Body
The Hollywood stereotype of a research scientist is a lonely individual working in a lab late at night surrounded by bubbling flasks. But contemporary scientific research is much more a team effort that relies heavily on data and statistics. In fact, the amount of data available for medical research has exploded in the past two decades as DNA sequencing has become less expensive and more routine, and electronic medical records have become the norm in U.S. hospitals and doctors’ offices.
That’s been a boon to medical researchers everywhere, including at the Center for Rehabilitation Outcomes Research (CROR) at Shirley Ryan 小恩雅, which focuses on research that improves the lives of people with disabilities. CROR’s researchers, who are looking at everything from Parkinson’s disease to employment levels of people with disabilities, frequently reach out to colleagues in the at Northwestern University’s Feinberg School of Medicine.
“Biostatistics are critically important to the success of all research projects,” says CROR Director Allen Heinemann, PhD. “Northwestern’s biostatisticians help our investigators anticipate data management and analysis issues from the launch of the project. I have confidence they are going to get the stats right.”
The Biostatistics Collaboration Center is involved in more than 300 research projects a year at Northwestern, where Heinemann is also a professor. CROR projects stand out, says Center Director and Professor Leah Welty, PhD. “CROR does fascinating, interesting, and important work,” says Welty. “It is often an exciting blend of clinical medicine engineering and health services. We try to do our part by helping them collect good data and analyze it.”
To get the most out of a collaboration, it’s important for researchers to involve biostatisticians from the beginning of the project, Welty notes, rather than when a problem arises during or after data are collected. “What works really well is when we’re involved from the very beginning in terms of designing the study and making sure there are enough participants,” she adds. “When you’re planning a study, you need to think ahead to what question you’d like to be able to answer – what you hypothesize. We can help make sure the study design will allow you to answer that question.”
To some extent, the Biostatistics Center has become a victim of its own success. During the pandemic, for instance, there was so much research that Welty’s staff had to turn away some projects: There just weren’t enough biostatisticians to go around. “During the early pandemic, when scientists and clinicians couldn’t go into their labs or offices, everyone pivoted to working on finishing their papers, so that created a spike,” Welty recalls. “Then there was so much energy going into Covid research, we just got slammed and we had to prioritize.” Fortunately, the center hired more people as Covid wound down. “We’ve gotten back to a more reasonable steady state,” Welty says. “We have a really good team and we can collaborate on research projects in a much more timely manner. Many projects are assigned a biostatistician within a day of request.”
One CROR researcher who has made extensive use of the Biostatistics Center is Dawn Ehde, PhD, a University of Washington rehabilitation researcher who has been studying whether a form of coaching known as cognitive behavioral therapy delivered by phone could help reduce chronic pain for people with a variety of conditions ranging from multiple sclerosis to spinal cord injuries. Biostatistician Elizabeth Gray, MS, helped Ehde’s team create the database that tracked recruitment of participants, record how often they received the coaching and their responses to surveys about pain levels. Other faculty and staff in the Biostatistics Center helped design the project and eventually transitioned it to Gray for her database expertise.
During the study, researchers sent out surveys at various times to participants asking them to rate their pain levels and whether they were still employed. As in almost all research studies, not all participants completed all the surveys, which meant some data were missing. In the research protocol, Ehde’s team had promised to do additional analysis if more than 5% of the data were missing. Gray’s assistance became critical at the end of data collection because slightly more than 5% of participants hadn’t responded to their 12-month follow-up survey, and those were the people who had reported the highest levels of baseline pain. “You want to make sure you take that into account,” Gray says. “If you’re missing the people with the worst pain, that could say something about the generalizability of your work.” Gray used a statistical method that estimates the missing data based on the available data. That allowed researchers to analyze the data and draw more generalizable conclusions.
Indeed, the study found that the virtual therapy did meaningfully reduce pain for the participants, and the team is now writing up the results. Gray is writing the part of the paper that will discuss the team’s methods and results and she will be named as a co-author.
The practice of including statisticians as authors has become much more common in the last few decades when their help has been crucial to the project. “I have to say there’s been a culture shift over time,” Gray says. “It’s usually a good thing to have a statistician on your paper because if there are questions about the stats from reviewers, we can answer them.” Ehde says Gray’s help was invaluable. “Lib Gray was instrumental in analyzing our clinical trial’s results. She approached it with excellent knowledge and attention to detail. That’s so critical to conducting clinical trials.”