|
|
Medical Assistant on the World Wide Web: A Prototype for an International Medical Database
在www上的医学辅助：一个国际化的医学数据库原型
Project Proposal
Rajesh C. Dash, M.D., John Haldi, Alan D. Proia, M.D., Ph.D
Duke University Medical Center, Durham NC
INTRODUCTION
Easy and reliable access to medical records is a necessity for patient care. Unfortunately, even within a single institution, such access is often suboptimal. Paper-based records are often damaged, misplaced, or lost entirely. While computer-based medical records offer a vast improvement in both reliability and access, many information systems in place today limit access to an individual department within a single institution. In the past few years, implementation of an integrated electronic patient medical record has been in the spotlight in the informatics community. Such a record would ideally encompass all information pertinent to patient care including diverse topics such as clinical history, radiology, laboratory data, and anatomic pathology and make such information available to all members of a health care system. 1
A step beyond the integrated electronic medical record, is the concept of an international integrated medical database. Such a system would be multi-institutional (or perhaps "non-institutional" is a better descriptor). A centralized international medical database would provide easy, reliable, and complete access to medical records regardless of location. It would also provide a revolutionary research tool for physicians and scientists by allowing both unparalleled access to "live" medical information and an invaluable resource for outcomes analysis.
The "Medical Assistant on the World Wide Web" project, while focusing on the field of autopsy pathology, will be a prototype for a full-fledged international medical database and will leverage the latest, cutting-edge Internet technologies. The breadth of this project was narrowed to include only autopsy pathology for several reasons. First, a streamlined, working model for autopsy data-entry and report formatting is already in place at Duke University Medical Center using a custom software tool called "Autopsy Assistant." Second, the scope of the project required limitation to allow completion within a reasonable timeframe. Third, masking patient identifiers in autopsy data is somewhat easier due to the structured nature of the autopsy report. Finally, autopsy data inherently provides an excellent source for rigorous outcomes analysis.
The proposed system will facilitate data entry through robust use of templates and automated SNOMED coding. Data analysis and data-mining tools will enhance retrieval and manipulation of entered data. Finally, rich dynamic HTML formatting (an Internet Web document format) will provide professional and flexible reporting of data, including multimedia content. Designed to be open for use by any institution around the globe with Internet access, "Medical Assistant on the World Wide Web" will provide a glimpse into the future of medical information systems.
PRIMARY GOAL
The primary goal of this project is to provide a centralized repository of autopsy data of an international scope. The benefits of such a system are numerous. Such a vast collection of data, unlimited by geographic boundaries, will provide a truly comprehensive view of autopsy pathology. The study of disease processes as they vary from locale to locale and from one point in time to the next should become a simple, almost automated task. The time required for sophisticated retrospective retrieval of dozens or even hundreds of cases of interest should be reduced to only a matter of minutes. Perhaps most importantly, collaborative studies spanning countries and continents will hopefully not only become more feasible but commonplace.
Successful interfacing of distributed, heterogeneous information systems using the Internet and the World Wide Web has been accomplished and described in the literature.2-4 While this project will leverage some of the same technologies and standards used in earlier projects, the goal and focus of this project is entirely different. Instead of enabling cross-communication of data between disparate, legacy information systems, the proposed system will centralize the data, allowing sophisticated analysis and providing universal access. Many standards for information encoding, storage, and communication are already in place today. These standards have matured over time and provide an elegant and sturdy foundation for an information system that is being designed "from the ground up". Existing open standards will be implemented for information encoding (Systematized Nomenclature of Medicine, SNOMED), image communication (Digital Image Communication Of Medicine, DICOM), and information system intercommunication (Health Level 7, HL7).5
DESIGN
Data Entry and Access
The design and implementation of such a system requires careful planning and execution. To meet the primary goal, four subgoals have been identified. The first subgoal is to provide a means for data access and entry. Use of a low-cost, ubiquitous technology would be ideal. The Internet and World Wide Web provide a consummate solution as most medical centers in developed countries already have this technology in place today. While the problem of universal access may be solved, use of the Internet and web-based technology raises additional issues. Foremost among them is the degree of security afforded over a publicly accessible network such as the Internet. The topic of web-based security and security in general will be addressed separately.
Once data access is available, efficient data entry must be permitted. Use of a computerized input system can greatly facilitate entry of autopsy data through the use of templates and "speed phrases." Templates provide pre-typed detailed descriptions for commonly occurring disease states, using a "fill-in-the-blank" approach to enter only the information specific to the current case. For example, part of a template for a description of the bone marrow may be predefined as "The bone marrow is (hyper/hypo)cellular overall averaging approximately ____% cellularity. There is (no / a mild / a moderate / a severe) (increase / decrease) in the myeloid to erythroid ratio which is approximately _:_ …". While the primary benefit of templates is the increased efficiency in which data is entered, use of templates also promotes systematic evaluation of issues pertinent to the particular topic. In addition, a standard presentation format inherently results from the use of templates.
"Speed phrases" are another mechanism by which data is more efficiently entered. Using this scheme, a short alphanumeric code, followed by a special command key results in expansion of that code into predefined text that corresponds to the entered code. For example, entry of "ned1" followed by the special command key would result in the automatic entry of the phrase "there is no evidence of dysplasia or carcinoma".
Data Storage, Retrieval and Analysis
Once a user has entered the data, it must be stored in such a manner that it is easily retrievable. To accomplish this objective, the SNOMED classification system will be employed to index textual information. SNOMED employs multiple codes spanning multiple categorical axes to encode a single medical concept. For example to encode the diagnosis of "squamous cell carcinoma of the right lung", separate codes are used to indicate topography or location (a T code), general morphology (an M code), and actual diagnosis (a D code). The efficiency with which SNOMED encodes medically related information and allows accurate and precise retrieval of information has been well documented. 6 Classification-based encoding systems have fared quite poorly in this regard.
Once SNOMED encoding is complete, the entered data along with the associated SNOMED codes are stored in a relational (SQL) database. Storage of image data requires special attention due to the inherently large storage requirements for digital image capture. The field of anatomic pathology (as well as a number of other medical fields) revolves around recognition of images and visual patterns as they provide a snapshot of a disease process. Standards for image data representation and communication have been developed (DICOM) and their application will be explored. Encompassing visual information is crucial to an anatomic pathology database. However, due to the tremendous storage requirements of image data, linking of externally stored image information to the database will be promoted rather than actual storing of image data within the central database. This will undoubtedly result in increased maintenance to ensure data integrity and performance degradation when image retrieval and manipulation is required. However, these deficits will be more than offset by the performance improvement in retrieval of non-image data and the data storage resources saved.
Retrieval of information will be allowed through multiple avenues. Retrieval of information based strictly on SNOMED coding will result in superior performance. If SNOMED coding fails to adequately capture the needed information, free-text searching will be permitted. In addition, Boolean operators such as AND, OR, and NOT and "wildcard" characters matching various data patterns will be available to generate sophisticated search queries.
The volumes of information that this system is designed to encode must be analyzed to be useful. Data-mining tools will be employed to automate this process. Automated data-mining tools have the potential to sift through tremendous quantities of information to discover new, previously unrecognized patterns of data associated with specific disease processes.7 A number of diverse technologies have been employed for such purposes in the past. These technologies include expert or rule-based systems, neural networks, and fuzzy logic. The effectiveness of various artificial intelligence motifs in the automated analysis and retrieval of data will be a notable and exciting area of research in this project.
Data Formatting, Reporting, Import and Export
The third aspect of the proposed project will involve presentation of the archived data to the outside world. The World Wide Web employs a document formatting language entitled HyperText Markup Language (HTML) for this purpose. In the past few years, this language has evolved from a static, simplistic and crude modality for data presentation into a dynamic, precise and robust one. The advent of enhanced server-side programmability and the development of client-side "Dynamic HTML" has dramatically increased the degree of control and flexibility with which web pages can be designed and presented to the end user.
Communication of data to the end user is just one part of the picture. The proposed system should have the facility to communicate with other computer systems. In particular, import and export of data from legacy information systems must be supported. The Health-Level-7 (HL7) communication protocol was designed for such a purpose. HL7 provides a standard by which disparate information systems may communicate using an array of standardized messaging protocols. "Medical Assistant on the World Wide Web" will utilize an HL7 interface to retrieve stored autopsy data from legacy information systems. This will allow rapid accumulation of data and immediate availability of a usable knowledge base. If possible, the system will attempt to store updated and newly entered data back into an institution’s information system.
Security
The final and perhaps most important aspect to consider for this project entails security. General security of data on an individual, institutional, and multi-institutional level needs to be considered in a project of this nature and scope. The security risks in a global information system have been elaborated in the literature and need to be addressed.8 On entry to the system, three key pieces of identification data will be required. These will include the user identification code, the institution identification code, and finally the password. Patient identification data will be masked from all users not a staff member of any given institution. Medical information, stripped of patient identification will only be made "public" to other physicians licensed to use the system. Furthermore, data will only be made available once the attending physician has consented that this should occur. Patient data will be present in a read-only format for staff located within the same institution but for whom editing is not required (for example, laboratory technicians, department secretaries, or nurses). Full access will only be granted to the attending and resident physicians on the case. Case data can only be verified or "signed-out" by the attending physician on the case. An intelligent audit trail will be generated for accesses to patient data that the system deems suspicious or out of the ordinary. These audit trails will be reported to the respective attending physicians on a periodic basis, and then purged from the system.
Internet and Web-based security are already somewhat advanced due to the prevalence of on-line commerce and the security needed for its use. The Internet is also being used more commonly as a "private virtual network" whereby the existing network hardware infrastructure is utilized for communication while heavy data encryption is employed to make that data undecipherable except by the intended recipient. The proposed project will leverage existing Internet security and encryption technologies, including secure socket layers and client certificates.
CONCLUSION
In summary the proposed project entails the design, development, and implementation of a centralized international autopsy database. Data entry will be facilitated through robust use of templates and automated SNOMED coding. Data analysis and intelligent data-mining tools will enhance retrieval and manipulation of entered data. Finally, rich dynamic HTML formatting will provide professional and flexible reporting of data, including multimedia content. Existing open standards will be implemented for information encoding (SNOMED), image communication (DICOM), and information system intercommunication (HL7).
There are a number of issues not touched upon in this document that will require further attention. For example, the question of data ownership arises. Who should be recognized as the owner of any particular data element and who should be credited for publications based on data spanning dozens of institutions? Are there security measures that can be employed to ensure that appropriate recognition is awarded to those individuals and institutions that contribute data to the centralized database?
In many ways this project redefines the way in which electronic information is viewed, processed and accessed by the health care community. It is hoped that it will result in more efficient data entry, storage, retrieval, and analysis and promote collaborative efforts between institutions and individuals that otherwise would not be feasible. "Medical Assistant on the World Wide Web" is a prototype for an international centralized medical database and will provide a glimpse into the future of medical information systems.
REFERENCES
1. McDonald CJ. The barriers to electronic medical record systems and how to overcome them. J.Am.Med.Inform.Assoc. 1997" 4:213-221.
2. van Wingerde, J, Schindler J, Kilbridge P, Szolovits P, Safran C, et al. Using HL7 and the World Wide Web for unifying patient data from remote databases. Proc.AMIA Annu.Fall Symp. 1996" 643-647.
3. Wang K, van WF, Bradshaw K, Szolovits P, Kohane I. A Java-based multi-institutional medical information retrieval system. Proc.AMIA Annu.Fall Symp. 1997" 538-542.
4. Halamka JD, Safran C. Virtual consolidation of Boston\'s Beth Israel and New England Deaconess Hospitals via the World Wide Web. Proc.AMIA Annu.Fall Symp. 1997" 349-353.
5. Aller RD. Software standards and the laboratory information system. [Review] [15 refs]. Am.J.Clin.Pathol. 1996" 105:S48-S53
6. Lussier YA, Bourque M. Comparing SNOMED and ICPC retrieval accuracies using relational database models. Proc.AMIA Annu.Fall Symp. 1997" 514-518.
7. McDonald JM, Brossette S, Moser SA. Pathology information systems: data mining leads to knowledge discovery. Arch.Pathol.Lab.Med. 1998" 122:409-411.
8. Waegemann CP. IT security: developing a response to increasing risks. Int.J.Biomed.Comput. 1996" 43:5-8.
来源：https://tissue.mc.duke.edu/maw3/docs/wpaper.html
|
|
发表于 2002-7-16 11:05:10