Katie S. Oliver (Programmer Analyst, Virginia Neurological Institute) Christine Remenyik (Graduate Research Assistant, University of Virginia) Alexander Crosby (Computer Systems Engineer, Virginia Neurological Institute) Jim Kazmer (Project Manager, Hewlett-Packard)
Hypertext servers were developed to access HTML documents located in file systems. A natural extension to hypertext servers is to enable access to dynamic data in addition to static files; intelligent hypertext servers must be able to provide access to heterogenous information sources (See Figure 1). The server will oversee the location and object type of all information available in order to generate the information domain. Intelligent hypertext servers no longer simply locate and retrieve existing files; rather, all data available on a given topic, regardless of location, format, database, or platform, is integrated and hyperlinked together to form an appropriate hierarchy from the information domain.
When gathering data, the intelligent hypertext server creates hyperlinks to important areas of the information domain creating a hypertext hierarchy of all information. This hierarchy simplifies access to data that would otherwise require many steps to reach and provides direct links to more detailed data on the current subject. The server creates an HTML document that includes this navigational information as hyperlinks prepended to the requested data. The server will also include links to manage the presentation of additional relevant data. Providing new tags for HTML to encode this additional information and modifying hypertext browsers to utilize this information, combined with the new intelligent hypertext servers, will promote the development of more sophisticated data-access applications.
The extra navigational and display management information will be encoded in new HTML tags that identify the organization of the information domain. These tags will be used by the browser and will separate the navigational and display management information from the document. Extending the current HTML will facilitate the implementation of this additional navigational and display management information by browsers.
This paper introduces HTML extensions and a browser that utilizes the extra information. The following sections describe the need to develop a new generation of browsers, present the extensions to HTML, present a dynamic hypertext browser that interprets the HTML extensions, describe a practical application utilizing these enhancements, and conclude with a summary and discussion of future work.
The traditional static hypertext file structure is an overview page with links to more detailed information. The user selects a link from the overview page, follows succeeding links, and transverses the information domain downwards. To pursue an alternative branch, the user must backtrack through previous documents to select a different link. This can be time consuming and confusing. Also, important sections of the information domain must either be repeated in each document displayed or forgotten; the hypertext browser can not retain links regarding important sections of the information domain without repeating the information in each document.
Navigation through hypertext documents relies on the placement of hyperlinks within the document. Requiring the combination of both data and hyperlinks to be displayed in a single window often obscures a clear presentation of information. The list of hyperlinks to relevant sections of a user's information domain, often found at the beginning of documents, aside from wasting valuable screen space, forces the desired information out of view. The user is required to choose which information to view: the hyperlinks or the document.
Finally, the ability for WWW hypertext browsers to control the synchronization of simultaneously displayed external windows does not exist. When information on a particular topic is retrieved from various databases, the user can not view all the data simultaneously in multiple windows with the guarantee that all windows contain information related to the desired topic. Should a user wish to view related pieces of information in two windows, for example, the user must clone the first window and navigate to the desired document in the second window. Display management mechanisms do not exist to prevent the display of unrelated information in the extra windows. Relying on the user to coordinate the presentation of information displayed across multiple windows introduces a potential danger when decisions are based on the information displayed.
Allowing HTML documents to bind the dynamic buttons permits an intelligent hypertext server (the creator of the HTML document) to moderate the buttons' functionalities and thus customize the browser. Bindings are contained within any document containing the new HTML tag
Similar to traditional hyperlinks, navigational buttons, when pressed, display the document located at the URL bound to the button. These buttons permit the removal of hyperlinks from the text window and allow links to remain visible while not interfering with the display of important data. The user is able to clearly see both navigation information (the buttons on the interface) and information content (the hypertext document).
Hypertext document developers often place a list of hyperlinks to related information near the beginning of each document inevitably wasting screen space, forcing the desired data out of view, and requiring the user to choose to view either the hyperlinks or the document. To prevent the user from forgetting the domain overview and getting lost, traditional hypertext viewers require this list to be repeated in every document.
The intelligent server can bind the navigational buttons once within any HTML document and the binding will remain throughout the display of other documents (until the buttons are re-bound). This capability combined with the ability to place links outside the document window provides a constant location for important links. The display of these links throughout the presentation of multiple documents helps user's avoid backtracking through previous documents when looking for a desired link. They also provide an overview of the information domain throughout the user's traversal of the information domain.
The hypertext browser should also guarantee that the information being displayed is a part of the desired data set. To prevent unrelated information from being simultaneously displayed, the additional windows must be restricted to the display of a single sub-tree within the information domain. The information server will ensure that this is true when constructing the information domain and generating the hyperlinks between individual sections of the domain.
Display management buttons, when pressed, display an additional hypertext browser window with limited functionality. In particular, these additional windows can not gain access to information that is not relevant to the individual topic. Therefore, neither display management buttons nor navigational buttons appear in any of the additional windows. This imposes a Main Window/Subwindow relationship among the multiple windows. The main window establishes the current topic and when the topic changes, all subwindows close. Similar to navigational buttons, display management buttons are bound by the intelligent server within any HTML document and retain their binding throughout the display of multiple documents until they receive a new binding.
Time is a critical issue for the clinicians, not only due to the time-critical nature of their jobs, but because the more time spent looking and waiting for patient information means less time available to spend treating patients. Therefore the interface to the VEMR must present the information in a clear and efficient manner.
The navigational buttons provided on the browser help accomplish this goal. These quick-jump buttons provide a designated area on the browser for the clinician to quickly select information that is relevant to his/her overall application. For example, these buttons take the clinician to a Patient Search screen, the Clinic Schedule, or the Inpatient Census. Placing these buttons on the browser interface, rather than within the document window, reserves the use of the document window for viewing data, guarantees to the clinician that the buttons will remain visible throughout his/her use of the patient record, and ensures navigation information will not interfere with the display of important data.
The display management buttons also improve the efficiency of the interface. These buttons provide subwindows to display various types of data available for an individual patient. For example, these buttons display patient data windows for transcriptions, lab reports, and radiology reports. The types of patient data differs from patient to patient depending on their medical record. Allowing the display management buttons to be defined by a the intelligent server, and bound within an HTML document, allows the buttons to reflect the current information available for a patient. The display management buttons also allow the clinician to simultaneously view particular types of information in individual subwindows. Navigation within particular patient information subwindows is restricted to available information on the specific topic (i.e. transcriptions).
To remove the potential for an incorrect diagnosis, it is very important that information contained in each window refer to the same patient. Therefore, when the clinician navigates out of an individual patient's record, all open subwindows are closed. This prevents the clinician from viewing lab reports from patient A and radiology reports from patient B.
This dynamic interface is the front end to an intelligent hypertext server. This intelligent server queries various hospital information systems to obtain all the current information on a patient. The resulting information is then properly structured, formatted, and served to the hypertext viewer. The server determines proper definitions for the navigational and display management buttons on the browser.
The Virginia Neurological Institute is using this extended HTML and dynamic browser to provide clinicians with a sophisticated hypertext tool that provides access to an information server that gathers and combines data from hospital information systems to create a dynamic hypertext patient record. A proper diagnosis is facilitated by the simultaneous display of all data available from multiple disparate hospital information systems in multiple windows with a guarantee that information presented in each window pertains to the same patient. An efficient diagnosis is facilitated by quicker navigation through patient data.
Future work focuses on extending capabilities to support movement among multiple domains. Development of a mechanism to represent a dynamic data view according to user is also being pursued.
Christine Remenyik received a Bachelor of Science degree in CS and a Bachelor of Arts degree in History from the University of Tennessee, Knoxville in 1993. She is currently working on her Computer Science Masters degree at the University of Virginia. Christine is involved in the NIIMS project as a graduate research assistant doing software development for the Virginia Neurological Institute at the University of Virginia.
Alexander Crosby received a Bachelors of Arts degree in The Growth and Structure of Cities from Haverford College, Pennsylvania in 1989. He is the Director of Informatics Projects at the Virginia Neurological Institute.
Jim Kazmer received a Bachelor of Science degree in Computer and Systems Engineering from Rensselear Polytechnic Institute, in Troy, NY in 1987. He currently working for Hewlett-Packard's Medical Products Group in the area of Health Care Re-engineering and Enterprise Integration.
The < NAV_INFO > tag provides the information needed to bind the navigational buttons to the appropriate URL. This information includes the label of the interface button.
The < DISPLAY_INFO > tag provides the information needed to bind the display management buttons to the appropriate URL. This information includes the label of the interface button and the title of the subwindow.
The < DOMAIN_ID > tag indicates the identity number of the document. This information tells the browser when to expect information regarding the navigation or display management buttons. This information also provides a means of controlling which documents belong to a particular topic.
Dynamic Navigation
The ability to navigate efficiently through hypertext is important when the amount of time required to reach specific information is critical. Dynamic navigational buttons provide a mechanism to support quick access to the entire information domain. Not only can movement be vertical, but horizontal movement across the information domain is facilitated with navigational anchors displayed on the browser to help the user avoid getting lost.
Multi-Window Display
When information is gathered from many information sources it may be desirable to view the multiple objects of information simultaneously. However, relying on the user to correctly coordinate the contents of the windows increases the risk of errors. The intelligent hypertext server requires a mechanism to automate the simultaneous display of information in multiple windows and guarantee that unrelated information can not be accessed through these extra windows. The introduction of display management buttons on the browser, see Figure 2, and HTML tags to define these buttons permits the hypertext server to name the specific documents that should be viewed together and automates the procedure to properly display the data in multiple windows.
4. Practical Application at the University of Virginia
The Virginia Neurological Institute at the University of Virginia and Hewlett Packard Health Care Re-engineering are using an intelligent information server, our extended HTML, and our enhanced NCSA Mosaic browser to provide clinicians with a sophisticated tool that gathers and combines data from diverse hospital information systems to create a dynamic hypertext patient record. This project, called the Virtual Electronic Medical Record (VEMR), will provide clinicians with simple, fast access to patient information.
5. Conclusions and Future Work
We have presented extensions to HTML to facilitate the representation of additional information regarding the content of the information domain and navigation through the information domain. This additional information will be used to improve navigation through applications and control over the display of related information. We have also presented an enhanced NCSA Mosaic browser that is able to interpret the additional HTML tags and react accordingly through the introduction of dynamic buttons on the browser. These new HTML tags and dynamic buttons allow the browser to react dynamically based on the information domain being accessed.
6. Acknowledgments
We would like to acknowledge the NCSA and the HTML standards working groups for their excellent foundation and documentation. We would also like to thank the NIIMS development team for their comments, suggestions, and contributions.
7. Author Biographies
Katie S. Oliver received a Bachelor of Science degree in Computer Science from St. Michael's College in Burlington, VT in 1992. She then attended the University of Virginia and received a Masters degree in Computer Science in 1994. She is currently working at the Virginia Neurological Institute at the University of Virginia as a programmer for the Neurosurgical Information Integration Management project (NIIMS), the parent project of the VEMR.
She may be contacted via e-mail at < oliver@virginia.edu >
She may be reached via e-mail at < remenyik@virginia.edu >
He may be contacted at < ac8n@virginia.edu >
He may be reached via e-mail at < kazmer@an.hp.com >
8. Appendix A
The extended HTML presented in this paper has three additional tags. The syntax for the tags and a short description of how the tags are used follows.
< NAV_INFO > button_name URL </ NAV_INFO >
< DISPLAY_INFO > button_name title URL </DISPLAY_INFO>
< DOMAIN_ID > number </DOMAIN_ID>