The current level of overlap of trials and systematic reviews (SRs) between bibliographic databases is uncertain.
To analyse the level of overlap of trials and SRs indexed from 2006 to 2015 in the Cochrane Library, Embase, LILACS and PubMed.
We performed a search in January 2017 in PubMed and LILACS to identify SRs and trials published over 10 years (2006 to 2015). Later years were excluded in order to discard incomplete indexing. We selected the search terms in each of the databases in order to favour their comparability (Box 1). We analysed trends for SRs and trials through a regression analysis performed in Stata® 14.1. To deduplicate records, we used EndNote X7.8 filtering by Author AND Year AND Title criteria.
In the 10-year series, the four databases showed a statistically significant increasing trend for SR and trial indexing, both separately and globally, after deduplication (Table 1, Figure 1).
The search retrieved 2,426,221 trials and 546,802 SRs. The order of contribution for both study designs was: 1) EMBASE, 2) PubMed, 3) Cochrane, 4) LILACS. The number of SRs retrieved exclusively by each database followed the same order (Table 1).
The proportion of overlap in the databases in the 10-year series was 18% (Figure 2).
The complete analysis will be presented at the Colloquium.
More than 60% of the SRs that were published in only one database were found in Embase. This figure rose to 75% when also considering PubMed. In addition, 16% more were published simultaneously in one of these two databases and in one or more others. As only 8% of the SRs were published in the Cochrane Library (7%) or LILACS (< 1%), a further analysis of the topics that these SRs evaluate could help us to better understand the real importance of searching for SRs in these databases.
Relevance to patients and consumers:
The search strategies for systematic reviews are a key quality component and therefore they are important in order for patients and consumers to be confident in the findings.
Patient or healthcare consumer involvement:
Healthcare consumers participated in the interpretation of the results.