Derivation and validation of the heart failure cardioscore: an AI diagnostic for heartfailure using novel biomarkers and clinical data - principal investigator/program director (last, first,…

The National Heart, Lung, and Blood Institute awarded Prenosis, Inc. $1,218,102 on September 1, 2026, under the Cardiovascular Diseases Research program (CFDA 93.837) to develop and validate the Heart Failure CardioScore, an artificial intelligence diagnostic tool for acute heart failure in emergency department patients.

Prenosis will combine novel biomarkers and clinical data with machine learning to improve diagnostic and prognostic accuracy for acute heart failure, addressing the challenge that approximately 20 percent of patients remain misdiagnosed despite natriuretic peptide testing. The work builds on Prenosis's prior NIH-funded development of the Sepsis Immunoscore, which received FDA authorization as the first AI diagnostic for sepsis in April 2024. To demonstrate feasibility, Prenosis has already recruited 2,682 patients with biospecimens at two partner hospital sites.

Performance occurs in Chicago, Illinois, with a period of performance through July 31, 2028. The project is classified as a Small Business Innovation Research (SBIR) Phase II grant, reflecting Prenosis's status as a self-certified small disadvantaged business. The $1,218,102 obligated amount represents the Phase II funding following successful completion of Phase I objectives.

Derivation and validation of the heart failure cardioscore: an AI diagnostic for heartfailure using novel biomarkers and clinical data - principal investigator/program director (last, first, middle): reddy, JR., bobby project summary: there is a substantial opportunity to improve the efficiency and accuracy of ED management of heart failure patients. Heart failure affects over 6 million adults in the US, costing the healthcare system upwards of $70 billion dollars per year. 80% of the ~676,000 ED visits for acute heart failure (AHF) lead to hospital admission, and 40% of those patients DIE within 1 year. The adverse prognosis and high healthcare costs associated with AHF are due in part to inaccurate ED diagnosis and disposition decision making, which are often challenging for AHF. Common symptoms, including shortness of breath (DYSPNEA) and fluid retention (CONGESTION) are non-specific, overlapping with other conditions, including pneumonia, chronic obstructive lung disease, and asthma. In addition, no single symptom or physical exam finding has sufficient diagnostic accuracy to confirm AHF. Natriuretic peptides (NPS), including BNP and NT-probnp, are the only FDA-approved and guideline- recommended circulating biomarkers for the diagnosis of HF. However, a considerable proportion (~20%) of patients remain incorrectly diagnosed in spite of NP testing, due largely to their NP's low specificity. Prenosis proposes to improve diagnostic and prognostic accuracy for AHF in ED patients by developing the heart failure cardioscore, an AI tool that combines novel biomarkers and clinical data with machine learning (ML). We will follow the approach we took for the sepsis immunoscore, using NIH funding to produce the first- ever FDA-authorized AI diagnostic for sepsis (DE novo granted in april 2024). To demonstrate feasibility, we began recruiting patients suspected of AHF at 2 of our partner hospital sites. We have already recruited 2,682 patients, including biospecimens and EHR data extraction. We also conducted a pilot study of 136 patients at 1 site, and showed that even with this small sample size we could improve upon the AUC of NPS by over 0.1 (0.85 VS 0.74). Model predictive performance was validated in an independent cohort at the second site (AUC 0.83). The overall objective of this investigation is to combine protein and clinical biomarkers with EHR data to create algorithms that improve the efficiency of ED management of patients with undifferentiated DYSPNEA. The overall approach is to leverage prenosis' network of hospital sites and precision medicine platform to derive (aim 1) and validate (aim 2) the heart failure cardioscore diagnostic algorithm. In aim 3, we will also train and test a prognostic score to identify patients with AHF who are likely to receive >24 hours of IV diuretics (an indicator that outpatient management is likely infeasible).