Table 4.

Studies related to ML-based discharge time prediction; the “*” denotes the best-performing model

Study	Predicted parameter	Patient population	Methodology	Main factors	Dataset size
Gholipour et al. [117]	LOS	Trauma patients	ANN*, Lagrangian regression	Mechanism of trauma, the site involved, vital signs and physical examination, laboratory findings	(n=125)
Barnes et al. [36]	Discharge time	General patients	Tree-based supervised ML models, Regression RF*	Admission and discharge times, demographics, basic admission diagnoses	(n= 8,852)
Elbattah and Molloy [99]	LOS	Elderly patients with hip fracture care	RF*, BDT, NN, Linear regression		(n=2,000)
Tsai et al. [118]	LOS	Cardiology patients	ANN*, Linear regression	Sex, age, location, main diagnosis	(n=2,377)
Turgeman et al. [112]	LOS	General patients	Regression tree	Demographics, outpatient and inpatient history, medication history, lab values and vital signs	(n=4,840)
Thompson et al. [135]	Prolonged LOS	Newborns	NB, Multi-layer Perceptron, Simple Logistic, SVM, DT, RF*, RT	Administrative data, minimal clinical data at the time of admission/birth	(n= 2,610)
Muhlestein et al. [119]	LOS	Brain tumor surgery patients	ML ensemble model	Nonelective surgery, preoperative pneumonia, sodium abnormality, race	(n= 41,222)
Kabir and Farrokhvar [113]	LOS	Surgical patients	ANN*, LR, SVM	surgical category	(n= 880,000)
Safavi et al. [114]	Discharge time	Surgical inpatients	Feedforward NN	Demographic, environmental, administrative, clinical	(n= 15,201)
Lazar et al. [37]	Discharge time	Surgical patients	RF	Age, sex, admission source, laboratory measurements, vitals	(n=10,904)
Bacchi et al. [107]	LOS	Stroke patients	LR, RF, DT, ANN*	Age, sex, estimated pre-stroke mRS	(n= 2,840)
Nemati et al. [115]	Discharge time	General patients	GB*, Fast SVM, Fast Kernel SVM	Age, sex	(n=1,182)
Bertsimas et al. [102]	Short-term discharges	General patients	Linear regression, CART DT, Optimal Trees, RF, GBT*	Demographics, provider orders, diagnosis codes, medications	(n= 63,432)
He et al. [121]	LOS, Flow	General patients	ANN	Age, sex, LOS, admit source department, medical features	(n=3,959)
Zhong et al. [122]	LOS	Ambulatory total hip arthroplasty patients	Multivariable LR, ANN*, RF*	Anesthesia type, BMI, age, ethnicity, white blood cell count	(n=63,859)
Gabriel et al. [136]	Discharge time	Surgical patients	Regression, RF*, balanced RF*, balanced bagging, NN, SVM	Patient’s surgical characteristics, age, sex, weight	(n=13,447)
Zeleke et al. [123]	LOS	General patients	Linear Regression, Ridge and Elastic-net regression, SVM, RF, KNN, XGB*	Demographic factors, mode of arrival/source of admission, risk categories, current problems	(n=12,858)