Drug Information Journal - March 2009 - (Page 134) 134 MEDICAL INFORMATION Cohen when a document management system is not used to manage the authoring workflow, as is common practice for protocols. STEP 3: INFORMATION RETRIEVAL Information stored in step 2 is physically retrieved in step 3. Such retrieval is not a straightforward task when the repository of information is large, as it is for large pharmaceutical or medical device companies that conduct many trials simultaneously. Information retrieval often requires a text search of keywords to find relevant material. Without semantic (ie, meaning-based) search capabilities, such searches are often inefficient, with a high return of irrelevant information. For the same reason, information conveyed in a document cannot easily be combined conceptually with information stored in related documents. STEP 4: INFORMATION INTERPRETATION Following retrieval, document-conveyed information must be interpreted in order to extract meaning from it. Ideally, the intention of the original authors will be transmitted to readers (users). Leaks at this step are most readily appreciated by example. Consider a clinical protocol whose schedule of activities includes the task “routine exam.” Is this an initial exam or a follow-up? Is it brief or comprehensive? Does it include a detailed history, a brief history, or no history? The reader cannot make these determinations from the term “routine exam” alone, necessitating additional work to determine the original authors’ intentions or else risk acting as the authors did not intend. Consider also the need to transmit documentconveyed information to dependent information systems, for example, the transfer of protocol information to the electronic data capture (EDC) system. It is not unusual for nonauthors of protocols to program the EDC system, adding a layer of communication to the interpretation of protocol information prior to its reuse. STEP 5: INFORMATION REUSE Information is reused when original concepts, though not necessarily original text, created at step 1 are employed for either the same or a new purpose. In the case of the protocol, information is frequently reused when creating new protocols and protocol-dependent documents, such as the case report form, investigator brochure, study report, and so on. Information reuse is typically manually controlled via document-to-document transmission, using desktop office tools like copy/paste and “file save as.” These processes allow human errors in process execution and judgment to cause some information loss and corruption. This type of leak also occurs when information systems reuse information created upstream. Using the example of the transfer of protocol information to an EDC system again, the protocol information must not only be interpreted, it must also be recoded from English or some other written natural language into a structured language, which can be understood by the EDC system to allow its transfer. IMPROVED KNOWLEDGE TRANSFER E F F I C I E N C Y W I T H S E M A N T I C A L LY MODELED STRUCTURED CONTENT Reducing information leaks during information reuse will necessarily reduce the activities needed to correct or compensate for them, potentially leading to productivity improvements. Relatively recent advances in computer science have created an opportunity for computers to play a more direct role in reducing such leaks. The information tools described earlier, such as word processors, have always been capable of transmitting unstructured content in its native form, what is usually referred to as natural language, with reasonably high fidelity. What these commonly used tools have not been capable of doing until more recently is transmitting the structure of information along with its form. Here, the term “structure” refers to the components of information and their interrelationships that, together with form and grammatical conventions, give language meaning (8). Natural language information conveyed in documents has an implicit structure, by virtue of its language rules and form, that is readily interpretable and analyzable by humans. Comput-
For optimal viewing of this digital publication, please enable JavaScript and then refresh the page. If you would like to try to load the digital publication without using Flash Player detection, please click here.