A Study on Predicting Clinical Prognosis of Patients with Chronic Acute Liver Failure Based on Disease Diagnosis Using Random Forest Algorithm Model
Objectives To analyze the clinical outcomes and related diseases affecting clinical prognosis of patients with acute-on-chronic liver failure in our hospital,and construct a clinical prognosis model based on disease spectrum using the random forest algorithm.Methods The clinical report data of patients with acute-on-chronic liver failure in a hospital from January 1st,2013 to December 31st,2023 was retrieved,the information on the patients as gender,age,admission time,discharge time,admission route,main diagnosis,other diagnoses,and exit method were collected.Inclusion of 1705 patients,and according to the clinical outcomes of patients,they were divided into two groups,as the inpatient death group(n=296)and the survival group(n=1409).Chi square or t-test methods were used to compare the differences between the two groups.Lasso regression was used to screen for disease characteristic variables that affected the outcomes.The dataset was divided into training and testing sets randomly in a 7:3 ratio,and the random forest algorithm was used to construct a clinical prognosis model of acute-on-chronic liver failure patients in the training set.It was evaluated in the testing set.Results The average age of patients in the death group was 59.57±12.71 year old,the survival group was 50.01±12.39 year old,the difference between groups was statistically significant(t=-12.01,P＜0.001).The trend chi square analysis showed an increase in hospitalization mortality rate with the increase of patient's age(Z=-10.83,P＜0.001).The in-hospital mortality rate of patients from emergency sources was higher than that of patients from outpatient sources(χ2=26.48,P＜0.001).Lasso regression analysis screened 31 diseases that had an impact on clinical prognosis.In the random forest algorithm model,the Gini index of hepatorenal syndrome(K76.7)was 40.2,which had the greatest impact on prognosis,followed by age,hepatic encephalopathy(K72.9),acidosis(E87.2),acute renal failure(N17.9),liver malignancy(C22.9),sepsis(A41.9),etc.The area under the curve(AUC)of the random forest algorithm model for predicting patient mortality outcomes in the test dataset was 0.87(95%CI:0.84-0.91),with a sensitivity of 0.70,specificity of 0.89,and Kappa value of 0.40.The calibration curve showed high prediction accuracy,and the clinical decision curve showed that the random forest algorithm model had clinical benefits in the threshold range of 0.1-0.6.Conclusions The random forest algorithm model based on disease spectrum had certain clinical value in predicting the clinical prognosis of acute-on-chronic liver failure patients.In order to improve the clinical prognosis of acute-on-chronic liver failure patients,elderly patients,those with concomitant hepatorenal syndrome,hepatic encephalopathy,acidosis,acute renal failure,and sepsis were the populations that need to be focused on in clinical practice.
NETL
NSTL
万方数据
