Record Details

Peer Review of Social Science Research in Global Health: A View Through Correspondence Letters to The Lancet

Harvard Dataverse (Africa Rice Center, Bioversity International, CCAFS, CIAT, IFPRI, IRRI and WorldFish)

View Archive Info
 
 
Field Value
 
Title Peer Review of Social Science Research in Global Health: A View Through Correspondence Letters to The Lancet
 
Identifier https://doi.org/10.7910/DVN/28144
 
Creator Fan, Victoria
Silverman, Rachel
Roodman, David
Savedoff, William
 
Publisher Harvard Dataverse
 
Description In recent years, the interdisciplinary nature of global health has blurred the lines between medicine and social science. As medical journals publish non-experimental research articles on social policies or macro-level interventions, controversies have arisen when social scientists have criticized the rigor and quality of medical journal articles, raising general questions about the frequency and characteristics of methodological problems and the prevalence and severity of research bias and error. Published correspondence letters can be used to identify common areas of dispute within interdisciplinary global health research and seek strategies to address them. These letters can be seen as a “crowd-sourced” (but editor-gated) approach to public peer review of published articles, from which some characteristics of perceived bias and error can be gleaned. In December 2012, we used the online version of The Lancet to systematically identify relevant correspondence in each issue published between 2008 and 2012. We summarize and categorize common areas of dispute raised in these letters. The five most-cited concerns are: measurement error (51% of papers); omitted variables and confounding (45%); implausibility and lack of external validity (43%); missing or low-quality data (32%); and lack of transparency of methods (30%). We offer several recommendations, including the use of checklists and guidelines to facilitate better documentation of areas of potential bias; deployment of econometric-specific reviewers where appropriate; and explicit online linkage between all correspondence letters and the original articles to which they refer. Most importantly, we recommend that The Lancet adopt the replication standard, whereby the data and the coding used to produce the estimates are provided at least to the journal, for reviewers to analyze and replicate the e
stimates reported by the authors, and ideally to the public, as the leading economics journals now require.
 
Date 2014-06