A contribution to the comparison of the usefulness of the data bases of Excerpta Medica and MEDLARS in biomedical literature retrieval

The results of searches on biological, medical and veterinary subjects by both the E.M. automated retrieval system and MEDLARS were compared. Of a total of 60 E.M. searches 23 were selected for the comparative study. Seven were eliminated because these needed specific terms not present in the MEDLARS terminology and 30 were omitted for other reasons. In 17 searches MEDLARS produced more relevant references than E.M. On the other hand in 12 searches the precision of E.M. was higher than that of MEDLARS. An average precision of 55 per cent was found for E.M. and of 38 per cent for MEDLARS. In 12 searches the cause of the failure to retrieve known relevant references was investigated. In MEDLARS 28 per cent was due to inadequate journal coverage and 72 per cent to indexing or searching failures (total 71 failures); in E.M. these percentages were 8 and 92 respectively (total 160 failures), the last percentage including an unknown proportion due to selective indexing of journal contents. Of 226 relevant references from E.M. and 467 relevant references from MEDLARS (the total retrieved in 15 searches), 94 references were duplicated. Recall figures were estimated: an average recall of 18 per cent was found for E.M. and 33 per cent for MEDLARS. Search strategies and indexing of overlapping references were compared. An estimate was made of the extension ratios of 12 searches. This measure averaged 1.9 for E.M. and 2.8 for MEDLARS.