Abstract Objective To determine whether targeted lesion-specific algorithms developed from data collected with accelerometer-based inertial measurement units worn by racing Thoroughbreds could retrospectively identify horses at high risk of incurring forelimb condylar or proximal sesamoid fractures. Methods Fracture-specific algorithms were generated from July 2021 to December 2024 with accelerometer data from 42,623 races by 15,755 horses, including 54 races by 23 horses that subsequently suffered condylar fractures and 90 races by 31 horses that later sustained proximal sesamoid fractures. Fracture-specific algorithms placed horses into relative risk groups: green (low risk; score = 1), amber (asymmetry while racing; score = 2), and red (high risk; scores of 3 to 5 5 = most at risk) based on inertial measurement unit data that were compatible with the subsequent occurrence of either fracture. Results The positive predictive values were 0.09 for condylar and 0.08 for sesamoid fractures, respectively, in the most-at-risk horses. For horses with ≥ 3 starts and mean risk scores > 3.0 over 3 consecutive starts, positive predictive value was 0.74 and OR was 133 (95% CI, 23.5 to 753) compared to horses with mean scores < 3.0. Eighty-nine percent of all 53 fractures occurred in the red-risk group. Greater than 98.8% of horses with a risk score < 5 did not suffer either fracture, while ≤ 1.2% of the nonfracture cases received a risk score of 5. Conclusions Most forelimb condylar and proximal sesamoid fractures were incurred by horses retrospectively identified as being at high risk by these fracture-specific algorithms. Having ≥ 3 risk scores/horse greatly increased the algorithms’ precision. Clinical Relevance Fracture-specific algorithms can identify horses at high risk of future condylar or proximal sesamoid bone fractures, thereby assisting veterinarians tasked with identifying and clinically evaluating them.
Hall et al. (2026) studied this question.