Study objective: We introduced self-service access to the medical literature database, MEDLINE, into clinical settings to assess the frequency, patterns, purposes, and success of use.
Design: Longitudinal descriptive study.
Setting: Inpatient and outpatient services of a university medical center.
Participants: All trainees and attending staff working at the service sites were invited to participate; 158 (84%) did so.
Interventions: Free online access was provided to MEDLINE through GRATEFUL MED software. Participants were offered a 2-hour introduction to online searching and 2 hours of free search time.
Measurements and main results: For each search, a computer program requested identification of the user and the question to be addressed. Search transactions were recorded automatically. Interviews were conducted after a random sample of searches, and search questions were given to more expert searchers to run for comparison with the original. Eighty-one percent of participants did searches on study computers, at a mean rate of 2.7 searches per month. On comparison searches, participants retrieved 55% of the number of relevant articles retrieved by reference librarians (P = 0.024) and 50% more irrelevant articles (P less than 0.001). Forty-seven percent of searches on patient problems affected clinical decisions, but often on scanty information.
Conclusions: MEDLINE searching from clinical settings is feasible with brief training and affects clinical decisions. However, inexperienced searchers miss many relevant citations and search inefficiently. Further studies are needed to assess the impact of searching on physician performance and patient care.