The sensitivity and precision of adverse effects search filters in MEDLINE, EMBASE and Science Citation Index (SCI)

Tags: Poster
Golder S1, Loke Y1
1Adverse Effects Methods Group, UK

Background: Systematic reviews should use search strategies which aim to identify as many relevant papers as possible. However, searching for information on adverse effects is challenging, not least because adverse effects terms (either specific such as 'venous thrombolismór 'weight gain’ or generic such as 'adverse event’ or 'side effect’) may not be included in the title, abstract or indexing of bibliographic records in databases such as MEDLINE and EMBASE.

Objectives: To assess the feasibility of using adverse effects terms when searching electronic databases to retrieve papers that report adverse effects data.

Methods: 242 included studies from 26 systematic reviews on adverse effects were analysed to ascertain whether the corresponding records in MEDLINE, EMBASE and Science Citation Index (SCI) included adverse effects terms in the title, abstract or indexing. In addition published adverse effects search filters devised for MEDLINE and EMBASE were tested to assess how many studies would have been missed had these filters been applied.

Results: Records in EMBASE (89%) were more likely to contain adverse effects terms in the title, abstract or indexing, than MEDLINE (80%) or Science Citation Index (SCI) (70%). The percentage of papers which would be missed with a combined search using adverse effects terms in MEDLINE and EMBASE was 8%. This figure is much lower than the 23% identified by a previous study in 2001. The sensitivity of published adverse effects search filters varied in MEDLINE from 1% to 93% and in EMBASE from 58% to 97%. Higher sensitivity was achieved when named adverse effects were included in the search strategies.

Conclusions: The proportion of records that include adverse effects terms in the title, abstract or indexing appears higher than in 2001. Although no adverse effects search filters captured all the relevant records, high sensitivity could be achieved, particularly in EMBASE.