In recent years, there has been an increased focus on head injuries (HIs) occurring in sports and recreation (SR) activities, largely driven by injuries sustained by professional athletes. As a result of elite athletes sustaining multiple HIs, the attention to the detection, treatment, and prevention of subsequent HIs occurring in SR has grown. Head injuries occurring in SR are known to affect memory,1–4 planning skills,2 balance,5 and reaction time.3 The cumulative effects of SR-HIs have also been investigated,1,3,6–15 however, with varying results. Research has also reported that repeated SR-HIs are associated with a number of long-term outcomes, including electrophysiological changes,16 prolonged recovery from subsequent HIs,3,6 depression,17 mild cognitive impairment,18 and chronic traumatic encephalopathy19; however, the magnitude of long-term effects from multiple SR-HIs is still unclear.
Head injuries are common in SR activities, and the odds of sustaining a subsequent HI is 1.4 to 11.1 times greater if a previous HI exists.6,20–29 Furthermore, 1% to 29% of HIs occurring in a single athletic season are subsequent HIs.6,23,30,31 To date, published findings have focused primarily on the elite athletes in contact sports; however, this fails to account for the majority of participants within the general population who participate in SR activities. The primary objective of this study was to determine the risk of presenting to an emergency department (ED) with a subsequent SR-related HI, using a large Canadian population-based sample. The second objective was to examine the duration between these subsequent SR-HIs when presenting to EDs.
This retrospective cross-sectional study was conducted in Edmonton and surrounding area, Alberta, Canada, between April 1, 1997, and March 31, 2008.
The population of the Edmonton Census Metropolitan Area was 862 597 in 199632 and 1 034 945 in 2006.33 There are 5 hospitals (2 tertiary care and 3 community care) with EDs within Edmonton and the adjoining community of St Albert. In this region, all ED physicians are full time and have either extensive experience and/or advanced training.
Instruments and Procedures
Data from ED records were accessed through the Ambulatory Care Classification System (ACCS), a provincial administrative health database that contains clinical data for all persons presenting to EDs. Demographic, health service, and clinical data, including 10 diagnosis and procedure fields, are included in the ACCS. Diagnoses and procedures were coded by ICD-9-CM 34 (before April 1, 2002) and ICD-10-CA 35 (after April 1, 2002). All ED charts are coded by trained nosologists from the physician diagnosis, ED chart, nursing notes, and consultation reports.
Computed tomography was performed on some patients; however, clinical decision rules for minor HI have been developed and implemented in this center. Moreover, use of advanced imaging is not reported in the ACCS. Additionally, advanced techniques, such as magnetic resonance imaging, are often needed to detect minor HI damage sustained through SR. Full neuropsychological testing is not performed in an ED or as an outpatient, except in selected cases.
The ACCS also contains a unique SR-coding system of more than 120 activities. To identify additional SR injury records that may not have been coded as such, all 10 diagnosis fields were searched for SR-related activities using ICD-9 and ICD-10 codes (Table 1). From the original coding in the ACCS, SR activities were categorized into groups of similar activities. Twelve groups contained more than 100 HIs and remained as unique activities, whereas all other activities formed the 13th group other sports and recreation (Table 1). Data were obtained for ICD-9 and ICD-10 codes pertaining to sprains, strains, dislocations, fractures, and acute injuries to internal organs for individuals younger than 36 years of age. Head injury was defined as any of the 10 diagnosis fields containing an ICD-9-CM or ICD-10-CA diagnosis of skull fracture (800, 801, 803, 804, S02-S02.1, S02.7-S02.9), intracranial injury (including concussion) (850-854, S06), crushing injury of the head (925, S07), and fractures involving the head and neck (T02-T02.01).
Records were excluded from the analysis if the record (1) was not for a resident of the Capital region, (2) was not identified as an SR injury, (3) had an ICD-9 diagnosis code of V00 to V89 (supplementary classification of factors influencing health status and contact with health services), (4) had an ICD-10 diagnosis code of Z00 to Z99 (factors influencing health status and contact with health services), (5) indicated a transfer to/between health care facilities, (6) was a repeat visit to an ED within a 14-day period from any injury presentation, or (7) was the only observation for the unique identifier associated with an ED record.
To identify the duration between HIs, Kaplan–Meier time-to-event curves (survivor functions) were constructed and then compared using the log-rank test.36 Risk factors of subsequent HIs were identified using a Cox proportional hazard model, which provided a hazard ratio for sustaining an SR-HI. The standard errors reported are clustered on the unique identifier.37 To compare time-to-HI and risk-of-HI for those with at least 1 previous SR-related HI to a baseline population, a comparison group of persons whose index (first) ED visit was not an HI were included (time-to-first HI group). The analysis time was recorded as the time from the first ED visit for any SR-related injury meeting the inclusion criteria until either (1) the first/subsequent HI, or (2) the last observed ED record. After calculating the time between events, the first ED record for all unique identifiers was removed from the analysis.
A sensitivity analysis was conducted to assess how eliminating observations occurring within 60 days of a previous ED visit affected the model, in contrast to the 14-day cutoff in the main analysis. A secondary analysis, using logistic regression, provides estimates for the odds of sustaining an HI from specific SR activities. Results were considered significant where P values were less than 0.05. Data were analyzed using Stata 10 (StataCorp LP, College Station, Texas).38
Ethical approval for this study was granted by the University of Alberta's Health Research Ethics Board. Patients were not contacted during this study.
From April 1, 1997, to March 31, 2008, there were 964 172 injury visits (29.8% of all visits) to the 5 EDs, of which 131 210 records (13.6%) were documented in the ACCS as a SR injury. There were 222 464 ED records extracted from the ACCS based on the inclusion criteria, which after the exclusion criteria, contained 9246 ED records for 8958 (97%) unique identifiers for the main analysis and 63 219 ED records for 50 461 (80%) unique identifiers for the secondary analysis (Figure 1).
The age of subjects with multiple visits to an ED because of SR-related injuries ranged from 1 to 35 years. Individuals who sustained subsequent HIs ranged from 4 to 34 years of age. Males accounted for 6724 (72.7%) of the subsequent visits and 737 (76.9%) of the HIs identified. There were 8958 persons seen in EDs for any subsequent SR injury that met the inclusion criteria. After an index presentation for an SR-related injury, there were 746 records (78%) for a first HI, 200 visits (21%) for a second HI, and 13 visits (1.4%) for a third HI. The median time-to-HI decreased as the number of documented HIs increased (Table 2). A test for trend of the survivor functions was highly significant (P < 0.001), indicating a linear decline in the number of days between SR-HIs as the number of SR-HIs increases.
Figure 2 illustrates the significant differences between time-to-HI curves for those at risk of a first, second, and third SR-HI (P < 0.0001). The median time-to-HI for those with a history of 1 SR-HI and 2 SR-HIs was 3179 days (8.7 years) and 1760 days (4.8 years), respectively. The median time-to a first SR-HI for those with no previous SR-HI was approached at 3938 days of follow-up (10.78 years). The probability of sustaining an HI at 1 year of follow-up was 2.5%, 9.0%, and 31.7% for those with 0, 1, and 2 previous SR-HIs, respectively. At 6 years of follow-up, the probability of sustaining an HI was 16.5%, 36.5%, and 70.5% for those with 0, 1, and 2 previous SR-HIs, respectively. Time-to-HI was significantly shorter for males (P = 0.048) and for those younger than 18 years (P < 0.0001).
A global test of the proportional hazards assumption based on Schoenfeld residuals39 indicated that the model did not violate the proportional hazards assumption (P = 0.10). Table 3 displays the results of the Cox regression analysis. Compared to those with no previous SR-HIs, individuals with 1 previous SR-HI were 2.62 times more likely [95% confidence interval (CI), 2.23-3.07] to sustain an HI, and this risk further increased to 5.94 (95% CI, 3.43-10.29) for those persons who sustained 2 previous SR-HIs. Males were 1.28 times more likely (95% CI, 1.09-1.50) to sustain an SR-HI, although age was also a significant risk factor. As seen in Table 3, individuals aged 7 to 13, 14 to 17, and 18 to 22 years were more likely to sustain an SR-HI than older persons (30-35 years).
The secondary analysis, using logistic regression, was performed to obtain estimates for the risk of sustaining an SR-HI from various SR activities when presenting to EDs (Table 4). Head injuries from animal-related activities comprised only 1.0% of the HI seen within this cohort, yet after adjusting for age and sex, it had the highest odds of HI [adjusted odds ratio (aOR) = 3.54; 95% CI, 2.84-4.40]. The sport that most frequently reported HIs was hockey, which had an aOR of 1.98 (95% CI, 1.79-2.20) for sustaining HI. The sport of basketball had the lowest odds of sustaining an HI (aOR = 0.38; 95% CI, 0.31-0.46).
This is the first study to exclusively investigate subsequent SR-related HIs in a population-based sample. Although other studies have almost entirely investigated subsequent HIs occurring in football,6,20,22–25,27,28 soccer,24,25 rugby,30,40 and hockey,21,29,31,41–43 our study identified subsequent HIs from numerous SR activities. Our findings over an 11-year observational period demonstrate an association between decreasing time-to-SR-HI ED presentations with each subsequent SR-HI. Although other confounders, such as the type of sport, the level of participation, medical comorbities, methods of training, season duration, and sex play important roles in determining HI, these findings emphasize the importance of concussion modifiers—factors such as the number of HIs sustained over time and HIs sustained within a close period, which may predict delayed recovery and subsequent HIs—and the graduated return-to-play (RTP) protocol outlined in the Zurich Consensus Statement on Concussion in Sport.44
In a previous population-based study of pediatric HIs, Swaine et al45 reported a similar increase in the odds for sustaining a subsequent HI. Using a comparable HI case definition, they reported a doubling in the odds of a subsequent HI 12 months after an initial HI. Although their study did not exclusively investigate SR-related HIs, it was noted that HIs occurring in recreation were a protective factor at 6 months after index injury but not at 12 months. This protective association that Swaine et al45 observed may be attributed to reduced exposure to SR activities during the first 6 months after injury; however, exposure was not reported in their study. Although our analysis did not compare SR with non-SR events, our findings demonstrate that a history of an SR-HI results in more than a 2.5-fold increase in the odds of sustaining future SR-HIs in a population-based sample.
Through our secondary analysis, we observed that the SR activities with the greatest odds of sustaining an HI were animal-related activities (ie, rodeo, horseback riding), rugby, and Off-Highway Vehicle/All-Terrain Vehicle activities. Although animal-related and OHV/ATV activities are not contact sports, they do expose participants to high-speed impacts and participants may not consistently wear helmets, which are required in sports such as hockey and football. Additionally, the odds of presenting to an ED with a HI related to skiing, snowboarding, or sledding was nearly identical to that of football players—athletes most commonly associated with SR-HI research.
Although most reporting on subsequent SR-HIs has focused on professional,17,18,28,30,46,47 college/university,6,22,26,27,40,48 and high school athletes,22,40,49 we used a population-based approach that included children younger than 7 years. Although children as young as 1 year of age were included in the analysis, these injuries were more due to recreation activities, such as general/unstructured play and playground activities, rather than participation in structured sport. We observed that individuals between 7 and 13 years of age had the greatest risk for sustaining an SR-HI when presenting to an ED with a subsequent SR-related injury, compared with those aged 30 to 35 years. Furthermore, 77% (10 of 13) of those who sustained a third SR-HI were aged 13 to 17 years at the time of ED presentation, yet there is little research on subsequent HIs or the cumulative effects of HIs for these younger age-groups. As the risk of subsequent SR-HIs increases and the duration between SR-HIs declines with successive HIs, clinicians and sport medicine practitioners are urged to provide conservative RTP advice. Given that there are no validated RTP guidelines for younger children,44 modifications of RTP advice for children and adolescents may include increasing the time of asymptomatic rest and graded physical exertion44 and should be individualized.44,50
Hypotheses for the inverse relationship that was observed between the time-to-HI and the number of SR-HIs sustained were explored. Anecdotal evidence suggests that as the number of SR-HIs sustained increases, the duration between subsequent events may decline. Empirical evidence has also shown that individuals with a greater number of SR-HIs experience more symptoms1,10,11,15 and take longer to recover3,6 than those experiencing a first SR-HI. Given that recovery times increase and the time-to-HI decreases as successive SR-HIs occur, there is a cause for concern, specifically in younger populations.
Excluding observations for the same individual within a 14-day and 60-day period were attempts to remove observations that were repeat ED presentations for the same injury and to control for different injuries, which may physically prevent participation in SR (eg, fractures). Given that a recent study observed an average RTP after concussion of 62 days in fourth-tier junior ice-hockey players,43 excluding cases that occurred within a 60-day period seems to validate the cutoff used in the sensitivity analysis (data not shown). However, even after removing these observations in the sensitivity analysis, those with a history of 1 or 2 SR-HIs were still more than 2.5 and 5 times more likely to sustain a subsequent SR-HI, respectively.
Strengths and Limitations
There are limitations to our study that should be considered. First, it is not possible to ascertain whether additional HIs were sustained before or between observed ED visits. Nonetheless, we have shown that as the number of SR-HIs increased, the hazard ratio for sustaining a subsequent SR-HI increased, along with a decrease in the number of days between ED presentations. Second, it was impossible to calculate rates of HI per athletic-exposure or player-game-hours as the time participating in SR activities is not documented in the ACCS or ED records. Therefore, comparisons between studies that measured time at risk of SR injury cannot be directly made. Third, data are limited to ED presentations only; therefore, the number of mild SR-HIs, which did not require immediate medical intervention, may be underestimated because data for individuals who sought treatment from a family physician, or at a sports medicine or walk-in clinic, were not captured in the ACCS database. However, severe SR-HIs would be referred to an ED if they had presented to general practitioner or walk-in-clinic with signs or symptoms of severe brain injury.51 Also, as the ACCS does not contain follow-up data, it was not possible to ascertain recovery time or time until returning to activities. For instance, while the duration between SR-HIs significantly decreased with successive SR-HIs, we cannot make recommendations for RTP given that we did not have clinical data to determine if individuals had fully recovered from a previous HI. Additionally, as the data are limited to the ED data entered into the ACCS database, we were not able to identify other potential risk factors that may contribute to sustaining subsequent HIs, such as the severity of HI (ie, Glasgow Coma Scale or clinical grading scales). Recently, Hamilton et al52 reported that typical epidemiological analyses for recurrent-events bias point estimates away from the null when investigating causal risk factors. Therefore, although we have shown that individuals with a previous HI have an increased risk of subsequent HI, there may be a number of variables, such as an individual's aggressiveness or style of play, that confound this relationship.
Although there are limitations to our study, there are also several strengths. The findings from this study contribute to the emerging evidence of subsequent HIs in SR activities. First, the methodological and analytical approaches used provide sound external validity. Second, unlike other approaches, the use of time-to-event analysis quantifies the duration between SR-HIs in days and not athlete-exposures or player-game-hours. For instance, if a hockey player experiences a concussion and is then unable to participate for a number of weeks or months and then sustains a subsequent concussion in their next ice session, the number of days between these events may be indicative of incomplete recovery. With this consideration, the number of days between HIs may be a better indication of subsequent HIs than athlete-exposures given a history of HI. Third, the use of an administrative database allowed us to retrospectively track time-to-HI for clinically diagnosed HIs over a decade. Although some studies have differentiated the odds of sustaining a subsequent HI between HIs occurring within and outside of the SR activity being investigated,24,25,28 this is rarely done; however, we were able to identify SR-HIs from a variety of activities by using this administrative database. Fourth, selection bias was minimized because the publically funded health care system that operates in Canada, which guarantees universal coverage for essential medical services, would not prohibit individuals with SR-HIs from seeking treatment at an ED.
Individuals with 1 or 2 previous SR-related HIs have an increased risk of sustaining a subsequent SR-HI, compared with those with no previous SR-HI. Additionally, as the number of SR-HIs sustained increases, time between successive SR-HI declines. Clinicians and sport medicine practitioners need to be aware that the duration between subsequent SR-HIs decreases while the risk of subsequent HIs increases after sustaining an HI. High-risk prevention programs specifically targeting those young persons who sustain an initial HI should be developed within both the community and medical settings. To aid in RTP protocols, additional research using time-to-event analysis is recommended. Further investigation is recommended to identify which of SR activities pose the greatest risk of HI to the younger populations and to investigate the long-term effects that multiple HIs may have on children's brain development. Given the added risk of subsequent HI, health promotion programs specifically designed to monitor RTP guidelines and injury prevention after an SR-HI are warranted for younger age-groups.
The authors thank the Alberta Health Services (AHS)-Edmonton Zone (formerly Capital Health) for providing the data for this research, as well as Brian Humeniuk at the AHS for extracting the data from the ACCS.
1. Iverson GL, Gaetz M, Lovell MR, et al. Cumulative effects of concussion in amateur athletes. Brain Inj. 2004;18:433–443
2. Matser EJ, Kessels AG, Lezak MD, et al. Neuropsychological impairment in amateur soccer players. JAMA. 1999;282:971–973
3. Covassin T, Stearne D, Elbin R. Concussion history and postconcussion neurocognitive performance and symptoms in collegiate athletes. J Athl Train. 2008;43:119–124
4. McCrea M, Kelly JP, Randolph C, et al. Standardized assessment of concussion (SAC): on-site mental status evaluation of the athlete. J Head Trauma Rehabil. 1998;13:27–35
5. McCrea M, Guskiewicz KM, Marshall SW, et al. Acute effects and recovery time following concussion in collegiate football players: the NCAA Concussion Study. JAMA. 2003;290:2556–2563
6. Guskiewicz KM, McCrea M, Marshall SW, et al. Cumulative effects associated with recurrent concussion in collegiate football players: the NCAA Concussion Study. JAMA. 2003;290:2549–2555
7. Guskiewicz KM, Marshall SW, Broglio SP, et al. No evidence of impaired neurocognitive performance in collegiate soccer players. Am J Sports Med. 2002;30:157–162
8. Macciocchi SN, Barth JT, Littlefield L, et al. Multiple concussions and neuropsychological functioning in collegiate football players. J Athl Train. 2001;36:303–306
9. Iverson GL, Brooks BL, Lovell MR, et al. No cumulative effects for one or two previous concussions. Br J Sports Med. 2006;40:72–75
10. Collins MW, Field M, Lovell MR, et al. Relationship between postconcussion headache and neuropsychological test performance in high school athletes. Am J Sports Med. 2003;31:168–173
11. Collins MW, Lovell MR, Iverson GL, et al. Cumulative effect of concussion in high school athletes. Neurosurgery. 2002;51:1175–1179 discussion 1180–1181
12. Collins MW, Grindel SH, Lovell MR, et al. Relationship between concussion and neuropsychological performance in college football players. JAMA. 1999;282:964–970
13. Bruce JM, Echemendia RJ. Hisotry of multiple self-reported concussions is not associated with reduced cognitive abilities. Neurosurgery. 2009;64:100–106 discussion 106
14. De Beaumont L, Lassonde M, Leclerc S, et al. Long-term and cumulative effects of sports concussion on motor cortex inhibition. Neurosurgery. 2007;61:329–336 discussion 336–337
15. Gaetz M, Goodman D, Weinberg H. Electorphysiological evidence for the cumulative effects of concussion. Brain Inj. 2000;14:1077–1088
16. De Beaumont L, Brisson B, Lassonde M, et al. Long-term electrophysiological changes in athletes with a history of multiple concussions. Brain Inj. 2007;21:631–644
17. Guskiewicz KM, Marshall SW, Bailed J, et al. Recurrent concussion and risk of depression in retired professional football players. Med Sci Sports Exerc. 2007;39:903–909
18. Guskiewicz KM, Marshall SW, Bailes J, et al. Association between recurrent concussion and late-life cognitive impairment in retired professional football players. Neurosurgery. 2005;57:719–726 discussion 719–726
19. McKee AC, Cantu RC, Nowinski CJ, et al. Chronic traumatic encephalopathy in athletes: progressive tauopathy after repetitive head injury. J Neuropathol Exp Neurol. 2009;68:709–735
20. Gerberich SG, Priest JD, Boen JR, et al. Concussion incidences and severity in secondary school varsity football players. Am J Public Health. 1983;73:1370–1375
21. Gerberich SG, Finke R, Madden M, et al. An epidemiological study of high school ice hockey injuries. Childs Nerv Syst. 1987;3:59–64
22. Zemper ED. Two-year prospective study of relative risk of a second cerebral concussion. Am J Phys Med Rehabil. 2003;82:653–659
23. Guskiewicz KM, Weaver NL, Padua DA, et al. Epidemiology of concussion in collegiate and high school football players. Am J Sports Med. 2000;28:643–650
24. Delaney JS, Lacroix VJ, Gange C, et al. Concussions among university football and soccer players: a pilot study. Clin J Sport Med. 2001;11:234–240
25. Delaney JS, Lacroix VJ, Leclerc S, et al. Concussions among university football and soccer players. Clin J Sport Med. 2002;12:331–338
26. Meeuwisse WH, Sellmer R, Hagel BE. Rates and risks of injury during intercollegiate basketball. Am J Sports Med. 2003;31:379–385
27. Hagel BE, Fick GH, Meeuwisse WH. Injury risk in men's Canada West University football. Am J Epidemiol. 2003;157:825–833
28. Delaney JS, Lacroix VJ, Leclerc S, et al. Concussion during the 1997 Canadian Football League season. Clin J Sport Med. 2000;10:9–14
29. Emery CA, Kang J, Shrier I, et al. Risk of injury associated with body checking among youth ice hockey players. JAMA. 2010;303:2265–2272
30. Kemp SP, Hudson Z, Brooks JH, et al. The epidemiology of head injuries in English professional rugby union. Clin J Sport Med. 2008;18:227–234
31. Echlin PS, Tator CH, Cusimano MD, et al. Return to play after an initial or recurrent concussion in a prospective study of physician-observed junior ice hockey concussions: implications for return to play after a concussion. Neurosurg Focus. 2010;29 E5
34. Puckett CD The Educational Annotation of ICD-9-CM: SoftCover Hospital Version. 19984th ed Reno, NV Channel Publishing Ltd:927–1124
35. Canadian Institute for Health Information. International Classification of Disease and Health Related problems, tenth revision. 2009 Ottawa, Ontario, Canada (ICD-10-CA):765–1002
36. Kleinbaum DG, Klein M Recurrent Event Analysis. Survival Analysis: A self-learning text. 20052nd ed New York, NY Springer
37. Rogers WH. Regression standard errors in clustered samples. Stata Techn Bull. 1993;13:19–23
38. Stata Statistical Software [computer program]. Release 10. 2007 College Station, TX StataCorp LP
39. Schoenfeld D. Chi-squared goodness-of-fit tests for the proportional hazards regression model. Biometrika. 1980;67:145–153
40. Hollis SJ, Stevenson MR, McIntosh AS, et al. Incidence, risk, and protective factors of mild traumatic brain injury in a cohort of Australian nonprofessional male rugby players. Am J Sports Med. 2009;37:2328–2333
41. Emery CA, Meeuwisse WH. Injury rates, risk factors, and mechanisms of injury in minor hockey. Am J Sports Med. 2006;34:1960–1969
42. Tegner Y, Lorentzon R. Concussion among Swedish elite ice hockey players. Br J Sports Med. 1996;30:251–255
43. Echlin PS, Tator CH, Cusimano MD, et al. A prospective study of physician-observed concussions during junior ice hockey: implications for incidence rates. Neurosurg Focus. 2010;29 E4
44. McCrory P, Meeuwisse W, Johnston K, et al. Consensus Statement on Concussion in Sport: the 3rd International Conference on Concussion in Sport held in Zurich, November 2008. Br J Sports Med. 2009;43:i76–i90
45. Swaine BR, Tremblay C, Platt RW, et al. Previous head injury is a risk factor for subsequent head injury in children: a longitudinal cohort study. Pediatrics. 2007;119:749–758
46. Pellman EJ, Powell JW, Viano DC, et al. Concussion in professional football: epidemiological features of game injuries and review of the literature—part 3. Neurosurgery. 2004;54:81–94 discussion 94–96
47. Pellman EJ, Viano DC, Casson IR, et al. Concussion in professional football: repeat injuries—part 4. Neurosurgery. 2004;55:860–873 discussion 873–876
48. Meeuwisse WH, Hagel BE, Mohtadi NG, et al. The distribution of injuries in men's Canada West university football. A 5-year analysis. Am J Sports Med. 2000;28:516–523
49. Collins CL, Micheli LJ, Yard EE, et al. Injuries sustained by high school rugby players in the United States, 2005-2006. Arch Pediatr Adolesc Med. 2008;162:49–54
50. Purcell L. What are the most appropriate return-to-play guidelines for concussed child athletes? Br J Sports Med. 2009;43:i51–i55
51. Sahai VS, Ward MS, Zmijowskyj T, et al. Quantifying the iceberg effect for injury: using comprehensive community health data. Can J Public Health. 2005;96:328–332
52. Hamilton GM, Meeuwisse WH, Emery CA, et al. Past injury as a risk factor: an illustrative example where appearances are deceiving Am J Epidemiol. 2011;173:941–948