Remodeling the model for end-stage liver disease for predicting mortality risk in critically ill patients with cirrhosis and acute kidney injury

A dynamic model to predict early occurrence of acute kidney injury in ICU hospitalized cirrhotic patients: a MIMIC database analysis

BACKGROUND

This study aimed to develop a tool for predicting the early occurrence of acute kidney injury (AKI) in ICU hospitalized cirrhotic patients.

METHODS

Eligible patients with cirrhosis were identified from the Medical Information Mart for Intensive Care database. Demographic data, laboratory examinations, and interventions were obtained. After splitting the population into training and validation cohorts, the least absolute shrinkage and selection operator regression model was used to select factors and construct the dynamic online nomogram. Calibration and discrimination were used to assess nomogram performance, and clinical utility was evaluated by decision curve analysis (DCA).

RESULTS

A total of 1254 patients were included in the analysis, and 745 developed AKI. The mean arterial pressure, white blood cell count, total bilirubin level, Glasgow Coma Score, creatinine, heart rate, platelet count and albumin level were identified as predictors of AKI. The developed model had a good ability to differentiate AKI from non-AKI, with AUCs of 0.797 and 0.750 in the training and validation cohorts, respectively. Moreover, the nomogram model showed good calibration. DCA showed that the nomogram had a superior overall net benefit within wide and practical ranges of threshold probabilities.

CONCLUSIONS

The dynamic online nomogram can be an easy-to-use tool for predicting the early occurrence of AKI in critically ill patients with cirrhosis.

Application of interpretable machine learning for early prediction of prognosis in acute kidney injury

Background

This study aimed to develop an algorithm using the explainable artificial intelligence (XAI) approaches for the early prediction of mortality in intensive care unit (ICU) patients with acute kidney injury (AKI).

Methods

This study gathered clinical data with AKI patients from the Medical Information Mart for Intensive Care IV (MIMIC-IV) in the US between 2008 and 2019. All the data were further randomly divided into a training cohort and a validation cohort. Seven machine learning methods were used to develop the models for assessing in-hospital mortality. The optimal model was selected based on its accuracy and area under the curve (AUC). The SHapley Additive exPlanation (SHAP) values and Local Interpretable Model-Agnostic Explanations (LIME) algorithm were utilized to interpret the optimal model.

Results

A total of 22,360 patients with AKI were finally enrolled in this study (median age, 69.5 years; female, 42.8%). They were randomly split into a training cohort (16770, 75%) and a validation cohort (5590, 25%). The eXtreme Gradient Boosting (XGBoost) model achieved the best performance with an AUC of 0.890. The SHAP values showed that Glasgow Coma Scale (GCS), blood urea nitrogen, cumulative urine output on Day 1 and age were the top 4 most important variables contributing to the XGBoost model. The LIME algorithm was used to explain the individualized predictions.

Conclusions

Machine-learning models based on clinical features were developed and validated with great performance for the early prediction of a high risk of death in patients with AKI.

The decreasing predictive power of MELD in an era of changing etiology of liver disease

The field of liver transplantation has shifted considerably in the MELD era, including changing allocation, immunosuppression, and liver failure etiologies, as well as better supportive therapies. Our aim was to evaluate the predictive accuracy of the MELD score over time. The United Network for Organ Sharing provided de-identified data on 120 156 patients listed for liver transplant from 2002-2016. The ability of the MELD score to predict 90-day mortality was evaluated by a concordance (C-) statistic and corroborated with competing risk analysis. The MELD score's concordance with 90-day mortality has downtrended from 0.80 in 2003 to 0.70 in 2015. While lab MELD scores at listing and transplant climbed in that interval, score at waitlist death remained steady near 35. Listing age increased from 50 to 54 years. HCV-positive status at listing dropped from 33 to 17%. The concordance of MELD and mortality does not differ with age (>60 = 0.73, <60 = 0.74), but is lower in diseases that are increasing most rapidly-alcoholic liver disease and non-alcoholic fatty liver disease-and higher in those that are declining, particularly in HCV-positive patients (HCV positive = 0.77; negative = 0.73). While MELD still predicts mortality, its accuracy has decreased; changing etiology of disease may contribute.

BDNF gene polymorphism and serum level correlate with liver function in patients with hepatitis B-induced cirrhosis

We investigate the correlation of serum brain-derived neurotrophic factor (BDNF) level and its gene polymorphism with liver function classification in patients with hepatitis B virus (HBV) induced liver cirrhosis. A total of 182 patients with HBV induced liver cirrhosis were collected as a case group, and 186 healthy subjects in the same period were used as the control group. ELISA measured serum BDNF levels. Polymerase chain reaction-restriction fragment length polymorphism was used to detect rs6265 (A/G) and rs10835210 (A/C) in the BDNF gene. The serum BDNF level was significantly lower in the case group than in the control group. With the elevation of Child-Pugh classification in patients with HBV induced liver cirrhosis, the decrease trend of serum BDNF level was even lower. The difference in frequency distribution between the case group and the control group was statistically significant regarding GG, GA, and AA genotypes, as well as G and A alleles in rs6265 (all P < 0.05). The frequency distribution of genotypes and alleles of rs6265 was statistically different in HBV induced liver cirrhosis patients with different liver function grades (P < 0.05). In patients with HBV induced liver cirrhosis, the AA genotype of BDNF gene rs6265 had the lowest level of serum BDNF. Our study suggests that serum BDNF plays an important role in the grading and early diagnosis of liver function in patients with HBV-induced liver cirrhosis, and AA genotype at rs6265 of BDNF gene is a negative factor for liver cirrhosis. Moreover, the polymorphism of this locus could affect the serum BDNF level.

Predicting in-hospital mortality of patients with acute kidney injury in the ICU using random forest model

OBJECTIVES

We aimed to construct a mortality prediction model using the random forest (RF) algorithm for acute kidney injury (AKI) patients in the intensive care unit (ICU), and compared its performance with that of two other machine learning models and the customized simplified acute physiology score (SAPS) II model.

METHODS

We used medical information mart for intensive care (MIMIC) III database for model development and performance comparison. The RF model uses the same predictor variable set as that of the SAPS II model. We also developed three other models and compared the RF model with the other three models in prediction performance. Three other models include support vector machine (SVM) model, artificial neural network (ANN) model and customized SAPS II model. In model comparison, the prediction performance of each model was assessed by the Brier score, the area under the receiver operating characteristic curve (AUROC), accuracy and F₁ score.

RESULTS

The final cohort consisted of 19044 patients with AKI in the ICU. The observed in-hospital mortality of AKI patients is 13.6% in the final cohort. The results of model performance comparison show that the Brier score of the RF model is 0.085 (95%CI: 0.084-0.086) and AUROC of the RF model is 0.866 (95%CI: 0.862-0.870). The accuracy of the RF model is 0.728 (95%CI: 0.715-0.741). The F₁ score of the RF model is 0.459 (95%CI: 0.449-0.470). The calibration plots show that the RF model slightly overestimates mortality in patients with low risk of death and underestimates mortality in patients with high risk of death.

CONCLUSION

There is great potential for the RF model in mortality prediction for AKI patients in ICU. The RF model may be helpful to aid ICU clinicians to make timely clinical intervention decisions for AKI patients, which is critical to help reduce the in-hospital mortality of AKI patients. A prospective study is necessary to evaluate the clinical utility of the RF model.