Developing efficient search strategies to identify reports of adverse effects in medline and embase

OBJECTIVE: This study aimed to assess the performance, in terms of sensitivity and precision, of different approaches to searching MEDLINE and EMBASE to identify studies of adverse effects. METHODS: Five approaches to searching for adverse effects evidence were identified: approach 1, using specified adverse effects; approach 2, using subheadings/qualifiers; approach 3, using text words; approach 4, using indexing terms; approach 5, searching for specific study designs. The sensitivity and precision of these five approaches, and combinations of these approaches, were compared in a case study using a systematic review of the adverse effects of seven anti-epileptic drugs. RESULTS: The most sensitive search strategy in MEDLINE (97.0%) required a combination of terms for specified adverse effects, floating subheadings, and text words for 'adverse effects'. In EMBASE, a combination of terms for specified adverse effects and text words for 'adverse effects' provided the most sensitive search strategy (98.6%). Both these search strategies yielded low precision (2.8%). CONCLUSIONS: A highly sensitive search in either database requires a combination of approaches, and has low precision. This suggests that better reporting and indexing of adverse effects is required and that an effective generic search filter may not yet be feasible.