AI helps to predict risk of a potentially fatal heart disease
Researchers at Stanford—and their collaborators—have developed an algorithm to identify patients’ susceptibility using EHR data.
An artificial intelligence algorithm is able to identify a patient's risk from a cholesterol-raising genetic condition that has a 20-fold increased risk of coronary artery disease if untreated.
Individuals with the potentially fatal heart disease—called familial hypercholesterolemia (FH)—carry a mutation that interferes with their bodies’ ability to clear low-density lipoprotein (LDL), the “bad” cholesterol that can lead to a harmful buildup in their arteries.
However, the challenge is that fewer than 10 percent of people with FH in the United States have been diagnosed with the condition, and without treatment, about 50 percent of men with FH have a heart attack by age 50, and approximately 30 percent of women have a cardiac event by age 60.
To address the problem, Stanford Medicine and the nonprofit Familial Hypercholesterolemia Foundation have joined forces to leverage machine learning and big data.
Specifically, researchers at the Stanford University School of Medicine—and their collaborators—have developed an AI algorithm trained to identify potential FH patients using electronic health record data.
“As part of the FH Foundation′s FIND (Flag, Identify, Network, Deliver) FH initiative, here we report the development and internal validation of a supervised machine-learning algorithm to identify probable FH cases based on EHR data from Stanford Health Care as well as the external validation on this classifier using EHR data from the Geisinger Healthcare System,” state the authors of a paper published online last week in npj Digital Medicine.
According to the study’s authors, the AI algorithm correctly identified 88 percent of the FH cases it screened in an internal test dataset, and it demonstrated 85 percent accuracy with an external Geisinger EHR dataset, which they contend “suggests that application of this classifier could lead to increased efficacy of targeting these high-risk patients for enhanced evaluation and intervention.”
Also See: AI algorithm predicts aneurysm risk from EHR, genetic data
“Theoretically, when someone comes into the clinic with high cholesterol or heart disease, we would run this algorithm,” said senior author Nigam Shah, associate professor of medicine and of biomedical data science at Stanford. “If they’re flagged, it means there’s an 80 percent chance that they have FH. Those few individuals could then get sequenced to confirm the diagnosis and could start an LDL-lowering treatment right away.”
Going forward, Shah and his colleagues are working in partnership with the FH Foundation to deploy the AI algorithm in clinical settings at Stanford Healthcare and at additional sites.
Nonetheless, at the same time, Shah acknowledges that “not everything can be solved by an algorithm,” adding that “we’re also thinking about how we can work with the FH Foundation to implement networks of family screening to reach more patients who might have the disease and not know it.”
Individuals with the potentially fatal heart disease—called familial hypercholesterolemia (FH)—carry a mutation that interferes with their bodies’ ability to clear low-density lipoprotein (LDL), the “bad” cholesterol that can lead to a harmful buildup in their arteries.
However, the challenge is that fewer than 10 percent of people with FH in the United States have been diagnosed with the condition, and without treatment, about 50 percent of men with FH have a heart attack by age 50, and approximately 30 percent of women have a cardiac event by age 60.
To address the problem, Stanford Medicine and the nonprofit Familial Hypercholesterolemia Foundation have joined forces to leverage machine learning and big data.
Specifically, researchers at the Stanford University School of Medicine—and their collaborators—have developed an AI algorithm trained to identify potential FH patients using electronic health record data.
“As part of the FH Foundation′s FIND (Flag, Identify, Network, Deliver) FH initiative, here we report the development and internal validation of a supervised machine-learning algorithm to identify probable FH cases based on EHR data from Stanford Health Care as well as the external validation on this classifier using EHR data from the Geisinger Healthcare System,” state the authors of a paper published online last week in npj Digital Medicine.
According to the study’s authors, the AI algorithm correctly identified 88 percent of the FH cases it screened in an internal test dataset, and it demonstrated 85 percent accuracy with an external Geisinger EHR dataset, which they contend “suggests that application of this classifier could lead to increased efficacy of targeting these high-risk patients for enhanced evaluation and intervention.” Also See: AI algorithm predicts aneurysm risk from EHR, genetic data
“Theoretically, when someone comes into the clinic with high cholesterol or heart disease, we would run this algorithm,” said senior author Nigam Shah, associate professor of medicine and of biomedical data science at Stanford. “If they’re flagged, it means there’s an 80 percent chance that they have FH. Those few individuals could then get sequenced to confirm the diagnosis and could start an LDL-lowering treatment right away.”
Going forward, Shah and his colleagues are working in partnership with the FH Foundation to deploy the AI algorithm in clinical settings at Stanford Healthcare and at additional sites.
Nonetheless, at the same time, Shah acknowledges that “not everything can be solved by an algorithm,” adding that “we’re also thinking about how we can work with the FH Foundation to implement networks of family screening to reach more patients who might have the disease and not know it.”
More for you
Loading data for hdm_tax_topic #better-outcomes...