Medical technology attempts to identify groups of patients at high risk of death post heart attack, but the current methods still fail, pretty badly. Physicians and scientists miss 70 percent of those who are at such high risk post heart attack that they die. This stat comes from Zeeshan Syed, assistant professor of electrical engineering at the University of Michigan. And Syed thinks he and a team of researchers have found a better way. And that their solution might save tens of thousands of lives.
Typically physicians detect high-risk heart attack patients based on overall health, blood work, the echocardiogram. But Syed says the echocardiogram is a “blunt metric” and physicians continue to miss the sickest patients.
The team of researchers from MIT, Harvard, and Brigham Women’s Hospital (Boston) along with Syed used one of the most common tools in cardiology, the electrocardiogram, (a.k.a EKG or ECG.) It’s best known as the machine that keeps constant watch over acute patients while in hospital, measuring the electrical activity of the heart. The team found subtle markers within hours of EKG measurements that can more accurately predict which heart attack patients are at high risk of dying. Most physicians miss these markers because they only check the EKG at specific moments. They don’t routinely examine the long history of EKG recordings.
The scientists analyzed the EKGs from more than 4,500 heart attack patients. And they found that the EKGs of patients who later died from cardiac-related issues had similar aberrant patterns.
Prior to this study such patterns were either thought of as insignificant noise or just impossible to detect because of the enormous amount of data analysis required. “But by using sophisticated computational techniques, we can separate what is truly noise from what is actually abnormal behavior that tells us how stable the heart is,” Syed is quoted in a press release. Their study is published today in the journal Science Translational Medicine.
The team notes that we cannot expect physicians to analyze 72 hours of EKG data. These days the ability to collect medical data has outpaced the expert’s capacity to analyze it.
Using data mining and machine learning techniques Syed and his team determined that there are three key biomarkers to be found EKG data. The first is so-called “morphologic variability” which is the amount of change in the shape of normal heartbeats over a long period of time. The second marker is “heart rate motifs” which are sequences of changes in heart rate that show whether and how the heart is responding to the body’s nervous system. The last marker is “symbolic mismatch” which compares how similar a patient’s EKG is to another patient with a similar medical history.
From their study the scientists found that those patients with one of the abnormalities, as noted by the biomarkers, were two to three times more likely to die within the year.