Searching chiropractic literature: a comparison of three computerized databases

To determine the efficiency of three computerized bibliographic databases in retrieving literature relevant to the chiropractor.

A cross-sectional design was used. English-language citations from 1990-92, on the topics of scoliosis, sciatica and neck pain, were searched in CHIROLARS, Index to Chiropractic Literature (ICL) and MEDLINE. Citations were assessed for relevance criteria by two assessors; a third assessor was used when both were unsure of relevance. Inter- and intraexaminer reliability of the relevance assessments was determined using the weighted Kappa statistic. The outcomes assessed were search time, search costs, number of citations, relevance, number of unique citations and the number of citations from refereed journals. Relative recall and cost per citation were used as primary measures of database efficiency.

A total of 846 citations were retrieved. After exclusions, 786 citations were assessed for relevance. Interexaminer reliability of the relevance assessments was moderate [K(w) (standard error) = 0.46 (0.03)]. Intra-examiner reliability was fair for each of the assessors [K(w) (SE) = .36 (.10) and .35 (.10), respectively]. Of the 385 relevant citations, CHIROLARS retrieved 88 (relative recall = 23%) at a cost of CAN$1.01 per relevant citation, ICL retrieved 37 (relative recall = 10%) at CAN$.65 per relevant citation, and MEDLINE retrieved 260 (relative recall = 68%) at CAN$.52 per relevant citation.

MEDLINE was found to be the most efficient database to search for literature relevant to the chiropractor; it retrieved the highest proportion of relevant citations and was the least expensive. CHIROLARS was the second most efficient of the three databases. No single database can function as a stand-alone source of information. For comprehensive searching, having an experienced reference librarian search MEDLINE in combination with at least one other database is recommended.