Multiple databases are often searched in Health Technology Assessment systematic reviews. However in rapid reviews, time and resources are limited and modifications to the search methodology may be necessary. In this retrospective study, the impact of searching fewer databases for three completed rapid reviews (i) Severe Mental Illness (SMI), (ii) Cannabis Cessation (CC), iii) Premature Ejaculation (PE) for the United Kingdom National Institute for Health Research was investigated.
The database coverage and indexing of the study references from the reviews were initially identified. The impact of fewer databases searched was then tested by (i) the number of studies that might be missed, (ii) the number of records for sifting and (iii) the overall rapid review conclusions.
A total of 178 included study references were found in the reviews (SMI n = 14 for 13 studies, CC n = 34 for 33 studies, PE n = 130 for 102 studies). Searching Medline only for SMI, Medline+Embase for CC, Medline+Embase+Cochrane Library for PE, would result in 1902 (74 percent), 466 (43 percent) and 240 (11 percent) fewer records needed to sift, respectively. There would also be a total of ten ‘would be missed’ references (SMI n = 1, CC n = 5 and PE n = 4). However, nine out of the ten references were found to have no or minimal impact on the overall findings of the reviews. The ten references were secondary reports of an included study, papers that lacked sufficient data for meta-analysis such as a conference abstract or an ongoing trial.
From the three reviews examined, limiting the search to fewer databases had no or minimal impact on the review conclusions despite the variable number of studies that would be missed and records needed to sift. More exploration during the scoping search prior to commencing the review will aid the decision on whether to limit the search to fewer databases.