The Cancer Letter: ACS is using an AI-powered platform in the largest-ever population study of Black women

Time and resources freed up from data management will go to outreach and access

VOICES of Black Women, the largest population study of Black women in the United States, will be the first of American Cancer Society’s large-scale population studies to be initiated using an AI-driven data management platform—promising to bring observational cancer research out of the age of Excel data files and email sharing.

In the VOICES study, initiated in May, ACS aims to follow over 100,000 Black women living across 20 states and in Washington, DC (The Cancer Letter, May 10, 2024).

The society is collaborating with Manifold Inc., a Massachusetts-based applied AI company, to build an AI-driven clinical research platform that streamlines and automates much of what used to be done by hand.

In previous ACS-led observational studies—the most recent of which was initiated in 2013, two years before the release of the first iPhone—ACS investigators relied on manual data collection, data entry, and management.

Such workflows have been sufficient in the past, but they are becoming cumbersome and time-consuming with the emergence of multimodal data, such as genomic information and environmental factors.

In previously initiated population studies, “we are having to manually bring together all these disparate data sources to be able to even understand whether we can ask and answer a scientific question,” Alpa Patel, senior vice president of population science at ACS, said to The Cancer Letter.

“As a researcher, if I want a simple number: ‘How many people in our cohort were diagnosed with breast cancer, have genomics data and/or have a blood specimen through which I can generate those genomics data, plus have these other exposures?’” Patel said. “We are having to manually bring together all these disparate data sources to be able to even understand whether we can ask and answer a scientific question.

“Then, it’s an extremely manual process to be able to conduct that kind of analysis. We have not modernized the way we work with data, even though the data available to us in the modern era are so vast. This has created this issue where we have a tremendous inefficiency.”

AI-driven technologies have the potential to address these inefficiencies.

“Imagine a future state where you get data—whether it’s generated from a biologic specimen or provided to us from a participant or through some other data source—those data are interconnected at the individual level, ready for analytic use. And then, on the backend, you have an analytic environment where you can safely have researchers come in and work with those data,” Patel said. “Think about the acceleration of science in that future state.”

ACS has implemented Manifold’s technology with its previous longitudinal observational studies, retroactively fitting the sometimes decades-old data into the new platform. However, the VOICES study is the first ACS study using the data system from the get-go.

“With VOICES of Black Women, we started in this environment, so there’s no retrofitting or required, which is phenomenal,” Patel said. “That integrated experience for not only us in study management and research, but also for the participant, is a more seamless interaction.”

Designed to better understand the multilevel drivers of cancer incidence, mortality, and resilience in Black women, the long-term VOICES study will gather valuable data from Black women between the ages of 25 and 55 from diverse backgrounds and income levels who have not been diagnosed with cancer.

The study will allow for a more nuanced exploration into the drivers of cancer disparities in Black women, said Lauren McCullough, co-principal investigator of the VOICES study and associate professor of epidemiology at the Rollins School of Public Health at Emory University.

McCullough co-leads the VOICES study with ACS’s Patel.

Said McCullough:

One of the things that I am excited about with VOICES is the ability to really dig into the lived experience of women. In health, we have largely compartmentalized things. Are you being screened? Do you see a healthcare provider? But trying to understand those answers in the context of other things, we haven’t really looked at. Those other things could be, “Well, no, I didn’t go to my screening, because I have dependent care issues.”

Being able to contextualize some of these things that we know happen around access to care and quality of care, but we can’t come up with an intervention because we don’t know the backstory behind it.

I’m excited for VOICES because I think we’ve done a really good job of being thoughtful about what are the lived experiences, what are the barriers, and then how can those things actually impact health? That is the goal.

Those things can range from social and communal community support to dependent care and elder care, to experiences of racism and discrimination. All of those things are important. We need to stop compartmentalizing people as “You are your health screening, you are your medical condition,” and start looking at more holistic approaches to understanding how all of these things intersect and interact to influence health outcomes.

The data collected for the VOICES study will be housed entirely within the data platform supported by Manifold’s technology, which allows for better data security, and in turn, could help build trust with participants.

“There is a higher level of data governance and data security when you are self-contained,” Patel said. “Being able to work within that integrated platform, being able to really explain to participants how we take sensitive identifying information and separate it from all of the other data that we may use analytically.

“Not having data live in a multitude of places absolutely adds to both the real and perceived trust in that data environment, because we’re able to explain to participants, ‘This is the environment in which your data are contained, this is how we segment your data, how we manage them, how access works, how we deal with security and protections around privacy,’” Patel said.

“The entire platform is built in a HIPAA-compliant environment. So, there’s a lot that we can say today that we weren’t able to say a decade ago, had we launched this study then.”

Public health registries can be easily and automatically integrated into the data environment, and so can data that is trickier to obtain.

“If we know a woman’s address, you can pull in variables like: what is the green space where they live? Where are the parks? Where are the sidewalks? You can pull in things about the built environment,” McCullough said. “You can pull in things about the toxic environment. We can assess air pollution by leveraging the existing monitors. Having the technology makes it easy to seamlessly integrate our survey data with data that are publicly available that also affect health.”

Not only does such integration create a richer dataset, it removes the need for self-reporting of those variables.

“It also allows us to alleviate that space on our survey,” McCullough said. “We don’t need to now ask you, ‘Do you have a sidewalk near your home?’ We can use that technology to actually look and see, ‘Do you have a sidewalk near your home?’ Which allows us the space to ask other questions that only participants can answer.”

Efficient use of survey questions has been top-of-mind for population researchers, since surveys have traditionally had to be filled out on paper, by hand. With this new data platform, participants can be asked only the questions that are relevant to them—and fill out the surveys online.

“When we talk about recruiting 100,000 women, particularly Black women who have multiple demands on their time, we want to think about things that they’re going to be engaged and excited to do, and things that are going to be super efficient; right?” McCullough said. “We don’t want anybody to sit down with a 50-page paper survey.

“By having the technology, it allows us to be able to recruit in a way that resonates with people and resonates with the times, but also allows for that efficiency,” McCullough said. “So, for example, if a woman hasn’t yet reached menopausal age, there are skip patterns, there’s logic built into the technology that allows them to skip that entire battery of questions.”

I’m excited for VOICES because I think we’ve done a really good job of being thoughtful about what are the lived experiences, what are the barriers, and then how can those things actually impact health? That is the goal. - Lauren McCullough

The time and resources saved by utilizing this more advanced technology will be redirected to better outreach and access.

“Our focus, now that we’ve stood up the environment, now that we’ve stood up everything that connects that participant experience, we can focus our attention on building that trust and bringing women to the study, versus, all of the manual data work that we may have had to do in terms of data checks and did these data get processed correctly and so on,” Patel said. “Rather than having to dedicate time and resources to that, we’re actually able to redirect that time and those resources to our enrollment efforts, which is a huge benefit that we’re immediately realizing.”

Moreover, features of Manifold’s new technology itself could help enrollment. Specifically, the HIPAA-compliant data environment allows for virtual consenting, eliminating physical barriers to study participation.

“With our Cancer Prevention Study-3 or CPS-3, we had to capture consent on paper, in writing. So, people had to go to a physical place,” Patel said. “That’s a barrier that we’ve immediately eliminated by having this be entirely online. You can join from wherever you are.

“Without that in-person component, we really can reach into communities across the 20 states and DC through a number of different ways. At the American Cancer Society, we have a number of partnerships, whether they be faith-based or with civic organizations, corporate entities, health systems, and so on. So, we work through our partners that are trusted entities in the communities in which they live, and they’re able to help us reach out into their communities.”

“There is a significant amount of R&D investment that needs to go into this. That’s why it’s hard for one institution alone, or even one technology company alone, and why it hasn’t happened yet,” Vinay Seth Mohta, CEO of Manifold Inc., said to The Cancer Letter.

Manifold announced $15 million in Series A funding in April.

“This is an investment,” ACS’s Patel said. “It takes a lot to be able to find those data solutions. But it was a solution that we really felt was necessary to create.”

Indiana University Melvin and Bren Simon Comprehensive Cancer Center and Winship Cancer Institute of Emory University have also utilized Manifold’s platform.

“It’s always scary to step out and do something new, especially with technology, because technology fails you; right?” McCullough said. “We saw with CrowdStrike. It’s never perfect. But what I think studies like VOICES do is it takes an opportunity, we work out the kinks, and it provides a template for other future studies.

“As we have these technologies and researchers step out on faith, it provides opportunity for future studies, bigger studies, and that’s how we get better and better research.”

Interested in learning more? Connect with us.

This article was originally published by The Cancer Letter on August 2, 2024. Visit The Cancer Letter for the original version.