| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
infoRAD |
1 From the Office of Clinical Informatics (D.T., C.E.K.) and the Department of Radiology, Section of Information and Decision Sciences (C.E.K.), Medical College of Wisconsin, 9200 W Wisconsin Ave, Milwaukee, WI 53226-3522; and the Center for Information-enhanced Medicine, National University of Singapore (D.T.). Recipient of a Certificate of Merit award for an infoRAD exhibit at the 1998 RSNA scientific assembly. Received April 6, 1999; revision requested May 18 and received May 28; accepted June 4. Address reprint requests to C.E.K.
 |
Abstract |
|---|
 Top Abstract INTRODUCTIONBACKGROUNDAPPROACHESCONCLUSIONSReferences |
|---|
Index Terms: Computers • Internet • Radiology and radiologists
|
INTRODUCTION |
|---|
|
TopAbstractINTRODUCTIONBACKGROUNDAPPROACHESCONCLUSIONSReferences |
|---|
Appropriateness criteria are a specific type of clinical practice guideline. Unlike algorithmic (flowchart-type) guidelines, which specify one intervention at each step of the management process, appropriateness criteria list the various interventions that can be applied. Each intervention carries a score that indicates its appropriateness to the clinical setting. In this article, we present the background of computer-based appropriateness criteria and describe approaches for making appropriateness criteria more widely available and useful; these techniques are demonstrated by using a set of appropriateness criteria for radiology.
|
BACKGROUND |
|---|
|
TopAbstractINTRODUCTIONBACKGROUNDAPPROACHESCONCLUSIONSReferences |
|---|
Paper-based versus Computer-based Appropriateness Criteria
Although there are relatively few health care sites where practitioners routinely use computer-based practice guidelines, there is no doubt that such computer-based systems can have a positive effect on patient care. In a review of 28 clinical trials that assessed the effects of computer-based clinical decision support systems on clinician performance and patient outcome, Johnston and colleagues (5) found beneficial effects from systems that provided drug dose information, preventive care reminders, or recommendations for the care of active medical problems. Few of the studies showed significant improvements in patient outcome, but the authors concluded that further studies are needed to assess clinical effects on patients.
Not only can computers improve the use of clinical practice guidelines, but the task of translating these guidelines into computer-based formats can highlight deficiencies and lead to revisions that make the guidelines more useful (2). Failure to develop rigorous approaches to practice guideline construction prompted Shiffman and Greenes (6) to apply decision-table methods to verify guideline completeness and to detect inconsistencies and redundancies. A knowledge representation scheme that promotes completeness and minimizes inconsistencies and redundancies is essential if one wants to use and share guideline information in computer-based applications. Specifying appropriateness criteria with sufficient detail for computer use can be difficult because greater precision may be required than is typically found in paper descriptions (7). Owing to errors in design, reliance on "common knowledge," or a reluctance to be specific when scientific evidence is not available, guidelines may omit details of the patient medical history, fail to specify appropriate interventions for certain conditions, or contain specific clinical scenarios that are very limited in scope. Conversion of paper-based guidelines into a computer representation tends to reveal these flaws and to require that they be corrected. Tierney and colleagues (7) found it difficult to incorporate the appropriateness criteria published by the Agency for Health Care Policy and Research into their local clinical information system. These researchers recommended that appropriateness criteria be written in a simple if-then-else format, with all parameters defined strictly on the basis of routinely collected clinical data (7).
Information systems offer several advantages in the use of clinical practice guidelines. In contrast to paper presentations, computer-based practice guidelines can include additional didactic material, such as images, videos, sounds, simulations, and links to bibliographic databases. Given patient data—obtained either from the user in response to queries or through direct access to an electronic medical record—information systems can select the most appropriate intervention automatically (8,9). Some systems can function autonomously, for example, by providing alerts to warn clinicians of drug interactions or vaccination schedules (10,11). Although there are advantages to use of computer-based appropriateness criteria applications, there are also limitations: Such applications often are difficult to develop and update, have access to limited data, and may not express all the nuances and uncertainties embedded in natural language. Several groups have experimented with implementing appropriateness criteria applications, but only a few such applications have been shared among institutions (12).
Types and Formats of Appropriateness Criteria
The 1992 Institute of Medicine report (3) lists five types of appropriateness criteria, which are categorized by purpose. We use these categories in our discussion of appropriateness criteria types implemented by existing systems. The appropriateness criteria types and examples provided by the Institute of Medicine report are as follows:
1. Screening and prevention (eg, mammography).
2. Diagnosis and prediagnosis care of patients (eg, evaluation of acute chest pain).
3. Use of therapeutic procedures (eg, indications for carotid endarterectomy).
4. Use of specific technologies and tests as part of clinical care (eg, use of autologous or donor blood for transfusions).
5. Care of clinical conditions (eg, care of patients after placement of a coronary artery bypass graft).
Another way to characterize appropriateness criteria is by the format in which the knowledge is presented. The 1992 Institute of Medicine report (3) defines effective formatting as "presenting appropriateness criteria in physical arrangements or media that can be readily understood and applied by practitioners, patients, or other intended user groups." The report presents 16 sample appropriateness criteria that demonstrate a wide variety of formats including the most common—narrative text, tables, and flowcharts—as well as graphs, maps, photographs, lists, critical pathways, and if-then-else statements.
Narrative text, tables, and other formats that work well for human comprehension may not be well suited for automated systems. If the computer system is expected to provide patient-specific recommendations in a timely fashion and direct the user's attention to only the most relevant portions of the appropriateness criteria, formats that provide greater structure and explicit data representations are required. Thus, effective formatting for computer-based appropriateness criteria systems must be not only appropriate for the end user but also effective for the intermediate representation used by the computer software to manipulate appropriateness criteria knowledge and patient data.
Radiology Appropriateness Criteria
In 1993, the American College of Radiology (ACR) determined that there was a need for nationally accepted, scientifically based appropriateness criteria to assist radiologists and referring physicians and that a system needed to be developed for the creation of criteria for imaging and treatment decisions (13). The determination of such explicit criteria is important because radiologists exhibit only modest agreement in their assessments of the appropriateness of individual requests for imaging procedures (14).
The ACR had received multiple inquiries from radiologists, hospitals, and payers concerning the availability of such criteria. These contacts emphasized the need for the discipline of radiology to take a leadership role in criteria development. The ACR's Task Force on Appropriateness Criteria recognized that setting criteria would require use of broad-based consensus techniques because data from scientific outcome and technology assessment studies are usually insufficient for this purpose. Participation of physicians from other medical specialties was sought.
|
APPROACHES |
|---|
|
TopAbstractINTRODUCTIONBACKGROUNDAPPROACHESCONCLUSIONSReferences |
|---|
Knowledge Representation
The method of representing knowledge about the appropriateness of medical interventions is critical to successfully integrate such information into computer-based systems and to make it useful to physicians. The knowledge representation scheme must capture information accurately, reflect faithfully the original intent of the authors, and allow the widest possible application of the information.
The Appropriateness Criteria Model Encoding (ACME) language was created to address these goals (15). ACME is defined by using the Standard Generalized Markup Language (SGML) (International Standards Organization [ISO] 8879; 1986), an internationally accepted, vendor-neutral standard for document interchange (16). Owing to its portability, popularity, and generality, SGML provides an effective architecture for knowledge representation in computerized patient records and for interchange of structured medical data (17).
The SGML format dictates that a document has a declaration, a Document Type Definition (DTD), and a document instance. The declaration and DTD define a markup syntax that is used in the document instance to encode the information content of the document. The declaration specifies which characters can appear in the document and how these characters can be used to form words. A standard "Reference Concrete Syntax" is the default declaration. The DTD defines the names of allowed elements, how often an element may appear in a document, the order in which elements may appear, and which elements may appear in relation to other elements. The DTD also specifies the attributes of the elements and their allowed values. The document instance is a text file consisting of a series of elements, which are defined by opening tags (eg, <term>) and closing tags (eg, </term>). Tags are keywords enclosed in angle brackets. Document elements consist of the opening tag, the content of the element, and the closing tag.
The ACME language is used to encode a set of appropriateness criteria. ACME documents consist of three sections: condition definitions (conddefs), procedure definitions (procdefs), and term definitions (termdefs). More information about the ACME DTD is available on the Internet at http://www.mcw.edu/midas/isis.
The conddefs section defines the clinical conditions for which appropriateness criteria are available, the terms by which the conditions are indexed, and the variants of each condition. Each condition and variant has a unique identification number corresponding to the number assigned by the ACR (eg, GI-5.3), which is specified by the attribute id. The query elements specify the clinical context in which the variants of a condition are to be applied. For example, clinical condition GI-4, "blunt abdominal trauma in adults," is indexed by the term "blunt abdominal trauma." This condition has two variants: GI-4.1 (adult patient in stable condition) and GI-4.2 (adult patient in unstable condition). The "patient condition" query has two possible answers—"stable" and "unstable"—which allow discrimination of the two variants:
The procdefs section defines the imaging procedures that apply to each clinical condition or variant. procedure elements have the attributes score and technique. score indicates the appropriateness score of the procedure for the specified clinical setting. In those instances in which no consensus has been reached, the score is encoded as 0. The technique attribute describes the technical specifications for the procedure, such as "non-contrast."
The termdefs section defines the semantic relationships among the concepts used in the appropriateness criteria. The scope attribute stipulates that the term is narrower than, broader than, or equivalent to the external concept; equivalence is the default. Each term element delineates a concept. The procdefs section determines the semantic type (condition, variant, procedure, or query) of a term element used. Thus, an example of the termdefs indicates that "back" is affected by "low back pain" and has the synonym "spine." Each index term or procedure term can be linked with a concept from an external vocabulary such as the Unified Medical Language System (UMLS) Metathesaurus (18). The purpose of the UMLS is to improve the ability of computer programs to "understand" the meaning in user inquiries and to use this understanding to retrieve and integrate relevant machine-readable information for users.
The radiology appropriateness criteria were encoded into a text file in the ACME language. An SGML validating parser (SP 1.1.1; James Clark, London, England; http://www.jclark .com/sp) was used to ensure the conformity of the ACME language output file (with its DTD and document instance) to SGML standards. From the knowledge base server, appropriateness criteria for specified clinical conditions can be retrieved via the World Wide Web and viewed remotely. Physicians who have Web access can use familiar Web client programs (browsers) to access radiology appropriateness information online during patient consultation.
Knowledge Management Tools
The Network-based Editor for Ontologies (NEON) system uses the World Wide Web as a platform-independent user interface for viewing and editing an appropriateness criteria knowledge model or ontology (19). NEON enables developers of appropriateness criteria to edit the indexing terms and the semantic network that form the abstract model for a set of appropriateness criteria. Ontologies built by means of the system can be exported and imported by using the ACME language.
Developers work with NEON via Web-based interfaces. Although it is used in a client-server environment, all modifications to the knowledge base are recorded immediately on the central server. Therefore, the server always has an up-to-date master copy of the appropriateness criteria ontology, even if many developers are accessing the system at the same time. NEON currently does not include "version control" mechanisms to track the changes made to a database by multiple developers; however, such facilities are provided by many commercial database systems, on which NEON can be overlaid.
NEON is written in the Perl programming language (version 5) (20). Two basic components make up NEON: the ontology viewer and the ontology editor. NEON's database consists of four tables: conditions, terms (index), links, and procedures. The conditions table stores all recorded valid clinical conditions and variants. The terms table records the valid concepts that are used as an index. For example, the identification code GI-1.2 for a clinical condition serves as a term name that is used in indexing; the full name of the term is recorded in the conditions table. The links table contains the relationships between terms. The procedures table specifies the appropriate radiology procedures for each clinical diagnosis or variant.
NEON operates in viewer mode or editor mode. The ontology viewer displays the appropriateness criteria knowledge base online. One can enter a word or word fragment to begin the search. For example, the search string "lu" gives a list of terms that begin with those letters so that the user can select a term by clicking on the hyperlink. Searches are case-insensitive. The term "lung" has links to several other conditions, which are listed in alphabetic order by relation and concept (Fig 1). A physician can navigate through the appropriateness criteria knowledge base by following related terms. The NEON viewer does not allow one to alter the knowledge base in any way.
|
Retrieval and Display of Appropriateness Criteria
The Intelligent Selection of Imaging Studies (ISIS) system provides a mechanism for physicians to view the ACR appropriateness criteria and incorporate them into clinical practice. ISIS uses the ACR appropriateness criteria, encoded in the ACME language, as part of its knowledge base. At present, ISIS includes a Web-based interface that allows physicians to search the ACR appropriateness criteria by clinical condition. Because the knowledge base incorporates a semantic network of related terms, the search engine identifies clinical conditions on the basis of related concepts. For example, a search for the term "aorta" finds such conditions as "pulsatile abdominal mass" and "chronic chest pain: suspected aortic dissection." After a search term is entered, ISIS finds the clinical conditions related to that term (Fig 2). A table for each clinical condition (and variant, if applicable) displays the imaging procedures, appropriateness scores, and textual comments (Fig 3). The values of the appropriateness scores for recommended procedures are displayed quantitatively and qualitatively by using color-coded bars that range from green (most appropriate) to red (inappropriate) (Fig 4).
|
|
|
|
CONCLUSIONS |
|---|
|
TopAbstractINTRODUCTIONBACKGROUNDAPPROACHESCONCLUSIONSReferences |
|---|
There are several ways to improve radiology appropriateness criteria:
1. Representation of clinical conditions: Currently, conditional expressions in the clinical condition are expressed as text strings. For example, clinical conditions might be "patient is less than 80 years old," "patient has proven postmenopausal bleeding," or "pulsatile abdominal mass." A formal syntax for the representation of conditional expressions, such as that afforded by the Arden Syntax (23), is needed.
2. Representation of clinical concepts: Computer-based patient record systems contain information about medical concepts, such as diseases, symptoms, physical findings, surgical procedures, and laboratory tests. Such concepts are essential for the use of eligibility criteria, patient conditions, and radiology appropriateness criteria. Unfortunately, different institutions do not always use the same clinical vocabulary or coding system for concept representation. The ability to share practice guidelines depends on the adoption of clinical vocabulary standards.
3. Representation of temporal information: Complex expressions that show temporal trends, such as "rapidly increasing fever," are expressed as narrative text. Shahar and Musen (24) discussed the problem of detecting temporal patterns, such as direction and rate of change of laboratory test results, and defined a rich notation for specifying temporal intervals that allows multiple time lines and uncertainty in starting time, ending time, and duration.
4. Representation of uncertainty: There are several types of uncertainty that affect the patient history and appropriateness criteria. A physician may doubt the truth of a given medical statement. For example, a clinician might believe that a patient probably had a pulmonary embolism in the past on the basis of the patient history; however, in the absence of clinical evidence, the clinician cannot prove this suspicion. Data required by radiology appropriateness criteria to choose a clinical variant may be impossible to obtain. For example, a clinician cannot document that a patient has Creutzfeldt-Jakob disease because the patient is still alive and cannot undergo an autopsy.
The use of standardized approaches, as emphasized in this article, is important to ensure that appropriateness criteria reach the broadest possible audience and that such efforts can be incorporated easily into automated systems. If appropriateness criteria could be encoded in a common knowledge representation language and shared among institutions electronically, there would be several advantages. First, a repository of shared radiology appropriateness criteria would avoid duplication of effort among institutions that wish to use common appropriateness criteria. Second, a common electronic format would allow amendments and updates of appropriateness criteria to be distributed effectively and conveniently. Third, a common format would encourage the commercial development of application tools to help health care practitioners download and use radiology appropriateness criteria information online. Fourth, a standardized representation format would help developers of radiology appropriateness criteria reduce ambiguity and misrepresentation.
|
Footnotes |
|---|
|
References |
|---|
|
TopAbstractINTRODUCTIONBACKGROUNDAPPROACHESCONCLUSIONSReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  G. Levy, A. Blachar, L. Goldstein, I. Paz, S. Olsha, E. Atar, A. Goldberg, and Y. B. Dayan Nonradiologist Utilization of American College of Radiology Appropriateness Criteria in a Preauthorization Center for MRI Requests: Applicability and Effects Am. J. Roentgenol., October 1, 2006; 187(4): 855 - 858. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. H. Gottlieb Imaging for Whom: Patient or Physician? Am. J. Roentgenol., December 1, 2005; 185(6): 1399 - 1403. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Tigges, D. Sutherland, and B. J. Manaster Do Radiologists Use the American College of Radiology Musculoskeletal Appropriateness Criteria? Am. J. Roentgenol., August 1, 2000; 175(2): 545 - 547. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOGRAPHICS | RADIOLOGY | RSNA JOURNALS ONLINE |
