Transparency · CHAI Applied Model Card

Liver Stress

CHAI Applied Model Card (v0.1) for the Liver Stress pattern in Zenlo Labs. This document is not CHAI-certified or CHAI-endorsed.

Header

Summary

The Liver Stress pattern identifies elevated hepatic transaminases and gamma-glutamyl transferase in adult laboratory panels, signaling potential hepatocellular injury or cholestatic stress. Deterministic rules compare ALT, AST, and GGT against sex-aware reference upper limits from the biomarker registry; any qualifying elevation triggers the pattern. Claude Haiku 4.5 produces a narrative summarizing which liver enzymes are abnormal for physician review. Designed for licensed functional-medicine physicians, the pattern is supportive clinical decision support — it does not stage fibrosis, diagnose NAFLD/NASH, or replace imaging or biopsy. Outputs highlight laboratory signals warranting correlation with alcohol use, medications, metabolic context, and viral hepatitis workup.

Uses & Directions

Intended use

Clinical decision support for licensed physicians reviewing hepatic enzyme elevations in adult panels.

Primary users

Licensed functional-medicine physicians and similarly qualified clinicians.

How to use

Correlate flagged enzymes with alcohol intake, medications, metabolic syndrome context, and viral hepatitis serologies as clinically indicated.

Target population

Adults aged 18 and older in the United States.

Out of scope

• Direct patient use without physician oversight
• Pediatric populations
• Standalone diagnosis of liver disease
• Fibrosis staging or elastography interpretation

Warnings

Clinical risk level

Low — supportive tool; the treating physician retains full clinical judgment and responsibility.

Known limitations

• Enzyme elevations are sensitive but not specific for etiology.
• Does not incorporate imaging, Fib-4, or NAFLD-specific scores.
• AST elevations may reflect non-hepatic sources (e.g., muscle injury).

Validation note

Validation pending — a Tier A NHANES validation run has not yet been completed for this pattern. Distribution, agreement, and fairness results will be published here when available.

Trust Ingredients

AI system facts

• Deterministic pattern detector (liver_stress) plus Claude Haiku 4.5 for narrative synthesis
• Primary inputs: ALT, AST, and GGT with sex-aware reference range comparison.
• Output: Pattern flag plus narrative on hepatic enzyme abnormalities.

Security & compliance

• Anthropic Business Associate Agreement with zero-data-retention configuration
• HIPAA-aligned design; no patient data used for model training

Ongoing maintenance

Versioned, transparent, and reproducible via a public independent audit harness (see Resources).

Transparency

Self-funded development; no third-party sponsor for this pattern card.

Key Metrics

Usefulness / Efficacy

Zenlo’s detection approach was benchmarked across five models in a separate study; see the medRxiv preprint in Resources. No per-pattern efficacy metric is published for this pattern. Source: 5-model benchmark, medRxiv MEDRXIV/2026/346284.

Fairness / Equity

Validation pending — a Tier A NHANES validation run has not yet been completed for this pattern. Distribution, agreement, and fairness results will be published here when available.

Safety / Reliability

• Supportive-only; not intended as a standalone diagnostic
• Physician authorization required before clinical use
• Deterministic detector is reproducible for inputs: ALT, AST, GGT

Resources

• medRxiv preprint MEDRXIV/2026/346284 — 5-model benchmark (system-level detector evaluation)
• JAMIA Open submission JAMIO-2026-0120 (under review)
• Independent audit harness: github.com/dimashibakov/zenlo-audit — reproducibility manifest, NHANES 2015–2016 cycle, harness commit c10afe8