Background: Overview identifies and synthesises systematic reviews (SRs) to answer an investigational question. However, it is necessary to identify redundant information.
Objectives: Investigate the overlapping in primary studies across the SRs included in an overview. The overview assessed nonpharmacological interventions to prevent adverse events in the intensive care unit (ICU).
Methods: We created a matrix of evidence as a grid, placing all the included SRs in the columns and their respective primary studies in the rows. We calculated the corrected covered area (CCA) for the whole matrix and for each pair of SRs. We repeated this process for each outcome, creating custom matrices including only the SRs and primary studies providing data for each specific comparison. We considered overlap to be low if the CCA was below 5%, moderate if the CCA was between 5% and 10%, high if the CCA was between 10% and 15%, and very high if the CCA was above 15%.
Results: A total of 37 SRs were included, and 27 nonpharmacological interventions were identified to prevent 11 adverse events. The SRs comprised a total of 246 individual primary studies. The overall CCA was 1.0% (slight overlap; see Figure 1); however, the overlap assessment at the outcome level showed a high and very high overlap for 12 comparisons. The most frequent adverse event assessed among the SRs was ventilator-associated pneumonia (VAP). One of its preventive interventions was subglottic secretion drainage (SSD) versus no drainage, which included four SRs with a total of seventeen individual primary studies. Seven of the seventeen trials were included in all four SRs, and the CCA was 52.9% (very high overlap; see Figure 2).
Conclusions: We found overlapping in some important outcomes assessing preventive nonpharmacological interventions for avoiding adverse events in the ICU; therefore, the findings should be interpreted carefully. Overlapping in an overview should be investigated globally but also at the outcome level. Patient, public, and/or healthcare consumer involvement: there was no involvement.