The biometric data that could predict racehorse injury

Imagine having the knowledge to prevent a catastrophic event from occurring ahead of time. In horse racing, this could be a possibility.

Our new research investigated whether the same systems used to help punters pick a winning racehorse could provide the data needed to protect that same racehorse from injury.

A problem with measurement

Efforts to date at predicting injuries in racehorses haven’t been much better than a coin flip. In the past, injuries in horse racing have been treated as a binary outcome – a horse is either injured or uninjured.

But injuries, largely a result of bone damage, have a gradual onset.

An injury may develop over weeks or months, unless it’s due to a traumatic event (like a horse running into a fence), so it rarely occurs the very day it is observed. We know this because the majority of catastrophic injuries in racehorses have shown evidence of pre-existing bone damage.

This damage accumulates during training and racing over time due to repeated loads being put on the musculoskeletal system. With every stride taken by a horse galloping at a moderate speed, loads of up to four tonnes on the fetlock joint surface have been measured.

Bone can only withstand a limited number of these loads and the strides taken at faster speeds produce greater loads.

Often the initial detection of injury is when the horse breaks down or shows signs of lameness, indicating that the threshold for bone damage has already been reached.

But rather than waiting until that injury becomes obvious, we realised that a way of measuring horses’ response to training and racing workloads was required.

The dawn of the data driven era

So, what if there was a way to measure the onset of injury using already established systems of data collection? As it happened there is.

It all started back in 2010 when the principal racing authority of Tasmania, Tasracing, partnered with StrideMASTER, a budding technology company developing training monitoring systems for the racing industry.

They developed one of the first race-day timing systems using GPS and precision motion sensor data. The initial purpose of this data was for real-time sectional timing and positional data of racehorses that was then intended for use as a broadcast and wagering product.

I had worked with Tasracing previously and in 2016, on my return to Australia, we reconnected.

Because Tasracing prides itself on being data driven, it’s this approach that provided us with a potential solution.

With access to biometric data – like racehorse speed, stride length and stride frequency – we now have the means of measuring changes over horses’ racing careers that might flag an injury occurring before it becomes too late.

A horse slowing down is a horse to watch

Our research firstly matched the StrideMASTER biometric data with other race field information and veterinary findings provided by Tasracing from between 2011 and 2016.

Using a statistical method that hadn’t been used before in this setting, we first modelled the changes in stride characteristics over successive race starts, then modelled the number of race starts before an injury occurred.

Finally, the two models were integrated in what is termed a ‘joint model’ to determine whether observed changes in stride characteristics predicted injury.

And they did.

For horses some way into their careers, those that sustained an injury during a race start slowed their racing speed and shortened their stride length more rapidly around six races beforehand.

My colleague, Professor Chris Whitton, noted that although we expected to see changes in speed and stride in races leading up to an injury, the fact that we were seeing those changes so long prior is surprising. Yes, we thought we might see changes maybe one or two races, but not six – that’s pretty astonishing.

For horses that sustained an injury earlier in their careers – if we look at just the first few races where there’s yet to be enough data about their normal stride characteristics – additional monitoring during training may be needed in order to predict these early career injuries.

A horse’s risk of injury increased by 18 per cent when decreasing their speed by 0.1 metres per second, or by 11 per cent when shortening their stride length by just 10 centimetres in the final section of each race.

And our findings held even after accounting for factors affecting speed and stride like race distance and track conditions.

Though these changes may sound small, within an individual horse that has its own unique stride, they are a sign that something may be off.

This may be because they cannot withstand the workload, they are experiencing pain or are otherwise compensating physiologically for the accumulated bone damage.

It’s at this point that veterinary advice should be sought.

Established systems used in meaningful ways

Our study is just the dawn of how data driven research could improve welfare and safety in the racing industry using new and meaningful ways to repurpose data.

Our findings also demonstrate the immense potential for identifying and preventing injuries in racehorses before they become catastrophic.

Racing authorities should take the lead and push for the wider implementation of motion sensor technology that can monitor both training and racing. This would cover the elements we’re still unclear on – like whether the same slowing of speed and shortening of stride also occurs during training.

With other jurisdictions now taking up similar systems collating biometric data, there’ll be more information, leading to refined algorithms, improved predictions and, ultimately, greater safety for horses.

This research was funded by The Grayson-Jockey Club Research Foundation, Tasracing, Racing Victoria, the Victorian Racing Industry Fund of the Victoria State Government, and the University of Melbourne.

Banner: Getty Images