Letter To The Editor: Beyond HISA's Star Metric–One Small Number Rewrites The Big One

I've read every HISA annual metrics report since the first one landed in 2024. All three of them. Cover to cover, footnotes included. And every year I've come away with the same feeling: something is missing.

Not from the effort. The effort is real. HISA has built the most comprehensive equine health database in the history of American racing. Seven million veterinary treatment records. Treatment histories that follow horses across state lines for the first time. The infrastructure is extraordinary.

What's been missing is statistical depth. Year after year, the report orbits a single number: racing-related fatalities per 1,000 starts, HISA's star metric. The number goes down, and the press release says progress. The number ticks up, and the language shifts to rolling averages and two-year windows. Every year, I've waited for the report to give us something more, something that tells a fuller story about what's happening to these animals.

This year, it did. And I don't think many will fully realize what it revealed.

Sarah Andrew photo

Eighty-Eight Percent
For the first time, HISA's 2025 annual report includes fatality subclassifications. It breaks down how horses die, not just how many. And the numbers are striking.

Of the 176 racing-related fatalities at HISA tracks in 2025, 88.1% were musculoskeletal. Bones. Tendons. Ligaments. Training-related fatalities told a similar story: 77.9% musculoskeletal. Combined, 82.2% of all equine deaths under HISA's reporting fell into that single category.

Here's what that means in plain language: nearly nine out of 10 horses captured by HISA's headline safety metric did not die spontaneously. They were euthanized. A veterinarian assessed a catastrophic injury and made a decision. In many of those cases, the decision would have been clear, the only humane option in front of an unsalvageable fracture. But in my experience, in others, it can be anything but.

If you own, manage, or train enough horses, you will find yourself on the receiving end of the euthanasia conversation with a veterinarian more times than anyone would like. I have. There is the horse. There is the injury. There is the vet telling you what they see. And then there is a cascade of factors that no metric captures: the severity of the fracture, the insurance situation, the surgical options available at that facility on that day, the owner's willingness to invest in a long and uncertain rehabilitation. Two vets looking at the same horse can reach different conclusions. Two owners facing the same prognosis will make different choices. A catastrophic injury does not mean instant death. It means a decision has to be made, and the outcome of that decision depends on circumstances at that moment.

Consider Barbaro. His breakdown in the 2006 Preakness was one of the most visible racing injuries in modern history. His connections had the resources, the will, and the veterinary team to attempt what few owners could. He fought for eight months. And ultimately, after complications from laminitis, the decision was made to euthanize him. By HISA's current definition, which captures deaths within 72 hours of a race, Barbaro would not have been counted in the headline fatality rate. One of the most famous racing-related deaths in the sport's history falls outside the metric's window. That's not a flaw in the definition. But it tells you something about what the number is designed to capture and what it isn't.

What 88.1% tells us is that HISA's headline metric is not measuring injury incidence. It is measuring the outcome of a decision-making process, shaped by multiple factors, and a 72-hour window that any determined connections could extend simply by choosing to try. Fatalities are the most objective endpoint available, and that's exactly why they became the default metric. But objectivity is not the same as completeness. What we are left with is a number that tells us how many decisions ended in death, not how many horses were injured.

The headline number is also inherently subject to chance. Tim Parkin, the Bristol epidemiologist most closely associated with the Equine Injury Database, once illustrated this by noting that adding a single additional death in one subgroup, one horse on a wet track, would have swung that category's rate from 1.18 to 1.57 per 1,000 starts. One horse. A 33% change. When you're measuring an event that occurs roughly once per thousand, chance dominates the year-to-year variation.

Then there's the denominator. Total starts across North American Thoroughbred racing have fallen from roughly 446,000 in 2009 to about 247,000 in 2025. That's a 44.6% decline over the same period the fatality rate fell 47%. The rate doesn't tell you how much of the improvement comes from the sport getting safer versus the sport getting smaller. Both are happening at once. A single number can't separate them.

So when we look at HISA's headline rate of 1.04 per 1,000 starts in 2025, down from 2.00 when the EID began tracking in 2009, the rate is going down. But what is it actually tracking? It could be measuring genuine improvement in horse safety. It could be measuring a shift in veterinary decision-making culture. It could be reflecting the statistical noise inherent in rare events.    It could be reflecting a sport with a shrinking footprint.

It is probably all four. And a single number can't tell you how much of each.

Expanding the Frame
So what should we measure? Before answering that, we need to reckon honestly with the nature of our sport.

Sarah Andrew photo

Even if we could somehow bubble wrap them, horses would still find ways to get hurt, because that is the nature of a 1,200-pound flight animal with legs built for speed. The only way the fatality rate reaches zero is if the number of starts reaches zero. We cannot eliminate risk. But we can stop chasing an impossible number and start building systems that detect and prevent harm in real time. That requires bigger data sets, and bigger data sets tell better stories.

When you track only fatalities, you're working with the smallest possible numerator, roughly 265 deaths across 247,000 starts in 2025. That's a sample so small that chance dominates the year-to-year signal. Expand the measurement, and the data set grows by orders of magnitude. Bigger data sets don't just reduce noise. They reveal patterns. And patterns are how we move from counting deaths to preventing them.

The most immediate expansion would be non-fatal injury rates. Track how many horses are injured, not just how many die, and you can see whether injuries are declining or simply being survived at higher rates. The British Horseracing Authority already does this. Their framework, called “A Life Well Lived,” tracks long-term injuries, falls, and fatalities as distinct categories, because their own data showed them that measuring deaths alone could mask what was happening to the horses that survived. HISA has the data to do the same. It should.

Career longevity matters too, because average starts per horse tells a story about soundness that a death count cannot. Time between starts and return-to-racing after injury both matter, because a horse that comes back to the track is a horse with ongoing veterinary oversight and a purpose. Instead of counting deaths retroactively, the sport could track the conditions that precede catastrophic injury and intervene before the breakdown occurs. That's the difference between a report card and a tool that saves lives.

And life after racing matters most of all. A horse's welfare is directly related to its utility. A horse that races, that has a purpose, is in many ways safer than a horse that has fallen out of the system entirely. There are more premature Thoroughbred fatalities outside the racing ecosystem than inside of it. Even the conservative estimate puts the number of former racehorses entering the slaughter pipeline annually at 4,000, a figure that dwarfs the 265 racing fatalities at HISA tracks in 2025.

The Jockey Club's Kristin Werner has said it directly: “Along the way, we lose the ability to track many, many horses.” If a horse's welfare is tied to its utility, then the sport has an obligation to invest in what happens when racetrack utility ends: retraining facilities, secondary markets where Thoroughbreds retain purpose beyond the track, pathways that keep horses visible and cared for rather than letting them disappear.

HISA's mandate may not extend beyond the racetrack. But HISA's metric is presented as a measure of how well we care for these animals. If the measurement stops at the gate, it should say so.

One Small Number
Come back with me to where we started. What triggered this piece was one line in a 22-page report. A subclassification that HISA added for the first time this year. 88.1% musculoskeletal. Most readers will scan past it. Most coverage didn't mention it. It's a small number, buried in the data.

But look at what that one small number revealed. It told us that the headline metric overwhelmingly measures decisions, not injuries. It raised questions about whether a declining rate reflects genuine safety improvement or a shifting landscape of veterinary practice, chance, and a shrinking sport. One number. One addition to the report. And suddenly the star metric that the entire sport has organized itself around looks incomplete.

Now imagine what a broader framework could tell us. If one small addition to the data could expose the limitations of the headline metric, a full set of welfare indicators could show us if and how the sport is actually getting safer and, more importantly, why.

To HISA's credit, the report is evolving, getting a little deeper every year. That trajectory matters. But the question now is whether that evolution continues to build outward, toward a measurement framework that captures welfare in full, or whether it circles back to the same headline fraction. The infrastructure HISA has built can support the first path. The database is there. The tools are there. What the sport needs now is the commitment to measure what matters, not just what fits on a single line.

I hope HISA chooses the first path. And I believe the industry should demand it. Because HISA has built something good enough to become something great, and the distance between good and great is the distance between one number and the full picture.

Sobhy Sonbol is the owner of Nile Bloodstock, a bloodstock and racing advisory service based in Kentucky. He has been involved with such notable runners as American Pharoah, Pioneerof the Nile, and Vyjack. He is also the president of Pioneerof AI, which services both racing and non-racing projects. He graduated from UC Berkeley with an engineering degree.