Workload and Noncontact Musculoskeletal Injury in Collegiate Swimmers: A Prospective Cohort Study

For this prospective cohort study, we recruited participants at a preseason team meeting of 1 National Collegiate Athletic Association (NCAA) Division III swim team. Athletes were eligible for inclusion if they were at least 18 years old and medically cleared to participate in sport. Volunteers who were non-English speakers and were unable to provide consent were excluded. A sample-size calculation for this study was based on the coach-reported injury incidence from the previous season (27%) and the need for a minimum of 10 injuries for logistic regression.¹²  This calculation suggested a minimum of 37 participants were required. Forty-one student-athletes provided written informed consent to participate during their 2018–2019 season and 37 completed the season (Table 1). This study, which the Drexel University Institutional Review Board approved, was part of a larger study in which intrinsic risk factors for injury were also examined.

The swim coach provided practice attendance, team training, and competition logs throughout the season. At the beginning of the season and during breaks in mandatory training (eg, Thanksgiving break, winter break), participants self-reported their preseason and self-directed workloads, respectively, via the Research Electronic Data Capture Web application (REDCap). Participants who did not respond to any questionnaire received 2 follow-up email reminders. The return rates for preseason and in-season REDCap workload questionnaires were 100% and 95.5%, respectively, resulting in 5 (0.6%) athlete-weeks of imputed data using team averages. Most training took place in a 25-yd (22.5-m) pool. We converted workloads in yards to kilometers. Our preliminary work verified the coach-reported workloads, with student-athletes completing an average of 97% ± 3% (range = 89%–103%) of the workloads indicated on the logs.

We defined injury as any noncontact musculoskeletal pain that resulted from team activities and prevented the swimmer from participating in a competition or at least 50% of 1 practice as prescribed. This definition is similar to that of interfering pain in a previous study⁸  but has a concrete threshold for interference of 50%. We chose this definition based on evidence that high-volume swimmers had high pain thresholds,¹³  and swimmers tended to continue training despite pain.²  As a result, only about 3 in 10 injuries caused time loss in swimmers.⁴  Thus, the 50% threshold included pain that did not result in complete time loss but still substantially interfered with training. Along with the practice attendance and training logs, the swim coach completed weekly injury reports. When the coach reported that a student-athlete had sustained an injury, that individual received a REDCap questionnaire in which he or she verified the injury, selected the mechanism (contact, noncontact, or overuse), described it in detail, and indicated the number of days affected (see Supplemental Figure, available online at http://dx.doi.org/10.4085/1062-6050-0135.21.S1). An injury episode ended when the swimmer returned to full sport participation. At the midpoint and end of the season, participants also described (via REDCap) any injuries they had not previously reported. They returned 93.2% of these questionnaires. Although using electronic medical records is the norm for NCAA injury surveillance,⁴  the college's athletic trainer discouraged querying these records because of the potential for underreporting. For sports with predominantly overuse injury mechanisms, the use of periodic self-report questionnaires has been encouraged.¹⁴ 

We calculated weekly workloads (Monday through Sunday) for each participant based on practice attendance and team training or competition logs. We categorized workload in 4 ways: (1) overall workload throughout the season, (2) acute workload, (3) chronic workload, and (4) ACWR. Overall workload was the number of kilometers swum over the entire season.¹⁵ Acute workload for each week was the number of kilometers swum that week. Chronic workload for each week was the average kilometers swum per week over that week and the preceding 3 weeks. The ACWR for each week was the quotient of the acute workload and chronic workload.¹⁰  During the first 3 weeks of the season, self-reported preseason workloads factored into the chronic workload and ACWR calculations. We calculated all workloads using Excel (version 16; Microsoft Corp).

The Shapiro-Wilk test and visual inspection of histograms were performed to determine the normality of all data. We used mean ± SD to describe normally distributed data and median (interquartile range [IQR]) to describe nonnormally distributed data. We conducted independent-samples t tests to assess differences in descriptive data between injured and uninjured student-athletes except for age and months per year swum (Mann-Whitney U tests) and sex and training groups (Fisher exact test).

For the first hypothesis, we used logistic regression to assess the association between the overall workload throughout the season (continuous independent variable) and injury (binary dependent variable). For the second hypothesis, we pooled the data for each of the other 3 workload categorizations (acute workload, chronic workload, and ACWR) for all athlete-weeks of training. We then dichotomized each workload categorization into high or typical. Because the pooled workload data were not normally distributed, outliers (third quartile + [1.5 × IQR]) signified high workloads and ACWRs. We denoted all other workloads and ACWRs as typical. We calculated 3 logistic regression models (1 for each workload categorization) using generalized estimating equations to assess the association between workload (the dichotomized independent variable) and injury during each athlete-week (the binary dependent variable). The generalized estimating equations accounted for repeated measures. Logistic regression provided odds ratios (ORs) with 95% CIs. Finally, the injury rate for a given range of acute workloads and ACWRs represented the number of injuries that occurred within that range divided by the total number of athlete-weeks in that range. The α level was set at .05. We completed all statistical analyses using SPSS (version 24; IBM Corp).

The 2018–2019 season spanned 21 or 22 weeks, depending on whether the swimmer qualified for the end-of-season championship meet. Competitions began in week 5. Final examinations and winter break interrupted mandatory team practice from weeks 12 to 15. After returning from winter break in week 16, the team went on a trip for 10 days of high-volume training. Before midseason, 4 student-athletes quit the team (reasons unknown), leaving 37 participants. Data from the 37 student-athletes who completed the season yielded a total of 809 athlete-weeks.

Of the 37 participants, 11 (29.7%) sustained 12 injuries. No differences existed in characteristics, training groups, months per year swum, or swimming experience between the uninjured and injured student-athletes (Table 1). Eight injuries were to the shoulder, 3 were to the back or sacroiliac joint, and 1 was to the knee. Participants described the mechanism of 11 injuries as overuse and 1 as overstretching. Ten injuries were incurred in the water; 2 were incurred on dry land. Participation time altered or lost ranged from 1 day to 2 weeks, with a median (IQR) of 2 (10) days. Nine of the 12 injuries affected participation for ≤4 days. One injury occurred in each of weeks 1, 3, 4, 5, and 18. Seven injuries (6 shoulder and 1 back, all self-reported overuse mechanisms) occurred in week 16.

A sample participant's workload throughout the season is shown in Figure 1. Across participants, the median (IQR) overall workload throughout the entire season was 532 (65) km, which corresponded to a weekly workload of 24.2 (3.4) km. No association existed between overall workload and injury (OR = 1.00; 95% CI = 0.99, 1.01; Figure 2). The demarcations for high acute and chronic workloads and ACWR were 37.2 km/wk and 1.56, respectively. Injury was associated with both a high acute workload and a high ACWR but not a high chronic workload (Table 2, Figure 3). The number and percentage of high workload and high ACWR athlete-weeks ranged from 28 (3.5%) to 49 (6.1%; Table 2). Of the 49 high ACWR athlete-weeks and 46 high acute workload athlete-weeks, 32 involved both a high ACWR and a high acute workload. Most high acute workload athlete-weeks (n = 35/46; 76.1%) and high ACWR athlete-weeks (n = 31/49; 63.3%) coincided with the winter training trip (week 16). All 7 of the injuries during week 16 occurred in the presence of a high acute workload and high ACWR. The injury rate was ≤1% for typical acute workloads and ACWRs, 15% for high acute workloads, and 14% for high ACWRs (Figure 4).

The purpose of our study was to prospectively determine the association between workload and noncontact musculoskeletal injury in collegiate swimmers. The data did not support our first hypothesis: we found no association between overall workload and injury. Our second hypothesis was partially supported. We found that a high ACWR was strongly associated with injury, whereas a high chronic workload was not. Contrary to our hypothesis, a high acute workload was also associated with injury.

The lack of association between overall workload and injury was consistent with that reported in a previous study¹⁶  of collegiate swimmers. Most researchers who examined college-aged swimmers along with younger and older swimmers also found no relationship between overall workload and injury.^8,9,16–20  The authors of only 2 studies^3,15  that included college-aged swimmers along with other age groups provided evidence for an association between high overall workloads and injury. The contrasting findings from those 2 studies could be due to differences in age ranges and injury definitions.⁶  The age ranges reported by Sein et al³  and Ristolainen et al¹⁵  were 13 to 25 and 15 to 35 years, respectively. In a recent systematic review, Feijen et al⁶  suggested that adolescent swimmers may be at high risk of load-related injuries because they have not reached full musculoskeletal and psychosocial maturity. Compared with youth swimmers, athletes who choose to continue swimming in college may be better able to withstand the rigors of high workloads (the “healthy athlete effect”).²¹  Meanwhile, 1 investigation²²  of masters swimmers (age >23 years) showed that those with shoulder pain had higher workloads than did those without shoulder pain. In addition, whereas we studied interfering pain anywhere in the body, Sein et al³  studied shoulder pain only and included all severities.

We observed that high ACWR was no more strongly associated with injury than was high acute workload alone. Although earlier research^23,24  supported the ACWR model of Gabbett,¹⁰  our data did not. Injuries tended to occur as often during weeks with high acute workloads as during weeks with high ACWRs. These results are consistent with a recent analysis²⁵  of data from 34 elite footballers in which the authors noted that ACWR was no more strongly associated with injury than was acute workload alone or even acute workload scaled by a random chronic workload. In our study, considerable overlap (>65%) existed between high ACWR weeks and high acute-workload weeks. Our findings could be attributed to the nature of this team's training program, which had small week-to-week fluctuations, except for the training trip in week 16. Another possible explanation for the lower-than-expected association between ACWR and injury could be self-report and recall biases. Two injuries occurred during the first 3 weeks of the season, during which time the participants' self-reported preseason workloads factored into calculations for the ACWR. Overestimation of preseason workloads could have resulted in underestimation of ACWRs.

Previous researchers¹¹  have expressed concerns about an elevated load-related injury risk during winter training trips. We are the first, to our knowledge, to corroborate these concerns with workload surveillance. Seven of the 12 injuries in this study occurred during the winter training trip. Swimming tradition dictates that the purpose of the training trip is to expose the student-athletes to high acute workloads.¹¹  In fact, it is common practice to plan the highest workloads of the season for 6 weeks before the end-of-season competition.²⁶  Based on supercompensation theory, these workloads are believed to facilitate enhanced performance after tapering for the end-of-season competition.¹  Fittingly, 35 (94.6%) of 37 student-athletes encountered a high acute workload (>37.2 km/wk) during the training trip, and 31 (83.8%) student-athletes encountered a high ACWR (>1.56). Five of the 7 injuries during the training trip affected ≤4 training days. However, the potential long-term effects of these injuries cannot be determined from these data. Regardless, these results raise concerns about whether the performance benefits of a traditional training trip outweigh the increased injury odds.

Because our workload data were not normally distributed, we demarcated high workloads and ACWR via outliers. A high ACWR was >1.56, which aligns with the danger zone of ≥1.5 that Gabbett¹⁰  proposed. A high acute workload was >37.2 km/wk; this value is similar to that reported by Sein et al³  (35 km/wk) for an increased risk of supraspinatus tendinopathy. However, their metric corresponded with the average weekly workload (akin to our overall workload) and not acute workload. Workload classifications based on SDs above the mean workload or ACWR have been used previously.²³  Had we used 1 SD above the mean, our demarcations would have been lower: 33.4 km for high acute workloads, 30.5 km for high chronic workloads, and 1.39 for ACWRs. These values would have reduced the associations with injury by classifying more athlete-weeks as high workload or ACWR without a commensurate increase in injuries. Given these data, we recommend adopting outlier-based values for high ACWR and high acute workload. Future authors should validate the demarcations from our study among larger, diverse samples of swimmers. Swim teams with workloads much different from those of the team in this study may require different values. Furthermore, we caution coaches against treating acute workloads >37.2 km or ACWRs >1.56 as values student-athletes should never exceed. Instead, coaches should continually monitor student-athletes' well-being and readiness to determine whether they are prepared for a certain workload.²⁷ 

Our injury definition yielded injury mechanisms, time loss, and injured body regions consistent with epidemiologic data.^4,28  Our methodological choice to rely on coach and student-athlete reports yielded 11 injured participants (29.7% of the sample). Although a student-athlete seeking medical attention did not constitute an injury in our study, we asked participants whether they had done so in our questionnaires. Only 4 of the 11 student-athletes reported seeking medical attention for their injury, which supports our methodological choice of coach- and self-report. Given our injury definition and reporting methods, overuse was the predominant injury mechanism, accounting for all but 1 injury. This finding is consistent with observations in a prior epidemiologic study⁴  of collegiate swimmers, in which overuse was the most common injury mechanism. In addition, only 3 injuries affected participation for >4 days. These outcomes also correspond with epidemiologic data showing that swimming had the lowest rate of severe injuries (based on time loss) across NCAA sports.²⁸  Because swimmers tend to continue training in the presence of pain, however, this low severe injury rate likely underrepresents their prevalence of pain.^2,3  Finally, also in keeping with epidemiology data, the injured body regions we observed (shoulder, trunk, and knee) align with those in a previous epidemiologic report on collegiate swimmers.⁴  Most of the swimming literature has been focused on the shoulder. Because nearly two-thirds of collegiate swimming injuries affect body regions other than the shoulder,⁴  we recommend future investigations of injuries to all body regions and not just the shoulder.

The small number of participants and injuries in this study meant that the 95% CIs surrounding the ORs for both acute workload (95% CI = 8.2, 89.8) and ACWR (95% CI = 7.7, 81.4) were wide. Based on these wide CIs, readers should interpret the magnitude of the associations we noted with caution. To improve precision, future studies would benefit from multiteam samples or data from 1 team over several seasons. Furthermore, in addition to swimming, participants engaged in strength training. However, we did not assess the possible influence of strength training on injury. Moreover, it was not feasible to measure the student-athletes' internal workload (ie, heart rate or rating of perceived exertion), which represents their physiological or psychological response to imposed external workloads (distance swum).²⁷  With that said, researchers²⁹  studying swimmers have shown a strong correlation between rating of perceived exertion and distance swum (r = 0.71). Nevertheless, the relationship between injury and combined internal and external workload in swimmers remains unknown. In addition, in this study, external swimming workloads were coach- and self-reported. In accordance with our preliminary work verifying coach-reported workloads, authors³⁰  found that student-athletes complied with coach-prescribed training volume at a high rate. A low percentage of missing data further strengthened our workload data. Thus, although our data were limited to coach- and self-reported external swimming workloads, high compliance and response rates bolstered the data fidelity. Finally, our analyses did not incorporate intrinsic risk factors for injury. Given the multifactorial nature of an injury, a comprehensive approach to reducing risk should consider both extrinsic and intrinsic factors.³¹ 

High acute workload and ACWR were both associated with injury in collegiate swimmers; high overall and chronic workloads were not. Therefore, collegiate coaches may prescribe high workloads spread out over the season without increased injury odds. However, they should be cautious in prescribing high acute workloads and high ACWRs. In this sample, acute workloads >37.2 km/wk and ACWRs >1.56 corresponded with increases in injury rates, from ≤1% for typical workloads and ACWRs to 15% for high acute workloads and 14% for high ACWRs. More than half of the injuries occurred during the team's winter training trip. These data raise concerns about the risk of these high acute-workload trips, particularly when they occur after final examinations and winter break, which are often periods of reduced workloads. Moreover, one-third of the observed injuries occurred during the first 5 weeks of the season. Considering the potential inaccuracy of the self-reported preseason workload in this study, swimmers may still benefit from focusing on building their chronic workload in the preseason.

We thank the student-athletes and coaches who made this study possible.