Disagreements in meta-analyses using outcomes measured on continuous or rating scales: observer agreement study

Ref ID 467
First Author B. Tendal
Journal BMJ
Year Of Publishing 2009
URL https://www.bmj.com/content/bmj/339/bmj.b3128.full.pdf
Keywords • Cochrane
• Author
• General medical
• Error
• Expertise
• Statistical
Problem(s) • Errors in effect estimate calculations or data synthesis
• Lack of statistical expertise in handling of quantitative data
• Data extraction errors and double counting
Article Type Empirical
Article Subtype Inter-rater agreement study
First Author Country Denmark
Aim To assess the inter-observer variation between 5 experienced methodologists and 5 PhD students for extraction and calculation of the first standardised mean difference related to extraction of continuous and numerical rating scale data from trial reports for use in Cochrane meta-analyses.
Level of Investigation Analytical
Summary of Findings Agreement was 53% at trial level and 31% at meta-analysis level. Reasons for disagreement reasons for disagreement were broadly related to: different choices, exclusion of a trial, and data extraction errors. In 14 out of the 100 standardised mean differences calculated at the meta-analysis level, individual observers reached different conclusions than the originally published review.
Number of systematic reviews included 10
Number of eligible systematic reviews assessed 32
Treatment impacted Yes
Treatment impacted description
Interpretation impacted Yes
Interpretation impacted description Ten per cent of the disagreements at the meta-analysis level amounted to an standardised mean difference of at least 1, which is far greater than the effect of most of the treatments we use compared with no treatment