New method predicts risk of pulmonary embolism in patients
Sheba Medical Center scientists develop algorithm that predicts likelihood of a newly admitted patient suffering a potentially life-threatening PE.
A hospital in Israel is predicting patients’ risk of suffering pulmonary embolism (PE) — a potentially life-threatening lung blockage and the third most common cause of death from cardiovascular diseases – based entirely on Al analysis of their medical records.
The condition is notoriously hard to diagnose because the symptoms – typically a sudden difficulty in breathing, chest pains, coughing blood and light-headedness – can easily be confused with many other illnesses.
But Sheba Medical Center in Ramat Gan, central Israel, has developed an algorithm that accurately predicts, when a patient is admitted to the hospital, the likelihood of that patient suffering a PE.
PE is primarily diagnosed through a CT scan. The algorithm instead uses only information gleaned from medical records: the patient’s age, sex, BMI (body mass index), past clinical PE events, chronic lung disease, past thrombotic events and use of anticoagulants.
PE happens when a blood clot (thrombosis) – usually dislodged from a patient’s legs — travels through the bloodstream and blocks a blood vessel in the lungs. This reduces blood and oxygen flow in the lungs, increases blood pressure in the pulmonary arteries, and potentially damages the heart or lungs.
The research team, writing in the Journal of Medical Internet Research, say late or under-identification of a blood clot in one or more arteries to the lungs seriously threatens patients’ lives and is “a major challenge confronting modern medicine.”
They analyzed data available prior to emergency department admission for 2,568 patients with PE and 52,598 patients in a much larger “control group” in which just 4 percent had PE.
Results from the study showed the algorithm was able to accurately identify and predict which patients were at high risk of PE upon hospital admission, allowing doctors to diagnose and begin treatment early.
“Early and timely diagnosis of pulmonary embolism is challenging, yet crucial, due to the condition’s high rate of mortality and morbidity,” said Prof. Gad Segal, head of the Sheba Education Authority, who conducted the study together with computational development researchers at Ben Gurion University in Beersheva, southern Israel.
“This study highlights the enormous potential of machine learning tools to support innovation in diagnostics. Even though the model only used data available from patients on arrival to the ER, it was still able to predict with high accuracy the likelihood a patient developing PE, a crucial advancement for patient care and outcomes.”
