In a National Health and Nutrition Examination Survey analysis of 1,838 stroke survivors followed for a median of 78 months, survival differed significantly across albumin-bilirubin (ALBI) quartiles, and higher ALBI independently predicted higher all-cause mortality. The study also found that adding ALBI to clinical and laboratory variables supported risk stratification, with machine-learning models used to estimate long-term mortality risk.

Why It Matters To Your Practice

• ALBI is a simple, objective score derived from albumin and bilirubin that may help identify stroke survivors at higher long-term risk.

• For clinicians using AI or predictive analytics in follow-up care, ALBI may be a practical input variable because it relies on routinely available labs rather than specialized testing.

• The signal persisted after adjustment for multiple demographic, clinical, and laboratory factors, suggesting ALBI may add prognostic information beyond standard risk markers.

Clinical Implications

• Consider whether liver-related biomarkers already in the chart could improve post-stroke risk assessment, especially when prioritizing follow-up intensity or secondary prevention efforts.

• Patients with higher ALBI scores may warrant closer surveillance, though the study supports prognostic use rather than treatment decisions based on ALBI alone.

• For teams evaluating AI tools, this is a reminder that models built from common data elements may be more scalable in practice than systems requiring novel inputs.

Insights

• The analysis used NHANES data from 1999-2018 and included Kaplan-Meier curves, Cox regression, restricted cubic splines, and five machine-learning models with 10-fold cross-validation.

• Boruta feature selection identified age, sex, race, marital status, body mass index, cardiovascular disease, lymphocyte count, liver enzymes, uric acid, polyunsaturated fatty acids, basophil percentage, and creatinine among key mortality-associated variables.

• SHAP methods were used to interpret the best-performing model, highlighting how explainability approaches can make AI outputs more clinically usable.

The Bottom Line

• Higher ALBI was linked to worse long-term survival after stroke in this cohort.

• For clinicians interested in AI, the takeaway is less about a specific algorithm and more about how low-friction biomarkers can strengthen risk prediction models that may fit real-world workflows.

• Before changing practice, clinicians should note that the summary provided does not include effect sizes for quartiles or external prospective validation.