Table 1.

Characteristics of ML algorithm prediction studies

S. No	Studies	Country	Data source	Centre	Outcome	Events/sample size (events per predictor)	Best performing algorithm
1	Melinte-Popescu et al. 2023 [59•]	Romania	Case–control	Single	Any-onset	116/233 (17)	Random forest
2*	Liu et al. 2022 [48•]	China	Retrospective cohort	Single	Any-onset	143/11,152 (14)	Random forest
3	Zhang et al. 2022 [50]	China	Retrospective cohort	Single	Any-onset	377/19,653 (126)	Light GBM
4	Gómez-Jemes et al. 2022 [56]	Slovenia	Medical record	Single	Any-onset	22/95 (7)	Decision tree
5	Bennett et al. 2022 [52]	USA	Prospective cohort	Multicentre	Any-onset	2743/31,431 (137)	Deep neural networks
6*	Ansbacher-Feldman et al. 2022 [60•]	UK	Prospective cohort	Multicentre	Preterm	484/60789 (35)	Neural network
7*	Chen et al. 2022 [61]	China	Case–control	Single	Any-onset	237/916 (40)	Random forest
8*	Li et al. 2021 [49•]	China	Retrospective cohort	Single	Any-onset	227/5243 (76)	XGBoost
9*	Wanriko et al. 2021 [62]	Kenya	Case–control	Single	Any-onset	88/352 (7)	Random forest
10*	Manoochehri et al. 2021 [63]	Iran	Case–control	Single	Any-onset	752/1452 (125)	SVM
11*	Marić et al. 2020 [51]	USA	Retrospective cohort	Single	Any-onset	561/5245 (80)	Gradient boosting
12*	Sufriyana et al. 2020a [54]	Indonesia	Nested case–control	Single	Any-onset	878/6734 (58)	Random forest
13	Sufriyana et al. 2020b [55]	New Zealand	Prospective cohort	Single	Any-onset	22/95 (4)	CVR
14	Marin et al. 2019 [53]	Romania	Medical record	Single	Any-onset	NR	Viterbi ML
15*	Jhee et al. 2019 [57]	South Korea	Medical record	Single	Any-onset	474/10,532 (67)	Gradient boosting
16*	Sandström et al. 2019 [58]	Sweden	Prospective cohort	Single	Preterm	497/58,276 (41)	Logistic regression

NB: XGBoost extreme gradient boosting, CVR classification via regression, SVM support vector machine, NR not reported

^*The studies that used the same sample to compare ML algorithms with classical regression models