Cardiac arrest AI that predicts outcomes
Cardiac arrest AI that predicts outcomes unknown
To address out-of-hospital cardiac arrest, Osaka Metropolitan University researchers developed a new scoring method that uses only data available from prehospital resuscitations to accurately predict neurological outcomes and enable clinicians to make better-informed decisions upon a patient’s arrival at the hospital.
WHY IT MATTERS
After patient transport, OHCA patients face unfavorable neurological outcomes, from disability to death.
Developed by researchers from Osaka Metropolitan University, the new model, R-EDByUS, is named the five key variables it is based on – age, duration to return of spontaneous circulation or time to hospital arrival, absence of bystander cardiopulmonary resuscitation, whether the arrest was witnessed, and the initial heart rhythm.
The model accurately predicted the neurological prognosis of cardiogenic OHCA upon hospital arrival, according to the research published in this month's issue of Resuscitation.
"We hypothesize that a scoring model consisting solely of prehospital factors in the algorithm would be easy to use and may predict prognosis at the earliest stage of medical care," the researchers said.
They leveraged the unfavorable features in the American College of Cardiology algorithm:
- Unwitnessed arrest.
- Initial nonshockable rhythm.
- No bystander CPR.
- Time-to-ROSC of > 30 min.
- Ongoing CPR.
- Blood pH of < 7.2, 7.
- Lactate level of > 7 mmol/l.
- Age > 85 years.
- End-stage renal disease.
- Oncardiac causes in cardiac arrest patients.
They used data OHCA gathered between 2005 and 2019 from the All-Japan Utstein Registry for 942,891 adults with presumed cardiac-origin. They categorized patients into two groups – those that achieved ROSC before arriving at the hospital or those still receiving CPR upon arrival. Then, they used detailed regression-based and simplified models to calculate R-EDByUS scores for each group.
They excluded patients under 18, patients who received CPR from bystanders and a few other factors.
In the prehospital ROSC group, 70.0% had unfavorable neurological outcomes while 55.7% experienced mortality. For those who had EMS continuing CPR upon arrival at the hospital, 99.4% had unfavorable neurological outcomes and 98.2% died.
"Our predictive model helps identify patients who are likely to benefit from intensive care while reducing unnecessary burdens on those with poor predicted outcomes," Takenobu Shimada, a medical lecturer at Osaka Metropolitan University's Graduate School of Medicine and lead author of the study, told MSN last week.
The article contends that the scoring model will become a valuable tool for healthcare providers, quickly helping to assess and manage patients undergoing resuscitation.
The researchers developed a web-based calculator they said is simple to use in a clinical setting, and has the potential for future validation.
THE LARGER TREND
AI-powered patient triage can potentially create proper channels of care, enhancing patient outcomes and experiences and optimizing resource use but require regulatory rigor to evaluate them, according to Piotr Orzechowski, founder and CEO of Infermedica, a health AI health company focused on preliminary symptom analyses and digital triage.
"AI tools are not authorized to diagnose patients," Orzechowski told Healthcare IT News in December during a conversation on how healthcare intersects with AI.
"Despite the remarkable progress in generative AI, we must remain cautious about their practical application in healthcare," he said.
ON THE RECORD
"Using this free calculation tool, the predictive probability for unfavorable neurological outcomes and mortality are easily estimated by checking each variable on the internet instead of calculating using nomograms," the researchers said.
Andrea Fox is senior editor of Healthcare IT News.
Email: afox@himss.org
Healthcare IT News is a HIMSS Media publication.
The HIMSS AI in Healthcare Forum is scheduled to take place September 5-6 in Boston. Learn more and register.