With $1.2 million in funding from the National Library of Medicine, researchers at the Medical University of South Carolina (MUSC) and Clemson University will establish a new training program that aims to make future data scientists more aware of health inequities. It will also build career development pipelines in biomedical data science for students from underrepresented minorities. The program will place special emphasis on using data science to address the toll chronic illness takes on rural communities.

South Carolina is the ideal location for a training program focused on addressing health inequities. Forty-three of its 46 counties, many of them rural, are designated as completely or partially medically underserved by the Health Resource and Services Agency (HRSA). South Carolina ranks 42^nd for life expectancy (Centers for Disease Control and Prevention), due in part to its high levels of chronic disease. The state has the eighth highest rate of diabetes (South Carolina Department of Health and Environmental Control) and sixth highest rate of stroke deaths (CDC) in the nation.

The leaders of the SC BIDS4HEALTH training program believe that harnessing big data could help to change that.

“Informatics and data science can be used to identify patients in need of extra health system resources. They can also help to identify areas within the health system where we are not as efficient in serving specific populations who are experiencing health inequities.” — Alexander Alekseyenko, Ph.D.

“Informatics and data science can be used to identify patients in need of extra health system resources,” said Alexander Alekseyenko, Ph.D., principal investigator of the new program. “They can also help to identify areas within the health system where we are not as efficient in serving specific populations who are experiencing health inequities.” Alekseyenko is a professor of Public Health Sciences in the MUSC College of Medicine and a member of the Biomedical Informatics Center.

“Programs such as SC BIDS4Health are essential in addressing the current lack of diversity in STEM fields. This program is a game-changer in showing brilliant young students the career possibilities that are open to them.”   — Marvella Ford, Ph.D.

The training program will build on and recruit from the MUSC and Clemson joint Biomedical Data Science and Informatics (BDSI) program as well as from historically black colleges and universities (HBCUs) statewide. When in full swing, it will train three predoctoral students and two postdoctoral fellows each year.

Brian Dean, Ph.D., professor in the School of Computing at Clemson, is a co-director of the training program. Marvella Ford, Ph.D., the associate director of Population Sciences and Cancer Disparities at the Hollings Cancer Center and the SmartState Endowed Chair in Prostate Cancer Disparities at South Carolina State University (SCSU), is the associate program director.

 

In addition to offering new courses on health inequities, the program will build on existing relationships with SCSU and other HBCUs and community groups across the state to build career pipelines and address health inequities using data science.

“Programs such as SC BIDS4Health are essential in addressing the current lack of diversity in STEM fields,” said Ford. “This program is a game-changer in showing brilliant young students the career possibilities that are open to them. We will leverage the training opportunities provided through the SC BIDS4Health program to produce more equitable career choice outcomes. The potential impact is enormous as the SC BIDS4Health participants will then serve as role models for younger learners in their communities.”

Trainees who will be mentored by seasoned data scientists will in turn teach data science to undergraduates at SCSU, introducing these students to careers in biomedical data sciences. Trainees will also spend time embedded in the community, working with local groups to use data science to tackle health inequities and chronic disease. For example, they can use artificial intelligence (AI) algorithms to predict which approaches to addressing chronic disease inequities are most likely to succeed.

“AI can create inequity if you’re not careful. So I think that an important component of the program is how you can leverage AI and data science in the right way to address these sorts of issues.” — Brian Dean, Ph.D.

“Exposing biomedical data science and informatics students and postdocs to formal training, mentoring and projects directed to rural and underserved communities across South Carolina, and beyond, will have a dramatic impact on improving access to care, health equity and health outcomes,” said Ron Gimbel, Ph.D., director of Clemson Rural Health, who is on the steering committee for the training grant. Gimbel is also associate director of the South Carolina Clinical & Translational Research Institute’s Dissemination and Implementation Science Collaborative. 

 

Overall, the program will help to ensure that communities of color reap the benefits of the transformative power of AI and its ability to turn big data into fresh insights into disease. AI algorithms learn by consuming large amounts of data. Once trained, they can “see” patterns in new data that a human eye might miss. AI algorithms can already recognize images of melanoma lesions more accurately than dermatologists. They can also help to predict future cardiovascular disease and speed drug discovery. 

Despite its great promise, AI also has the potential for harm. The accuracy of AI algorithms depends in large part on the quality of the data used to train them. If that data is incomplete, flawed or biased, AI will pick up and amplify those biases and will not detect disease as well in minority patients or groups. For example, an AI algorithm trained on data from largely White patients with cardiovascular disease might not detect disease appropriately in people of other races.

“Exposing biomedical data science and informatics students and postdocs to formal training, mentoring and projects directed to rural and underserved communities across South Carolina, and beyond, will have a dramatic impact on improving access to care, health equity and health outcomes. — Ron Gimbel, Ph.D.

If medicine comes to rely more on these AI algorithms, as is expected, such biased data sets could result in chronic disease going undiagnosed and untreated in communities of color.

“AI can create inequity if you’re not careful,” said Dean. “If you have a smart medical decision-making system that’s trained on data and the training data doesn’t represent the population in the right way, then, the actual output of the decision-making system can actually create bias. So I think that an important component of the program is how you can leverage AI and data science in the right way to address these sorts of issues.”

The new program will help to guard against bias in AI by raising awareness about racial disparities in trainees and teaching them techniques for minimizing them, said Dean. The close collaboration with community organizations will enable trainees to see firsthand the consequences of health inequities and apply their newly developed skills to address them.

The SC BIDS4HEALTH program will enroll its first trainees on July 1. More information is available on the BDSI website or by emailing Alekseyenko.