Surveillance search techniques identified the need to update systematic reviews

Abstract

Objective: This article reports on literature surveillance methods to identify new evidence eligible for updating systematic reviews.

Study design and setting: Five surveillance search approaches are tested in the context of identifying studies that would signal major or invalidating new evidence for existing systematic reviews of health care interventions. Recall for each search approach was assessed as proportion of a composite yield of relevant studies across all search approaches that were identified by that approach. Screening burden was the number of studies that would need to be reviewed to identify the evidence that would necessitate updating.

Results: Searches were tested in a cohort of 77 systematic reviews. No one method yielded consistently high recall of relevant new evidence, so combinations of the strategies were examined. A search algorithm based on PubMed's related article search in combination with a subject searching using clinical queries was the most effective combination, retrieving all relevant new records in 68 cases. Screening burden was a median of 71 new records per review (inter-quartile range: 42-161).

Conclusion: Surveillance for emerging evidence that signals the need to update systematic reviews is feasible using a combination of subject searching and searching based on the PubMed's related article function.