Article Text

Download PDFPDF

Epidemiology of injury and illness in 153 Australian international-level rowers over eight international seasons
  1. Larissa Trease1,2,
  2. Kellie Wilkie3,4,
  3. Greg Lovell5,6,
  4. Michael Drew7,8,
  5. Ivan Hooper9
  1. 1 Orthopaedics ACT, Woden, Australian Capital Territory, Australia
  2. 2 School of Medicine, Healthcare in Remote and Extreme Environments Program, University of Tasmania, Hobart, Tasmania, Australia
  3. 3 BODYSYSTEM Physiotherapy, Hobart, Tasmania, Australia
  4. 4 gROWINGbodies, Canberra, Australian Capital Territory, Australia
  5. 5 Sports Medicine, Australian Institute of Sport, Belconnen, Australian Capital Territory, Australia
  6. 6 UCRISE, University of Canberra Research Institute for Sport and Exercise, Canberra, Australian Capital Territory, Australia
  7. 7 Department of Physiotherapy, Australian Institute of Sport, Canberra, Australian Capital Territory, Australia
  8. 8 Australian Collaboration for Research into Injury in Sport and its Prevention, Federation University Australia, Ballarat, Victoria, Australia
  9. 9 Queensland Sports Medicine Centre, Brisbane, Queensland, Australia
  1. Correspondence to Dr Larissa Trease, Orthopaedics ACT, Woden, ACT 2600, Australia; dr.larissa.trease{at}


Aim To report the epidemiology of injury and illness in elite rowers over eight seasons (two Olympiads).

Methods All athletes selected to the Australian Rowing Team between 2009 and 2016 were monitored prospectively under surveillance for injury and illness. The incidence and burden of injury and illness were calculated per 1000 athlete days (ADs). The body area, mechanism and type of all injuries were recorded and followed until the resumption of full training. We used interrupted time series analyses to examine the association between fixed and dynamic ergometer testing on rowers’ injury rates. Time lost from illness was also recorded.

Results All 153 rowers selected over eight seasons were observed for 48 611 AD. 270 injuries occurred with an incidence of 4.1–6.4 injuries per 1000 AD. Training days lost totalled 4522 (9.2% AD). The most frequent area injured was the lumbar region (84 cases, 1.7% AD) but the greatest burden was from chest wall injuries (64 cases, 2.6% AD.) Overuse injuries (n=224, 83%) were more frequent than acute injuries (n=42, 15%). The most common activity at the time of injury was on-water rowing training (n=191, 68). Female rowers were at 1.4 times the relative risk of chest wall injuries than male rowers; they had half the relative risk of lumbar injuries of male rowers. The implementation of a dynamic ergometers testing policy (Concept II on sliders) was positively associated with a lower incidence and burden of low back injury compared with fixed ergometers (Concept II). Illness accounted for the greatest number of case presentations (128, 32.2% cases, 1.2% AD).

Conclusions Chest wall and lumbar injuries caused training time loss. Policy decisions regarding ergometer testing modality were associated with lumbar injury rates. As in many sports, illness burden has been under-recognised in elite Australian rowers.

  • athlete
  • injury
  • rowing
  • epidemiology

Statistics from


Rowing is a physically demanding pursuit requiring many hours of preparation on and off the water for success.1–4 For athletes in crew boats, time training together is essential to master the technical aspects of the sport.5 Previous literature on injury in elite rowers has emanated from cross-sectional studies conducted at events such as World Championships and Olympic Games.1 6–8 These studies have a high risk of survivor bias—injured athletes are not present at major events and these events are only for a short period in training and competition year. Limited prospective surveillance reports also exist with two single season studies in national and international elite rowing programmes.2 9 Due to the variability in coaching and athletes selected, single season studies have lower external validity beyond that season. Australian researchers completed a 10-year retrospective medical record review of rowers in the 1990s.10 Compared with large team sport such as football,11 there is a paucity of prospective multiseason data to guide injury prevention programmes in rowing.

We aimed to report the epidemiology of injuries—and illnesses—in elite rowers over two Olympic cycles (8 years). Our secondary aim was to explore contextual factors related to these health problems. This is the first epidemiological longitudinal study focused on rowing.


Study period and process

In elite Australian Rowing during the London and Rio Olympic training cycles, athletes trained in their home state or territory until the selection regatta for the National Team. Once athletes were selected into Australian Rowing Team (ART) crews, they relocate to train in their crew until the World Championships or Olympic Games. Every athlete selected to the ART between 2009 and 2016 was prospectively followed and all injuries and illnesses resulting in lost training time were recorded and followed until a return to full training (defined as on-water training as prescribed by their coach without medical restrictions.) Every athlete provided consent for the collection of their data as part of their nomination for selection.

Population and exposure

During the study period, all 153 individual athletes selected were followed, for a total of 48 611 athlete days (ADs). Each year the number of monitored athletes varied as determined by the National team’s selection policy. The demographic data of the ART are presented in online supplementary table 1 .

Supplemental material

Exposure was measured by ADs available within each season.12–14 ADs were calculated as the number of days between the first day following the annual selection trials to the last day of the World Championships or Olympic Games regatta multiplied by the number of athletes selected to the team and followed for that season. For example, 20 athletes followed for a season of 150 days would equal 3000 AD.

Data collection

Data were recorded by two research teams (IH 2009–2013; KW/LT 2013–2016) using consistent methodology across the surveillance period. During the period of the study, a Medical Management Policy required all athletes to present promptly to medical staff and report any injury or illness to the Sports Medicine Coordinator (SMC) of the ART. This was a condition of the athlete selection agreement. Medical staff across Australia also completed a report to the SMC whenever an injured or unwell athlete was assessed.

Over the 8-year observation period, 11 sports medicine doctors and 15 sports physiotherapists contributed to the medical record. Each practitioner had a minimum of 5 years of clinical sports medicine experience. This group met annually to review the data at the conclusion of each season. A standardised report form captured, date of injury or illness and for injuries, the activity engaged in and timeframe in which the injury occurred, diagnosis, injury cause (acute or overuse; new or recurrent), body area and side injured (online supplementary 2, online supplementary table 2). The diagnosis was classified according to V.10.1 of the Orchard Sports Injury Classification System (OSICS-10).15 The injury and illness surveillance data were recorded in a commercial database. Each injury and illness were followed for the entire rehabilitation period, with the date of return to full on-water training determined by the managing sports medicine practitioner and recorded in the study database.

Supplemental material


An injury or illness was defined as any episode that resulted in the athlete being unable to fully participate in on-water rowing training or competition as planned by coaching staff for a period in excess of 24 hours. This definition aligns to ‘sports incapacity’ definitions published after this study was initiated.16 All OSICS-1015 codes were converted to a body region which corresponded to the first letter of the code. The term lumbar ‘injury’ has been used in this paper for consistency with the sporting epidemiological literature; however, our cases include non-specific lumbar region pain as captured by the first letter (body region) of the 2007 OSICS-10 code.15 For surgical codes, OSICS-10 codes were matched to the body region that corresponded to the injury.

An acute injury was defined as any injury with an acute onset as a result of a specific event. While this is similar to the definition of trauma used in subsequent Olympic epidemiology papers,17 we use the terminology that was recorded throughout our study period to accurately represent our methods at the time. The definition of ‘overuse injury’ was a simplified version described by Waldén et al 18 as any pain syndrome of the musculoskeletal system of insidious onset. A recurrent injury was defined as an identical injury (same area, side, type) as a previously recorded, but completely resolved injury, where the athlete had returned to full on-water training.19–21 The sports medicine clinician who diagnosed the injury determined whether the injury was classified as a new or recurrent overuse at the time of assessment, based on the athletes’ previous medical record. During the collection period, injuries were classified as (1) new onset acute, (2) recurrence of an acute injury (after the rower had returned to full training/competition), (3) new onset overuse or (4) recurrence of an overuse injury (after the rower had returned to full training/competition). The activity undertaken by the athlete at the time of injury was recorded. Activities were classified as: racing, on-water rowing training, rowing on an ergometer, weights training, cycling, running and other. Incidence was defined as any new recordable event within the surveillance period. Person-period incidence rates were defined as events per 1000 AD.

Fixed and dynamic ergometer testing

In 2009, Rowing Australia mandated that ergometer testing change from the fixed Concept II ergometer to the dynamic Concept II (on sliders). In 2013, this policy was reversed, and testing was again conducted on a fixed ergometer. The fixed ergometer, the international standard, remained in use for the rest of our study period.

Statistical analysis

All statistical analyses were undertaken in Stata V.15 (Stata 15 IC, StataCorp). To compare the incidence of injury and illness across the surveillance period, Poisson regression with robust errors (to account for overdispersion) was used with season as the fixed effect and ADs as the exposure variable. Incidence rate ratios, adjusted for exposure (person days), for each year were generated with 95% CIs that did not include 1.0 indicating statistical significance.

To assess whether demographic data (height and weight) changed across the surveillance period, a linear mixed model (restricted maximum likelihood) with year as a fixed effect with random intercepts for sex and weight categories was used. To compare the rate of injury between the sexes in each of the main body areas, we calculated a risk ratio (RR) with 95% CIs. We used interrupted time series analysis22 (ITSA) to evaluate policy interventions in a real-world setting.23 24 All ITSA were adjusted for the year within the Olympiad as the cohort of athletes is different across the years, that is, younger athletes are typically selected following an Olympic year and the best athletes are selected in the Olympic year. We used the Strengthening the Reporting of Observational Studiesin Epidemiology cohort checklist as a reporting guideline.25


Population and exposure

During the study period, all 153 individual athletes selected were followed, for a total of 48 611 ADs. The mean±SD season length was 142±32 days with an average surveillance period of 6076±1894 ADs/season. Each year the number of monitored athletes varied as determined by the national team’s selection policy.

The cohort included 61 athletes who were only selected to 1 team (40.0%), 19% were selected to 2 teams (29 athletes) and 13.7% to 3 teams (21 athletes). The number of athletes selected to 4 and 5 teams in the study period was 16 in each group (10.5%) and 10 athletes were selected to 6 or more teams (6.6%) in the 8-year period (online supplementary table 3). The demographic data of the ART is presented in online supplementary table 1. There was no statistically significant difference in the height or weight of the team members in any year studied (online supplementary table 1).

Supplemental material

Injuries and illness

A total of 9% of the available ADs were lost to injury and illness over two Olympiads with an incidence of 4.1–6.4 injuries and 0.9–9.4 illnesses per 1000 AD. We recorded 270 injuries and 128 illnesses. The number of injuries per athlete ranged from 0 to 12 with illnesses between 0 and 6 per athlete across the surveillance period. The highest incidence of injuries occurred in 2009 (table 1: 6.4 injuries/1000 AD); there were no statistically significant differences between years of our study. The highest incidence of illness occurred in 2016—a 10-fold increase over preceding years (table 1: 2016 9.4/1000 AD, IRRadj=10.5, 95% CI 1.3 to 82.2).

Table 1

Surveillance period and injury/illness incidence

The number of injuries of each body region, and ADs lost are presented in table 2. The lumbar region accounted for the greatest number of injuries (n=84, 21.1%) followed by chest wall (n=64, 16.1%), forearm (n=20, 5.0%) and knee (n=17, 4.3%). Chest wall pain resulted in a disproportionally large total number of ADs lost, equating to 2.6% of all ADs in the 8-year period. Lumbar injuries resulted in the loss of 1.7% of ADs followed by illnesses (1.2%). Illness represented 32% of all causes of lost training time (table 2).

Table 2

Injuries by body region, missed athlete days

In 10 athletes, injury or illness prohibited ongoing rowing during the season. The causes were: rib stress fractures (n=4), lumbar disc injury (n=1), wrist intersection syndrome (n=1), knee bursitis (n=1), atrial fibrillation (n=1), hip surgery (n=1) and road cycling trauma (n=1). The training time lost in these cases was counted as the number of days remaining in the campaign until the World Championships or Olympic Games in that year.

Sex comparison of injury and illness experience

The comparative epidemiology of the injury and illness of the different sexes is presented in figures 1 and 2 and table 3. There was a significant difference in the risk of chest wall injuries in women (RR 1.4, 95% CI 1.2 to 1.7) and women were 50% less likely to develop low back pain (RR 0.5, 95% CI 0.4 to 0.6) than men. There was no significant difference in return to sport timelines between men and women for these injuries. There was no difference in the rate of illness between female and male athletes (RR 1.1, 95% CI 1.0 to 1.3).

Figure 1

Comparative incidence and burden (athlete availability) for top six diagnoses (injured body area or illness) in male and female rowers over the 8-year surveillance period. AD, athlete days.

Table 3

Injuries and illness, athlete days (AD) affected, by sex

Figure 2

Lumbar, chest and forearm injury across the 8-year surveillance period for male and female rowers. Note: missing scatter plots represent zero incidence in the respective year. ADs, athlete days.

Cause of injury and injury activity

New onset overuse injuries were the most frequently reported cause (n=117, 43%) followed by overuse injuries of a recurrent nature (n=107, 40%), new onset acute injuries (n=39, 14%), acute injuries of a recurrent nature (n=3, 1%) and unknown (n=4, 2%). Rowing on-water training sessions represented the activity where injury was most frequently reported 68% (n=191). Injuries attributed to other activities included weight training (n=22, 8%), other (n=20, 7%), road cycling (n=14, 5%), rowing on an ergometer (n=17, 6%), racing (n=12, 4%) and running (n=4, 1%). Some injuries were related to more than one training modality, such as when an athlete had undertaken two different training sessions immediately prior to the onset of symptoms. One-third of acute injuries were attributable to road cycling accidents (n=13/39, 33.3%.) with a mean AD burden of 31 days (±39 SD) per injury.

Association between type of ergometer testing and injuries

We report a significant decrease in lumbar injuries across the London Olympiad, following the introduction of dynamic ergometers in 2009 (figure 3). This representing a decrease of 0.36 injuries per 1000 AD per annum (β=−0.36, 95% CI −0.44 to −0.27; equivalent to approximately two new injuries in an average season length being prevented) and a decrease of 7.3 days per 1000 AD each season (β=−7.3, 95% CI −9.89 to −4.7; equivalent to approximately 44 days gained in an average season). The return to fixed ergometers in 2013 was associated with an increased incidence of lumbar injury by 0.78 injuries per 1000 AD per annum (β=0.78, 95% CI 0.64 to 0.92; equivalent to an increase of approximately five injuries per year in an average season) and an average increase of 7.7 days per 1000 AD increase each season being unavailable due to injury (β=−7.3, 95% CI −9.9 to −4.7; equivalent to approximately 47 extra days lost in an average season length), with the effect of the policy on both incidence (F(3,4)=116.34, p<0.001) and burden (F(3,3)=147.5, p<0.001) being statistically significant. There was no statistically significant difference in rib injuries between the two ergometer testing modality periods (, online supplementary figure 1).

Supplemental material

Figure 3

Interrupted time series reporting the effect of ergometer testing policy on lumbar injury. ITSA, interrupted time series analysis.


Our study, the largest prospective longitudinal epidemiological investigation in Olympic-level rowers highlights three recognised clinical issues and points to one previously underappreciated mechanism of injury. Chest wall pain was the leading cause of interruption (injury burden) to on-water training—an important take home message. Our study underscores previous findings2 9 26–32 that the lumbar region accounted for the greatest number of injuries. Traumatic injury is uncommon in Olympic rowers but we highlight the acute injury risk of road cycling for cross training in this cohort. Our study design allowed us to report illness more rigorously than in many other studies and the incidence of illness was higher than the incidence of lumbar spine or chest wall injuries.

Chest wall pain

Chest wall pain was responsible for the greatest number of lost training days over two Olympiads with almost 20 days lost per injury (±18.5 days). Previous literature has suggested that women are more likely to have interrupted training due to chest wall injuries10 and our data support this finding with women 1.4 times more likely to develop a chest wall injury. There was no difference in the burden of lost training time between the sexes. There was no difference observed in chest wall pain or rib bony injuries between the Olympiad where athletes were tested on dynamic ergometers or fixed ergometers (online supplementary 4, online supplementary figure 1). Chest wall pain is a major challenge for rowers, coaches and sports medicine staff. Further research into the causes of rib stress injuries,33 particularly in female athletes, will assist in developing targeted prevention programmes to reduce the burden of time lost to training from this injury.

Lumbar spine

The lumbar spine accounted for the greatest number of injuries in our rowing cohort (84, 21.1%) and the second highest burden in ADs (1.7%, 10 days per case±14.5). The incidence of lumbar spine injuries was lower in female rowers than male rowers (RR 0.5, 95% CI 0.4 to 0.6). It has been postulated that as female rowers have greater hip and pelvic range of motion than male rowers, their lumbar spines may be subjected to less load in the flexed positions of the rowing stroke.34 35 That hypothesis is not something our study can inform.

Between the London (2012) and Rio (2016) Olympic Games, Rowing Australia changed the ergometer testing modality in the elite rowing programmes. The change was informed by the biomechanical34 36–43 and medical literature2 31 34 44–48 and as a result, coaches and athletes were observed to train in each Olympic cycle on the ergometer type used for testing. The use of dynamic ergometers in the London Olympic cycle (2009–2012) was associated a lower incidence and burden of low back pain.

In prospective cohort studies of rowers, there was an association between both total ergometer training time per month, and individual ergometer training sessions of more than 30 min, and low back pain.2 47 49 50 We did not monitor the time athletes spent training on the ergometer but our clinical impression (LT, KW, IH, GL) was that there was more ergometer-based training and testing completed in the 2009–2012 cycle—where the incidence of back pain was relatively low. If our impression is correct, other factors, such as ergometer type, need to be considered as possibly contributing to lumbar injuries. Vinther and Thornton33 51 52 have suggested a causal link between ergometer type and chest wall pain in rowers. Our findings neither support nor negate a possible causal function, but encourage further research into the injury rates—lumbar region and chest wall—associated with ergometer type.

Acute (traumatic) injuries

While acute (traumatic) injuries are uncommon in rowers, for the small number of new-acute traumatic injuries sustained in our study, one-third were attributable to road cycling accidents with a mean AD burden of 31 days per injury. This is an association only previously reported by our Australian colleagues in a state institute setting53 and may be specific to our training environment, but we encourage coaches and athletes to consider this risk in the planning of cross-training modalities.


Medical illness was the most frequent cause of lost training time (burden) with 128 cases accounting for one-third (32.2%) of all presentations. The mean time lost to each illness was 4.7 days which is comparable to the 6.5 days reported by Budgett.9 In 2016, there was a spike in respiratory illnesses among the rowers during the Australian late autumn and early winter. This was the only year that the team was centralised to train and live in one location, highlighting the risk of illness transmission in athlete group training environments. At the Olympic Games in London and Rio de Janeiro, rowers had a similar rate of illness to competitors in other sports (London n=40, 7.3%; Rio 5%).7 8 Unlike in periods of injury, where an athlete can often cross train to sustain a level of fitness, or indeed increase their volume as found by rowing clinician Fiona Wilson,2 illness often precludes any training. Illness may be undervalued by both clinicians and coaches as a source of lost training time due to the relatively short duration of each episode.

Limitations and conclusions

Our 8-year longitudinal analysis of injury and illness has ecological validity of the dataset but we acknowledge several limitations. The data were collected by 11 doctors and 15 physiotherapists across all states and territories in Australia. To reduce the bias of an individual practitioner’s diagnosis, we limited our analysis to the body part injured (OSICS score first letter), rather than a specific structural diagnosis (eg, we reported lumbar spine as a body region, instead of facet joint or disc-related injury). Previous research has shown this to be reliable among different professional backgrounds.54

No training exposure data (hours or modality) were available for the analysis. To mitigate this limitation, we measured available training days. This affects risk calculations55 and this is an area of future improvement for studies in elite rowing populations. Our data were collected between 2009 and 2016. Since then, many papers have advanced the methods in sport and exercise medicine epidemiology.56 Lastly, we note that rowing policy decisions (eg, ergometer testing) were associated with health outcomes (lumbar back pain). Our findings in that regard are associations—our study design does not allow us to infer causality. Causal investigation requires directed acyclic graphs and a purposeful study design.57

In summary, our 8-year longitudinal study of Australian elite rowers highlights that chest wall and lumbar injuries are the greatest challenge and focus areas for future injury prevention for the rowing athlete and clinician. Cross-training modalities and policy decisions can significantly influence injury rates. Illness is an underappreciated burden in rowers.

What are the findings?

  • Chest wall pain had the greatest injury burden (athlete days lost) and was 1.4 times more common in female than in male rowers.

  • Low back pain was the most common injury (incidence); this was two times more common in male than female rowers.

  • During years when rowers were tested on dynamic ergometers, there was a lower incidence and burden of lumbar injuries than during the years the rowers were tested on fixed ergometers.

  • The most common cause of traumatic injuries in rowers was road cycling accidents while cross training

  • Illness represents over 30% of cases presenting to a rowing medical team—as a category the incidence of illness was higher than any single injury (eg, higher than lumbar or chest wall injuries).

How might it impact on clinical practice in the future?

  • Priority areas for injury prevention in rowing: chest wall injuries (particularly among female rowing athletes) and low back pain (particularly among male rowers).

  • Coaches and athletes should be aware that road cycling for cross training is associated with traumatic injury and consider this in their training prescription.

  • Team clinicians should consider illness prevention programmes in rowers.


The authors would like to thank Rowing Australia and the Australian Institute of Sport for their support throughout the study. They acknowledge and thank the doctors and physiotherapists of the Australian Rowing Team and Australian provider network for their diligence in the care of Australian rowers and the completion of their medical records. They would like to acknowledge Peter Blanch for his assistance in the initial advice on designing the surveillance parameters.



  • Twitter @DrLarissaTrease, @KellieWilkie, @_mickdrew

  • Contributors IH formalised the national medical network for Rowing Australia in 2006 that subsequently collected the data which was compiled by IH/LT/KW. IH was involved in the original design, data collection and approving the manuscript. LT and KW collected data from 2013 to 2016 and were involved in the interpretation of the statistics and were primarily responsible for writing the manuscript. GL was involved in interpretation of the statistics and drafting and approving the manuscript. MD undertook all statistical analyses and was involved in the interpretation of the statistics and drafting and approving the manuscript.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

  • Patient consent for publication Not required.

  • Ethics approval Ethical approval for this project was obtained through the Australian Institute of Sport Ethics Committee (Approval number 20170601).

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data availability statement No data are available. Given the confidential nature of the medical records that comprise this data set, no data are available to be shared.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.