By Dan Ross
A new HISA computer model to identify at-risk horses is being used by Churchill Downs officials to complement existing track safety nets, representatives from the Horseracing Integrity and Safety Authority (HISA) and Horseracing Integrity and Welfare Unit (HIWU) announced during Tuesday's press conference on equine safety protocols in place for this year's GI Kentucky Derby and GI Kentucky Oaks.
“It's essentially a model that looks at 44 potential risk factors that might cause a horse to be at risk for injury,” said HISA CEO Lisa Lazarus, who added that the model assigns a “numerical” value to the horses analyzed.
HISA is currently beta-testing the system with regulatory veterinarians at roughly 10 tracks around the country, said Lazarus.
“But it is being provided to Churchill Downs' head of the Derby to be one of the tools that they are able to use with their sophisticated veterinary team and practices to help determine whether there are any horses at the Derby or on Derby Day that are at risk of being injured,” she said.
Though the system as a whole is still being beta-tested, “I would expect by the end of this quarter that it's available to any regulatory veterinarian,” a HISA representative explained on background during a follow-up conversation.
“Right now, it's grabbing the things that we already know are concerning,” the HISA representative added. “It's the 'what's-going-to-happen' that's so exciting, because we've never been able to think about using data before in this way.”
Experts have long identified specific factors that place a racehorse at increased risk of injury, including fatal injury. Some of these long-understood factors include sex of the racehorse, type and distance of race, and number of previous non-fatal injuries.
The 44 factors woven into the computer model, said Lazarus, include vets' list histories, if the horse has changed hands, any drops in class, and time between races.
According to the HISA representative, the model also looks at possible associations with the horse's veterinary treatment history.
“One of the things HISA can do that the state commissions couldn't necessarily do at this scale is look for associations with treatment records,” said the HISA representative, explaining how even in fairly closed racing populations like California, such processes were difficult to accomplish for logistical reasons.
“We can look at associations between this type of treatment pattern in the past and a poor outcome,” said the representative.
As to how the system works, a daily report is generated on every horse entered to race, said the HISA representative. This report assigns a number between 0 and 44 depending on the number of individual risk factors flagged by the system.
“It's got the horse's name, it has an aggregate number, and then it has the breakdown of that within the 44 risk factors,” the HISA representative explained.
When asked if a certain numerical threshold triggers a specific set of follow-up protocols, Lazarus said that regulators have broad discretion as to how they weave the model into their existing safety and screening practices.
“It's not a tool that's designed to say this horse has to be scratched. It's a tool that's designed to help the veterinarians on the ground do the most effective veterinary inspection they can,” said Lazarus.
“These are the horses you may want to spend extra time on when you're out on your rounds,” the HISA representative further explained. “Or, if you're in a jurisdiction that does post-entry screening and then goes and looks at the horses, like in California, maybe there's an extra horse you put on your list to go and look at before race day.”
The model, said Lazarus, is still being tweaked to ensure that horses at no increased risk of injury are inadvertently flagged by the system. “The reason why we haven't deployed it everywhere yet is because we're still improving it and making sure that it's at an optimal level,” she said.
“There's always a tension between a false positive and a false negative,” agreed the HISA representative.
“There are horses where there's a predictive value here [that] go out and don't just run credibly but run off the charts and win,” the HISA representative added. “But the harm in being wrong by saying 'this horse needs extra scrutiny' is a much lesser harm than in missing a horse.”
When asked about the accuracy of the model, the HISA representative said that HISA had got the percentage of “poor outcomes” down to the low single-digits. Poor outcomes, the HISA representative further explained, includes horses scratched on race day and other injury records. “It's not just fatalities.”
Currently, horsemen are not privy to the information generated by the model, said the HISA representative, while leaving the door open to such an eventuality. “That's just a policy discussion we haven't got to yet.”
Not a subscriber? Click here to sign up for the daily PDF or alerts.
