Researchers Validated AI Model for SBP Diagnosis

A machine learning tool demonstrated high negative predictive values for identifying spontaneous bacterial peritonitis.

Updated on Sept. 22, 2026 in Asthma

Isometric editorial illustration featuring a medical specimen vial and a geometric data lattice on a laboratory table. — Researchers validated a new machine learning model designed to effectively identify and stratify risk for spontaneous bacterial peritonitis in patients with cirrhosis. AI Illustration. Upload story photo >

Live Poll

Should medical professionals rely on machine learning models to non-invasively diagnose patient conditions?

Researchers validated a machine learning model designed to exclude spontaneous bacterial peritonitis in patients with cirrhosis and ascites. The study utilized patient data collected between 2020 and 2023 within the Veterans Health Administration system.

Why it matters

The re-validation of clinical models has become essential due to significant changes in health care delivery and patient populations following the COVID-19 pandemic. This tool aims to assist clinicians in more effectively stratifying risk for this serious infection.

The validation cohort consisted of 4,192 patients, with 630 diagnosed with spontaneous bacterial peritonitis. The model relied on 20 distinct clinical and laboratory values to achieve its predictive results.

The players

Veterans Health Administration

This is the integrated health care system that provided the patient data and the hospital setting for the model validation study.

The details

Researchers performed manual chart reviews at two tertiary-care Veterans Health Administration hospitals to confirm the model's accuracy. The model demonstrated a 100% negative predictive value at a 5% probability threshold within the validation subgroup, which contained 7 confirmed cases out of 107 patients.

The patient cohort admission period spanned from 2020 to 2023.

The Big Picture

This study leverages the longitudinal data held within the Veterans Health Administration's electronic health record database to refine clinical decision-making protocols. By validating models against this large repository, researchers can better address post-pandemic shifts in patient health care delivery.

For patients with cirrhosis, this model could lead to more accurate and rapid diagnostic assessments for bacterial infections. This may help in reducing unnecessary diagnostic procedures while improving the speed at which clinicians identify high-risk cases.

The takeaway

The study highlights how machine learning can be optimized to assist clinical risk-stratification for patients with complex liver conditions. Adopting validated automated models may allow medical professionals to allocate resources more efficiently during diagnostic processes.