gms | German Medical Science

Background/research question: In healthcare decision-making, systematic reviews (SRs) are essential to guide practitioners and policy makers. The development of carefully designed search strategies is crucial for conducting reliable and efficient SRs. Systematically developed and validated search filters can enhance this process. Geographic search filters are particularly useful in addressing research questions with regard to the local context. While geographic search filters exist for some countries (e.g., the UK or US), none are available for Germany. Our aim was to design and validate a highly sensitive geographic search filter for MEDLINE (PubMed) to identify studies conducted in Germany.

Methods: We adopted a stepwise approach for the objective development and validation of search filters [1]. First, we created a gold standard set (GS) of German studies from SRs analyzed in a recently updated methodological review [2] that we divided into a 'development' and 'testing' set. Second, candidate search terms were identified using (i) term frequency analyses in the 'development set' and a random set of MEDLINE records; and (ii) a list of German geographic locations, compiled by our team. Third, the search filter was iteratively developed, assessing its sensitivity in both the 'development' and 'testing' set. Fourth, we conducted case studies (CS) to validate the search filter using a sample of SRs that included studies conducted in Germany, but did not geographically restrict their search strategy.

Results: The search filter demonstrated sensitivity of 99.85% and 100% within the 'development' and 'testing' set, respectively. When validating the filter in 17 CS by combining it with the original search strategies of the SRs, the median precision was 2.64% (interquartile range [IQR]: 1.34–6.88%) vs. 0.16% (IQR: 0.10–0.49%) without the filter. The median number-needed-to-read was reduced from 625 (IQR: 211–1042) to 37 (IQR: 15–76). The filter identified 173 out of total 178 relevant records. A sensitivity of 100% was achieved in 13 CS, 85.71% was reached in 2 CS, while in 2 remaining CS the filter demonstrated the sensitivity of 87.50% and 80%.

Conclusion: We developed and validated a geographic search filter for MEDLINE (PubMed) that reliably identifies studies conducted in Germany and improves efficiency of the screening process. The search filter can be applied in SRs and other evidence syntheses that aim to identify studies that were conducted in Germany.

Competing interests: The authors declare that they have no conflicts of interest