Using GRADE to develop an evidence-based benchmark for a rate: the use case of the patient safety indicator ‘postoperative venous thromboembolism’

Article type
Authors
Burnard B1, Januel J1
1Lausanne University Hospital, Switzerland
Abstract
Objectives: To adapt the GRADE methodology for a rate in the framework of a systematic review aimed at developing a benchmark for in-hospital symptomatic venous thromboembolism (VTE) in hip or knee arthroplasty patients (TPHA, TPKA) under recommended VTE prophylaxis.

Methods: We assessed the suitability of each GRADE component. We included randomized clinical trials (RCT) testing efficacious VTE prophylaxis regimens and observational studies of patients receiving VTE prophylaxis, which reported confirmed post-operative symptomatic VTE that occurred before hospital discharge following TPHA and/or TPKA in adult inpatients. We developed ad hoc criteria to rate inconsistency and imprecision of pooled estimates.

Results: We considered GRADE elements about study design, outcome assessment, and sources of potentials biases. We included allocation concealment, blinding, sparse data, attrition bias, indirectness, potential measurement bias, and potential conflict of interest at study level. At group level, we included risk of bias, imprecision, inconsistency, indirectness, and publication bias. We summarized the quality assessment of the included studies in five categories: consistency, imprecision, generalisability to population of interest, publication bias and other limitations (allocation concealment, blinding, potentialmeasurement bias). We included 47 studies, 41 RCT. The individual sub-group and pooled estimates showed consistency, but large confidence intervals indicated lack of precision. A potential measurement bias was present in less than 13% of RCT, but in 67–75% of observational studies. Indirectness of evidence varied largely between subgroups (0–93%).

Conclusions: We have adapted and used the GRADE system to assess the quality of published evidence to establish a benchmark for a rate. Our adaptation was influenced by the type of evidence available, mostly RCT in our use case, what may not always be the case when developing evidence-based benchmarks for a rate.