Etiology and Prognostic Analysis of Acute Liver Failure in Chinese Children

Data were organized and analyzed using WindRays statistical software. Quantitative data that followed a normal distribution were expressed as mean ± standard deviation (x ± s), and intergroup comparisons were performed using t-tests. For non-normally distributed quantitative data, median (interquartile range) [M (P25, P75)] was used, and intergroup comparisons were conducted using the Mann-Whitney U test. Qualitative data were presented as frequencies (n) and percentages (%), with intergroup comparisons made using χ² or Fisher's exact test. Multivariate logistic and Cox regression analyses were performed to adjust for potential confounding factors in the association between PALF etiology, treatment, and prognosis. Results were presented as odds ratios (ORs) with 95% confidence intervals (CIs). A multivariate-adjusted restricted cubic spline model was constructed to examine possible non-linear dose-response relationships between PALF etiology, treatment, and prognosis. Variables were selected using stepwise regression, lasso regression, and best subset regression. Risk scoring models were established using regression analysis, nomograms, scoring systems, decision trees, and machine learning algorithms. Sensitivity analyses were conducted using propensity score matching (PSM), regression analysis, and stratified analysis. The predictive efficacy of the model for PALF prognosis was evaluated using the area under the receiver operating characteristic (ROC) curve (AUC), with AUC ≥ 0.80 indicating good predictive value. Data with missing values less than 5% were deleted directly, and data with missing values between 5% and 20% were imputed using multiple imputation methods.

Handling of lost to follow-up: Multiple contact information and emergency contacts were collected at the beginning of the study to maintain regular communication with the patients and their guardians to reduce the risk of lost follow-up. If follow-up failed, multiple attempts were made to contact participants via phone, email, or other methods. Once lost to follow-up, multiple imputation methods were used to estimate missing data, and sensitivity analyses were conducted (if possible).

Inclusion criteria

• According to the diagnostic criteria for PALF, the following conditions must be met:

① Severe biochemical liver abnormalities occurring within 8 weeks;

② Coagulation disorders uncorrectable by vitamin K: if hepatic encephalopathy is present, prothrombin time (PT) > 15s or international normalized ratio (INR) > 1.5; regardless of the presence of hepatic encephalopathy, PT > 20s or INR > 2.0.

• For patients who were admitted twice, only the information from the first hospitalization will be used.
• Patients with a history of chronic liver disease (e.g., genetic metabolic factors, drug-induced factors) or biliary obstruction that led to the acute episode are categorized under acute liver failure.
• The child's legal guardian has signed a written informed consent form.

Exclusion criteria

• Age > 18 years;
• Significant data missing (>10%);
• The child's legal guardian refuses participation.