|
|
|
Index - Major Sections
Site Map
Product and Services _______________ Index - Same Level Subject
Index - Child Subjects | Introduction
"The Problem"The problem with sharing information is the problem of distributed processing. See an example at Health Informatics/Distributed Systems/"The Problem" Reasons to Share DataThere are several reasons to share data.
Today, it is generally only for the first two reason that systems are being developed. Very few, if any, are addressing the needs for the development of the third. Potentially this is the most important reason to share information. OperationsOperations or transactional processing is the continuous processing of day to day data. It is the creation and updating of the patient and operating information. This type of processing needs quick access to the patient records and is of limited duration. Networking is generally limited to local LANS or for larger organizations a WAN. Single viewer needs for a single patientThis type of application is self describing. It concerns the need of an individual user to view data concern a patient. This data can be local to the user or it may cross several organizations with different data structures and operating systems. Very little alteration, if any, of records should be allowed. Access times required are somewhat moderate, depending on the user and his needs. Emergency situations may require fast responses. Networking may be over a WAN. Decision Support or ResearchInstead of looking a single patient record, Decision Support and Research requires the ability to look across all patients, in all locations, for all times. Access times are not as important and may occur over several minutes or hours. Networking is over a WAN. IntegrationBecause of the highly heterogeneous nature of the present systems, attention has been focus on methods to integrate these systems. Methods:
BackgroundGenerally discussions concerning sharing of patient information focuses on the ability to communicate across systems with other systems. However, very few of these discussions addresses the issue of how this information is storage across these systems. From a practical point, organization's should be stored in two different types of databases:
These type types of databases are designed differently for very good reasons. They support two very different functions. develop In a working environment transactional data should be organized for rapid assess. Transactional systems should not allow outside access. Data records that are not being accessed should be stored in a separate database. There is very little reason, if any, why distributed transactional systems must carry on inter-system communications. There are many reasons why they could not. At the heart of the medical records system is that changes should not be made directly to a patient record once that record is created. Therefore after a medical record is created is should be "read only." These records can be easily stored in a data warehouse without complications. Modern data warehouses present interfaces that can be accessed by all forms of operating systems. Data Warehousing has proved very effective in other industries. Since the data is "standardized" in the data warehouse, this data warehouse is "the patient record." Data Warehousing can effectively "map," bi-directional, one coding system to the other since there is only a two-way communication. There is no need for local systems to conform to the standards of the data warehouse.
ArchitectOnce a patients record is created is can be replicated to the data warehouse. When any authorized viewer needs the information for that patient, he can easily access a "read only" copy of the patients history. There is no reason for that viewer to update the information that he has obtained from the data warehouse. Once the authorized viewer has created his own information, then that information can be replicated to the data warehouse. Since modern data warehouses accepts any data structure, including text and video, each an every health care application, from the laboratory to imaging can easily be added to this database. By the same reasoning this data can be accessed. Application
RulesOperational data should never be queried by outside the system. No locks should be placed on operational data any longer than necessary. Operational data should never be added to or updated by outside the system. Operational systems are designed for fast response times (1-3 sec). No records should be updated once those records have been loaded. All additional data added to the data warehouse should come from replication of that data from the transactional data. IssuesSecuritySecurity issues are the biggest problems to overcome in a distributed system. No organization will let access (or should not let access) their transactional data by outside users. As the Spanish language permits more than one negatives, this is a good place for it. Nadie y nunca y ninguna y jamas. As patient privacy issues are make more public and as legal penalties for unauthorized disclosure of this information increases it will become even more difficult to share information with others across a network. Although both operational and decision support databases need strong security, different security issues are involved. Operational data should only be accessible by users that have direct needs for adding and updating of the information. Problem of transactional systems and authorization of users. High overhead. Auditing access. Large turnover in staff especially in university hospitals where students regularly access patient information. Data DictionaryBefore a user is able to use data, he must know where that data is located. If the user does not know where it is located he can not access that data. Example a patient has attended several clinics. A data dictionary provides information about the information. Quality of DataAssessing information across distributed systems is dangerous when the quality of that information at the other end can not be verified. ProductivityAccessing a transactional data system always increases response times. Locks must be maintained. Data MappingMapping of data revolves around standards. Information cannot be obtained unless the user understands the data that he is viewing. If all systems have different names for a piece of data, then some type of language-directory must be available to make the conversion. DevelopmentPatient RecordsAt the heart of the development of a model for health care information is the Electronic Health Care Record (EHCR) or the Computerized Patient Record (CPR). MessagingThe message based approach is the most prevalent system in health care organizations in the U.S. This system allows different information systems to exchange standardized messages through an interface. These systems have limitations on scalability. When the number of communicating systems increase the systems breaks down. Data WarehousingData warehousing is the process of storage all information off-line in a separate database. Although data is duplicated, this is not a serious problem as the cost of data storage devices decrease... develop. Queries to this data warehouse can take any form that the viewer wishes to view that information. Able to except information from many formats. For example Microsoft SQL7 Databases can except date from diverse structures as text files, Excel spreadsheets, desk top data bases such as Access, dBase, Paradox, and FoxPro. It can expect data from other SQL systems such as Oracle, Informix, and Sybase. It can except data from mainframe frames systems such as xxx. Federated DatabasesLike the Data Warehouse, Federated Databases must agree on a standardized data format.
Distributed Component-Based Systems
|
|
|