A new study by some researcher conducted at the Stanford University School of Medicine and their collaborators reports that a new algorithm can identify if a patient is likely to have a cholesterol-ringing genetic disease which causes early, and at times, fatal heart problems.
Familial hypercholesterolemia as the disease is called often misdiagnosed as garden-variety high cholesterol.
As Joshua Knowles, MD, Ph.D., assistant professor of cardiovascular medicine at Standford said, they had a belief that less than 10 percent of the people with FH in the United States know that they have it. As it turned out, it was a critical oversight because an FH patient with high cholesterol has three times more likelihood of developing early heart disease than another person with high cholesterol but have no FH. Someone with FH has ten times more the risk of heart attack disease than someone with normal cholesterol.
Joining forces with another associate professor of medicine and biomedical data science, Nigam Shah, Knowles came up with a solution to help grab hold of more cases of FH. They came up with a computer algorithm that flags patients who are likely to have the disease. When they ran some tests with the algorithm, the Al correctly recognizes 88 percent of the cases it screened. In theory, if the algorithm were to be used in a clinic, any patient it flagged as having FH could undergo further genetic testing to verify the calculation of the Al.
When there is no intervention, about 50 percent of men with FH have a heart attack by the age of 50, and by the age of 60, around 30 percent of women have a heart attack. Early and swift diagnosis and treatment of the disease, however, can essentially neutralize this threat. It is also important to catch the disease before it is too late and this is where Shah and Knowles believe their algorithm could make an impact.
After they have trained the algorithm, the team then moved on to the testing phase and ran it on a set of roughly 70,000 de-identified patient records it has never encountered. Among the flagged patients, the group reviewed 100 patient charts, extrapolating that the Al has detected patients who have FH with 88 percent accuracy.
Then, they teamed up with the Geisinger Healthcare System to test the algorithm on 466 FH patients and 5,000 non-FH patients. Shah explained that the predictions came back with 85 percent accuracy, and they knew that several of the patients in Geisinger had a confirmed FH diagnosis with genetic sequencing and they have a complete conviction that the Al worked.
Both Shah and Knowles are working towards how they can implement the Al in the offices of the doctors, and they are actively pursuing FH clinic at Standford.
