Can a Psychologic Profile Predict Successful Return to Full Duty After a Musculoskeletal Injury?

Abstract

Background

Psychologic variables have been shown to have a strong relationship with recovery from injury and return to work or sports. The extent to which psychologic variables predict successful return to work in military settings is unknown.

Questions/purposes

In a population of active duty soldiers, (1) can a psychologic profile determine the risk of injury after return to full duty? (2) Do psychologic profiles differ between soldiers sustaining injuries in the spine (thoracic or lumbar) and those with injuries to the lower extremities?

Methods

Psychologic variables were assessed in soldiers returning to full, unrestricted duty after a recent musculoskeletal injury. Most of these were noncombat injuries from work-related physical activity. Between February 2016 and September 2017, 480 service members who were cleared to return to duty after musculoskeletal injuries (excluding those with high-velocity collisions, pregnancy, or amputation) were enrolled in a study that tracked subsequent injuries over the following year. Of those, we considered individuals with complete 12-month follow-up data as potentially eligible for analysis. Based on that, approximately 2% (8 of 480) were excluded because they did not complete baseline surveys, approximately 2% (11 of 480) were separated from the military during the follow-up period and had incomplete injury data, 1% (3 of 480) were excluded for not serving in the Army branch of the military, and approximately 2% (8 of 480) were excluded because they were not cleared to return to full duty. This resulted in 450 soldiers analyzed. Individuals were 86% (385 of 450) men; 74% (331 of 450) had lower extremity injuries and 26% (119 of 450) had spinal injuries, including soft tissue aches and pains (for example, strains and sprains), fractures, and disc herniations. Time-loss injury within 1 year was the primary outcome. While creating and validating a new prediction model using only psychological variables, 19 variables were assessed for nonlinearity, further factor selection was performed through elastic net, and models were internally validated through 2000 bootstrap iterations. Performance was deciphered through calibration, discrimination (area under the curve [AUC]), R², and calibration in the large. Calibration assesses predicted versus actual risk by plotting the x and y intersection of these values; the more similar predicted risk values are to actual ones, the closer the slope of the line formed by the intersection points of all subjects is to equaling “1” (optimal calibration). Likewise, perfect discrimination (predicted injured versus actual injured) presents as an AUC of 1. Perfect calibration in the large would equal 0 because it represents the average predicted risk versus the actual outcome rate. Sensitivity analyses stratified groups by prior injury region (thoracic or lumbar spine and lower extremity) as well as the severity of injury by days of limited duty (moderate [7-27 days] and severe [28 + days]).

Results

A model comprising primarily psychologic variables including depression, anxiety, kinesiophobia, fear avoidance beliefs, and mood did not adequately determine the risk of subsequent injury. The derived logistic prediction model had 18 variables: R² = 0.03, calibration = 0.63 (95% confidence interval [CI] 0.30 to 0.97), AUC = 0.62 (95% CI 0.52 to 0.72), and calibration in the large = -0.17. Baseline psychologic profiles between body regions differed only for depression severity (mean difference 1 [95% CI 0 to 1]; p = 0.04), with greater mean scores for spine injuries than for lower extremity injuries. Performance was poor for those with prior spine injuries compared with those with lower extremity injuries (AUC 0.50 [95% CI 0.42 to 0.58] and 0.63 [95% CI 0.57 to 0.69], respectively) and moderate versus severe injury during the 1-year follow-up (AUC 0.61 [95% CI 0.51 to 0.71] versus 0.64 [95% CI 0.64 to 0.74], respectively).

Conclusion

The psychologically based model poorly predicted subsequent injury. This study does not minimize the value of assessing the psychologic profiles of injured athletes, but rather suggests that models looking to identify injury risk should consider a multifactorial approach that also includes other nonpsychologic factors such as injury history. Future studies should refine the most important psychologic constructs that can add the most value and precision to multifactorial models aimed at identifying the risk of injury.

Level of Evidence

Level III, prognostic study.

Introduction

Musculoskeletal injuries are the leading medical cause, by far, for limited duty days for military service members [25, 37]. Each year, approximately 50% of individuals in a military unit experience a musculoskeletal injury [17, 31, 45]. Because injuries are bound to happen in this physically active population, tactics to promote successful recovery and prevent subsequent injuries are needed to save time and resources spent on rehabilitation. Many studies have examined the associations of psychologic variables and return to physical activity with the risk of reinjury in the context of sports injuries [7, 12, 19, 27, 32, 44]. Such findings are encouraging, because soldiers are tactical athletes who experience injuries because of physical activity similar to that of civilian athletes. Similar to injured athletes with recurrent injuries that continue to remove them from important games, soldiers with recurrent injuries cannot deploy or work at full capacity. This can impact the readiness of the individual, unit, and whole military. However, there is little to no research examining how psychologic variables relate to successful return to work (termed “duty” in the military) after a musculoskeletal injury. Medical clearance for full duty indicates that a soldier’s injury is no longer limiting their ability to perform their unique job requirements. If they are cleared to return to work after insufficient recovery, physically or mentally, they may be less prepared to avoid a subsequent injury.

In the context of injury and recovery, psychologic variables have been recognized for the impact they have on a person’s readiness to return to physical activity and their risk of reinjury [21, 23, 48]. Clinical assessments for return-to-play decisions often emphasize physical function and pain [41]. Because these variables alone are not comprehensive enough to determine the risk of subsequent injury, psychologic factors might provide additional insight. Unlike previously identified injury risk factors such as age [8, 18, 33, 47], gender [4, 38], and prior injury [9, 44, 51], which are nonmodifiable [39], psychologic variables have been known to fluctuate across different stages of the health injury continuum [48], similar to physical function, and could be future targets of intervention to aid in successful return to full duty. There is no universal consensus on the defintion of “psychologic readiness” to return to sports [32]; however, relationships between several different psychologic constructs and return to sport [27], future injury [44], rehabilitation adherence [12], and functional outcomes [7] after injuries have been reported. For instance, fear of reinjury has been associated with poor rehabilitation outcomes, difficulties with returning to sport, or even with function and distraction when returning to sport [19]. Further, psychologic interventions (such as cognitive behavioral therapy and stress management) have been used for injury prevention programs [24] and recovery programs for injured athletes [26, 40]. In the current study, we explored whether a profile comprising primarily psychologic variables could predict subsequent injury after return to full duty within the subsequent year. We further examined differences in the psychologic states of soldiers cleared to return to full duty after sustaining injuries in two of the most common areas experienced in this population: the back and lower extremities [37].

We therefore asked, in a population of active duty soldiers, (1) can a psychologic profile determine the risk of injury after return to full duty? (2) Do psychologic profiles differ between soldiers sustaining injuries in the spine (thoracic or lumbar) and those with injuries to the lower extremities?

Patients and Methods

Study Design and Setting

This was a prediction model derivation study using a previously defined cohort of US soldiers [35]. Recruitment occurred in three large military hospitals, with enrollment from February 2016 to September 2017. Injury surveillance occurred prospectively for 1 year after enrollment. The Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis was used to guide reporting [6]. Training and testing splits are not advised in small- or moderate-sized datasets [13] because this reduces power for developing a prediction model and potentially biases the models, depending on the split. It is recommended to internally validate the model through bootstrapping or crossvalidation, which is what was performed in this study.

Participants

The established cohort included full-duty soldiers 18 to 45 years old [35]. Individuals were excluded if they were planning to leave the military in the next 12 months, had a concomitant injury they were already seeking or planning to seek medical care for, had permanent limited-duty restrictions, were pending Medical Evaluation Board, had trauma resulting in amputation, had injuries from high-velocity collisions, or were pregnant or recently pregnant within the past 6 months. Potentially eligible participants were those who had recently experienced musculoskeletal injuries of the lumbar or thoracic spine or lower extremity (n = 480), confirmed by electronic health record. Of those, approximately 2% (8 of 480) were excluded because they did not complete baseline surveys, approximately 2% (11 of 480) because they were separated from the military during the follow-up period and had incomplete subsequent injury data, approximately 1% (3 of 480) were excluded for not serving in the Army branch of the military, and approximately 2% (8 of 480) were excluded because they had not been cleared to return to full, unrestricted duty. This resulted in 94% (450 of 480) of the soldiers available for inclusion in the model analysis (Fig. 1).

Fig. 1.

This is the flow diagram for the patients who were enrolled in this study. Injury region reflects the body region of prior injury before return to duty at the study baseline. Three individuals from non-Army service branches were excluded for sample homogeneity. ^aSix individuals with thoracic or lumbar spine injuries also had lower extremity injuries.

Description of Experiment, Treatment, or Surgery

Participants were eligible to participate after they had completed routine care with a medical provider and been cleared for full, unrestricted duty. After enrollment, a variety of psychologic variables were collected. Full inclusion and exclusion criteria are available in the published protocol [35].

Description of Follow-up Routine

Participants were followed for 1 year to see whether they sustained a musculoskeletal injury that resulted in time lost from full duty. Surveillance occurred through monthly surveys and validation against electronic medical records and the eProfile data source that tracks the number and reasons for individual limited duty days.

Descriptive Data

Soldiers (n = 450) returning to full duty after spine (n = 119) or lower extremity injuries (n = 331) had a mean age of 27 ± 6 years, 86% (385 of 450) were men, and 60% (271 of 450) had been in the military fewer than 3 years (Table 1). During the 1-year follow-up period, 38% (169 of 450) sustained another musculoskeletal injury resulting in days of limited duty, with a median of 41 days per injury (49 injuries classified as moderate and 115 as severe). Of note, 43 of the original injuries were treated operatively (42 lower extremity, one spine) and 407 were treated nonoperatively, with 37% (16 of 43) of the operative and 38% (153 of 407) of the nonoperative group experiencing a time-loss injury.

Table 1.

Baseline data for soldiers returning to duty after thoracic or lumbar spine or lower extremity musculoskeletal injuries

Variable	All (n = 450)	Thoracic or lumbar spine (n = 119)	Lower extremity (n = 331)
Age in years	27 ± 6	28 ± 6	27 ± 6
Height in m	1.7 ± 0.1	1.7 ± 0.1	1.7 ± 0.1
Weight in kg	83 ± 14	83 ± 14	84 ± 14
BMI in kg/m2	27 ± 4	27 ± 4	27 ± 4
Women	14 (65)	15 (18)	14 (47)
Ethnicity, Hispanic or Latino	24 (109)	27 (32)	23 (77)
Race
American Indian/Alaska Native	1 (5)	1 (1)	1 (4)
Asian	2 (9)	1 (1)	2 (8)
Pacific Islander	1 (4)	1 (1)	1 (3)
Black or African American	26 (115)	21 (25)	27 (90)
White or Caucasian	59 (264)	62 (74)	57 (190)
Other	12 (53)	14 (17)	11 (36)
Education level
High school/GED	40 (181)	40 (48)	40 (133)
Some college or 2-year degree	38 (170)	40 (48)	37 (122)
College or 4-year degree	14 (64)	13 (15)	15 (49)
Some postgraduate courses	4 (17)	3 (3)	4 (14)
Postgraduate degree	4 (18)	4 (5)	4 (13)
Marital status
Single	36 (160)	34 (40)	36 (120)
Married	60 (270)	60 (71)	60 (199)
Divorced	4 (20)	7 (8)	4 (12)
“In general, would you say your (physical) health is:”
Fair	5 (24)	5 (6)	5 (18)
Good	24 (110)	29 (35)	23 (75)
Very good	49 (220)	51 (61)	48 (159)
Excellent	21 (96)	14 (17)	24 (79)
“I do not feel that I need any more medical care before returning to my full military duties”
Strongly agree	12 (56)	11 (13)	13 (43)
Agree	71 (321)	73 (87)	71 (234)
Disagree	14 (63)	15 (18)	14 (45)
Strongly disagree	2 (9)	1 (1)	2 (8)
Sleep qualitya
Normal	59 (264)	56 (67)	60 (197)
Abnormal	41 (186)	44 (52)	41 (134)
Regular smoker	12 (56)	14 (17)	12 (39)
Self-reported physical activity level
Inactive	1 (4)	1 (1)	1 (3)
Average	22 (97)	24 (29)	21 (68)
Active	45 (203)	43 (51)	46 (152)
Very active	32 (146)	32 (38)	33 (108)
Rank category
Enlisted	88 (395)	84 (100)	89 (295)
Officer	12 (55)	16 (19)	11 (36)
Time in service
< 1 year	37 (167)	32 (38)	39 (129)
At least 1 year, < 3 years	23 (104)	21 (25)	24 (79)
At least 3 years, < 5 years	17 (76)	19 (23)	16 (53)
5 + years	23 (103)	28 (33)	21 (70)
Proportion of job time spent on physical tasks
0% to 25%	41 (185)	48 (57)	39 (128)
26% to 50%	33 (148)	30 (36)	34 (112)
51% to 75%	18 (83)	15 (18)	20 (65)
76% to 100%	8 (34)	7 (8)	8 (26)
Deployed > 2 months in the past 5 years	36 (163)	39 (47)	35 (116)
Sustained a subsequent injury during 1-year follow-up	38 (169)	39 (47)	37 (122)

Data presented as mean ± SD or % (n).

^aClinically poor sleep on Epworth Sleepiness Scale (< 16) or Pittsburgh Sleep Quality Index (≤ 5) was considered “abnormal.”

Variables, Outcome Measures, Data Sources, and Bias

Demographics of age, gender, and years of military service were collected for model inclusion to control for their known relationships with injury risk. The psychologic profile included depression (clinically rated and self-perceived), anxiety, anger, fear, frustration, exercise enjoyment, job, military, life satisfaction, kinesiophobia, fear avoidance beliefs, pain catastrophizing, stress, and mood. Psychologic constructs were assessed via self-report, and raw group medians and means were observed (Supplemental Tables 1 and 2; http://links.lww.com/CORR/B265). Depression severity was measured with the Patient Health Questionnaire-9 (PHQ-9) [22, 42], and a single-item rating was used to quantify the perceived level of usual depression: “Rate your usual level of the following: depression 0 (none) to 100 (most severe imaginable).” Response options were 0 to 100, in increments of five. Anxiety, frustration, anger, and fear were also measured using this same single-item rating. Personal satisfaction with their job (“duties in the armed forces”), the military (“being in the armed forces”), and life were assessed on a scale from 0 to 100. A single item asked, “Do you enjoy exercising?,” with anchors ranging from 0 (completely do not like to exercise) to 100 (completely enjoy exercise). A total Tampa Scale for Kinesiophobia score was calculated from 11 items indicating the extent to which participants feared movement on 4-point Likert scales ranging from 1 (strongly disagree) to 4 (strongly agree) [52]. Total Pain Catastrophizing Scale scores were summed from 13 items with individual responses rated on a 0 to 4 scale, with possible scores ranging from a minimum of 0 to a maximum of 52 [43]. The Fear-Avoidance Beliefs Questionnaire total score [50] assessed participants’ perceptions of the extent to which certain physical activities exacerbate their back or lower extremity pain. Subjective stress level and mood over the past 2 weeks were measured on 5-point scales ranging from “highly stressed” (1) to “very relaxed” (5), and “highly annoyed/irritable/down” (1) to “very positive mood” (5), respectively [28].

Primary Study Outcome

The primary outcome was defined as any musculoskeletal injury in the year after return to full duty that resulted in at least 1 day of limited duty (time-loss injury). Limited duty days associated with musculoskeletal injury were identified from the eProfile source in the Medical Operational Data System profile database [35]. Absence of an injury served as a proxy for successful return to full duty. We built a model to try to predict how well the expected occurrence of musculoskeletal injury, based on psychologic variables, could determine the actual occurrence of injury. After testing, models with better performance had calibration and area under the curve (AUC) values that approached 1 and intercept values closer to 0.

Statistical Analysis, Study Size

To assess potential psychologic differences between body regions (thoracic or lumbar spine versus lower extremity), analyses of covariance were performed and are reported as mean difference (95% confidence interval [CI]). All analyses of covariance were controlled for age, gender, BMI, and years of military service at the initial profile, with statistical significance set at α = 0.05.

Sample Size Calculation for Developing the Prediction Model

An a priori sample size calculation was performed [36] that used R² = 0.64 from the primary study [34, 35], with a prevalence of 0.34, indicating that 27 allowable parameters could be included to reduce the risk of overfitting.

Missing Data

Before model development, data missingness was assessed. Five percent of the data were missing and consequently imputed, most notably the Pain Catastrophizing Scale (5%) and PHQ-9 (3%) scores, with a mechanism of missing at random. Multiple imputation with chained equations with 20 iterations per imputation, for a total of 20 imputations, was performed to control for missing data.

Model Development

We developed a logistic model with injury as the primary outcome from variables that were included in a psychologic survey that was used in a warrior athlete population [50]. These variables encompassed the biopsychosocial constructs of job and life satisfaction, depression, anxiety, fear, frustration, anger, stress, mood, exercise enjoyment, pain catastrophization, kinesiophobia, and pain-related fear-avoidance beliefs. Originally, we assessed 19 predictor variables (16 psychologic and three demographic: age, gender, and military years) for a potential nonlinear association via restricted cubic splines. Anxiety was the only factor with a nonlinear relationship to the primary outcome of injury and was transformed for inclusion in the model using restricted cubic splines. We used a data-driven approach for nonlinear assessments, and the lowest Akaike information criterion was used for selecting the optimal predictor of outcome association. Bivariate correlation among all variables was assessed. For any variables with moderate or strong relationships, we computed variance inflation factors to assess multicollinearity. Further predictor selection was performed through elastic net regularization, resulting in the exclusion of “fear” from the model. All models were internally validated through 2000 bootstrap iterations to reduce optimism bias. Prediction performance was deciphered through calibration (1 = best), discrimination (AUC), Cox-Snell R² (overall fit of the model), and calibration in the large. Calibration is the assessment of the predicted versus actual risk, with a slope of 1 equaling optimal calibration. Discrimination is the assessment of patients predicting versus actually having the outcome, with an AUC of 1 equaling perfect discrimination. Calibration in the large is the average predicted risk versus the actual outcome rate; perfect calibration in the large (intercept) equals 0 [3]. The primary analysis examined the performance of the prediction model. Sensitivity analyses examined the efficacy of the model in subgroups including injury severity (days of limited duty) defined as moderate (7 to 27) or severe (28 or more) and those returning from recent injury to the back (thoracic and lumbar spine) versus lower extremity body region.

Ethical Approval

Ethics approval was granted by the institutional review board at Madigan Army Medical Center, Tacoma, WA, USA, and all participants provided signed informed consent.

Results

Can a Psychologic Profile Determine the Risk of Injury After Return to Full Duty?

A profile comprised of psychologic variables, such as depression, anxiety, and mood, did not perform well when used to determine the risk of subsequent injury for soldiers who had recently returned to full duty. The final prediction model had 18 variables, with no concern for multicollinearity (Table 2). The final model resulted in R² = 0.03, calibration = 0.63 (95% CI 0.30 to 0.97) (slope of 1 = optimal calibration), AUC = 0.62 (95% CI 0.52 to 0.72) (1 = perfect discrimination), and calibration in the large = -0.17 (0 = perfect) (Fig. 2). Sensitivity analyses revealed weak performance of the model for those who had experienced an initial thoracic or lumbar spine injury compared with those with a lower extremity injury before returning to duty (AUC 0.50 [95% CI 0.42 to 0.58] and 0.63 [95% CI 0.57 to 0.69], respectively) or those with moderate versus severe duty limitations (AUC 0.61 [95% CI 0.51 to 0.71] and 0.64 [95% CI 0.64 to 0.74], respectively) (Fig. 3).

Table 2.

Final prediction model equation coefficients

Variable	Coefficient
Age	-0.04
Gender	0.79
Years of military service	-0.17
Exercise enjoyment	0.005
Depression	-0.01
rcs(Anxiety, 4) anxiety	0.04
rcs(Anxiety, 4) anxiety′	-0.23
Frustration	0.01
Anger	0.01
Fear
PHQ-9, total score	0.01
TSK-11, total score	0.04
PCS, total score	0.02
FABQ, total score	0.005
Stress	0.30
Mood	-0.02
Job satisfaction	-0.01
Military satisfaction	0.02
Life satisfaction	-0.002

Fear was dropped from the model after elastic net regularization (shrunk to zero). Intercept = -3.4. The intercept is the overall prevalence of the outcome in the sample, depicted in log odds. The intercept value in the model equation helps correct the model to make it more accurate. Models that perform well to accurately predict events of interest, in this case, “time-loss injury,” would have intercepts closer to 0 (because the variables included in the model would contribute more predictive capacity on their own). PHQ-9 = Patient Health Questionnaire-9; TSK-11 = Tampa Scale for Kinesiophobia-11; PCS = Pain Catastrophizing Scale; FABQ = Fear-Avoidance Beliefs Questionnaire.

Fig. 2.

This graph shows the final model after multiple imputation and elastic net and bootstrapping 2000 iterations. Calibration is the relationship between the predicted and actual probability of the event. The calibration slope plots the predicted risk graphically against the observed outcome, displaying the calibration intercept and calibration slope. Perfect calibration would result in a 45° line. In this calibration plot, the original developed (apparent) calibration and after 2000 bootstrap internal validation (bias corrected) are reported on this plot. Both calibrations' slopes are above the ideal 45° line for the entire risk spectrum. This is interpreted as follows: The predicted risk is below the actual (ground truth) risk. The clinical implication is that a person’s outputted risk from this prediction model may change clinical decisions away from providing injury prevention methods to soldiers across the risk spectrum.

Fig. 3.

This is a calibration of the model for individuals returning from (A) lower extremity or (B) spine injuries. Calibration is the relationship between the predicted and actual probability of the event. The calibration slope plots the predicted risk graphically against the observed outcome, displaying the calibration intercept and calibration slope. Perfect calibration would result in a 45° line. In this calibration plot, the original developed (apparent) calibration and after 2000 bootstrap internal validation (bias corrected) are reported on this plot. Both calibrations’ slopes diverge from the ideal 45° line below 0.20 risk and 0.5 risk. This is interpreted as follows: The predicted risk is below the actual (ground truth) risk for risk below 0.2 and above the actual risk for risk above 0.5. The clinical implication is that a person’s outputted risk below 0.2 from this prediction model may be higher and change clinical decisions away from providing injury prevention methods to soldiers for this specific risk threshold. Risks above 0.50 would not be altered on a clinical level from this model.

Do Psychologic Profiles Differ Between Lower Midspine and Lower Extremity Injuries?

Comparing the cohort with initial spine injuries with those who had lower extremity injuries, baseline psychologic profiles differed only slightly. Specifically, PHQ-9 total scores were generally low but greater for individuals returning after a spine injury than for those with lower extremity injuries (mean difference 1 [95% CI 0 to 1]; p = 0.04). Mean scores on all other psychologic variables assessed at the time of return to duty did not differ between soldiers with initial spine injuries and those with lower extremity injuries (Table 3).

Table 3.

Differences in psychologic variables between thoracic or lumbar spine and lower extremity injuries

Variable	Number	Thoracic or lumbar spine	Number	Lower extremity	Mean difference (95% CI)	p value
Exercise enjoyment (0 to 100)	119	83 ± 19	329	84 ± 19	-1 (-4 to 3)	0.82
Depression (0 to 100)	119	7 ± 15	331	6 ± 15	1 (-2 to 4)	0.61
Anxiety (0 to 100)	119	11 ± 19	331	10 ± 18	1 (-3 to 5)	0.54
Frustration (0 to 100)	119	25 ± 23	331	21 ± 23	4 (-1 to 9)	0.09
Anger (0 to 100)	119	17 ± 21	329	15 ± 21	3 (-2 to 7)	0.23
Fear (0 to 100)	119	5 ± 11	330	5 ± 11	1 (-2 to 3)	0.62
PHQ-9 total (0 to 27)	116	3 ± 3	320	2 ± 3	0.7 (0.0 to 1.4)	0.04
TSK-11 total (11 to 44)	118	17 ± 6	330	17 ± 6	0 (-1 to 1)	0.93
PCS total (0 to 52)	111	3 ± 6	314	3 ± 6	0 (-2 to 1)	0.62
FABQ total (0 to 96)	119	22 ± 16	331	23 ± 16	-1 (-5 to 2)	0.53
Stress (1 to 5)	118	3 ± 1	330	3 ± 1	-0.1 (-0.2 to 0.1)	0.48
Mood (1 to 5)	119	4 ± 1	331	4 ± 1	0.0 (-0.2 to 0.2)	0.97
Job satisfaction (0 to 100)	119	69 ± 24	331	71 ± 24	-3 (-8 to 3)	0.34
Military satisfaction (0 to 100)	119	77 ± 22	331	79 ± 22	-3 (-7 to 2)	0.24
Life satisfaction (0 to 100)	119	86 ± 16	331	88 ± 16	-2 (-5 to 1)	0.28

Data presented as estimated marginal mean ± SD unless otherwise stated; covariates included age in years, BMI in kg/m²,

gender (man or woman), and time in service in years. Higher scores on exercise enjoyment, mood, job satisfaction, military satisfaction, and life satisfaction reflect more positive emotions. Higher scores on the Patient Health Questionnaire-9 (PHQ-9) and depression, anxiety, frustration, anger, and fear measures reflect greater severity of those constructs, while scores of 0 reflect the absence of them. Higher scores on the PHQ-9, Tampa Scale for Kinesiophobia-11 (TSK-11), Pain Catastrophizing Scale (PCS), and Fear-Avoidance Beliefs Questionnaire (FABQ) reflect greater levels of depression, kinesiophobia, pain catastrophizing, and fear avoidance.

Discussion

Competitive and tactical athletes experience musculoskeletal injuries because of their physically active lifestyles, and these injuries cause substantial burden to individuals, teams, and military forces. The high incidence of spine and lower extremity injuries in the military makes it exceptionally important to better understand how to successfully return people to full duty without the risk of sustaining another injury. Psychologic factors have long been considered related to injury and recovery. Yet, these findings have not been validated in the context of a prediction model. Helping to determine which individuals are at risk for subsequent injury could make the existence of such relationships useful for informing clinical decisions about return to work or play. For this study, we derived a model comprising 15 psychologic and three demographic variables to assess injury risk after return to full, unrestricted work. This model demonstrated poor performance in predicting 12-month injury, although it was marginally better in individuals returning from lower extremity injuries and when predicting injuries in individuals who had longer durations of limited duty (that is, more severe injury from a time-loss perspective). Because of the model’s inadequate ability to reliably distinguish between individuals who would experience subsequent injury and those who would not, it is not recommended for clinical use.

Limitations

The first limitation of this study is the definition of subsequent injury. We did not examine how similar the subsequent injuries were compared with the original injuries. However, a subsequent injury can still be related to the original injury regardless of whether it was the same injury [8, 53]. Hallegraeff et al. [15] successfully predicted chronic low back pain 12 weeks after medical care that was sought for new episodes of acute low back pain, by capturing concerns about chronicity of symptoms, locus of control, impact on life, and knowledge and emotions surrounding the condition. Their study examined outcomes after a single injury type, whereas our study included injuries with heterogenous pathologies. This limits our ability to know how well the model would work for specific subpopulations, making external validation difficult. However, regardless of the type of injury experienced or rehabilitation care received, clinical decisions had already been made to clear soldiers for return to duty, and the purpose of this study was to examine the ability of this prediction model in patients who represented most injured soldiers. Hallegraeff et al. [15] also included different psychologic constructs, which points to the second limitation of this study. Certain psychologic constructs and skills were not included in the existing profile, such as self-efficacy, confidence, locus of control, coping, trait sport confidence [49], or psychologic readiness to return to sport [11], but may be relevant to predicting injury. Alternative measures exist for almost all constructs that were assessed in this profile, and this variability could impact predictive capability. A consensus on preferred psychologic metrics in this setting is lacking. Others have also attempted to test relationships of composite scores from a series of questionnaires to reinjury, with and without success [5, 44]. For instance, Christakou et al. [5] assessed a number of different but related psychologic constructs (worry, confidence, and attention) as competitive athletes returned to sport after injury and tested the utility of these measures to predict the total count of reinjuries during the next season. Their profile accounted for 80% of the variance in the total number of reinjuries (1, 2, or 3) at the beginning of the season, and about 66% for those occurring midseason. This could indicate more relevant constructs, or it may reflect differences in populations or respective physical activities, methods of assessment, or injury types (because back pain was not examined in their sample). Alternatively, subscales might offer more predictive capacity than the total scores we chose for model inclusion. Perhaps certain psychologic constructs are more informative for individuals with certain types of injuries because certain risk factors may better predict injury in specific subpopulations [30]. This could partially explain the poor performance of our model. However, the exact type or number of subsequent injuries that occurred may be less relevant because any injury could have occurred due to insufficient recovery before returning to duty. That is partially what drove this investigation initially; very little is known about the psychologic status of someone cleared to return to duty, much less how that psychologic state relates to the subsequent risk of injury.

Other limitations to this work stem from the military population we examined. This could limit generalizabiltiy to civilian athletes, leaving opportunity to explore model performance in a nonmilitary setting. A variety of military occupations were represented in this sample. Physical activity exposure was not assessed and likely differed within and between individuals, both preinjury and over the course of follow-up. Although this may have influenced the relative contribution of psychologic variables to the prediction of future injury, it did not impede our ability to discover how well psychologic variables could predict it. Nevertheless, if individuals reduced their exposures on return to activity, relative to preinjury, that would be a confounder. Lastly, the current model was not tested for the ability to predict injury latency (that is, how soon someone was injured after returning to duty).

Can a Psychologic Profile Determine the Risk of Injury After Return to Full Duty?

In isolation, the psychologic profile in this study was not useful for determining the risk of subsequent injury in an active-duty military patient population. This may seem surprising because prior studies showed that some psychologic variables have been associated with injury [5, 16, 44]. However, the greatest difference between past studies and this study was that we attempted to develop a prediction model. Previous studies demonstrated evidence of relationships between psychologic factors and future injury, but they had not taken this next step. Terry et al. [44] found that individuals who expressed any concern at baseline about sustaining an injury in the next 9 months were two times more likely to experience an injury than those who expressed no concern. This “concern” could represent an individual’s self-awareness of their own risk level, based on their training loads, psychologic stress, recovery habits, and injury history. Mechanistically, fear or distress might prohibit smooth movement or promote body tension. Fear of lost balance, falling, or tripping can result in apprehensive gait patterns that alter lower leg muscle contractions from that of normal gait [14]. Additionally, higher perceived stress in elite college athletes has been associated with greater likelihood of injury (OR 0.89 [95% CI 0.85 to 0.94]) [16]. Although our model included perceived stress, timing of the assessment differed and deserves consideration. Their study [16] assessed stress at multiple points over the academic year without regard to prior injury status, while our study only assessed stress at the time of return to duty. People may have been relatively optimistic at this point and in good mental health compared with how they may have felt immediately after injury, and greater optimism has been associated with lower likelihood of injury over the next 2 years [51]. Our model’s performance may also have been limited in its ability to discriminate future injury because there was homogeneity of psychologic profiles at this timepoint. Ailliet et al. [1] found that participants with spine pain reported generally positive mental states, with only few reporting high scores on negative psychologic variables, resulting in poor prediction of recovery with their model consisting of such factors. Even so, we cannot be sure that psychologic scores closer to the point of injury would have any better capacity to predict future injury. Similar to physical function and socioenvironmental factors, psychologic variables can be transient [48]. Models that incorporate change scores, rather than point measures, may better predict subsequent injury [29]. Increases in tension, depression, anger, fatigue, and mental confusion have been demonstrated in elite soccer players from preseason to midseason, when training load and competitive stressors are higher [2]. Injuries might occur in chronologic proximity to major events, such as sports competitions, when anxiety and emotional distress might be heightened [20] or when training load increases [16]. Regular monitoring of psychologic health in relation to training and deployment cycles may be necessary to recognize fluctuations indicating susceptibility to injury. The performance of our model was stronger for individuals returning to duty after lower extremity injuries than for those with back injuries; however, this particular array of psychologic variables did not reliably discriminate which individuals would experience subsequent injury in the subsequent year, nor the severity of those injuries. Clinicians should not rely on this profile, if assessed at the point of return to duty, to inform them on subsequent injury risk. However, this profile does not exhaust all psychologic variables, and researchers should continue to examine whether more informative assessments exist.

Do Psychologic Profiles Differ Between Lower Midspine and Lower Extremity Injuries?

Psychologic profiles differed only slightly between individuals with initial spine injuries and those with lower extremity injuries. Differences were noted on depression severity, as measured by the PHQ-9, where individuals returning from thoracic or lumbar spine injuries scored 0.73 points higher, but the same difference was not present when comparing self-perceived depression on a scale from 0 to 100. Although higher depression in military service members with low back pain is consistent with the evidence [10], scores in our sample only reflected normal or minimal depression.

Conclusion

Psychologic variables alone were unable to adequately predict who would have a subsequent injury after an initial musculoskeletal injury. This study does not minimize the value of assessing psychologic profiles of injured athletes, but rather suggests that models looking to identify injury risk should consider a multifactorial approach that also includes other nonpsychologic factors such as injury history. These findings do, however, provoke curiosity surrounding which outcomes best capture relevant psychologic constructs and which assessment timepoints or change scores could be the most informative. Future research should examine the timing of assessment and choice of constructs for model inclusion, as well as model performance in groups with diverse psychologic profiles to refine injury prediction. Psychologic variables are likely still an important component of injury prevention, but because return to work is complex and influenced by multiple dimensions, psychologic variables should be combined with other known risk factors [46] such as injury history for a more comprehensive approach.

Supplementary Material

abjs-482-617-s001.docx ^{(28.4KB, docx)}