Background: developing trustworthy practice guidelines requires the use of systematic approaches to identify relevant evidence. Some guideline projects address a large number of PICO (population, intervention, comparison, outcome) questions. This requires the use of efficient approaches for identifying eligible studies. One such approach consists of running a literature search common to all included questions. However, it becomes a challenge to screen and match studies to a large number of questions at the same time.
Objectives: to describe an approach to searching for, screening and matching studies for a large number of PICO questions.
Methods: we developed the approach in the context of updating the American College of Rheumatology (ACR) guidelines for the management of rheumatoid arthritis (RA). This update addresses 102 PICO questions and considers evidence from both randomized and non-randomized studies. The literature review team is composed of 17 members. We developed the approach through discussion among team members, piloting and iterative revisions. In addition, we developed a plan to validate the approach.
Results: the proposed approach consists of six distinct parts:
2) searching the literature;
3) title and abstract screening;
4) full-text screening and tagging;
5) matching eligible studies to questions; and
6) validating the matched studies.
First, we categorized the ‘populations’ of the different PICO questions into subgroups of the RA population. We similarly categorized the ‘interventions’ and ‘controls’. Then, we grouped PICO questions into sets with non-overlapping categories of interventions. Second, we performed distinct literature searches for two sets of PICO questions: one with vaccine types of interventions, and one with ‘other’ types of interventions. We conducted the subsequent parts separately for these two sets. Third, for the title and abstract screening, we judged the studies as eligible if they met study design criteria, as well as the population and interventions/controls categories relevant to the set. Fourth, using an online DistillerSR form, we screened the full-texts using the same eligibility criteria as above. Concurrently, we tagged the included studies for the different categories identified in the PICO-analysis. Fifth, the matching process aimed to identify all studies eligible for each PICO question. To conduct an efficient matching, we started by grouping PICO questions into subsets for which we identified eligible studies; for that, we captured the studies with tags matching the PICO categories relevant to the subset. Subsequently, we allocated the studies to the individual PICOs within each subset. Sixth, we aim to validate our results against those obtained in L·OVE (by Epistemonikos).
Conclusion: the proposed approach has shown to be feasible and efficient for identifying eligible studies for a large number of PICO questions. Its value will depend on proving its validity.
Patient/consumer involvement: none