More rural, minoritized people get amputations – AI gets closer to why
Researchers hope the model will help avoid preventable amputations from vascular disease
University of Maryland
image:
Dr. Paula Strassle, Assistant Professor of Epidemiology at the University of Maryland School of Public Health.
view moreCredit: UMD
Why do rural adults and racial and ethnic minorities with vascular disease get major leg amputations more often? A new study out today in Epidemiology uses AI to solve the mystery, finding an unaccounted-for factor that researchers think points to implicit bias in the clinical decision-making process.
“The AI model allowed us to distinguish among the many reasons behind these much higher rates of amputation among certain groups of people with vascular disease,” said Paula Strassle, lead author and assistant professor of epidemiology at UMD’s School of Public Health. “We found that, after accounting for everything else, people’s unconscious biases are likely behind why some groups receive amputation instead of alternative treatment that preserves their limb.”
“We hope our results will be a catalyst to create evidence-based guidelines that help vascular surgeons and other providers who make this life-changing decision do so objectively.”
More than 12 million adults in the US live with a vascular disease called Peripheral Artery Disease (PAD), a chronic circulation condition that restricts blood flow to the limbs. It results in leg pain, numbness and in severe cases, limb loss. About 10% of people with PAD develop Chronic Limb-Threatening Ischemia (CLTI) at which point either they receive a procedure to restore blood flow to their lower leg or their limb must be amputated. Revascularization is a surgical procedure that can save the limb, but it also requires intensive follow-up and is a relatively expensive surgical procedure. Vascular surgeons are also in short supply.
After accounting for known differences in clinical presentation, the study found higher rates of amputation among Black, Hispanic, Native American, and white people in rural areas as well as among Black and Native American people in urban areas. After further accounting for differences in hospital and neighborhood resources, higher amputation rates persisted among Black, Hispanic, and Native American people in rural areas, and Black and Native American people in urban areas.
“We found a substantial unexplained portion that would suggest an implicit bias in clinical decision-making occurring at the physician and hospital level,” Strassle said.
The study examined hospitalizations between 2017 and 2019 of people under 40 with PAD or CLTI, across five states (Florida, Georgia, Maryland, Mississippi and New York) using State Inpatient Databases from the Healthcare Cost and Utilization Project.
Researchers programmed an AI model to consider a huge number of variables (70+) that contribute to known reasons for differences in leg amputations of people with PAD. Variables included clinical factors such as age and other health conditions, healthcare system capacity to perform revascularization and limb amputations, legal and regulatory climate, and the physical environment such as a person’s distance to the nearest emergency room and ZIP code median income.
“This AI model will allow us to easily assess intersectionality across race, sex, income and rurality, and offers us the ability to indirectly study hard-to-measure causes of disparities, like implicit bias and stereotyping,” said Strassle.
Limb-threatening conditions are often the result of decades of difficult-to-control diseases like diabetes, high cholesterol and nicotine dependence. For surgeons, who know these conditions lead to worse surgical outcomes, this can make the decision to pursue a complex limb-saving surgery even trickier.
“As vascular surgeons we have surgical guidelines, but we don’t have detailed guidelines to help us make the decision between amputating someone’s leg and limb-saving surgery in patients who are not medically ready. Given the number and complexity of variables involved, we need more information describing the optimal treatment for each person in different conditions. We need to know we can perform a successful vascular operation, and also not increase the risk of dying,” said Katharine McGinigle, a vascular surgeon, associate professor of surgery at the University of North Carolina and senior author of the paper.
“There are so many medical, surgical, and social factors that contribute to disease progression, limb-loss and even death. Surgeons and others making treatment recommendations deserve evidence-based guidance that will help us avoid unconscious biases and make the right decision at the right time for each person based on their unique clinical and social needs. AI methods, similar to the one used in this research, can help us achieve that goal,” said McGinigle.
Strassle and McGinigle hope that their findings will inform comprehensive guidelines and health policies that help clinicians avoid unconscious bias and other unjustified differences in the quality of care provided, to safely save limbs of people living with advanced vascular disease.
To request an interview with Dr. Strassle, please contact sph-comm@umd.edu.
***
Funding for this project was provided by the Division of Intramural Research, National Institute on Minority Health and Health Disparities, National Institutes of Health. The opinions expressed in this article are those of the authors and do not necessarily reflect the views of the National Institutes of Health, the Department of Health and Human Services, or the United States government.
Journal
Epidemiology
Method of Research
Computational simulation/modeling
Subject of Research
People
Article Title
Disaggregating health differences and health disparities with machine learning and observed-to-expected ratios: An applied example using major lower limb amputation
Article Publication Date
9-Jul-2025
