Comparison of Search Design and Output of Embase Using Two Different Search Interfaces: Embase.com and OVID

OBJECTIVES: Different search interfaces use different search functionality and indexing. This research aimed to investigate the variances between the two most commonly used search interfaces and their impact on search results.

METHODS: We compared access to Embase through Embase.com and OVID interface in the following ways: coverage, results of step-by-step searching using different approaches (e.g., general syntax, combinations, truncations, phrase and field search), and keywords (n=10), and operational characteristics rated on a scale of 1-10 by different (n=4) researchers.

RESULTS: Embase, through their Embase.com interface, covers citations from 1947, indexes 3,086 journals, and 3.6 million conference abstracts from 2010 onwards. On the other hand, coverage via OVID includes citations from 1947, 3,584 journals, and 2.1 million conference abstracts from 2009 onwards. Even with a higher number of journals indexed by OVID, citations were consistently lower than those retrieved from Embase.com for identical search approaches. Apart from coverage, the variability may be due to differences in indexing. E.g., a search for mesothelioma in OVID showed twelve relevant terms indexed as sub-headings, while Embase.com indexed those same terms as appropriate sub-headings (n=3), synonyms (n=7) and under dissimilar disease sub-headings (n=2). However, the search for diabetes mellitus resulted in exactly the same sub-headings (n=13) across both interfaces. Still, OVID showed lower hits in both examples. Interestingly, in searches using proximity operators, the difference in citations retrieved across interfaces was reduced but still lower in OVID. The search by journal name (e.g., Nature) also retrieved different results. Regarding operational characteristics, our researchers rated Embase.com as relatively easier to use, less complex with fewer search boxes, and more persistent for longer search history recall after inactivity (mean scores: 9.1 vs. 8.3).

CONCLUSIONS: These findings will have implications for search protocols using an indexed term approach, emphasizing the need to include multiple sources and scrutinize indexing (sub-headings, synonyms).