According to a new research, a computer screening program can accurately diagnose a heart attack even when some information required by the program is missing. In a study of more than 1,000 adults who went to the emergency department for chest pain, the diagnostic tool identified more than 80% of people who were having a heart attack.
The findings suggest that emergency department physicians may be able to rely on the computer program to diagnose patients with chest pain in "real time," the study's authors say. In the study, Dr. William G. Baxt and colleagues at the Hospital of the University of Pennsylvania in Philadelphia tested a computer program called an artificial neural network, which is designed to mimic the human thinking process by adjusting its decision-making based on prior experience.
In the program tested in the study, a doctor enters a variety of information, including a patient's risk factors for heart disease, characteristics of the patient's chest pain and other symptoms, medications, vital signs, results of a physical exam and a measure of the heart's electrical activity called an ECG.
Although doctors are able to identify most cases of heart attack without using this technology, the computer program is intended to reduce the percentage of heart attacks that are misdiagnosed. About 2% to 5% of patients with a heart attack are mistakenly discharged from emergency departments each year.
Several studies have shown that the network can detect heart attack in patients with chest pain, but Baxt and his colleagues set out to see how well the technology worked when some of the information it uses to make a diagnosis was missing.
The researchers applied the technology to 1,400 adults who arrived at the emergency department complaining of chest pain. On average, 5% of the information to be entered into the program was missing.
Despite the missing information, the artificial neural network was highly accurate at identifying patients who were having a heart attack, Baxt's team reports in the April issue of the journal Annals of Emergency Medicine. Based on telephone calls made to patients a month after they had been treated in the emergency department, the screen was almost 95% accurate at detecting heart attack, identifying100 to 150 cases.
Thus the author concludes that the artificial neural network has the potential to be used as a real time aid to identify the presence of (heart attack) in patients presenting to the emergency department with anterior chest pain.