3 keys to tackling bias in health data and algorithms and why it matters
Newswise – Underserved and marginalized populations in the United States and around the world have long suffered from unequal access to basic human health care needs due to factors that should not be relevant: race, gender, origin ethnicity, geographic location and income, says Maryland Smith. Ritu Agarwal in a recent American Medical Association”go around” Podcast.
The podcast explored a wide range of questions, says Agarwal, that medical students, researchers and practitioners should think about as the digitalization of healthcare accelerates. “When you juxtapose this presence of health inequities with the digital transformation of healthcare – with a growing reliance on data and algorithms – ‘boom,’ we have the potential for a very serious problem that we need to be very, very vigilant and repair,” she adds. “It’s essential to increase the diversity of the datasets that researchers use and that the algorithms are based on.”
One dimension of this question is illustrated in a 2021 study racial disparity in COVID-19 vaccine distribution co-authored by Agarwal, University Professor Emeritus, Dean Robert H. Smith Chair in Information Systems and co-director of the Center for Health Information and Decision Systems (CHIDS) at the Robert H. Smith School of Business at the University of Maryland.
In the podcast, Agarwal says that when vaccines first started coming out, black people were vaccinated at much lower rates than white people and this – through the media – was widely attributed to distrust of the healthcare system. . “I disagreed with this narrative, and felt we needed to debunk it or else find the data to support it. I had a hunch that it was the structural barriers and other determinants health institutions more than anything else that were causing the disparity in vaccination rates between different races… We collected a unique data set and addressed this question.In a nutshell, we found that politics and privilege , socio-economic status and education matter.
Agarwal’s work – published in the Proceedings of the National Academy of Sciences of the United States – “exposed in explicit terms the racial and socio-economic disparities in [COVID] distribution and use of vaccines,” noted his AMA interviewer and University of Toledo medical student Kristofer Jackson. “This scholarship made a significant difference in how we rolled out the vaccine for the underserved,” he added. “It was a tool to really drive change.”
How can such a change be reinforced? Agarwal described three pathways for students and physician-researchers to health equity:
Ask the right questions for any automated algorithm used for diagnosis and frequent decisions. “On what data were they trained? What is the quality of the data? What is the representativeness [the data] of the patient population? In the genome-wide association studies78% of the data comes from people of European descent and only 2% from blacks and 1% from Hispanics… You check the quality of your instruments to ensure that they are not malfunctioning, but you must be literate and informed about algorithms and data too.
In clinical work, capture all of the social determinants of health in your unstructured notes. “This will be the gold mine of data that will drive future algorithms and inferences. Our research at the center (CHIDS) has shown that EHR data captures these factors about marginalized populations at significantly lower rates than the majority. It will be your responsibility to ensure that you understand the full range of health care influences in which your patients’ clinical conditions are presented.
Engage vigorously with the NIH AIM FORWARD (Artificial Intelligence/Machine Learning Consortium to Advance Health Equity and Researcher Diversity). And commit to its mission to uncover biases in data and algorithms to ensure that the scientists and individuals who help transform healthcare through AI and machine learning come from all walks of life, all walks of life, all genders, all races, all ethnicities”.
“Health informatics, AI and ML play a very important role in addressing inequalities,” says Agarwal. “Even though we often refer to ‘biases in AI and ML’, it’s also essential to keep AI and ML in mind as instruments of positive change, too, for us. help overcome our prejudices.”