Which database for which search?



By ranking electronic databases according to the percentage of questions answered, Brian S. Alper and colleagues1 imply that all electronic databases are alike and can be compared in this fashion. They omit the key second step in the process of efficiently finding good answers to one’s clinical questions: to devise a search strategy that answers the question, “Which database is most likely to efficiently yield the answer I’m seeking?”

Evidence-based medicine (EBM) databases, such as TRIP, and journals of secondary publication that catalogue high-quality, predigested information are small, but the answers found are more likely to approximate the best available evidence. However, searching them will be futile if one’s question pertains to an issue for which good evidence on the topic does not exist. This explains the low ranking of TRIP in contrast with databases such as MDConsult, which are information rich and produce sensitive searches (yield more answers) but lack built-in quality filters (the answers don’t necessarily represent the best available evidence).

Take, for example, the following question: “What is the safest and most effective drug treatment for a 2-month-old infant who is not gaining weight because of severe gastroesophageal reflux disease (GERD)?” MDConsult yielded 3 textbook references, 595 journal articles, and 13 guidelines. The sheer volume of citations precluded me from satisfactorily answering my question.

In contrast, TRIP yielded 5 evidence-based links, one of which was relevant to my patient but provided an incomplete answer (less sensitive, more specific). For this question, DynaMed proved to be the most appropriate database, confirming the lack of efficacy of cisapride; recommending ranitidine as first-line therapy, based on its safety and efficacy; and providing the appropriate dose and a reference.

Unfortunately, there is currently no “one-size-fits-all” electronic database. Different questions are best answered by different databases. This situation requires clinicians to develop a pragmatic search strategy that that will direct them to the database most likely to yield the answer to each question as it arises.

Eamon C. Armstrong, MD
Lehigh Valley Hospital
Allentown, Pennsylvania

Drs Alper, Stevermer, White, and Ewigman responded as follows:

Dr Armstrong makes several excellent points. We agree that individual electronic databases have unique strengths and weaknesses. The ideal database would be valid, relevant, convenient, fast, and affordable. Since we lacked the resources to evaluate all these factors, investigators and consultants from the Family Practice Inquiries Network (FPIN) Consortium agreed that the ability to obtain an adequate answer was the single most important factor for this initial study. We defined an answer as adequate if it provided a reasonable course of action for a family physician. This approach is used by most clinicians; research shows that convenient and relevant resources are selected over evidence-based resources.2-4

We also agree that clinicians need pragmatic search strategies. An individual’s specific strategy depends on available resources, types of information sought, experience and skill of the searcher, and an understanding of what can be found. We are currently conducting research to better define such strategies. As members of FPIN, we share the goal of creating a single database using the best existing evidence and designed for the practicing family physician, thereby simplifying this process.

In the meantime, for searches that maximize both validity and efficiency we currently recommend initial searches using evidence-based databases followed by searches using highly referenced databases. Evidence-based databases have been developed through systematic literature searching (eg, Clinical Evidence or the Cochrane Library) or through systematic literature surveillance (eg, ACP Journal Club Online, DynaMed, InfoRetriever, or TRIP). Highly referenced databases provide reference support specifically linked to included statements. Among these databases are DynaMed, Praxis.MD, Scientific American Medicine, and UpToDate.

Brian S. Alper, MD, MSPH
James J. Stevermer, MD, MSPH
David S. White, MD
Bernard G. Ewigman, MD, MSPH
University of Missouri School of Medicine

Next Article: