Probabilistic photonic computing based on chaotic Kerr optical frequency combs.

PURPOSE: To develop and validate a nomogram using routine admission indicators to predict total bilirubin on day 7 (TBil-Day7) in hepatitis B virus-related liver cirrhosis (HBV-LC) patients, enabling early identification of inadequate short-term TBil response and providing a reference for intensive treatment.

MATERIAL AND

METHODS

We retrospectively enrolled 284 HBV-LC patients, randomly assigned 7:3 to training (n=198) and internal validation (n=86) cohorts. Candidate variables were clinical and laboratory parameters available within 24 hours of admission, with TBil-Day7 as the outcome. Predictors were selected via least absolute shrinkage and selection operator (LASSO) regression, and a multiple linear regression model was built and visualized as a nomogram. Model performance was evaluated using the coefficient of determination (R2), root mean square error (RMSE), mean absolute error (MAE), and the proportion of predicted values falling within ±20 μmol/L of the observed TBil-Day7 values.

RESULTS

LASSO regression identified four predictors: admission TBil, direct bilirubin (DBil), aspartate aminotransferase (AST), and international normalized ratio (INR). The prediction equation was: TBil-Day7 (μmol/L) = -23.0159 + 0.5721×TBil + 0.5847×DBil + 0.0601×AST + 14.0707×INR. The training cohort had an R2 of 0.932. In the internal validation cohort, R2=0.75, RMSE=38.06 μmol/L, MAE=21.27 μmol/L, and 74.4% of predictions were within ±20 μmol/L of actual values. The calibration curve showed good agreement between predicted and observed TBil-Day7.

CONCLUSION

This nomogram, incorporating four routine admission indicators (TBil, DBil, AST, INR), can predict TBil-Day7 in HBV-LC patients with reasonable accuracy. It facilitates early identification of high-risk patients with insufficient TBil decline and allows estimation of short-term treatment response at admission.