A nomogram for predicting short-term parenteral nutrition-associated liver disease in hospitalized adults ‒ A population-based analysis

Introduction: Background: parenteral nutrition-associated liver disease (PNALD) is a common complication in patients receiving total parenteral nutrition (TPN). Our study aimed to establish a nomogram to predict short-term occurrence of PNALD patients and implement intervention early as far as possible. Methods: 187 patients who met the eligibility criteria were retrospectively screened May 2020 to May 2022. They were further divided into the training cohort and validation cohort. The univariate and multivariate analysis were used to select risk factors for the predictive model which was presented with a nomogram. Model building and validation were done using R version 4.4.0. The discrimination and calibration of the nomogram were evaluated using receiver operating characteristic (ROC) curve analysis and calibration plots. The net benefits of the nomogram at different threshold probabilities were quantified using decision curve analysis (DCA). Results: the incidence of PNALD in patients who received TPN was 22.99 %. The logistic regression model equation was constructed based on the three factors of fasting time, glycolipid ratio and ALT by the univariate and multivariate analysis: Logit (p) = 0.056 x fasting time - 3.721 x glycolipid ratio + 0.044 x ALT + 0.518. The AUC value (0.695) indicated satisfactory discriminative ability of the nomogram. The calibration plots showed favorable consistency between the prediction of the nomogram and actual observations in both the training and validation cohorts. Furthermore, DCA showed that the nomogram was clinically useful and had discriminative ability to recognize patients at high PNALD risk. Conclusions: a nomogram was developed and validated to assist clinicians in predicting possibility of PNALD patients in short term.