Automated Landing Error Scoring System Performance and the Risk of Bone Stress Injury in Military Trainees Open Access

Participants were first-year cadets at the US Military Academy (USMA). All participants were enrolled in a large cohort study of risk factors for lower extremity injury in 2015 and 2016. Cadets who elected to participate in the parent study signed an informed consent document approved by the Institutional Review Board of the University of North Carolina at Chapel Hill. Cadets were included if they were part of the incoming classes entering the USMA in the summer of 2015 or 2016 and were free of any health conditions precluding the completion of at least 3 successful trials of the LESS jump-landing task.⁹  To attend USMA, cadets must meet Department of Defense Medical Evaluation Review Board accession standards. Any potential participants who did not successfully perform 3 trials of the LESS jump-landing task or complete and sign the informed consent form were excluded. Participants with incomplete datasets were excluded from all analyses using the variable or variables for which data were missing (Figure).

Participants provided demographic and self-reported lower extremity injury history information via a standardized, multiple-choice questionnaire. Sex was reported as female or male. Participants were asked if they had experienced a stress fracture in the 6 months before their arrival at the USMA; this variable was coded as dichotomous (yes or no). Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared (kg/m²) from the height and weight data recorded during each cadet's standardized Army Physical Fitness Test, which was administered the same week as the automated LESS-22 assessment.

We operationally defined the lower extremity as the hip, knee, and ankle joints and the bones of the pelvis, thigh, lower leg, and foot. A BSI was defined as any stress reaction or stress fracture in the bones of the lower extremity, diagnosed by a medical provider (physician, nurse practitioner, physical therapist, physician assistant, or certified athletic trainer), and entered into a participant's electronic medical record (EMR). Traumatic and other types of fractures were excluded. Injury data were obtained from the Cadet Illness and Injury Tracking System (CIITS) and Armed Forces Health Longitudinal Technology Application (AHLTA) EMR system. The CIITS system generates data for each cadet extracted from clinical encounters documented in the AHLTA EMR system. Using standardized methods and criteria,^20,21  we searched the CIITS and AHLTA records for each participant for encounters with a diagnosis of a stress reaction or stress fracture of the lower extremity that occurred during the study period. Because all USMA cadets receive care in a closed health care system that uses the AHLTA EMR system, it is highly likely that all lower extremity BSIs that occurred during the study period were included in the dataset.

Each participant's days at risk were calculated as the total number of calendar days a cadet was on restricted lower extremity activity status due to injury or illness subtracted from 368 days (the length of the first academic year at the USMA). For example, if a cadet missed a total of 5 days due to injury or illness, that cadet had a total of 363 days at risk.

We computed descriptive statistics for all variables. Crude injury incidence rates (IRs) per 1000 person-days (IR = number of incident injuries/number of person-days) and incidence rate ratios (IRRs; exposed IR/unexposed IR) were calculated for the automated LESS-22 score and each potential covariate, namely, previous stress fracture (dichotomized to yes or no), BMI (continuous), and sex. We elected to use the IRR rather than the injury proportion ratio because the latter considers only those individuals who incur an injury rather than the overall population and is thus used to look for patterns within individuals who experience an injury.²²  Because the LESS is designed to be used as a first-line screening tool,⁹  the IRR is the superior choice. Additionally, the IRR was used by Cameron et al,¹³  which allowed us to compare our results with theirs.

An initial multivariate Poisson regression model using the natural log of calendar days as an offset variable was conducted to compare the association between LESS-22 score and BSI incidence. Risk factors were included in the final model as covariates if the 95% CI for the crude IRR did not contain 1.00. Participants with missing data were excluded from all multivariable analyses using the variable with missing data. Each Poisson model was assessed for distributional appropriateness using the Deviance chi-square test with a P value of ≥.05, indicating a Poisson distribution.

A total of 2235 participants (510 females [22.8%]) enrolled in the study. The proportion of participants from the 2015 entering cohort was 49.9% (n = 1116). Thirty-two (1.43%) participants reported experiencing a stress fracture in the 6 months before basic training. Mean BMI was 22.94 for females and 24.67 for males. A total of 54 lower extremity BSIs occurred in 42 cadets during the study period, resulting in an overall incidence rate of 0.07 BSI per 1000 person-days (95% CI = 0.05, 0.09). Thirty-four BSIs occurred in males and 20 in females. The mean time to injury was 105.54 ± 109.23 days, with a median of 43 days. The mean number of exposure days was 345.4 ± 61.12 (range = 3–368 days). The mean LESS-22 score was 5.22 ± 1.99 (range = 0–14). Complete demographic data and BSI location counts are presented in Table 1.

In univariable analyses, the total LESS-22 score was significantly associated with the BSI incidence (IRR = 1.08; 95% CI = 1.02, 1.15; P = .01). For their respective crude IRRs, the 95% CIs for sex and BMI did not contain 1.00 and thus were included in the final model along with the LESS-22 score (Table 2). The final model yielded an adjusted IRR for a LESS-22 score of 1.06 (95% CI = 1.002, 1.13; P = .04), indicating that each additional LESS-22 item documented at baseline was associated with a 6.0% increase in the incidence rate of BSI during the follow-up period, adjusted for sex and BMI. Additionally, 6 individual LESS-22 items were significantly associated with the incidence rate of BSI in univariable and multivariable analyses (Table 3). These items were trunk flexion at initial contact (IRR = 1.45; 95% CI = 1.05, 2.01), excessive trunk-flexion displacement (IRR = 1.52; 95% CI = 1.06, 2.19), ankle plantar-flexion angle at initial contact (IRR = 1.59; 95% CI = 0.97, 2.59), asymmetrical foot contact (IRR = 2.43; 95% CI = 1.07, 5.55), knee wobble (IRR = 2.96; 95% CI = 1.30, 6.75), and overall impression (IRR = 3.08; 95% CI = 1.30, 7.33).

To the best of our knowledge, we are the first researchers to examine the association of performance on the expanded LESS-22 with the BSI incidence in a military training population, as well as the first to do so using an automated version of any iteration of the LESS. Our results demonstrated an association between performance on the automated version of the LESS-22 and the risk of subsequent BSI, with each additional LESS item present associated with an increase of 6.0% in the incidence rate of BSI, adjusted for sex and BMI.

Our findings are similar to those for the manually scored LESS-17 used by Cameron et al.¹³  Those investigators reported a significant univariable association between manually scored LESS-17 scores and the incidence rate of BSI in 2 earlier USMA cohorts (2013 and 2014). However, they did not find an association between performance on the LESS-17 and the incidence rate of BSIs once the LESS-17 score was adjusted for sex and cohort.¹³  Two key differences between the Cameron et al¹³  study and ours that may account for this discrepancy are the exclusion of individuals with a history of BSI from the earlier study and the recent expansion of the LESS to include 5 additional items¹⁵  used in this study. Of the 6 individual LESS-22 items associated with increased incident BSIs in this study (Table 3), 2 (excessive trunk displacement and knee wobble) are new items recently added to create the LESS-22 version of the test.¹⁵  This information indicates that the expanded LESS-22 may have an enhanced ability to detect movement patterns associated with an elevated risk of BSIs compared with the original LESS-17.

The LESS-22 items demonstrating the strongest associations with incident BSI were knee wobble, defined as “one or both of participant's knees appears to ‘wobble' or demonstrate a quick varus/valgus motion”²³  during jump landing (IRR = 2.96; 95% CI = 1.30, 6.75; P = .01) and overall impression (IRR = 3.08; 95% CI = 1.30, 7.33; P < .01), defined as poor “if the subject displays a stiff landing and at least some frontal or transverse plane lower extremity motion or if there is a large frontal or transverse plane lower extremity motion” and excellent if an individual “displays a soft landing and no frontal or transverse plane motion.”⁹  This finding is notable because a plurality of the observed BSIs occurred in the tibia (40.7%, n = 22; Table 1). Tibial stress fractures are hypothesized to be the result of cantilever bending stresses resulting from large vGRFs.²⁴  The magnitude of cantilever bending moments is elevated by the presence of medial collapse, or dynamic knee valgus, during gait, which causes lateral migration of the vGRF relative to the midline of the tibia, increasing the cantilever bending moment experienced by the tibia.²⁵  These results, in combination with recent reports regarding the validity of the automated LESS-22,^17,18  support the criterion and construct validity of the automated version of this test.

Our outcomes for individual LESS items differed somewhat from those in the previous study with manual scoring of the LESS-17. The following 3 LESS items were significantly associated with BSI in both studies: trunk-flexion angle at initial contact, plantar-flexion angle at initial contact, and asymmetrical foot contact. The greatest adjusted IRR values in Cameron et al¹³  were reported for ankle plantar-flexion angle at initial contact (IRR = 2.33; 95% CI = 1.36, 3.97) and asymmetrical timing (IRR = 2.53; 95% CI = 1.34, 4.74). We observed similar magnitudes for these variables (Table 3), with even stronger associations for knee wobble (IRR = 2.96; 95% CI = 1.30, 6.75) and overall impression (IRR = 3.08; 95% CI = 1.30, 7.33).

A closer examination of the 6 LESS-22 items associated with increased BSI incidence in our work revealed that 3 items (trunk-flexion angle at initial contact, excessive trunk-flexion displacement, and ankle plantar-flexion angle at intial contact) are unique to the sagittal plane, 1 (knee wobble) is unique to the frontal plane, and 2 are multiplanar items (asymmetrical foot contact and overall impression).²⁶  In 2017¹⁵ and 2021,¹⁸  Mauntel et al reported that the Physimax system provided more reliable and valid measures of joint kinematic variables in the sagittal plane than those in the frontal and transverse planes. This finding indicates that some additional frontal- and transverse-plane LESS items may be associated with the BSI incidence but were not detected by the Physimax system. However, we do not believe this is likely because Cameron et al¹³  found only 1 frontal- or transverse-plane item (wide stance width) associated with increased BSI incidence in their sample.

The automated LESS-22 for screening in military populations has several major strengths, including the system's low cost, portability, and minimal personnel training and involvement required for testing and scoring.¹⁸  These strengths overcome the relative drawbacks of moderate validity¹⁸  and strength of association with injury, particularly when one considers that most incoming military personnel currently receive little or no movement-quality or injury-risk screening at entry.¹⁴  The automated LESS-22 shows promise for continued development as a first-line screening tool^14,27  to assess incoming military personnel en masse and identify individuals who have movement patterns associated with an increased risk of BSI. These individuals could then undergo second-level, in-depth screening by a movement-quality specialist, such as an athletic trainer or physical therapist, or a laboratory-based movement analysis using a 3-dimensional motion-capture system, such as that produced by Vicon or Qualisys. This 2-tiered strategy would help direct the limited resources for injury-risk reduction to those individuals at greatest risk for BSI.¹⁴  Identifying movement patterns associated with an increased risk of BSI, such as the 6 LESS-22 items displayed in Table 3, can also greatly aid in creating effective injury-prevention programs²⁸ ; movement quality is a modifiable risk factor for lower extremity injury^29–31  that can be used to inform the creation of injury-prevention programs^28,32  based on neuromuscular retraining through corrective exercises.^29,33,34 

Use of the automated LESS-22 also holds 2 important advantages over the traditional means of BSI risk assessment, namely, gait analysis. First, poorer performance on the LESS was associated with additional lower extremity injury risk, most notably involving the ACL.^12,35  Second, performance on the LESS is associated with several of the biomechanical variables used in gait analysis,¹  including ground reaction forces,⁹  but with fewer equipment, time, and space constraints. Furthermore, performance on the LESS improved in response to a movement-quality intervention consisting of corrective exercises.²⁹  These features make it a more useful tool than traditional gait analysis for screening large groups such as incoming military personnel. Researchers should determine if performance on the automated LESS-22 is associated with a future risk of other lower extremity injuries, such as ACL and lateral ankle sprains, and if performance on the automated LESS-22, in conjunction with other explanatory variables, is associated with the BSI risk at the individual, in addition to the aggregate, level. Additionally, investigators should assess the prognostic accuracy of the automated LESS-22 with a goal of identifying thresholds for preemptive interventions to enhance movement quality.

This study had several limitations. Despite including many participants, we showed that the number of incident BSIs (n = 54, 2.4% incidence proportion) was still relatively low. Additionally, missing data on 248 participants (11.1%), primarily due to missing LESS-22 scores, may have resulted in a systematic error in our analyses. However, this error was unlikely, as missing LESS-22 data were primarily due to training schedule conflicts (ie, an entire company of cadets missed the LESS-22 assessment in order to stay on schedule and complete training requirements). Incoming cadets are randomly assigned to companies, so this was not likely to have influenced our results. A final limitation was the possibility that some incident BSIs were not captured. Yet we believe this was highly improbable given that all cadets are provided with medical care in a local, closed medical system.

Our work also had several strengths. First, the inclusion of males and females in our sample allowed us to examine the association of LESS performance and BSIs in both sexes. Second, the large sample size resulted in adequate power for accurately evaluating the performance of the automated LESS. Finally, the closed health care system in the study environment likely resulted in the inclusion of virtually all incident BSIs during the study period.

We provided evidence that performance on the expanded LESS-22 assessed using an automated markerless motion-capture software system was associated with the incidence rate of lower extremity BSIs in a military training population. Each additional LESS item present was associated with an increase of 6.0% in the incidence rate of BSIs. Coupled with the low cost and time constraints associated with the LESS-22, these results support further investigation of the automated LESS-22 as a scalable solution for first-line screening of the BSI risk in military training populations.

Supplemental Content. LESS-17 scoresheet.

Found at DOI: http:doi.org/10.4085/1062-6050-0263.21.S1