DISCUSSION
This case study investigated the prospective implementation of different deprescribing search filters in deprescribing systematic reviews search strategies. The maximized sensitivity search filters and the USDeN filter were developed with alternative approaches. USDeN filter was developed as part of a larger search strategy that include older people filter and specific drug name deprescribing filter. These alternative approaches partially explain difference between the filters. The maximized sensitivity deprescribing filter for MEDLINE included 18 free-text terms, phrases and MeSH terms against 9 free-text terms and MeSH terms for USDeN’s. Only 3 terms were in common and 2 differs only by truncation applied. This explained observed differences in sensitivity and precision.
In both systematic reviews, using the maximized sensitivity search filters for MEDLINE and Embase either increased or maintained the strategy’s sensitivity. These results are consistent with the aim of developing maximized sensitivity filters.
Sensitivity is critical in order for systematic reviews to exhaustively summarize the evidence7. At equal sensitivity, the maximized sensitivity search filter was more precise than the original search strategy in systematic review 2. However, higher sensitivity (compared to the original search strategy) in systematic review 1 was associated with lower precision. On the other hand, the USDeN deprescribing filter had higher precision, thus allowing a significant decrease in the number needed to read of articles, at the cost of a lower sensitivity. Precision and number needed to read reflect the research burden of a search strategy7: the lower the precision, the higher the number of articles to read to identify one supplemental relevant article. These performances are crucial for researchers who have limited resources. From that perspective, the USDeN deprescribing filter is a valuable tool for scoping reviews or rapid reviews which are not intended to be systematic.
The main limitation of our case study is the small number of systematic reviews included. This reduces the extent to which the results can be generalized. Another limitation is the lack of systematic review focusing on deprescribing interventions in the included reviews.
To conclude, using maximized sensitivity deprescribing filters may increase the exhaustivity of deprescribing systematic reviews. On the other hand, the precision offered by the USDeN deprescribing filter is a convenient alternative for non-systematic reviews.
Acknowledgments: TM and JPF were involved in the conception of the study. TM, CH, LZ, EH and AC was involved in data collection; TM and JPF were involved in data analysis and interpretation. TM drafted the first version of the manuscript. TM, CH, LZ, EH, AC and JPF read and approved the final manuscript.
The authors thank Wade Thompson for his precious help in disseminating the protocol.
Funding: This research did not receive funding.
Reprints and correspondence: Dr Jean-Pascal Fournier, Département de Médecine Générale, Faculté de Médecine, Université de Nantes, France. Tel: 33 (0) 24 041 1129, Fax: 33 (0) 24 041 2879, e-mail:jean-pascal.fournier@univ-nantes.fr
Conflict of interest statement: The authors declare that they have no competing interest.