Background: An analysis of searches conducted for the NICE guideline surveillance programme indicated that only about 3% of studies are included following sifting. Currently, population only searches are used with the aim of retrieving all relevant articles. This can result in a high number of results with low precision. Although machine-learning techniques may offer a mechanism for improving precision of surveillance searches in the future, they are not commonplace currently. An alternative approach involves utilising search techniques to increase the precision of the searches used for surveillance, without losing the recall of the search.

Objectives: To conduct a retrospective analysis comparing the impact of a modified search approach on the surveillance decision.

Methods: Five guidelines were selected for inclusion in this retrospective analysis using the following criteria:
• Surveillance decision was to update the guideline
• Large database of results from the search strategy (>3000 studies)
• Low number of included studies summarised (<5%)

The searches for those topics were re-run using additional search techniques including:
• Major/focused subject headings
• Frequency operators
• Subheadings
• Truncation amendments
• Methodological filters with higher precision

Results: The impact of the modified search approach will be compared with the original search approach employed and the following factors considered:
• Impact on number of results
• Impact on identification of key papers
• Impact on surveillance decision

Conclusions: The implications of using additional search techniques in guideline surveillance will be discussed with a particular focus on which techniques optimise the balance between precision and sensitivity, with the least impact on surveillance decisions.