| Tim Parkin, Ph.D. - Senior Fellow in Clinical Research, University of Glasgow; Consultant, Equine Injury Database
Ogden Mills Phipps: When The Jockey Club and Grayson-Jockey Club Research Foundation organized and underwrote the first of our Welfare and Safety of the Racehorse Summits, we felt confident that good things would come from it, and many have.
One of the most successful and high-profile initiatives is the Equine Injury Database.
The Equine Injury Database is a shining example of industry cooperation and of doing right by the horse.
Dr. Tim Parkin has studied equine injury data and served on prestigious veterinary committees in racing jurisdictions throughout the world. We are fortunate to have him as a consultant on this project.
Some of you met him at the last Welfare and Safety of the Racehorse Summit at Keeneland, but this is his first appearance at the Round Table Conference. He is going to share some very interesting information about his recent findings.
Dr. Parkin, welcome to the Round Table...
Tim Parkin, Ph.D.: Thank you very much. Good morning everyone.
In the fall of 2006, the inaugural meeting of the Welfare and Safety of the Racehorse Summit identified as one of its key objectives the development of a database to facilitate the analysis of key data elements related to racing injuries.
Five years later and with 90% of racetracks participating, the Equine Injury Database is truly a shining example of the progressive thinking our industry is capable of when faced with challenges to better understand and reduce injuries to our athletes.
Sophisticated databases for recording injury information already exist in Australia, Hong Kong, Great Britain and Japan. The information that emerges from these large warehouses of data has provided terrific insight into the whys and whats of racing injuries, providing a scientifically sound basis for decision-making.
In March, June and December of 2010, we published the results of our statistical survey of the database looking at the relationship of a number of single variables to the prevalence of injury in the racehorse. These single variables included sex, distance raced, weight carried, type and condition of racing surfaces.
However, as is the case with many complex systems, the whole story is only truly understood when multiple variables are considered rather than trying to find a single magic bullet.
With over 1 million race starts represented, the EID is capable of supporting these very complex analyses where we can look at two, three and even more variables to better understand racing injuries; and today, I want to share the preliminary results of these complex analyses. As we add another year of data to the database this winter, we will continue to refine our analysis and further update you.
The power of data resides in its ability to assist us in evaluating the risk associated with various management decisions. Our objective is to identify horse and race profiles which present the highest risk of injury, and today we will limit our discussion to a very specific injury, catastrophic lower limb fracture. So in other words, fractures that require euthanasia of the horse.
I will touch briefly on some background to the analysis, use human examples to illustrate what we are trying to achieve and finally present you with two messages that we feel confident will help toward our ultimate goal of improving equine, and jockey, safety and welfare.
Before we start I would like to briefly acknowledge the assistance of members of our group in Glasgow, Richard [Reardon] and Claire [Walls], who helped with a lot of the data manipulation, and also The Jockey Club staff and Dr. Mary Scollay, who has maintained a passion and commitment to the success of the EID.
There is a clear determination in North America and other racing jurisdictions around the world to minimise the risk of injury on the track.
Epidemiologists help to achieve this goal through the analysis of large databases that provide substantial amounts of information about every start, made by every horse, over significant periods of time. Large databases such as the EID in combination with race data provided by Equibase provide the maximum opportunity to identify factors that increase or indeed decrease the risk of injury.
I have performed similar analysis for racing jurisdictions in Hong Kong, Australia, and the UK and continue to work closely with a common aim of reducing injury with individuals in all of those countries.
So what is it that we are trying to do?
It is probably easier to think of human examples. So we know, through very similar work to that which we are conducting with the U.S. Jockey Club, that smoking, alcohol and body weight are risk factors for heart disease and that smoking is a predominant risk factor for lung cancer.
When one arrives at the doctor he or she will attempt to make an assessment of your likely risk of suffering a heart attack. This assessment will be based on a number of things that are directly measurable such as your blood pressure and cholesterol level, but he or she will also take account of your lifestyle, i.e., drinking and smoking habit, and will no doubt encourage you to moderate your consumption.
The ultimate aim of our work is to produce risk profiles for different types of horses or races such that decisions can be made about the risk each horse is exposed to when entered into a particular race.
There is one methodological item I want to cover before we get to the results of this first round of analysis.
This concept is extremely important and distinguishes the current analysis from our initial analysis of the data presented at the Welfare and Safety Summit last year.
I want to describe confounding to you, and to show why it is so important to take account of as many variables that may or may not be associated with the outcome as possible.
If the only bit of information we had about everyone in this room was the colour of their hair, and we were required to identify individuals at greater risk of heart disease, we may decide that we would class everyone with grey hair in the "at risk" group and everyone else would be low risk. This would be a rather crude method of risk classification but would on average perform better at identifying those at risk rather than randomly dividing into two groups the people in this room.
It is of course ridiculous to say that grey hair causes heart disease. Grey hair is a red herring. It is simply a proxy measure of some other factor that truly is associated with heart disease, i.e., high blood pressure, which tends to increase with age as does the number of grey hairs on one's head.
In our situation with racehorse injuries, if we were to ignore all possible variables we may inadvertently identify proxy measures of the risk of injury without identifying the true causal pathway.
It is for that reason that we often start these analyses by constructing causal webs such as this that attempt to map out, not only the relationships with the outcome of interest but also the inter-relationships that may exist between these factors so that we can best ensure against the identification of proxy measures or red herrings. When looking at this slide, it is clear that the genesis of racing injury is complex and oversimplification may result in a source of risk being clearly overlooked.
We have performed complex analyses of the data that included more than 1.5 million race starts to identify risk factors for catastrophic lower limb fracture in race data collected between November 2008 and October 2010.
At this time, we have identified eight different factors that are associated with this outcome.
So let's see what we have learned, and more importantly, how we can use this information.
First, we identified particular profiles of previous racing histories that are associated with an increased risk of a start ending in catastrophic lower limb fracture.
The highest risk profile included:
A review of past performances affords the ability to compare this high risk profile with a given horse — and perform an improved risk assessment. This, in turn, could result in more effective decision-making when entering a horse, or indeed when implementing more sophisticated pre-race examination procedures.
This table illustrates the two extremes of each of the risks I just described.
As you might imagine there are very few horses that fit the very highest risk profile — under "Horse A" here on the left. This horse started more than 10 times in the past six months without any starts in the past 15 days and with its first start occurring in the last nine months. Equally there are very few horses that fit the very lowest risk profile as represented by "Horse Z" in this table. The vast majority of horses fall somewhere between A and Z with more of them clustered around the middle, say J to P, than anywhere else.
As we continue to add data to the EID, our aim is to provide refined guidance on how these risk profiles change when these variables undergo subtle changes themselves.
Nevertheless, it is important to show you the potential difference in risk associated with different profile horses. Here I have compared the highest and the lowest risk profiles and calculated that the highest risk horse is on average approximately at 10 times greater risk than the lowest risk horse.
It is also possible to change a single variable one at a time to identify by how much risk would be reduced or increased if horse profiles, or racing histories, were altered slightly.
In terms of what we can talk to trainers about and what "type" of horse may pose a greater risk, we have identified that colts, horses that start racing later in life and those that are currently older form a "high risk" type of horse. This is on top of any risk profile associated with racing history.
Comparing the risk associated with the highest risk horses and the lowest risk horses — i.e., females or geldings, that are currently 2-years of age and those that started racing as 2-year-olds — shows that on average there is a 40-fold difference in the risk of catastrophic lower limb fracture.
This is a very large difference, but we should remember that there are very few horses — in fact only one in the current dataset — that display the highest risk profile. As with racing history the vast majority of horses will have risk profiles that place them in the medium risk category. Again, this identification of the at-risk profile means that an opportunity to intervene has been identified. The trainer with a 4-year-old colt making his first start needs to exercise a higher level of vigilance. This horse can race safely and successfully. But the increased risk must be mitigated by more mindful management.
It is obviously possible to combine all of the high risk profiles together to identify the combination of factors resulting in the greatest cumulative risk of catastrophic lower limb fracture. Doing this and comparing it with the lowest risk profile indicates that the risk difference would be approximately 250-fold — a very large and significant difference.
As I have already mentioned, these very high risk profiles — and indeed very low risk profiles — are extremely rare, but I can see that these types of analyses provide us with an exciting opportunity to identify the top, say, 25% of risk profiles and focus our interventions on those 25% of race starts.
In this way we are likely to have the greatest impact on the overall population prevalence of injury without wasting time and effort on starts that are extremely unlikely to result in injury anyway. In fact, the data indicates that 50% of cases occurred in the top 25% of high-risk profiles.
By profiling we can identify the population of horses at markedly increased risk — and implement measures to mitigate that risk.
Any decisions on where our focus would be best placed will be helped by further analysis that will refine these risk profiles such that we should be able to target interventions at fewer races with greater impact.
It is important to note that what we have presented today is simply the start of a series of analyses that will gradually build in complexity.
In November there will be yet another year of data from EID available to us. With each additional year of data, our ability to identify more subtle relationships with injury will increase.
We will also aim to include workout data, where available, so that we can better model the relationship between high-speed exercise and injury.
All of this will result in more accurate profiles of high- and low-risk horses and races to enable more targeted interventions and advice to improve equine welfare and safety.
We anticipate that ultimately these types of profiles will play a part in decision-tree analysis used to optimise pre-race examinations, stage safe races, or place horses in appropriate races.
As presented at the Welfare and Safety Summit, data from existing international databases has been used successfully to identify strategies to reduce risk of catastrophic injury in racehorses around the world. The EID is just starting to fulfill its potential in North America, and we look forward to continuing these analyses with the health of the horse as our priority.
Thank you very much.
Ogden Mills Phipps: Thank you very much, doctor, for those insights into the Equine Injury Database. We appreciate your efforts to get here from Scotland.