligent thesauri). Cheaper mass storage is likely to be a significant part of the solution to the problem of making digital information permanently available.

Access to the Internet is increasing rapidly among the U.S. population, but ensuring that everyone who needs health information has appropriate information technology remains an important challenge. A range of federal programs has been designed to promote connectivity, including the NLM's Internet connections grants for health institutions.²⁰ After a slow start, the implementation of the universal service provisions of the Telecommunications Act of 1996 is now making access more affordable for some rural health care providers and libraries.²¹ Nonetheless, reliable high-speed access to the Internet remains an unsolved technical or economic problem in some geographic areas.²²

As yet, the public Internet cannot deliver the quality of service needed for many routine health applications. There is reason to hope that engineering developments will make the Next Generation Internet better able to meet the requirements of health care, public health, and biomedical research. The National Research Council's Computer Science and Telecommunications Board recently released a report of an NLM-commissioned study of the technical capabilities required to make the Next Generation Internet suitable for many health applications.²³ According to this report, health applications do not have unique technical requirements, but they do demand complex and variable combinations of technical capabilities and a higher degree of security than other types of applications. To ensure that health requirements are addressed in network engineering research and development, the biomedical community must increase its interaction with the Internet engineering community. The NLM has recently funded the implementation phase of a series of innovative medical applications of Next Generation Internet capabilities that may help achieve this goal.²⁴

Organizational Connections

Organizational connections are the mutual agreements both in and between institutions to provide or obtain access to information on different systems. Even in single organizations, linking systems from different departments or facilities can be difficult, although institutions that have embraced IAIMS (e.g., Vanderbilt University²⁵ and the Oregon IAIMS²⁶) have shown that it can be done. Lack of incentives for interorganizational cooperation is certainly one of the primary obstacles to forming aggregate health databases that cross institutional boundaries, as the case histories of many attempts to develop community health information networks illustrate.

Perhaps the biggest interorganizational problem currently affecting connections between individual health records and digital libraries is the difficulty of arranging access to electronic publications produced by many different organizations, including commercial companies and scientific societies. Fearing loss of revenue, many publishers wish to use restrictive licence agreements that in essence eliminate the “fair use” rights available under U.S. copyright law and guidelines. In addition, publishers' customary license agreements may not support remote access by faculty, students, and practitioners—let alone patients—in clinics, offices, and homes. Working directly with publishers or through third-party consolidators, individual libraries and consortia can be successful in negotiating arrangements that support the uses that are most valuable to their users, but the negotiations may be time-consuming and expensive. To date, most biomedical journals available in electronic form continue to be available in print. Often, access to the electronic version means that a library must obtain a print subscription as well. One way or another, providing access to electronic journals often costs libraries more than purchasing print journals, many of which are already staggeringly expensive.

Further complicating the licensing picture is the Uniform Computer Information Transactions Act (UCITA),²⁷ recently proposed by the National Conference of Commissioners on Uniform State Laws. This document, which is intended to provide a uniform commercial framework for the licensing of information in all 50 states, presents numerous potential problems related to shrink-wrap and click-on licenses, user rights under the Copyright Act, consumer protection issues, and transaction costs. As a proposed model state law, UCITA will be debated, adopted, modified, or rejected by individual state legislatures.

In the meantime, a major national policy debate about electronic intellectual property rights continues as society attempts to grapple with the impact of new technologies on the ability to reproduce, distribute, control, and publish information.²⁸ The 1998 Digital Millennium Copyright Act²⁹ ostensibly preserves fair use of electronic intellectual property, but under fairly murky conditions, which will probably be clarified in subsequent court cases. There is also a strong push to legislate new sui generis rights for databases, to protect the investment of those who have expended resources in aggregating facts or data that are themselves not covered by current copyright law.³⁰ Current law protects the entire compilation, but not individual facts in it. The range of data affected could be enormous, including stock quotes, phone numbers, baseball scores, gene sequences, names of procedures in coding systems, and clinical values in aggregated health databases.

Rising concerns about costs and restrictive access to electronic journals have led to serious efforts to develop alternative publication avenues. The Association of Research Libraries' Scholarly Publishing and Academic Resources Coalition (SPARC)³¹ seeks to promote the creation of lower-cost, peer-reviewed electronic journals to compete with specific commercial journals that are thought to be significantly overpriced. The National Institutes of Health have announced plans to host a free permanent repository of full-text scientific articles in the life sciences, called PubMed Central.³² (Earlier versions of this proposal were referred to as e-Biomed and e-Bioscience.) PubMed Central will be built by NLM's National Center for Biotechnology Information, with advice from an advisory board of nongovernment experts. The full-text articles it contains will be linked to their corresponding bibliographic citations in the PubMed database, which will be expanded to include any non-MEDLINE journals or articles deemed to be within the scope of PubMed Central. PubMed Central may become a primary dissemination mechanism for some existing and some new electronic journals. Since PubMed Central will permanently archive all the material it accepts, including high-resolution graphics, some publishers are expected to deposit articles on a delayed basis after their original electronic publication. As recent counter-moves by some publishers indicate, PubMed Central and other alternative electronic publishing and archiving mechanisms are bound to have an effect on scientific publishing.

Conceptual Connections

As any current Web user knows, technical connections and permission to use electronic resources mean little unless an appropriate conceptual connection is made between the user's information need and pertient content in different digital resources. The NLM's UMLS project is focused on enabling these conceptual connections. It builds intellectual “middleware”—electronic knowledge sources and related lexical programs—to help system developers build applications that can interpret user inquiries and find and integrate relevant information from different information sources.² The NLM continues to expand the UMLS products to improve their utility and to make them easier for system developers to use. Some success has been achieved, and the UMLS data and programs are now in production and research use in a wide variety of applications at NLM and elsewhere.^33,34 Many challenging informatics research questions remain to be answered, however, before flexible and seamless linking becomes a reality and before systems become capable of delivering the specific piece of information needed when and where it is needed.³⁵

Linking electronic health records and digital library resources will become easier when each resource of interest adheres to content conventions or standards, such as a defined set of data elements and the use of controlled vocabularies or classifications in some of these data elements. This is not to say that different types of resources will ever have the same set of content standards, just that it is easier if each follows some set of standards that can be factored into system development. Reliance on standards has certainly contributed to the success of the National Center for Biotechnology Information's efforts to integrate different genome resources.

A 1990 comparison³⁶ found considerable similarity in the types of content standards that existed for printed literature (which constituted the vast bulk of what was available in 1990) and those needed in the clinical arena. These types include:

• Standards for what constitutes a description of a particular entity or event, e.g., a report of an injury or a catalog record for a book

• Standards that specify the data elements to be included in various types of interinstitutional transactions, e.g., a laboratory test order or an interlibrary loan request

• Standards for allowable values for specific data elements, which might be any value present in a large formal coding system or controlled vocabulary, e.g., the International Classification of Diseases for a discharge diagnosis or the Medical Subject Headings (MeSH) for subject descriptors in a MEDLINE record.

Although the types of required standards are similar, the extent to which standardization had been achieved and the impetus for standardization in the health care and library arenas were very different in 1990, and they remain very different today.

Health Data Standards

In 1990, there was almost no standardization of clinical information and transactions, although de facto quasi-standard formats for some administrative and billing transactions were in wide use, primarily because of Medicare requirements. Today, there is substantially more standardization of the format and content of clinical transactions, particularly for laboratory tests and drugs, and a fair likelihood that standardization of other types of clinical information will increase. Factors that have contributed to the move toward standardization include the general trend toward electronic commerce, increasing automation in health care institutions, the spread of managed care, pressures to measure health care quality and outcomes, and heightened awareness of the potential importance of content standards, including controlled vocabularies. Some of these factors played into the enactment of the administrative simplification provisions of the Health Insurance Portability and Accountability Act of 1996 (HIPAA), which has had a significant effect on the development of content and messaging standards for both administrative and clinical data.

In addition to its privacy requirements, HIPAA's administrative simplification provisions instruct the Secretary of Health and Human Services to establish national standards for a variety of electronic administrative health transactions, for coding systems and unique identifiers to be used in these transactions, and for ensuring the security of electronic health data. Although HIPAA's focus is administrative transactions, the line between administrative and clinical data is not a sharp one, particularly for claims attachments, which can contain detailed clinical information. HIPAA also expanded the role of the National Committee on Vital and Health Statistics (NCVHS) to provide wide-ranging advice to the Secretary and the Congress on health data policy, and specifically charged the NCVHS to develop advice on any necessary federal action to promote complete electronic medical records by August 2000.³⁷

HIPAA's direct impact on clinical standards will come in the establishment of standards for claims attachments. The proposed rule for claims attachments will designate the HL7 (Health Level 7) clinical transactions standard³⁸ and LOINC (Logical Observation Identifiers, Names, and Codes)³⁹ as HIPAA standards. The expectation that LOINC will become a HIPAA standard, plus its recent selection as a target system for the exchange of laboratory test data between the Department of Defense and the Department of Veterans Affairs in the Government Computer-based Patient Record (GCPR) project, has prompted an interagency federal contract to support LOINC's ongoing maintenance and free electronic distribution. Federal agencies are also exploring government-wide licensing as a method of facilitating use of clinical vocabularies.

HIPAA has undoubtedly increased the pace of development of health data standards. The Act requires the government to consult with existing consensus standards groups and to select currently available standards whenever possible. The HIPAA standards selection process has encouraged existing standards groups to cooperate to complete or modify existing standards or to develop new ones to meet HIPAA requirements. Speculation that the recommendations made by the NCVHS in 2000 may lead to additional clinical data standards has also speeded development and fostered collaboration between previously competing groups. Although there is probably still a long, hard road to travel before a complete set of standards for electronic health data is achieved, its likelihood is much greater today than it was a decade ago.

Digital Library Standards

Meanwhile, the digital library has yet to achieve the level of standardization that applies to the organization, management, and sharing of printed materials. To state the obvious, there is substantial overlap in the printed materials acquired by different libraries, and considerable cost savings result if a cataloging or indexing record created by one library can be used by another. A relatively small number of large players, including the Library of Congress and the NLM, create and disseminate bibliographic records for significant segments of the printed literature. If local libraries use the same standards employed by these large players, locally produced records can be easily integrated with records obtained from outside sources. As a result, adoption of a standard by key producers or processors of bibliographic records for printed literature in essence guarantees the adoption of that standard by many other institutions.

As of 1999, the factors that promoted standardization in records for the printed literature do not yet apply to most components of the digital library. Although there is great interest, discussion, and exploration of the issues involved in describing, organizing, obtaining access to, and archiving electronic information, consensus on a reasonable strategy for accomplishing these functions has yet to emerge. Some promising collaborative tests are underway, but libraries are not yet generating or sharing large numbers of descriptive records for electronic resources. Among the many reasons for this are the rapidly evolving nature of electronic information sources; uncertainty about whether separate descriptive records (distinct from standard descriptive data embedded in the electronic resources) will be necessary in the long term; slow progress in developing efficient mechanisms for integrated searching of Web documents and records in large existing catalog databases; and unwillingness to expend resources describing information sources to which access cannot be guaranteed, either because licenses will become prohibitively expensive or the sources will just disappear from the Web.

Under these circumstances, it is not surprising that the development of standards for electronic information sources and the necessary harmonization of these standards with those for printed literature are both in their infancy. Given the NLM's mission to improve access to the results of research, it is fair to ask what the Library is doing to promote the standardization and cooperation that will improve current access to digital health information and ensure future access. The NLM's strategy is to use its own products and services as test-beds for technical and organizational approaches to organizing and managing digital information. The Library's goal is to gain experience from concrete experiments so that it can contribute to the development of workable national standards and strategies and also provide useful advice to other publishers of electronic information. Current NLM activities are focused on identifying the metadata and mark-up tagging needed for the NLM's own electronic publications and services, and on developing workable models for collaborative arrangements for organizing electronic resources, with MEDLINEplus,⁴⁰ the Library's new consumer health Web resource, as a test case for such collaboration.

The proposed Dublin Core set of metadata for electronic resources is the focus of considerable serious activity and publicity in the Web and library communities.⁴¹ As yet, it falls short of a reasonable standard because its extreme flexibility works against the creation of consistent metadata by different content producers. Like many other organizations, the NLM found it necessary to create additional metadata tags and to impose additional restrictions to promote consistency of application when adapting the Dublin Core for use in Profiles in Science.¹³ Metadata may also be a good way for Web publishers to announce their intentions regarding the permanence of the documents they are publishing. The NLM has initiated work to define and convey appropriate levels of permanence for the different types of electronic publications it produces and to develop the policy and procedures needed to guarantee those levels of permanence.

An important theme in discussions about organizing and archiving Web-based information is the need for persistent, unique identifiers for electronic resources that provide shorter, more permanent links than current URLs. Of the several efforts to establish an infrastructure to support assignment and look-up of permanent unique identifiers for digital entities, the Digital Objective Identifier (DOI) initiative may be the most prominent.⁴² Neither the DOI nor any other system has yet achieved broad acceptance, and today it appears likely that there will be several different systems for different types of digital objects, including some built on existing national bibliographic systems for print publications.

The success of PubMed and its approach to linking MEDLINE citations to the full text of articles available on publishers' Web sites has alerted scientific publishers to something that medical librarians have known for years. The NLM's article citation numbers are permanent unique identifiers for journal articles, and MEDLINE itself is a publicly accessible directory that includes the unique identifier, a standard description of the article to which the identifier refers and, in an increasing number of cases, the URL for the full text of the article. Many biomedical publishers now use NLM identifiers to link journal article references in their electronic publications to the corresponding MEDLINE records and, if available, to the electronic full text of the articles. The Library's formats for SGML/XML submissions of citations, abstracts, and now full text for PubMed Central probably also have a standardizing effect on electronic journal publishing.

A New Environment for Achieving an Old Goal

“Electronic health record meets digital library” is the Web-era reformulation of the long-standing informatics goal of seamless integration of automated clinical data and relevant knowledge-based information to allow informed decisions. When this goal first gained currency, the assumption was that health care professionals were the decision makers. Clinicians remain a primary target for integrated “just-in-time” information services, but these services are also needed by public health professionals and, in an era in which individuals are assuming more responsibility for their health, by patients and the well public.

Technical developments during the last decade have set the stage for achieving rapid delivery of specific knowledge relevant to an individual person or community at the time and place of need. Systems that can link electronic health data and knowledge-based information in limited settings or limited ways are here today. Their existence partly explains the increased public interest in resolving relevant policy issues, including health data privacy, the “digital divide,” electronic intellectual property rights, and standards for health data and for electronic publishing. While the environment is highly favorable for progress, many important informatics research problems remain to be solved before electronic health records and digital library resources are linked “seamlessly” to support decisions and actions that improve individual and community health. Solutions are likely to come from the work of multidisciplinary teams of health professionals, medical informaticians, librarians, computer scientists, linguists, and behavioral scientists and from continuing serious effort to achieve positive resolution of policy issues. These are worthy priorities for the beginning of this millennium.