Denise Mc Sweeney, Yuan Wang, Scott E. Palmer, Mikael Holmströem, Kevin D. Donohue, Kelly D. Farnsworth, Macarena G. Sanz, David H. Lambert, Warwick M. Bayly
Background
Fatal musculoskeletal injuries (FMIs) remain a significant welfare and sustainability issue in Thoroughbred racing, despite reductions in race day fatalities. Approximately 93% of FMIs involve preexisting bone pathology that develops unnoticed during training. Visual detection of subtle gait abnormalities at racing speed is nearly impossible. Accelerometer-based inertial measurement unit (IMU) sensors have recently been used to identify horses at increased risk of FMI by detecting subclinical biomechanical deviations. This study aimed to determine whether algorithm-derived risk scores from IMU data could predict FMI risk and to assess associations with horse demographics and track characteristics.
Methods
A retrospective analysis of IMU data from 28,481 races involving 11,834 Thoroughbreds (July 2021–May 2024) across 10 U.S. racetracks was performed. Sensors mounted on saddlecloths recorded three-dimensional acceleration (800 Hz sampling) and GPS data. A machine learning algorithm, trained on stride data from Grade 1–2 stakes winners, generated risk scores from 1 (lowest risk) to 6 (highest risk) based on deviations from biomechanical norms. Logistic regression models evaluated associations between fatal injuries and variables including risk score, age, gender, race distance, and track surface. The 120 days preceding each fatal injury were used for temporal analysis.
Results
Seventy-four horses suffered FMIs, primarily fetlock fractures (65%). Horses with risk scores of 6 (0.4% of starts) had a 44.6-fold higher probability of FMI compared with those scoring 1. The probability of injury increased exponentially with risk score (R²ₘcF = 0.67).
-Incidence: 4.2% of horses with a score of 6 sustained a fatal injury within 120 days.
-Gender: Males had higher FMI rates than females.
-Age: Not significantly associated with injury risk.
-Distance: Higher risk was associated with shorter races.
-Surface: Fatality rates were highest on dirt (0.0035/start), followed by turf (0.0023/start), and lowest on synthetic tracks (0.0005/start; P < .05).
-Horses with repeated IMU monitoring exhibited improving risk prediction accuracy, suggesting cumulative data enhance algorithmic reliability.
Limitations
The study lacked histopathologic confirmation of preexisting lesions and relied on retrospective injury reporting, potentially underestimating total FMI events. Sensor data were not uniformly collected across all U.S. tracks, limiting generalizability. Additionally, the proprietary algorithm was developed primarily from dirt-track data, which may bias predictions by surface type.
Conclusions
Thoroughbred racehorses assigned a risk score of 6 by IMU-based algorithms were significantly more likely to suffer a fatal musculoskeletal injury within 120 days than horses with lower scores. Implementing IMU-based screening in combination with targeted lameness examinations and advanced imaging could prevent up to 19% of racehorse fatalities. IMU data offer a practical early warning system—akin to a “check engine light”—for detecting biomechanical abnormalities before catastrophic failure, potentially transforming racehorse safety and welfare.
Relationships between fatal musculoskeletal injury rates and risk scores assigned to 11,834 Thoroughbred racehorses from 28,481 race starts from July 25, 2021, to May 4, 2024, according to the track surfaces on which the races were run. On dirt tracks, the odds of a horse with a risk score of 6 suffering a fatal injury were 43.4 times greater than for a risk score of 1. The odds of horses with risk scores of 3 to 5 suffering a fatal injury were greater than for a risk score of 1 but not for a risk score of 2 on dirt. There was no difference in injury probability between risk scores of 1 and 2 on dirt surfaces. On turf tracks, there was insufficient statistical power to allow valid comparisons between different risk scores with respect to the odds of a fatal injury occurring. The odds of a horse racing on the all-weather synthetic (AW) surface suffering a fatal musculoskeletal injury were not different for risk scores from 1 to 6.
How did we do?
Disclaimer: The summary generated in this email was created by an AI large language model. Therefore errors may occur. Reading the article is the best way to understand the scholarly work. The figure presented here remains the property of the publisher or author and subject to the applicable copyright agreement. It is reproduced here as an educational work. If you have any questions or concerns about the work presented here, reply to this email.
