Dealing With Data Biases and Overwhelm in Clinical Research

By Allison Proffitt 

March 29, 2023 | At a panel presentation during the ViVE event in Nashville, Tenn., on Monday, Kulleni Gebreyes of Deloitte, hosted a discussion about unbiased data in health care.  

Alisahah Jackson, ​President Lloyd H. Dean Institute for Humankindness & Health Justice, CommonSpirit Health, challenged the foundational idea right away. “There’s a myth around unbiased data,” she said. Data are generated and input by humans with implicit biases, and so can never be wholly unbiased, she contended.  

At CommonSpirit Health, Jackson said she is working to counteract those biases by diversifying their internal team—everyone from IT to nursing teams to the clinical workforce—and engaging partner organizations to help further stretch thinking. She highlighted an expanded partnership with Morehouse School of Medicine to address the underlying causes of health equity, including underrepresentation of Black clinicians.  

Diversifying data inputs certainly helps balance implicit biases, but—in healthcare—that comes with challenges as well. John Kraus, Executive Vice President, Chief Medical Officer, Otsuka America Pharmaceuticals, explained that often in diseases, we don’t know what we don’t know.  

The disease data we start with may not be representative, he explained. “The epidemiology may or may not reflect the diversity of the general population,” he said. “We have to make sure we have the right population in the trial,” he added, so that we are getting the full picture of a disease.  

He outlined some of the tools Otsuka uses to do that. Among them: e-consent to better explain trials, decentralized trials, and trials conducted outside of traditional sites, for instance through pharmacies. “The whole goal is to try and make [clinical research] less restrictive… more inclusive,” he said.  

Pharmacy Role 

Ramita Tandon, Chief Clinical Trials Officer at Walgreens, certainly shares Kraus’s vision for pharmacies’ role in trials. Walgreens has close access to 78% of the U.S. population, she said, and she reiterated the pharmacy chain’s success in patient education and care delivery around covid-19 vaccines.  

“How do we improve access? We’re in a very unique position that we’ve got access to 120 million lives. We’ve all seen the statistic… but the lion’s share of this nation has not participated or been invited to clinical research. More work needs to be done.”  

Walgreens has converted some of their locations to clinical trial centers, and Tandon says the sites are well-equipped to host “low-complexity visits.” The pharmacy chain is open to partnerships with health systems, payers, and advocacy groups, Tandon said, emphasizing existing strong relationships between pharmacists and their patients. She also highlighted Walgreens’ investment in home health. (See, “Walgreens’ Decentralized Clinical Trial Future”).  

But the challenges are not only getting more information; in some cases we need to refine the information we have and review our processes for gathering it.  

Accurate Data 

“Let’s challenge our assumptions that we are already collecting accurate data on some basic demographic data,” Jackson added. “When you look at our accuracy in collection and even our rates of collection around race, ethnicity, language, sexual orientation, gender identity data—they are pretty poor.” Jackson volunteered her own anecdote: when taking two of her children for well-child visits, she realized they had been coded as different races—all without any questions from the healthcare provider. 

Her experience is not an outlier, Jackson warned. The data we are collecting are inherently biased, messy, and incomplete because they are collected by people. But Jackson was not hopeless.  

“When we look at the research on what impacts health, 80-90% of it happens outside of clinical care walls. It happens in the communities where we all live, work,… play, and worship. So how are we really being intentional about collecting data around social determinates of health, not only to help improve [care for] the people who are in front of us that we are providing services to, but the communities to which they are going back.”  

Accurate and complete data are empowering, said Shannon West, Chief Product Officer, Datavant. “Without a complete picture of either an individual or a complete picture of a population that we’re looking at…, we miss out on the moments that create experiences—both within clinical trials and care delivery and, quite frankly, payments—to make sure we’re making equitable decisions in healthcare.”  

Datavant builds privacy-preserving record linkages to support clinical trials as well as a network of connected providers that enable both patients and providers to access records. Real patients don’t actually want all the technological bells and whistles, West reported. They want control over their data and the assurance that their anonymity is protected. 

“As we’re building products and as we’re thinking about increasing data access, it means that we’re supporting anyone in that journey in moving toward fully-digitized medical records.”  

Gebreyes did an astute job summarizing the full panel. We must gather data that are accurate and fit-for-purpose. “You have to be clear on the question you’re asking,” she said, when you are collecting data. Those data need to be returned to physicians or patients in ways that are usable and that distinguish signal from noise. Finally, partnerships and collaborations are essential to build diverse, complex datasets.