Table 2. A brief summary of the included studies.

Reference	Sample (Feature) size	Outcomes	Predictors/variables/features	Missing values handled	Hyperparameter selection	Validation	Calibration	Best Algorithm	Compared algorithms
Al Taleb et al. 2017	358 (15)	Length of Stay	At admission	Single imputation	Not reported	10-fold CV	No	Bayesian Network	DT (C4.5)
Asadi et al. 2014	107 (8)	90-day binary and 7 scale mRS	At admission	Not reported	No	Training, test, validation for ANN, Nested CV for SVM	No	SVM	ANN, Linear Regression
Liang et al. 2019	435 (4)	90-day binary mRS	Admission, laboratory data	Not reported	Not reported	Training and test split	No	ANN	LR
Heo et al. 2019	2604 (38)	90-day binary mRS	Admission	Complete case analysis	No	Training and test split	No	DNN	RF, LR
Konig et al. 2007	3184 (43)	100-day Bathel Index	first 72h after admission	Complete case analysis	Yes, Grid search	Temporal and external validation	Yes for LR	-	RF, SVM, LR
Celik et al. 2014	570 (22)	10-day mortality	At admission	-	Yes, grid search	5-fold CV	No	LR	ANN
Ho et al. 2014	190 (26)	Discharge mortality	Admission and interventions	Complete case analysis	Not reported	10-fold CV	No	SVM	Naïve Bayes, DT, RF, PCA+SVM, LR
Cox et al. 2016	2580 (72)	Post stroke spasticity	Not clear	Not reported	Not reported	Training, test and validation split	No	RF	DT (CART), Adaboost
Kruppa et al. 2014	3184 (43)	100-day Bathel Index	First 72h after admission data	Complete case analysis	Yes, For KNN, bNN and RF	Temporal and external validation	Yes, Brier score	SVM and LR	K-NN, b-NN, RF
Easton et al. 2014	933 (-)	Short/very short mortality	Not clear	Not reported	Yes, DT is pruned	Training and test split	No	-	Naïve Bayes, DT, LR
Mogensen and Gerds 2013	516 (12)	3-month/1-year/3-year/5-year mortality	Admissiondata	Complete case analysis	No, manually set up	Bootstrap CV	Yes, Brier score	-	Pseudo RF, Cox Regression, and Random survival forest
Van Os et al. 2018	1383 (83)	Good reperfusion socre, 3-month binary mRS	Admission, laboratory and treatment data	Multiple imputation by chained equations	Yes, nested CV with random grid search	Nested CV	No	-	RF, SVM, ANN, super learner, LR
Peng et al. 2010	423 (10)	30-day mortality	Admission, laboratory, radiographic data	No missing values	Yes, empirically	4-fold CV	No	RF	ANN, SVM, LR
Tokmakci et al. 2008	70 (6)	Quality of life	Admissiondata	Not reported	Not reported	Training and test split	No		ANFIS
Monteiro et al. 2018	425 (152)	3-month binary mRS	Admission/2 hours/24 hours/7 days data	single imputation	Yes, Grid search	10-fold CV	No	RF and Xgboost	DT, SVM, RF, LR (LASSO)
Tjortjis et al. 2007	671 (37)	2-month mortality	Admission data	Cases discarded with missing outcomes	Yes, pruned	Training and test split	No	DT (T3)	DT (C4.5)
Lin et al. 2018	382 (5)	Neurologic deterioration	Admission and laboratory data	Not reported	Yes, CV	Training and test split	No	-	SVM
Tanioka et al. 2019	95 (20)	Delayed cerebral ischemic after SAH	Admission/1-3 days variables	Complete case analysis	Yes, Grid search	Leave one out CV	No	-	RF