Stroke can be difficult to diagnose because patients don’t always present with the classic symptoms and other conditions can mimic it. developed a machine learning model that accurately predicts and facilitates diagnosis.
Diagnostic errors pose a significant public health problem, leading to preventable patient harm and wasted health care costs. Preventable deaths from misdiagnosed stroke are 30 times more common than misdiagnosed heart attacks.
Stroke can be particularly difficult to diagnose because its signs and symptoms can resemble other conditions, such as seizures, migraine headaches, psychiatric disorders, and drug or alcohol intoxication. In addition, stroke can present with atypical symptoms. About 25% of stroke patients do not present with the usual speech problems, drooping face, and weakness of the extremities, further complicating the ability of health care providers to make an accurate diagnosis.
Researchers at Carnegie Mellon University, Florida International University, and Santa Clara University have used machine learning techniques to develop an automated screening tool to take the guesswork out of diagnosing stroke.
“Machine learning techniques have been used to help detect stroke by interpreting detailed data such as clinical records and imaging results,” said Rema Padman, corresponding author of the study. increase. “However, when patients are first triaged in hospital emergency departments, such information may not be readily available, especially in rural or underserved areas.”
To develop a stroke prediction algorithm, researchers used records from more than 143,000 individual patients admitted to acute care hospitals in Florida between 2012 and 2014. We also incorporated data from the American Community Survey conducted by the US Census Bureau, including demographics such as age. , gender, race, and pre-existing medical conditions.
A machine learning model predicted stroke with 84% accuracy. It was also more sensitive and superior to existing diagnostic models, which tend to miss up to 30% of strokes.
“The moderate sensitivity of existing models raises concerns that they miss a significant proportion of stroke patients,” said Min Chen, the study’s lead author. “In hospitals where medical resources and clinical staff are scarce, our algorithms will complement current models to help quickly prioritize patients for appropriate intervention.”
The findings suggest that the machine learning model can accurately predict the likelihood that a person has had or is likely to have a stroke, even before imaging and laboratory confirmation. .
“Because our model does not require clinical records or diagnostic test results, it may be particularly useful in addressing the challenge of misdiagnosis when working with walk-in patients with mild and atypical symptoms,” he said. study. “Emergency departments in small or non-stroke centers where health care providers have limited daily exposure to stroke, and rural areas with limited access to sensitive diagnostic tools, may also be useful.”
However, the researchers point out that their algorithm is not intended as a standalone model. It should be used in conjunction with existing models of stroke diagnosis.
The researchers recommend incorporating their stroke prediction algorithm into an automated, computer-assisted screening tool accessible at the time of admission.
This research Journal of Medical Internet Research.
Source: Carnegie Mellon University
