← Issue №13/ week of Sep 27, 2026/Hepatology

Validation of a machine learning model to non-invasively exclude spontaneous bacterial peritonitis.

From GI Signals issue №13: what this paper found, what it changes, and where it sits against the current standard of care, reviewed by Simon Mathews, MD.

Hepatology prospective cohort · n=4,192 · Sep 22, 2026 · Clin Transl Gastro · IF 3.4

Validation of a machine learning model to non-invasively exclude spontaneous bacterial peritonitis.

New evidenceartificial intelligencebiomarker
Clinical takeawayConsider using this ML model as an adjunctive tool to non-invasively exclude SBP in cirrhotic VA patients with ascites, particularly in resource-limited settings, but do not replace guideline-mandated paracentesis when indicated.
What it foundA machine learning model excluded SBP with negative predictive values of 93.6%, 92.0%, and 90.9% at <5%, <10%, and <15% probability thresholds in a national VA cohort, and 100%, 97.5%, and 95.4% in a manual confirmation subgroup.
ContextThis validates a previously developed ML model in a contemporary VA cohort post-2020, confirming its utility in post-COVID-19 healthcare settings.
Measured against the standard of care· AASLD 2021 Practice Guidance (Biggins et al., Hepatology) / AGA Clinical Practice Update 2025 (Orman et al., Gastroenterology 2025;169(7):1547-1557)

No single passage of this standard matched the paper closely enough to quote, so none is shown. The standard is cited above.

Our full summary of this standard

Perform a diagnostic paracentesis as soon as possible in any patient with new-onset ascites that can be sampled. Also perform one in any patient with cirrhosis and ascites who is admitted to hospital. AASLD 2021 says this applies to emergent admission for any reason. AGA 2025 BPA 2 names admission for ascites- or encephalopathy-related symptoms. The tap rules out SBP even without signs of infection, because up to one-third of SBP is asymptomatic. Repeat it whenever signs, symptoms or labs suggest infection, and in tense ascites with AKI. Paracentesis is safe with coagulopathy: INR >1.5 or platelets <50×10^9/L is not a contraindication, and routine FFP or platelet transfusion is not recommended. Possible exceptions are DIC, and uremia with thrombocytopenia, where desmopressin may be considered. Initial fluid testing: neutrophil (PMN) count, total protein, and ascitic albumin with a same-day serum albumin to calculate the SAAG. AGA 2025 adds Gram stain and culture to the initial panel for new-onset ascites and hospitalized patients. AASLD 2021 treats culture as part of the SBP workup rather than the routine panel. Either way, inoculate ≥10 mL into aerobic and anaerobic blood-culture bottles at the bedside before antibiotics. Amylase, cytology, glucose, LDH, AFB/ADA and triglycerides are ordered only for a clinical reason. SAAG ≥1.1 g/dL means portal hypertension (~97% accuracy). SAAG <1.1 g/dL points to a non-portal-hypertensive cause: carcinomatosis, TB peritonitis, nephrotic syndrome, pancreatic or chylous ascites. Within high-SAAG ascites, total protein <2.5 g/dL suggests cirrhosis (also massive liver metastases or late Budd-Chiari). Total protein ≥2.5 g/dL suggests a cardiac or postsinusoidal cause (heart failure, portal vein thrombosis, SOS, early Budd-Chiari), so work up with echocardiography, hepatic-vein Doppler or right-heart catheterization. SBP is diagnosed by ascitic PMN >250/mm^3, with or without a positive culture, once a surgically treatable source has been excluded. SBE is diagnosed by pleural fluid PMN >250/mm^3 in a hepatic hydrothorax after pneumonia and other secondary causes are excluded. It is treated with antibiotics and no chest tube.

AASLD 2021 Practice Guidance (Biggins et al., Hepatology) / AGA Clinical Practice Update 2025 (Orman et al., Gastroenterology 2025;169(7):1547-1557) · reviewed 2026-10-11 ↗
Tsai SY … Bajaj JS · Clinical and Translational Gastroenterology · IF 3.4 · PubMed ↗Permalink
← Read the whole of issue №13 Every paper GI Signals surfaces gets a page like this one. All issues