Skip to main content

Table 1 Study characteristics

From: Machine learning methods in sport injury prediction and prevention: a systematic review

Authors, Year

Outcome Variable

Predictor Variables

Participants (Age Mean ± sd)

Period

Study Design

Unit of Observation

Number of Observations

Total Amount of Injuries / No. Of Injured Athletes (N =)a

Number of Features

AYALA ET AL., 2019 [2]

Occurrence of Hamstring strain injury

Individual (sport-related background, demographic, previous hamstring strain injury), psychological and neuromuscular measurements

96 Male professional soccer players from 4 teams in 1st and 2nd league in Spain

6 players that did not complete the tests and 4 players that left their teams were removed

1 season (2013–2014)

Prospective cohort

Player

86

NR/18

229

CAREY ET AL., 2018 [11]

Occurrence of non-contact injury, non-contact causing time loss injury and hamstring injury

Training load variables (+ Exponential Weighted Moving Average features and Acute Chronic Workload Ratio features)

75 male professional players from 1 team in the Australian Football League in Australia

3 seasons (2014–2016)

Prospective cohort

Player matches and player training sessions

13,867

Non-contact: 388/NR

Non-contact causing time loss: 198/NR

Hamstring: 72/NR

58

LĂ“PEZ-VALENCIANO ET AL., 2018 [25]

Occurrence of lower extremity muscle injury

Individual (sport-related background, demographic, previous injury), psychological and neuromuscular measurements

132 Male professional players in handball (34) and soccer (98) in the first three National Leagues in Spain

6 players that did not complete the tests and 4 players that left their teams were removed

1 season (2013–2014)

Prospective cohort

Player

122

32/29

151

MCCULLAGH ET AL., 2013 [27]

Occurrence of injury and injury type (contact or non-contact)

Workloads, squeeze test data, soft tissue scores, stress level, mood, sleep score, ankle flexibility, fatigue and player perceived performance, years played, player durability, age

39 male professional players from the Australian Football League in Australia

1 season (2010)

Prospective cohort

Player weeks

1210

163/NR

30

OLIVER ET AL., 2020 [32]

Occurrence of non-contact lower limb injury

Personal data (age, Body Mass Index, etc.) and neuromuscular control tests data

355 Male youth soccer players (age 14.3 ± 2.1) from Premier League and Championship clubs in England

1 season (2014–2015)

Prospective cohort

Player

355

NR/99

20

RODAS ET AL., 2019 [34]

Occurrence of Tendinopathy

Genetic markers

363 Male (89%) and female (11%) professional soccer, futsal, basketball, handball and roller hockey players (age 25 ± 6) from FC Barcelona in Spain

10 years (2008–2018)

Case–control

Player

363

199/199

1 419 369

ROMMERS ET AL., 2020 [35]

Occurrence of injury and type of injury (acute and overuse)

Anthropometric measurements, motor coordination and physical fitness

734 Male U10 to U15 youth soccer players (age 11.7 ± 1.7) of 7 premier league clubs in Belgium

1 season (2017–2018)

Prospective cohort

Player

734

NR /368

29

ROSSI ET AL., 2018 [36]

Occurrence of injury

Personal, Workload features from GPS Tracking data, previous injury

26 Male professional soccer players (age 26 ± 4) in Italy

1 season (2013–2014)

Prospective cohort

Player training session

952

23/13

55

RUDDY ET AL., 2018 [38]

Occurrence of hamstring strain injury

Age, previous hamstring strain injury, low levels of eccentric hamstring strength

362 Male professional players from the Australian Football League in Australia: 186 in 2013 (age 23.2 ± 3.6) and 176 in 2015 (age 25.0 ± 3.4)

2 seasons (2013, 2015)

Prospective cohort

Player

2013: 186

2015: 176

2013: NR/27

2015: NR/26

3 or 8

THORNTON ET AL., 2017 [41]

Occurrence of Injury

Training intensity

25 Male professional rugby players from Australian National Rugby League in Australia. Athletes were included in the dataset if they sustained more than 3 injuries in total

3 seasons (2013–2015)

Prospective cohort

Player days

NR

156/25

NR

WHITESIDE ET AL., 2016 [46]

Occurrence of ulnar collateral ligament reconstruction

Demographic and pitching performance

208 Male professional baseball pitchers from the Major League Baseball in the USA and Canada: 104 cases (age 27.3 ± 3.8) and 104 controls (age 27.8 ± 3.7)

5 years (2010–2015)

Matched Case–control

Player

208

NR/NR

14

  1. a The for the analysis relevant number is put in bold. If the unit of observation is player, then the number of injured players is relevant, since one only detects if a player gets injured at least once. If there are multiple observations per player, the total number of injuries is relevant for the analysis