An analysis of data sources and study registries used in systematic reviews

BACKGROUND: Reporting standards for data sources in systematic reviews (SRs) have been developed, yet research shows varying compliance in the methods section. When this happens, replication of search results is difficult and creates ambiguous and biased data sources. AIMS: This study captured author practices in choosing English and non‐English‐language databases, listing all the databases searched, and incorporating study registries as part of the search strategy. METHODS: Using an analytic, cross‐sectional, study design, volunteer data collectors (n = 107) searched one of two assigned English language platforms for SRs on specified health conditions. All the data sources found in the methods section of each SR were documented and analyzed for patterns using bibliographic techniques. RESULTS: The final sample size of the SRs reviewed was N = 199. The mean number of data sources seen in the SRs was 3.9 (SD 2), with a range of 1–10. Eighteen records (9%) used a single data source to conduct the SRs. Four leading language platforms were seen in the SRs: English (100% of occurrences), up to 8% used Chinese data sources, and 4% included Spanish or Portuguese. The four most frequently used data sources were: (1) Medline (98%), (2) Embase (65%), (3) Cochrane Library (56%), and (4) Web of Science (33%). The percentage of SRs listing study registries was 30%. LINKING EVIDENCE TO ACTION: Strategies to reduce bias and increase the rigor and reliability of SRs include comprehensive search practices by exploring non‐English‐language databases, using multiple data sources, and searching study registries. By following PRISMA‐S guidelines to report data sources correctly, reproducibility can be accomplished.