Sepsis is an organ failure disease caused by an infection resulting in extremely high mortality. Machine learning algorithms XGBoost and LightGBM are applied to construct two processing methods: mean processing method and feature generation method, aiming to predict early sepsis 6 hours in advance. The feature generation methods are constructed by combining different features, including statistical strength features, window features, and medical features. Miceforest multiple interpolation method is applied to tackle large missing data problems. Results show that the feature generation method outperforms the mean processing method. XGBoost and LightGBM algorithms are both excellent in prediction performance (AUC: 0.910∼0.979), among which LightGBM boasts a faster running speed and is stronger in generalization ability especially on multidimensional data, with AUC reaching 0.979 in the feature generation method. PTT, WBC, and platelets are the key risk factors to predict early sepsis.

Artificial intelligence (AI) technology has huge scope in developing models to predict the survival rate of critically ill patients in the intensive care unit (ICU). The availability of electronic clinical data has led to the widespread use of various machine learning approaches in this field. Innovative algorithms play a crucial role in boosting the performance of models. This study uses a stacked ensemble model to predict mortality in ICU by incorporating the clinical severity scoring results, in which several machine learning algorithms are employed to compare the performance. The experimental results show that the stacked ensemble model achieves good performance compared with the model without integrating the severity scoring results, which has the area under curve (AUC) of 0.879 and 0.862, respectively. To improve the performance of prediction, two feature subsets are obtained based on different feature selection techniques, labeled as SetS and SetT. Evaluation performances show that the SEM based on the SetS achieves a higher AUC value (0.879 and 0.860). Finally, the SHapley Additive exPlanations (SHAP) analysis is employed to interpret the correlation between the risk features and the outcome.

Some theories are explored in this research about decision trees which give theoretical support to the applications based on decision trees. The first is that there are many splitting criteria to choose in the tree growing process. The splitting bias that influences the criterion chosen due to missing values and variables with many possible values has been studied. Results show that the Gini index is superior to entropy information as it has less bias regarding influences. The second is that noise variables with more missing values have a better chance to be chosen while informative variables do not. The third is that when there are many noise variables involved in the tree building process, it influences the corresponding computational complexity. Results show that the computational complexity increase is linear to the number of noise variables. So methods that decompose more information from the original data but increase the variable dimension can also be considered in real applications.

In the status forecasting problem, classification models such as logistic regression with input variables such as physiological, diagnostic, and treatment variables are typical ways of modeling. However, the parameter value and model performance differ among individuals with different baseline information. To cope with these difficulties, a subgroup analysis is conducted, in which models’ ANOVA and rpart are proposed to explore the influence of baseline information on the parameters and model performance. The results show that the logistic regression model achieves satisfactory performance, which is generally higher than 0.95 in AUC and around 0.9 in F1 and balanced accuracy. The subgroup analysis presents the prior parameter values for monitoring variables including SpO${}_2$, milrinone, non-opioid analgesics and dobutamine. The proposed method can be used to explore variables that are and are not medically related to the baseline variables.