The original 1979 resolution for Sigma Theta Tau International's (STTI) Virginia Henderson International Nursing Library (VHINL) called for "a national nursing library resource offering services to nurses and those interested in nursing," and soon after that, an additional call for "a national clearinghouse for information regarding nurse researchers and nursing research" emerged. Ten years later, the first computer was purchased for the library, enabling the beginning of an electronic library. With that development and with the establishment of a database that stored the findings of research studies, variables, research design, and researcher information, nursing knowledge was made available in an electronic format to the nursing community worldwide.1 In 2001 a "revisioning the library" meeting was convened, and participants framed their vision in the context of the dramatic technology innovations and changes in nursing care delivery that occurred in the 10 years since the VHINL purchased its first computer. Today, a new call has emerged-the call to extend the functionality of the VHINL from a library to a Web-based portal for nursing knowledge resources to support global nursing practice, education, and research while retaining and enhancing the rich legacy of knowledge modeling.
The design or redesign of any interactive browser-based information system such as a digital library requires the consideration of (1) humans needing to perform the search, (2) information systems containing the desired content, and (3) the interactions between the human being and the system. Indeed, any interactive browser-based information system not only needs to provide useful content but also must present the content in efficient, effective, and satisfying ways for the user. These principles guided the development of a "tell and ask" functional interface, in which the professional-in this case, the nurse-communicates with the knowledge base by making logical assertions (tell) and posing questions (ask)2 based on her/his professional knowledge and experience.
Formative usability evaluations of the previous version of the library enabled the detection of a certain number of usability defects, such as difficulty in learning and navigating the system and a high error rate in retrieving information. The usability issues resulted in low stickiness: in 74.24% of the cases, the sessions lasted only a few seconds and the page views totaled one or less. Search strings were ill-constructed and the search strategy was often ineffective. Search terms were limited: only 93 distinct search terms were used; single terms were the most frequently used search string, and variables, theory, or research design were seldom used as a search string. Boolean searches occurred only in two instances. Unsuccessful searches were restarted at another location, such as journal and case study, with the same search string.
After an in-depth analysis of the usability issues, a prototype interface was developed and feedback was obtained from key nurse user groups. The formal knowledge model that currently exists within the library was examined, needed expansions to the database were identified, and opportunities for a controlled terminology for indexing abstracts and research studies were recognized. These analyses motivated revisions to the existing data model and the development of the specification of an information model. The goal was that the knowledge model and data model are complementary and that both complement the search behavior or search strategies of nurse users.
During the redesign, it also became evident that formal usability evaluation did not capture how well the interface supports nurses in their search tasks, nor did it take into consideration the overall environment in which search tasks are performed.3,4 Hence, an extended evaluation of the usability and utility of the redesigned library was proposed. Human-computer interactions and search behaviors are studied in their natural environment. Web logs are used to detect patterns of cognitive activities and behaviors that occur within human-computer interactions. Random pop-up surveys complement the Web log analyses to better evaluate whether the revised knowledge model as mapped to the search behavior model provides a level of consistency and framework for searching the library.
From the analyses of the online survey during a 2-month period after the launch of the redesigned interface and database, it can be assumed based on the users' comments that the interface has improved; however, more observations are needed. Furthermore, the results of the survey need to be merged with the Web log analyses, which were not available at the time of writing.
As mentioned earlier, controlled nursing terminologies were proposed for indexing abstracts, variables, and keywords of research studies. The ideas behind this approach are to (1) create a context-free method of indexing, (2) provide a universal framework for searching consistent with nursing practice, (3) provide links to other published nursing and health research indexed in online databases such as PubMed, and (4) build evidence-based practice knowledge bases for nursing. This phase of the project is still ongoing. Preliminary results show that mapping variables and keywords of clinical research studies to the ICNP terminology returns the best lexical and semantic matches. The match rate is lower, however, for nonclinical concepts such as education and administration. Other terminologies, such as SNOMED-CT, NMMDS, and the other UMLS thesauri, will be explored for mapping nonclinical concepts.
ACKNOWLEDGMENT
The author would like to thank the project team at Sigma Theta Tau International: Marcelline Harris, Cheryl B. Thompson, Amy Coenen, Linda Finke, Margie Wilson, Rick Olson, and Dimas Gutierrez. Support for this project came from the Ruth Lilly Nursing Informatics Scholarship, Sigma Theta Tau International.
REFERENCES