Application of machine learning in predicting hospital readmissions: a scoping review of the literature

Table 2 Descriptive statistics of AUC by ML category

ML category	Number of Studies^a	Mean (STD)	Median	Min	Max	IQR
NN	15	0.71 (0.07)	0.71	0.61	0.81	0.64–0.78
Boosted tree^a	17	0.70 (0.06)	0.7	0.59	0.81	0.66–0.75
RF	16	0.68 (0.09)	0.64	0.53	0.9	0.63–0.72
DT	9	0.70 (0.10)	0.67	0.59	0.88	0.63–0.77
Regularized Logistic Regression^b	16	0.69 (0.08)	0.65	0.58	0.84	0.64–0.75
SVM	10	0.70 (0.11)	0.68	0.5	0.86	0.65–0.78
Other ML algorithms^c	10	0.68 (0.04)	0.68	0.62	0.77	0.66–0.71

Abbreviations: ML machine learning, NNs neural networks, RF random forest, DT decision tree, SVM support vector machine, STD standard deviation, IQR the interquartile range. ^athe total number of studies is larger than total number of included studies, because some studies used more than 1 ML algorithms. ^aIt includes adaboost, gradient boosting, gradient descent boosting, boosting, XGBoost; ^bIt includes Lasso (L1 regularization), ridge regression (L2 regularization), and elastic-net algorithms; ^cIt includes: DT ensembled with SVM, RF combined with SVM, tree-augmented naïve Bayesian network

ISSN: 1471-2288