Addressing missing participant data in systematic reviews: Continuous outcomes—Part 2

Article type
Authors
Ebrahim S1, Akl EA2, Johnston B3, Alonso P4, Mustafa RA1, Sun X5, Walter SD6, Heels-Ansdell D6
1McMaster University, Canada
2American University of Beirut, Lebanon
3Hospital for Sick Children Research Institute, Canada
4Iberoamerican Cochrane Center, Spain
5Center for Clinical Epidemiology and Evidence-based Medicine, Xinqiao Hospital, China
6Department of Clinical Epidemiology & Biostatistics, McMaster University, Canada
Abstract
Objectives:

To describe how to use an innovative approach to addressing missing participant data for continuous outcomes in systematic reviews.

Description:

The workshop will consist of the following: [1] An interactive presentation of the proposed approach: a complete case analysis and four imputation strategies for making plausible assumptions about the outcomes of participants with missing data. These strategies increasingly challenge the robustness of the pooled estimates of the intervention effect in the systematic review. We will also describe how to calculate the proportion of patients who have an important treatment effect for each strategy, using the minimally important difference (MID) threshold, and provide guidance on how to use the results to judge the impact of missing participant data on risk of bias. [2] A hands-on exercise using an Excel sheet to apply the approach. The participants will use their own laptops and use data provided by facilitators. We will use two example systematic reviews to illustrate the application of the approach: one review restricted to one instrument measuring the outcome, and another using different instruments to measure the same construct. [3] An open discussion of the advantages and limitations of the proposed approach.