OBJECTIVES: Search filters are predetermined combinations of phrases that have been developed to obtain a subset of records based on specific topics of interest. The objective was to create and test two filters supporting the identification of observational studies: one that could be used when running rapid reviews (RRs), containing only the main terms related to these types of studies (’basic’), and the second one, more sensitive, that could suit well systematic literature reviews (SLRs), named ‘comprehensive’.

METHODS: The basic filter was created by combining keywords used in the observational search filters published by 3 identified sources: Scottish Intercollegiate Guidelines Network, British Medical Journal Knowledge Center, and OVID expert searches. In the comprehensive filter, keywords published by the Canadian Agency for Drugs and Technologies in Health were additionally included. The filers were tested using Cochrane’s SLRs as a ‘gold standard'.

RESULTS: The performance of the filters was satisfactory – for the comprehensive filter, the coverage of available data was over 92%, while for the basic filter, which could be implemented in RRs aiming to synthesise key trends in a cost-efficient manner, the coverage was 77%, with most of the studies falling in the range between 84-96%. The use of the basic filter was associated with a significantly higher reduction of workload compared to the comprehensive filter – between 2 and 7 workdays. Studies were omitted mostly due to poor indexing and not using relevant keywords in abstracts. 63% of studies missed by the comprehensive filter were published prior to the year 2000.

CONCLUSIONS: Effective evidence retrieval is critical for performing high-quality evidence synthesis to support healthcare decision-making. Implementing the methodological filters into the search strategy significantly reduces the workload, and therefore generates savings in time and costs. However, it is important to underline that restricting the strategy may lead to omitting relevant data sources.