We created SmartSearch (http://SmartSearch.UTHSCSA.edu) in order to automate much of the search for medical evidence by clinicians. The clinician only needs to know how to: 1) construct a search query using best search terms and Boolean logic (Boolean logic is the proper use of 'ands' and 'ors'), and 2) recognize and distinguish between higher and lower quality articles retrieved by SmartSearch. For example, the clinician needs to recognize randomized controlled trials.

SmartSearch examines the clinician’s query then automatically chooses best resources to search, formats the search for each resource, and executes the searches. Based on the search results, SmartSearch may search alternative resources when too few articles are retrieved, or SmartSearch may reformat a more restrictive search when too many articles are retrieved.

In addition, SmartSearch analyzes searches that find no articles and suggests how to correct the search. SmartSearch also can manage passwords to journal web sites.

Although SmartSearch has not yet been formally announced outside of our institution, our medical students and residents are encouraged to use SmartSearch and are instructed in how to use it.

We used two Internet server logs to measure the impact of SmartSearch. First, we analyzed the frequency of selection of SmartSearch and other resources by clinicians who use our clinical web site. Second, we analyzed the server log of SmartSearch to assess the frequency and characteristics of successful searches. Successful searches are defined as retrieving at least one articles from MEDLINE.

In the last one month, SmartSearch was queried 1767 times. Users of the clinical web site requested SmartSearch 449 times and requested OVID, the next most frequent resource, 69 times. In the last month, 81% of queries were successful while 19% of searches failed. The was no difference in the number of search terms between successful and failed searches. Successful searches, as compared to failed searches, were more likely to use two or less words within all search terms (83% versus 51%). When clinicians used more than two words within a search term, 41% of searches failed. Forty-four percent of searches for etiology questions and 33% of searches for adverse effects failed, whereas only 14% of searches for therapeutic questions failed.

Clinicians who use our clinical web site use SmartSearch more often than any other medical resource. Eighty-one percent of searches find at least one article. Successful searches are more likely to use search terms that are two words or less in length and more likely to address therapeutics

Clinicians at our institution may be using SmartSearch to replace MEDLINE as the preferred method for seeking medical evidence. Although our definition of successful searches is certainly an overestimate, this definition is still able to identify characteristics of better search strategies. SmartSearch can then be programmed to warn clinicians when they are about to submit poor strategies such as using three words in a search term.