The model, a successor to VUNO Med-DeepCARS (DeepCARS)—an AI system that generates a 0-100 risk score for inpatient cardiac arrest based on vital signs—was validated in a study published in March in eClinicalMedicine. ICU readmission is a key metric for hospital efficiency and patient safety.Yet, in the absence of standardized protocols, discharge decisions often rely on subjective judgment. Traditional scoring systems, such as early warning scores and severity indices, have been criticized for their inconsistent accuracy.

To close this gap, VUNO partnered with researchers at Seoul National University Hospital (SNUH) to develop iREAD, which evaluates readmission risk based on ICU patient vital signs and blood test results. The study analyzed 70,842 patient records from SNUH spanning 14 years and validated the model against two major U.S. ICU datasets: MIMIC-III (43,237 patients, Beth Israel Deaconess Medical Center) and eICU-CRD (90,271 ICU admissions across 208 hospitals). The model’s predictive performance was assessed using the area under the receiver operating characteristic (AUROC) curve.

VUNO’s iREAD model demonstrated superior predictive performance, achieving an AUROC of 0.820—a statistically significant improvement over traditional ICU readmission scoring systems, which scored 0.768 in MIMIC-III and 0.725 in eICU-CRD datasets. Kaplan-Meier analysis further revealed that over 40 percent of patients classified as high-risk by iREAD required ICU readmission within 48 hours, a more than fourfold increase in correctly identifying high-risk patients compared to existing benchmarks.

VUNO’s existing DeepCARS system is already deployed in 120 general wards across Korea, assisting in predicting cardiac arrest, mortality, and septic shock. The company is now seeking regulatory approval for iREAD, with plans to integrate it into ICU workflows for discharge decision-making.

Source: https://www.koreabiomed.com/news/articleView.html?idxno=26869