Effective study design and analysis are critical to any research project, and statistical literature reflects many methodological advancements for a range of study designs in the last ten to fifteen years. But scholarly articles about statistical methods are rarely published in journals commonly read by public health and global health researchers.
Liz Turner, assistant professor of biostatistics, bioinformatics and global health, wanted to change that. Last year, she approached David Murray, director of the Office of Disease Prevention at the National Institutes of Health, about updating his widely-cited review of methodological developments in group-randomized trials, published in the American Journal of Public Health in 2004. Murray embraced the idea.
“It’s very common for public health and global health researchers to conduct group-randomized trials, but design and analysis of these projects can be tricky,” said Turner. “We wanted to bring some of the advances that have been documented in the statistical literature to a forum that these researchers are more likely to access, so they’re better equipped to conduct effective studies.”
Two-Part Series Focuses on Design and Analysis
Last month, Turner and her colleagues, including Murray, published a two-part update in the same journal. Other Duke authors include John Gallis, a biostatistician at the Duke Global Health Institute (DGHI) and the department of biostatistics and bioinformatics, and Fan Li, a PhD student in biostatistics.
The series focuses on group-randomized trial methodologies and developments since Murray’s 2004 paper. In group-randomized trials, also known as cluster-randomized or community-randomized trials, the unit of randomization is a group of individuals—such as a village, school or hospital—and outcomes are measured for members of the group.
Part One outlines developments in the design topics addressed in the earlier review, such as clustering, matching and sample size. It also covers new topics, such as constrained randomization and alternative designs that can be used to overcome challenges that traditional design solutions may not be able to address. These alternative designs include stepped-wedge cluster-randomized trials, network-randomized trials and individually randomized group-treatment trials.
Part Two covers problems that arise in the analysis of group-randomized trials. As with Part One, the article provides updates on the topics in the earlier review, such as methods for parallel-arm cluster-randomized trials and missing data, as well as sharing insights on new topics, such as alternative estimation methods.
Turner says the series was partly inspired by complicated research questions that arose during collaboration with Wendy Prudhomme-O’Meara, associate professor of medicine and global health, on a cluster-randomized trial on the targeting of anti-malarial drugs in western Kenya. The series was partially supported by the same National Institute of Health grant that’s funding Prudhomme-O’Meara’s study.
Articles Are Extension of DGHI’s Research Design and Analysis Core
Turner directs DGHI’s Research Design and Analysis Core (RDAC), a group of biostatisticians and global health researchers who collaborate with the Institute’s faculty and students in designing and analyzing studies and developing grant proposals that incorporate a wide range of research methods.
“We serve as translators of the more technically complex methodologies,” Turner said. “Our goal is to help our faculty and students find answers to critical research questions around global health issues.”
In addition to assisting faculty and students one-on-one with their research projects, RDAC members develop translational and educational resources. These pieces range from one-page guides on topics such as how to implement logistic regression to articles in peer-reviewed journals—like this two-part series—that discuss statistical methods in detail.
“We’re continuously looking for opportunities to share the lessons we’ve learned to help educate researchers in the field, whether they’re experienced faculty members or graduate students, and these articles are an example of that,” Turner said. “I hope public health and global health researchers will find them helpful in enhancing the quality of their research.”
We wanted to bring some of the advances that have been documented in the statistical literature to a forum that public health and global health researchers are more likely to access.Liz Turner, assistant professor of biostatistics, bioinformatics and global health