Limited research was conducted to quantify risk factors associated with race injuries in racehorses until the 1990s. Injuries were generally attributed to bad racetrack conditions, typically on the basis of opinion or hearsay rather than objective facts.1 Furthermore, associations between various risk factors and injuries were usually evaluated with only univariable analyses, and the use of multivariable techniques that can account for confounding was largely neglected. Age,2–6 sex,2,3,7,8 race distance,4,5,7,9–12 racetrack surface type and conditions,4,6,10,11,13 race type,3–6,12,14 field size (number of horses in race),10,11 season,5 prior racing history,4,5,7,9–11 distance galloped during training,5,9,13,15–18 exercise history,11,13,16,18 prerace inspection by regulatory veterinarians,19 and horseshoe characteristics20,21 have been associated with fatal injuries in Thoroughbred racehorses in previous studies.
The Jockey Club was established in New York in 1894.22 Its mission is to improve Thoroughbred breeding and racing primarily in the United States, Canada, and Puerto Rico.22 In 2008, The Jockey Club initiated the EID.22 The purpose of the EID is to identify the frequency, types, and outcomes of racing injuries in Thoroughbred racehorses competing in flat racing in a standardized format so that valid statistics can be generated in the hope that factors associated with specific injuries can be identified and appropriate measures implemented to prevent such injuries and improve the safety of Thoroughbred racing.22 The EID contains information for most Thoroughbred races that take place in the United States and Canada and serves as a near-census collection of available data. Despite differences in geography and Thoroughbred racing practices among countries and racing jurisdictions, we believe that the EID database can be used to identify risk factors associated with fatal injuries, and that information can be used to improve racing conditions in the United States and Canada as well as other Thoroughbred racing jurisdictions.
The purpose of the study reported here was to identify risk factors associated with fatal injuries in Thoroughbred racehorses competing in flat racing in the United States and Canada. Specific factors such as age, sex, race distance, racetrack surface type and conditions, race type, field size, and prior racing history that were associated with fatal injuries in racehorses in previous studies2–21 were evaluated by use of multivariable analysis.
Materials and Methods
Data collection
Data were obtained from the EID for the 5-year period from 2009 through 2013. Race start data were provided by The Jockey Club and included information for all racetracks that voluntarily contributed data to the EID during the years of interest. The race starts reported to the EID represented the starts for 93.9% of all official Thoroughbred racing events in the United States and Canada during the 5-year observation period. Injury reports were submitted to the EID by veterinarians at the participating racetracks. The EID database also contained information for approximately 11,000 anonymized trainers and 3,000 anonymized jockeys associated with the recorded races, which enabled us to analyze trainer- and jockey-related risk factors. Because the EID did not contain information regarding extent of training exercise, we used the number of races started by a horse prior to the race of interest as a proxy for cumulative training exercise. Also, the EID did not contain information regarding prerace inspection by regulatory veterinarians; therefore, we used the number of times a horse had been withdrawn from a race (scratched) prior to the race of interest as the proxy for prerace regulatory inspection.
Data analysis
The unit of analysis for the study was race start. For each race start, horses that died or were euthanized within 3 days after sustaining an injury during a race (ie, fatal injury) were classified as cases and all other horses were classified as controls. All official race starts were used to calculate historical information for each horse, but only the 1,891,483 official starts from the racetracks that voluntarily reported injuries to the EID were used in the regression analyses. The outcome of interest, or outcome variable, for all regression analyses was fatal injury. Forty-four plausible risk factors, or independent variables, for fatal injury were selected for consideration in our analysis. Initially, the respective associations between each risk factor and the outcome variable were assessed by use of univariate logistic regression. Only horses that had been competing in flat racing for at least 6 months were included in the analyses to assess risk factors that summarized historical racing information prior to the race of interest; this facilitated the assessment of those relationships because data for horses that had only begun racing within the preceding 6 months were excluded from those analyses.
Risk factors with values of P < 0.20 on univariate analysis were eligible for inclusion in a multivariable logistic regression model. A threshold of P < 0.20 was chosen to prevent exclusion of a potentially significant risk factor that only becomes evident when a confounder has been controlled for in a multivariable analysis.23 An automated stepwise selection process was used to build the multivariable model. Potential confounders were identified by use of a forward bidirectional elimination approach and assessment of the AIC. The AICs for competing models were compared, and the model with the lowest AIC was preferred.24 Only risk factors with values of P < 0.05 were retained in the final model.
The potential effect of horse in the data analyses was evaluated by creating a mixed-effects model that included horse as a random effect. Results were nearly identical (< 0.01 change in ORs and no meaningful changes in P values) to results obtained with models that did not include random effects. Therefore, only results for models based on race starts without inclusion of horse as a random effect are reported.
Results
Between 2009 and 2013, Thoroughbred flat races took place at 144 racetracks, of which 37 had < 1,000 race starts recorded during the observation period. Collectively, data were extracted from the EID for 154,527 Thoroughbred racehorses that competed in 234,577 races at 89 racetracks. Those horses accounted for 1,891,483 race starts, of which 3,572 resulted in a fatal injury; thus, there were 1.9 fatal injuries/1,000 race starts. Univariate logistic regression results of the respective associations between the selected risk factors and the incidence of fatal injury for all horses (Table 1) and only those horses that had been competing in flat racing for ≥ 6 months (Table 2) were summarized.
Results of univariate logistic regression for assessment of risk factors associated with fatal injuries in Thoroughbred racehorses competing in flat racing in the United States and Canada during the 5-year period from 2009 to 2013.
| Risk factor | Control | Case | OR (95% CI) | P value |
|---|---|---|---|---|
| Sex | ||||
|  Mare or gelding | 1,648,276 (87) | 2,986 (84) | Referent | — |
|  Stallion | 239,635 (13) | 586 (16) | 1.35 (1.24–1.48) | < 0.001 |
| Age (y) | 4.516 (4.514–4.519) | 4.596 (4.546–4.646) | 1.034 (1.013–1.056) | 0.002 |
| Age at first race start (y) | 3.428 (3.426–3.429) | 3.514 (3.473–3.554) | 1.055 (1.029–1.082) | < 0.001 |
| Country | ||||
|  Canada | 155,510 (8) | 188 (5) | Referent | — |
|  United States | 1,732,401 (92) | 3,384 (95) | 1.616 (1.395–1.872) | < 0.001 |
| On vet list | ||||
|  No | 1,877,766 (99) | 3,531 (99) | Referent | — |
|  Yes | 10,145 (1) | 41 (1) | 2.149 (1.579–2.926) | < 0.001 |
| First race start | ||||
|  No | 1,740,519 (92) | 3,375 (94) | Referent | — |
|  Yes | 147,392 (8) | 197 (6) | 0.689 (0.597–0.796) | < 0.001 |
| Field size (No. of horses) | 8.521 (8.519–8.524) | 8.548 (8.486–8.611) | 1.007 (0.990–1.024) | 0.404 |
| Finish percentage of previous race (%) | 54.047 (54.025–54.069) | 52.470 (51.963–52.977) | 0.993 (0.991–0.995) | < 0.001 |
| No. of months in racing | 13.254 (13.237–13.274) | 13.178 (12.780–13.576) | 0.100 (0.997–1.002) | 0.711 |
| No. of previous injuries | 0.0242 (0.0240–0.0244) | 0.0414 (0.0343–0.048) | 1.612 (1.388–1.872) | < 0.001 |
| No. of previous scratches | 1.246 (1.243–1.248) | 1.369 (1.304–1.434) | 1.030 (1.015–1.046) | < 0.001 |
| Odds at start of race | 17.571 (17.541–17.602) | 15.266 (14.642–15.890) | 0.994 (0.992–0.996) | < 0.001 |
| Odds rank in race | 4.761 (4.757–4.765) | 4.375 (4.287–4.463) | 0.946 (0.934–0.958) | < 0.001 |
| Starting (post) position | 4.761 (4.757–4.765) | 4.801 (4.712–4.890) | 1.006 (0.993–1018) | 0.374 |
| Low purse race (≤ $7,500) | ||||
|  No | 1,585,978 (84) | 3,029 (85) | Referent | — |
|  Yes | 301,933 (16) | 543 (15) | 0.942 (0.859–1.032) | 0.197 |
| Purse ($) | 23,905 (23,920–23,989) | 21,011 (19,467–22,555) | 0.998 (0.997–0.999) | 0.001 |
| Change in purse since previous race | ||||
|  No change | 851,954 (45) | 1,528 (43) | Referent | — |
|  Decrease | 438,357 (23) | 959 (27) | 1.220 (1.125–1.322) | < 0.001 |
|  Increase | 382,306 (20) | 693 (19) | 1.011 (0.924–1.106) | 0.817 |
|  Sharp decrease | 116,107 (6) | 223 (6) | 1.071 (0.930–1.233) | 0.340 |
|  Sharp increase | 99,187 (5) | 169 (5) | 0.950 (0.810–1.114) | 0.527 |
| Race distance (fur) | 6.703 (6.701–6.705) | 6.559 (6.515–6.603) | 0.919 (0.896–0.943) | < 0.001 |
| First start with new trainer | ||||
|  No | 1,710,891 (91) | 3,155 (88) | Referent | — |
|  Yes | 177,020 (9) | 417 (12) | 1.277 (1.153–1.415) | < 0.001 |
| Training with first trainer | ||||
|  No | 826,388 (44) | 1,737 (49) | 1.216 (1.134–1.298) | < 0.001 |
|  Yes | 1,061,523 (56) | 1,835 (51) | Referent | — |
| Time with same trainer (mo) | 6.368 (6.356–6.379) | 5.542 (5.300–5.784) | 0.986 (0.982–0.991) | < 0.001 |
| No. of days between race starts since change of trainer | 3.144 (3.085–3.204) | 1.979 (0.526–3.432) | 0.999 (0.998–1.000) | 0.097 |
| First start with new jockey | ||||
|  No | 914,758 (48) | 1,639 (46) | Referent | — |
|  Yes | 973,153 (52) | 1,933 (54) | 1.109 (1.038–1.184) | 0.002 |
| Time with same jockey (mo) | 0.845 (0.842–0.848) | 0.785 (0.726–0.844) | 0.985 (0.968–1.002) | 0.083 |
| Racetrack surface | ||||
|  Synthetic | 235,878 (12) | 289 (8) | Referent | — |
|  Off-dirt* | 248,168 (13) | 528 (15) | 1.737 (1.504–2.005) | < 0.001 |
|  Dirt | 1,152,404 (61) | 2,350 (66) | 1.664 (1.473–1.881) | < 0.001 |
|  Turf | 251,461 (13) | 405 (11) | 1.315 (1.130–1.529) | < 0.001 |
Values for controls and cases were the mean (95% CI) or the number (%) of horses for all race starts. Data were obtained from the EID. A fatal injury was defined as an injury sustained during a race that resulted in the euthanasia or death of the horse within 3 days. There were 1,891,483 official race starts by 154,527 horses; 3,572 starts resulted in a fatal injury (cases), all other starts were considered controls. On vet list refers to horses that were placed by a veterinarian on the vet list and entered a race while on the list. Finish percentage in previous race was calculated as the finish position of a horse/field size of the previous race (ie, a horse that finished fourth in a field of 8 would have a finish percentage of 50%, whereas a horse that finished eighth in a field of 8 would have a finish percentage of 100%). Within a variable and category (control or case), percentages may not total 100 because of rounding.
Dirt in any nonfast (sloppy or muddy) condition.
— = Not applicable. CI = Confidence interval.
Results of univariate logistic regression for assessment of risk factors associated with fatal injuries for the horses of Table 1 that had been racing for ≥ 6 months.
| Risk factor | Control | Case | OR (95% CI) | P value |
|---|---|---|---|---|
| Sex | ||||
|  Mare or gelding | 1,078,244 (90) | 1,945 (86) | Referent | — |
|  Stallion | 117,862 (10) | 311 (14) | 1.463 (1.298–1.649) | < 0.001 |
| Age (y) | 4.942 (4.940 to 4.945) | 5.001 (4.940 to 5.062) | 1.028 (0.999–1.057) | 0.055 |
| Age at first race start (y) | 3.328 (3.326 to 3.330) | 3.410 (3.362 to 3.457) | 1.061 (1.025–1.097) | 0.001 |
| Country | ||||
|  Canada | 94,696 (8) | 105 (5) | Referent | — |
|  United States | 1,101,410 (92) | 2,151 (95) | 1.761 (1.448–2.143) | < 0.001 |
| On vet list | ||||
|  No | 1,188,425 (99) | 2,229 (99) | Referent | — |
|  Yes | 7,681 (1) | 27 (1) | 1.874 (1.281–2.741) | 0.001 |
| Field size (No. of horses) | 8.394 (8.390 to 8.397) | 8.465 (8.388 to 8.541) | 1.019 (0.998–1.042) | 0.079 |
| Finish percentage in previous race (%) | 53.205 (53.183 to 53.227) | 51.418 (50.925 to 51.912) | 0.988 (0.984–0.991) | < 0.001 |
| No. of months in racing | 19.658 (19.639 to 19.678) | 19.376 (18.911 to 19.840) | 0.998 (0.994–1.001) | 0.223 |
| No. of previous injuries | 0.0357 (0.0353 to 0.0361) | 0.0603 (0.0497 to 0.0709) | 1.600 (1.368–1.871) | < 0.001 |
| No. of previous scratches | 1.775 (1.773 to 1.779) | 1.915 (1.822 to 2.007) | 1.027 (1.009–1.044) | 0.003 |
| No. of race starts by horse | ||||
|  < 30 d prior to race | 0.837 (0.835 to 0.838) | 0.758 (0.731 to 0.785) | 0.845 (0.795–0.899) | < 0.001 |
|  30–60 d prior to race | 0.966 (0.965 to 0.967) | 0.100 (0.969 to 1.031) | 1.053 (1.001–1.108) | 0.047 |
|  61–90 d prior to race | 0.868 (0.867 to 0.870) | 0.954 (0.922 to 0.987) | 1.134 (1.080–1.192) | < 0.001 |
|  91–180 d prior to race | 2.198 (2.195 to 2.201) | 2.502 (2.432 to 2.572) | 1.098 (1.073–1.123) | < 0.001 |
| No. of race starts by trainer | ||||
|  < 30 d prior to race | 19.757 (19.717 to 19.797) | 19.778 (18.923 to 20.634) | 1.000 (0.998–1.002) | 0.963 |
|  30–60 d prior to race | 17.160 (17.122 to 17.197) | 17.204 (16.398 to 18.010) | 1.000 (0.998–1.002) | 0.920 |
|  61–90 d prior to race | 16.090 (16.052 to 16.128) | 16.059 (15.231 to 16.888) | 1.000 (0.998–1.002) | 0.946 |
|  91–180 d prior to race | 42.091 (41.984 to 42.197) | 42.476 (40.179 to 44.772) | 1.000 (0.999–1.001) | 0.759 |
| No. of trainer wins | ||||
|  < 30 d prior to race | 3.061 (3.052 to 3.070) | 3.075 (2.879 to 3.271) | 1.001 (0.992–1.009) | 0.898 |
|  30–60 d prior to race | 2.684 (2.676 to 2.692) | 2.740 (2.558 to 2.921) | 1.002 (0.994–1.011) | 0.574 |
|  61–90 d prior to race | 2.528 (2.520 to 2.537) | 2.580 (2.392 to 2.769) | 1.002 (0.994–1.011) | 0.602 |
|  91–180 d prior to race | 6.644 (6.621 to 6.667) | 6.840 (6.329 to 7.350) | 1.001 (0.998–1.004) | 0.469 |
| Winning percentage | ||||
|  < 30 d prior to race | 0.0858 (0.0854 to 0.0862) | 0.0926 (0.0810 to 0.1042) | 1.096 (0.944–1.272) | 0.228 |
|  30–60 d prior to race | 0.0945 (0.0941 to 0.0949) | 0.1234 (0.1109 to 0.1359) | 1.423 (1.242–1.630) | < 0.001 |
|  61–90 d prior to race | 0.0854 (0.0850 to 0.0858) | 0.1097 (0.0979 to 0.1215) | 1.385 (1.201–1.598) | < 0.001 |
|  91–180 d prior to race | 0.0972 (0.0968 to 0.0976) | 0.1208 (0.1118 to 0.1298) | 1.642 (1.374–1.962) | < 0.001 |
| Mean speed during previous race (m/s) | 16.210 (16.208 to 16.213) | 16.259 (16.197 to 16.321) | 1.033 (0.995–1.072) | 0.086 |
| Mean change in speed from previous race (m/s) | −0.006 (−0.009 to −0.003) | 0.020 (−0.066 to 0.107) | 1.008 (0.985–1.032) | 0.482 |
| Time since previous start (d) | 43.639 (43.517 to 43.762) | 41.843 (38.902 to 44.784) | 0.9995 (0.9989–1.0002) | 0.212 |
| Odds at start of race | 16.947 (16.910 to 16.984) | 14.347 (13.610 to 15.084) | 0.993 (0.990–0.995) | < 0.001 |
| Odds rank in race | 4.691 (4.686 to 4.696) | 4.291 (4.183 to 4.400) | 0.942 (0.927–0.958) | < 0.001 |
| Starting (post) position | 4.698 (4.693 to 4.703) | 4.773 (4.662 to 4.884) | 1.011 (0.995–1.027) | 0.178 |
| Low purse race (≤ $7,500) | ||||
|  No | 982,380 (82) | 1,893 (84) | Referent | — |
|  Yes | 213,726 (18) | 363 (16) | 0.881 (0.788–0.986) | 0.028 |
| Purse ($) | 22,683 (22,574 to 22,792) | 19,014 (17,989 to 20,038) | 0.997 (0.995–0.998) | < 0.001 |
| Change in purse since previous race | ||||
|  No change | 492,517 (41) | 905 (40) | Referent | — |
|  Decrease | 299,774 (25) | 641 (28) | 1.164 (1.052–1.288) | 0.003 |
|  Increase | 263,620 (22) | 471 (21) | 0.972 (0.870–1.087) | 0.622 |
|  Sharp decrease | 77,497 (6) | 141 (6) | 0.990 (0.829–1.183) | 0.913 |
|  Sharp increase | 62,698 (5) | 98 (4) | 0.851 (0.690–1.048) | 0.129 |
| Race distance (fur) | 6.792 (6.790 to 6.794) | 6.589 (6.533 to 6.644) | 0.889 (0.861–0.978) | < 0.001 |
| First start with new trainer | ||||
|  No | 1,063,968 (89) | 1,938 (86) | Referent | — |
|  Yes | 132,138 (11) | 318 (14) | 1.321 (1.173–1.488) | < 0.001 |
| Training with first trainer | ||||
|  No | 739,211 (62) | 1,523 (68) | 1.284 (1.176–1.403) | < 0.001 |
|  Yes | 456,895 (38) | 733 (32) | Referent | — |
| Time with same trainer (mo) | 9.057 (9.041 to 9.074) | 7.647 (7.297 to 7.998) | 0.981 (0.976–0.986) | < 0.001 |
| No. of days between race starts since change of trainer | 4.162 (4.068 to 4.255) | 1.932 (−0.346 to 4.209) | 0.999 (0.998–1.000) | 0.043 |
| First start with new jockey | ||||
|  No | 518,291 (43) | 946 (42) | Referent | — |
|  Yes | 677,815 (57) | 1,310 (58) | 1.059 (0.974–1.151) | 0.180 |
| Time with same jockey (mo) | 1.050 (1.046 to 1.055) | 0.930 (0.841 to 1.018) | 0.978 (0.960–0.997) | 0.021 |
| Racetrack surface | ||||
|  Synthetic | 131,402 (11) | 150 (7) | Referent | — |
|  Off-dirt* | 162,002 (14) | 353 (16) | 1.909 (1.577–2.311) | < 0.001 |
|  Dirt | 739,765 (62) | 1,500 (66) | 1.776 (1.502–2.101) | < 0.001 |
|  Turf | 162,937 (14) | 253 (11) | 1.360 (1.111–1.665) | 0.003 |
There were 1,198,362 official race starts by 101,256 horses that had been racing ≥ 6 months; 2,256 starts resulted in a fatal injury (cases), and all other starts were considered controls. Winning percentage was calculated as the number of wins/number of race starts. See Table 1 for remainder of key.
The final multivariable logistic regression model for all horses included 17 risk factors (Table 3). Horse-related risk factors included sex, age at first race start, whether the race was the horse's first race start, finish percentage (finish position/field size; eg, a horse that finishes fourth in a field of 8 would have a finish percentage of 50%, whereas a horse that finishes eighth in a field of 8 would have a finish percentage of 100%) in previous race, whether the horse has been placed by an association or regulatory veterinarian on the vet list (on vet list), number of previous injuries, number of previous scratches, and odds rank in the race (a horse with an odds rank of 1 is the horse favored to win the race). The risk of a fatal injury was increased for stallions, compared with that for mares and geldings. The risk of a fatal injury also increased if the horse was on the vet list and as its number of previous injuries and scratches increased. Protective horse-related risk factors (factors that decreased the risk of fatal injury) included first race start and odds rank in race (risk decreased as horses become less favored to win). The risk of fatal injury also decreased as the finish percentage in the previous race increased. Racetrack-related risk factors included country where the race took place, racetrack surface, race distance, and race purse. The risk of a fatal injury was increased for horses that raced on racetracks in the United States, compared with racetracks in Canada, and on tracks with a turf, dirt, or off-dirt (dirt in any nonfast [sloppy or muddy] condition) surface, compared with tracks with a synthetic surface. The risk of a fatal injury decreased as the race distance and purse increased. Trainer-related risk factors included time with the same trainer and number of days between races after a change in trainer. Both of those factors were protective; the risk of fatal injury decreased as the number of months the horse was trained by the same trainer increased and as the number of days between races after a change in trainer increased. The only jockey-related risk factor included in the final model was first start with a new jockey; the risk of a fatal injury for a horse in its first race with a new jockey was greater than that for a horse in a race in which it was ridden by a jockey who had ridden it in previous races.
Final multivariable model for the horses of Table 1.
| Risk factor | OR (95% CI) | P value |
|---|---|---|
| Intercept | 0.0024 (0.0018–0.0032) | < 0.001 |
| Age at first race start (per y) | 1.076 (1.049–1.104) | < 0.001 |
| Sex | ||
|  Mare or gelding | Referent | — |
|  Stallion | 1.430 (1.305–1.566) | < 0.001 |
| Country | ||
|  Canada | Referent | — |
|  United States | 1.376 (1.185–1.598) | < 0.001 |
| On vet list | ||
|  No | Referent | — |
|  Yes | 1.766 (1.292–2.413) | < 0.001 |
| First race start | ||
|  No | Referent | — |
|  Yes | 0.633 (0.529–0.743) | < 0.001 |
| Finish percentage in previous race (per 1%) | 0.993 (0.991–0.995) | < 0.001 |
| No. of previous injuries (per injury) | 1.550 (1.332–1.804) | < 0.001 |
| No. of previous scratches (per scratch) | 1.029 (1.013–1.046) | < 0.001 |
| Odds rank in race (per rank) | 0.966 (0.953–0.980) | < 0.001 |
| Low purse race (≤ $7,500) | ||
|  No | Referent | — |
|  Yes | 0.830 (0.752–0.916) | < 0.001 |
| Purse (per $1,000) | 0.998 (0.997–1.000) | 0.022 |
| Change in purse since previous race | ||
|  No change (between −$1,000 and +$1,000) | Referent | — |
|  Moderate decrease ($1,000–$15, 000 decrease) | 1.135 (1.042–1.236) | 0.004 |
|  Moderate increase ($1,000–$15,000 increase) | 0.941 (0.856–1.035) | 0.212 |
|  Large decrease (> $15,000 decrease) | 1.057 (0.912–1.225) | 0.464 |
|  Large increase (> $15,000 increase) | 1.034 (0.861–1.242) | 0.717 |
| Race distance (per furlong) | 0.916 (0.892–0.940) | < 0.001 |
| Time with same trainer (per mo) | 0.983 (0.979–0.988) | < 0.001 |
| No. of days between race starts since change of trainer (per d) | 0.999 (0.998–1.000) | 0.014 |
| First start with new jockey | ||
|  No | Referent | — |
|  Yes | 1.076 (1.003–1.154) | 0.042 |
| Racetrack surface | ||
|  Synthetic | Referent | — |
|  Off-dirt* | 1.582 (1.367–1.830) | < 0.001 |
|  Dirt | 1.524 (1.345–1.726) | < 0.001 |
|  Turf | 1.295 (1.110–1.512) | 0.001 |
See Table I for key.
The final multivariable logistic regression model for only those horses that had been competing in flat racing for ≥ 6 months also included 17 risk factors (Table 4). Risk factors included in both final multivariable models (ie, the models for all horses and only horses that had been racing for ≥ 6 months) included age at first race start, sex, country where race took place, on vet list, finish percentage in previous race, number of previous injuries, number of previous scratches, odds rank in race, race purse, race distance, time with same trainer, number of days between races after a change in trainer, and racetrack surface. The respective effects (increase or decrease risk of fatal injuries) of those risk factors were the same for both models. Risk factors included in the final model for horses that had been racing for ≥ 6 months but not in the final model for all horses were number of race starts prior to the race of interest and the winning percentage (number of wins/number of race starts). The risk of a fatal risk injury increased as the winning percentage up to 180 days before the race of interest increased. Interestingly, the risk of a fatal injury decreased as the number of race starts within 30 days before the race of interest increased but increased as the number of race starts within 61 to 180 days before the race of interest increased.
Final multivariable model for the horses of Table 2.
| Risk factor | OR (95% CI) | P value |
|---|---|---|
| Intercept | 0.002 (0.001–0.003) | < 0.001 |
| Age at first race start (per 1 y) | 1.078 (1.041–1. 116) | < 0.001 |
| Sex | ||
|  Mare or gelding | Referent | — |
|  Stallion | 1.544 (1.366–1.744) | < 0.001 |
| Country | ||
|  Canada | Referent | — |
|  United States | 1.304 (1.066–1.594) | 0.010 |
| On vet list | ||
|  No | Referent | — |
|  Yes | 1.526 (1.038–2.243) | 0.032 |
| Field size | 1.060 (1.035–1.085) | < 0.001 |
| Finish percentage in previous race (per l%) | 0.991 (0.987–0.996) | < 0.001 |
| No. of previous injuries | 1.550 (1.323–1.815) | < 0.001 |
| No. of previous scratches | 1.020 (1.002–1.039) | 0.032 |
| Odds rank in race | 0.951 (0.934–0.968) | < 0.001 |
| Low purse race (≤ $7,500) | ||
|  No | Referent | — |
|  Yes | 0.817 (0.725–0.922) | 0.001 |
| Purse (per $1,000) | 0.996 (0.994–0.998) | 0.001 |
| No. of race starts | ||
|  < 30 d prior to race | 0.801 (0.748–0.857) | < 0.001 |
|  61–90 d prior to race | 1.080 (1.023–1.140) | 0.006 |
|  91–180 d prior to race | 1.064 (1.038–1.092) | < 0.001 |
| Race distance (per fur) | 0.895 (0.865–0.925) | < 0.001 |
| Time with same trainer (per mo) | 0.985 (0.980–0.990) | < 0.001 |
| No. of days between race starts since change of trainer (per d) | 0.998 (0.997–0.999) | 0.002 |
| Winning percentage (per l%) | ||
|  31–60 d prior to race | 1.248 (1.084–1.437) | 0.002 |
|  91–180 d prior to race | 1.227 (1.007–1.494) | 0.042 |
| Racetrack surface | ||
|  Synthetic | Referent | — |
|  Off-dirt* | 1.752 (1.442–2.128) | < 0.001 |
|  Dirt | 1.618 (1.363–1.920) | < 0.001 |
|  Turf | 1.322 (1.075–1.627) | 0.008 |
See Table 2 for key.
Discussion
Results of the present study suggested that the risk for fatal injuries (euthanasia or death of a horse within 3 days after sustaining an injury during a race) in Thoroughbred racehorses that competed in flat racing in the United States and Canada during the 5-year period of 2009 to 2013 was increased for horses that were examined by a veterinarian immediately before a race (on vet list) or had sustained a previous injury. Racing a horse that was on the vet list was one of the risk factors with the highest magnitude of effect in the present study. Additionally, the risk that a horse would sustain a fatal injury increased approximately 3% for each scratch (withdrawal from a race). Those findings concur with the results of another study19 in which horses identified as being at risk by regulatory veterinarians during prerace examinations were 5 to 14 times as likely to sustain a musculoskeletal injury or injury of the suspensory or SDFT as horses that were identified as not at risk by regulatory veterinarians during prerace examinations.
In the present study, the risk of a fatal injury was greater for stallions than for mares and geldings. Results of a case-control study25 of deceased Thoroughbred racehorses in California indicate that stallions are more likely than mares and geldings to fracture the proximal sesamoid bones of the forelimb. Greater than 80% of the fatal injuries assessed in the present study were fractures. Stallions are also at greater risk of sustaining a catastrophic musculoskeletal injury,3 nonfatal SDFT injury,26 or indeed any form of fatal injury,2 compared with mares and geldings. Interestingly, results of a case-control study27 involving Thoroughbred racehorses in Australia suggest that sex is not significantly associated with serious musculoskeletal injures. However, the investigators of that study27 compared the incidence of serious musculoskeletal injures between males and females; it is possible the results were confounded, or biased, by the inclusion of geldings with stallions in the male category.
Successful racehorses appeared to be at an increased risk of fatal injury in the present study. Races with a purse < $7,500 appeared to be protective against fatal injury, although the risk of fatal injury appeared to decrease as the race purse increased. Races with a low (< $7,500) purse might have been a proxy for the quality of racehorses or lack of competitiveness in those races, which may explain their apparent protective effect against fatal injury. Another indicator that successful and competitive horses are at increased risk of fatal injury is the apparent protective effect of the odds rank; the risk of fatal injury decreased as the odds rank for a horse increased (ie, as the horse became less favored to win). In the case-control study27 of Thoroughbred racehorses in Australia, the risk of serious musculoskeletal injury also increased as the competitiveness or race purse increased.
In the present study, the risk of a fatal injury decreased as the number of race starts for a horse within 30 days prior to the race of interest increased, but increased as the number of race starts for a horse increased within 61 to 90 and 91 to 180 days before the race of interest. This finding suggested that the risk of fatal injury increases for horses with a > 1-month break between races, and is consistent with results of another study19 in which the risk of a horse sustaining a nonfatal SDFT injury increased 8-fold after a > 2-month break in racing. Results of another study28 also indicate that the risk of an SDFT or suspensory apparatus injury was greater for horses with no race starts than for horses with multiple race starts. Investigators of a case-control study17 of Thoroughbred racehorses in California reported that the risk of a fatal skeletal injury was increased as the cumulative exercise and race distance during the 2 months prior to the race of interest increased. The number of workouts, or exercise sessions, for racehorses is positively associated with the risk of proximal sesamoid bone fractures.25 The EID did not include information regarding the extent of training or exercise for individual horses evaluated in the present study, but the number of race starts during certain periods (eg, 61 to 90 days and 91 to 180 days prior to the race of interest) could be used as a proxy measure for cumulative exercise, in which case our findings are similar to those of the other studies17,25 in which the risk of injury increased as cumulative exercise increased. Another interesting finding of the present study was that the risk of fatal injury decreases as the time that a horse is trained by the same trainer increases.
Racetrack surface was also associated with the risk of fatal injuries in the Thoroughbred racehorses of the present study. Compared with racetracks with synthetic surfaces, the risk of fatal injury was greatest for racetracks with off-dirt and dirt surfaces followed by racetracks with turf surfaces. That finding was consistent with the results of another study29 in which the race-associated fatality rate for racehorses in California was greatest for dirt tracks followed by turf and then synthetic tracks. In that study,29 the race-associated fatality rate for racetracks with dirt surfaces was significantly greater than that for racetracks with synthetic surfaces but did not differ significantly from that for racetracks with turf surfaces, and the fatality rate for racetracks with turf surfaces did not differ significantly from that for racetracks with synthetic surfaces. The reason we were able to identify a statistical difference between synthetic and turf surfaces in the present study is likely a reflection of the large cohort available for analysis, which provided substantially greater power than the smaller study population in that other study.29 In the present study, as in a previous study,1 the risk of fatal injury on racetracks with dirt surfaces was significantly greater than that for racetracks with turf surfaces but did not differ significantly from that for racetracks with off-dirt surfaces.
The risk of fatal injury decreased as race distance increased and field size decreased in the present study. Results of a case-control study10 of fatal fractures in the distal portion of the limbs of Thoroughbred racehorses in the United Kingdom also indicate a positive association between the risk of fatal injury and field size; however, unlike the present study, there was a positive association between risk of fatal injury and race distance. The reason for the discrepancy in the nature of the association between risk of fatal injury and race distance between that study10 and the present study is most likely associated with differences in distance ranges for flat races in the United Kingdom, compared with the distance ranges for flat races in the United States and Canada. Thoroughbred flat races in the United States and Canada tend to be shorter than those in the United Kingdom. Consequently, races in the United States and Canada are run at a faster pace than races in the United Kingdom, and that fast pace likely contributed to the negative association between risk of fatal injury and race distance observed in this study.
Investigators of multiple studies10,15,17,19,27,28 have reported a positive association between age and injury risk in Thoroughbred racehorses. Although age was eligible for inclusion in the multivariable models of the present study, it was not maintained in the final models, likely because of multicollinearity. Age was highly correlated with age at first race start, first race start, and number of previous injuries; however, correlation does not imply causation and inclusion of age in the models did not substantially improve the information gain.
The incidence rate of fatal injuries in Thoroughbred racehorses competing in flat racing at 89 tracks in the United States and Canada during the 5-year period from 2009 to 2013 was 1.9/1,000 race starts. That incidence rate was fairly consistent with the incidence rates for fatal musculoskeletal injuries in Thoroughbred racehorses reported in other US studies,17,12,30,31 (1.2 to 1.96 fatal or catastrophic injuries/1,000 race starts) and a South African study32 (1.4 catastrophic injuries/1,000 race starts). The incidence rate of fatal and nonfatal injuries in Thoroughbred racehorses was 3.18/1,000 race starts in a study33 that involved 27 racetracks in Minnesota from 1987 to 1992; unfortunately, the number of fatal injuries was not reported separately from the number of nonfatal injuries in that study. Regardless, the incidence rates of fatal injuries in Thoroughbred racehorses in the United States appear to be much higher than those for racehorses in the United Kingdom13,34,35 (0.38 to 0.9 fatal injuries/1,000 race starts) and Australia36,37 (0.3 to 0.44 fatal injuries/1,000 race starts). Moreover, the incidence rates within each jurisdiction, or country, have remained fairly constant since the 1980s.
To our knowledge, the present study was the first to extensively evaluate data obtained from the EID. The data evaluated in the present study represented 93.9% of all official Thoroughbred racing events in the United States and Canada during the 5-year observation period. Therefore, we are confident that our results are unbiased effect estimates for the risk factors evaluated because the data for the approximately 6% of racing events that were not reported to the EID represent a small source of bias.
It is important to note that we did not make any attempt to differentiate the causes of fatal injury in the present study. Risk factors vary among types of fractures,13,25 and it is likely that some of those risk factors were not identified in the present study. The types of injuries sustained and the reason for euthanasia have been accurately reported to the EID only recently. Thus, future analyses will be able to use more specific outcome variables to identify risk factors associated with the most common reasons for euthanasia of Thoroughbred racehorses following race-induced injuries.
We concede that statistical significance does not necessarily translate to clinical significance. Although we identified several risk factors that were significantly associated with fatal injuries in Thoroughbred horses competing in flat racing, it is important to point out that the vast majority (1,887,911/1,891,483 [99.8%]) of race starts evaluated in the present study did not result in a fatal injury. Because of the extremely large number of race starts evaluated and the resulting high statistical power of this study, the magnitude of effect for some of the risk factors was very small, and this should be considered during discussions regarding the selection and implementation of measures expected to have the greatest effect on minimizing the number of horses that sustain fatal injuries during flat races in the United States and Canada. Nevertheless, the results of the present study can be used as a guideline for the identification of racehorses at high risk of sustaining a fatal injury during a race.
Acknowledgments
Supported by an Industry Partnership Doctorate of Philosophy Program provided by The Jockey Club and the University of Glasgow.
Presented as an abstract at the 20th International Conference of Racing Analysts and Veterinarians, Republic of Mauritius, September 2014.
ABBREVIATIONS
| AIC | Akaike Information Criterion |
| EID | Equine Injury Database |
| SDFT | Superficial digital flexor tendon |
References
1. Mohammed HO, Hill T, Lowe J. Risk factors associated with injuries in Thoroughbred horses. Equine Vet J 1991; 23:445–448.
2. Estberg L, Stover SM, Gardner IA, et al. Fatal musculoskeletal injuries incurred during racing and training in Thoroughbreds. J Am Vet Med Assoc 1996; 208:92–96.
3. Estberg L, Stover SM, Gardner IA, et al. Relationship between race start characteristics and risk of catastrophic injury in Thoroughbreds: 78 cases (1992). J Am Vet Med Assoc 1998; 212:544–549.
4. Henley WE, Rogers K, Harkins L, et al. A comparison of survival models for assessing risk of racehorse fatality. Prev Vet Med 2006; 74:3–20.
5. Lyle CH, Blissitt KJ, Kennedy RN, et al. Risk factors for race-associated sudden death in Thoroughbred racehorses in the UK (2000–2007). Equine Vet J 2012; 44:459–465.
6. Williams RB, Harkins LS, Hammond CJ, et al. Racehorse injuries, clinical problems and fatalities recorded on British racecourses from flat racing and National Hunt racing during 1996, 1997 and 1998. Equine Vet J 2001; 33:478–486.
7. Boden LA, Anderson GA, Charles JA, et al. Risk factors for Thoroughbred racehorse fatality in flat starts in Victoria, Australia (1989–2004). Equine Vet J 2007; 39:430–437.
8. Hernandez Ja, Hawkins DL, Scollay MC. Race-start characteristics and risk of catastrophic musculoskeletal injury in Thoroughbred racehorses. J Am Vet Med Assoc 2001; 218:83–86.
9. Estberg L, Gardner IA, Stover SM, et al. A case-crossover study of intensive racing and training schedules and risk of catastrophic musculoskeletal injury and lay-up in California Thoroughbred racehorses. Prev Vet Med 1998; 33:159–170.
10. Parkin TD, Clegg PD, French NP, et al. Race- and course-level risk factors for fatal distal limb fracture in racing Thoroughbreds. Equine Vet J 2004; 36:521–526.
11. Parkin TD, Clegg PD, French NP, et al. Risk factors for fatal lateral condylar fracture of the third metacarpus/metatarsus in UK racing. Equine Vet J 2005; 37:192–199.
12. Peloso JG, Mundy GD, Cohen ND. Prevalence of, and factors associated with, musculoskeletal racing injuries of Thoroughbreds. J Am Vet Med Assoc 1994; 204:620–626.
13. Parkin TD, Clegg PD, French NP, et al. Horse-level risk factors for fatal distal limb fracture in racing Thoroughbreds in the UK. Equine Vet J 2004; 36:513–519.
14. Parkin TD, Clegg PD, French NP, et al. Analysis of horse race videos to identify intra-race risk factors for fatal distal limb fracture. Prev Vet Med 2006; 74:44–55.
15. Cohen ND, Berry SM, Peloso JG, et al. Association of high-speed exercise with racing injury in Thoroughbreds. J Am Vet Med Assoc 2000; 216:1273–1278.
16. Estberg L, Gardner IA, Stover SM, et al. Cumulative racing-speed exercise distance cluster as a risk factor for fatal musculoskeletal injury in Thoroughbred racehorses in California. Prev Vet Med 1995; 24:253–263.
17. Estberg L, Stover SM, Gardner IA, et al. High-speed exercise history and catastrophic racing fracture in Thoroughbreds. J Am Vet Med Assoc 1996; 57:1549–1555.
18. Hernandez JA, Scollay MC, Hawkins DL, et al. Evaluation of horseshoe characteristics and high-speed exercise history as possible risk factors for catastrophic musculoskeletal injury in Thoroughbred racehorses. Am J Vet Res 2005; 66:1314–1320.
19. Cohen ND, Peloso JG, Mundy GD, et al. Racing-related factors and results of prerace physical inspection and their association with musculoskeletal injuries incurred in Thoroughbreds during races. J Am Vet Med Assoc 1997; 211:454–463.
20. Kane AJ, Stover SM, Gardner IA, et al. Horseshoe characteristics as possible risk factors for fatal musculoskeletal injury of Thoroughbred racehorses. Am J Vet Res 1996; 57:1147–1152.
21. Kane AJ, Stover SM, Gardner IA, et al. Hoof size, shape, and balance as possible risk factors for catastrophic musculoskeletal injury of Thoroughbred racehorses. Am J Vet Res 1998; 59:1545–1552.
22. The Jockey Club. Available at www.jockeyclub.com. Accessed Jun 30, 2016.
23. Dohoo I, Martin W, Stryhn H. Veterinary epidemiologic research. Charlottetown, PEI, Canada: AVC Inc, 2003;321–322.
24. Bozdogan H. Model selection and Akiake's Information Criterion (AIC): the general theory and its analytical expressions. Psychometrika 1987; 52:345–370.
25. Anthenill LA, Stover SM, Gardner IA, et al. Risk factors for proximal sesamoid bone fractures associated with exercise history and horseshoe characteristics in Thoroughbred racehorses. Am J Vet Res 2007; 68:760–771.
26. Takahashi T, Kasashima Y, Ueno Y. Association between race history and risk of superficial digital flexor tendon injury in Thoroughbred racehorses. J Am Vet Med Assoc 2004; 225:90–93.
27. Bailey CJ, Reid SW, Hodgson DR, et al. Risk factors associated with musculoskeletal injuries in Australian Thoroughbred racehorses. Prev Vet Med 1997; 32:47–55.
28. Perkins NR, Reid SW, Morris RS. Risk factors for injury to the superficial digital flexor tendon ans suspensory apparatus in Thoroughbred racehorses in New Zealand. N Z Vet J 2005; 53:184–192.
29. Arthur RM. Comparison of racing fatality rates on dirt, synthetic, and turf at four California racetracks, in Proceedings. 56th Annu Conv Am Assoc Equine Pract 2010;405–408.
30. Hill T, Carmichael D, Maylin G, et al. Track condition and racing injuries in Thoroughbred horses. Cornell Vet 1986; 7:361–379.
31. Dibbern DK. Use of the post parade examination as the primary method of minimising catastrophic injury, in Proceedings. 11th Int Conf Racing Anal Vet 1996;275–277.
32. MacDonald DM, Toms TS. Survey into the incidence of race track injuries in the Transvaal racing district of southern Africa, in Proceedings. 9th Int Conf Racing Anal Vet 1994; 262–264.
33. Wilson JH, Robinson RA. Risk factors for equine racing injuries. Compend Contin Educ Pract Vet 1996; 18:682–690.
34. McKee SL. An update on racing fatalities in the UK. Equine Vet Ed 1995; 7:202–204.
35. Wood JL, Harkins L, Rogers K. A retrospective study of factors associated with racehorse fatality on British racecourses from 1990–1999, in Proceedings. 13th Int Conf Racing Anal Vet 2000;274–277.
36. Bourke JM. Wastage in Thoroughbreds, in Proceedings. Annu Sem Equine Br N Z Vet Assoc 1995;107–119.
37. Boden LA, Anderson GA, Charles JA, et al. Risk of fatality and causes of death of Thoroughbred horses associated with racing in Victoria, Australia: 1989–2004. Equine Vet J 2006; 38:312–318.
