New AI tools help predict recovery time for patients after CTEPH surgery
Analysis specifically looked at accuracy of platform's hospital stay predictions
Written by |
Artificial intelligence (AI) tools developed by Etiometry may help to predict, among other outcomes, the length of a patient’s hospital stay following surgery among people with chronic thromboembolic pulmonary hypertension (CTEPH).
As such, these tools could potentially be used to improve patient care, and better monitor post-surgery recovery, for individuals with this rare form of pulmonary hypertension (PH).
Those are the findings of a new study conducted at the University of Texas Southwestern Medical Center (UTSW) in Dallas, which were shared over the weekend at the 2026 Society of Thoracic Surgeons Annual Meeting in New Orleans.
According to Etiometry, by monitoring physiological data immediately after surgery and identifying individuals at risk of longer hospital stays, the tool may guide decisions for optimizing post-surgery care.
“Our rapidly growing adult data [demonstrate] how Etiometry’s platform consistently delivers value across patient populations and care settings,” Shane Cooke, president and CEO of Etiometry, said in a company press release detailing the presentation.
“This study is one example of how our physiologic indices can be applied in adult critical care to help teams recognize emerging risk sooner, guide management decisions and optimize care pathways,” Cooke said.
CTEPH is a form of pulmonary hypertension in which blood clots result in abnormally elevated pressure in the pulmonary arteries, which carry blood from the heart to the lungs. The gold standard treatment for CTEPH is a surgical procedure to remove disease-driving clots.
Following surgery, patients are carefully monitored, with doctors tracking measures such as blood oxygen levels and heart rate. Etiometry’s platform uses AI algorithms to calculate specific risk scores from this type of physiological data.
According to the company, these scores “may help bring attention to changes in patient status that warrant the clinician’s review.”
AI tools used to assess 71 CTEPH patients having surgery
In the study, researchers at UTSW used Etiometry’s platform to assess 71 people with CTEPH who underwent surgery at their clinic between January 2022 and March 2025.
Specifically, the scientists looked at three AI-based measures: total pulmonary resistance index (TPRI), which is indicative of the amount of pressure in the lungs’ blood vessels; inadequate oxygen delivery index (IDO2), which suggests the body’s tissues are not getting enough oxygen; and hyperlactatemia index (HLA), which indicates high levels of a molecule called lactate that is associated with poor oxygen delivery in the body.
The researchers specifically wanted to see if these three AI-based measures could help predict the length of hospital stays following surgery. For the analysis, patients were split into two groups — the half with the longest hospital stays, and the half with the shortest — which were compared using the AI-based measures.
Similar analyses looking specifically at the length of time spent in an intensive care unit (ICU) after surgery were also conducted.
Differences found for patients with shorter vs. longer hospital stays
The results showed that the median TPRI and HLA over the first 24 hours after surgery were significantly higher among patients with longer hospital stays. IDO2 also tended to be higher among patients who were hospitalized longer, though the difference wasn’t statistically significant.
The data also showed that TPRI and IDO2 were both significantly higher among patients with longer ICU stays, while HLA tended to be higher among patients with longer ICU stays; however, the the difference wasn’t statistically significant.
“Continuous monitoring with the Etiometry platform identified higher post-operative TPRI, IDO2, and HLA in patients with prolonged ICU or hospital stays after [surgery]. These indices may serve as early markers of adverse outcomes and resource utilization,” the researchers concluded. They added that these findings warrant further testing to determine whether this platform can be used to proactively guide management and improve outcomes.
