Pre- and Post-neuromusculoskeletal Injury Risk Evaluation for Return-to-Duty Enhancement

Non-combat related musculoskeletal injuries (MSKI) are the leading cause of morbidity and disability in the United States Military, eroding combat readiness more than any other single disease or health condition. The MSKI problem is challenging to address, in part because it is difficult to comprehensively assess all of the factors that increase MSKI risk and those factors that influence post-MSKI outcomes, value-based care must have the necessary infrastructure to capture and process actionable information, and then implement relevant changes (i.e., learning healthcare system). However, there are no comprehensive Military Health System (MHS) recommendations regarding the collection and use of objective and subjective patient-centric data to inform MSKI risk assessment and mitigation strategies. PREPARE will develop efficient pre- and post-MSKI risk assessments that incorporate objective and subjective measures. The overall study objective is to develop comprehensive clinical assessments that identify the Service member specific factors that contribute to MSKI risks and undesired post-MSKI outcomes.

These objectives will be achieved by the following Specific Aims (SA):

Specific Aim #1: Determine the intra-rater, inter-rater, and inter-day reliability of the study related measurements within a (pilot) cohort of healthy active duty Service members.

Hypothesis SA#1: There will be good-to-excellent intra-rater, inter-rater, and inter-day reliability for study related measurements.

Specific Aim #2: Determine the pre-MSKI physical and psychosocial traits of Service members that differentiate between individuals who go on to sustain an MSKI and those who do not within 1 year of study enrollment.

Hypothesis SA#2: A common set of field- expedient physical (e.g., movement assessments, gait, joint range of motion) and psychosocial (e.g., National Institutes of Health Patient-Reported Outcomes Measurement Information System [PROMIS], TSK-11) assessments can identify the pre-MSKI factors that contribute to greater MSKI risks.

Specific Aim #3: Determine the post-MSKI physical and psychosocial traits of Service members with non-surgically managed MSKIs that differentiate between individuals who go on to have undesired outcomes and those who have expected outcomes.

Hypothesis SA#3: A common set of field- expedient physical (e.g., movement assessments, gait, joint range of motion) and psychosocial (e.g., NIH PROMIS, TSK-11) assessments can identify the post-MSKI factors that contribute to undesired patient outcomes (e.g., increased MSKI risk, symptom/condition chronification, delayed return-to-duty/activity [RTD/A]).

Specific Aim #4: Create optimized (parsimonious) pre- and post-MSKI clinical assessments for the identification of physical and psychosocial factors that provide the information needed to improve pre- and post-MSKI risk mitigation and rehabilitation strategies.

Hypothesis SA#4: A common set of semi-automated field-expedient assessments can be structured to correctly inform clinical decision-making and inform Service members and healthcare providers about likely patient outcomes (e.g., MSKI risk, time to RTD/A).

Specific Aim #5: Validate the optimized versions of our pre- and post-MSKI assessments by demonstrating their abilities to predict MSKI risks and outcomes in new Service Member cohorts.

Hypothesis SA#5: The optimized pre and post-MSKI assessments will accurately identify Service Members at the highest risk for sustaining MSKIs and undesired post-MSKI outcomes across Service member populations.

This is a multi-site observational study to identify the pre- and post-MSKI physical and psychosocial factors contributing to greater MSKI risk (Specific Aim #2) and undesired post-MSKI outcomes (Specific Aim #3) and to identify the minimum set of clinical assessments needed to provide healthcare providers with the data required to optimize MSKI care (Specific Aim #4) and to validate the optimized pre- and post-MSKI assessments by demonstrating their abilities to predict MSKI risks and outcomes in new Service Member cohorts. For Specific Aim #1, 10 healthy active duty Service members will serve as a pilot cohort for participation in the study. These individuals will complete all study related activities so that the study team can establish the best time estimate possible for the completion of study related activities. These individuals will repeat the same study related activities 7-10 days following the initial visit; this will allow the study team to establish the intra-rater, inter-rater, and inter-day reliability of the study related measures. For Specific Aim #2, Service members (n=560) who are assigned to a military unit will be prospectively enrolled. Participants will undergo a comprehensive pre-MSKI clinical assessment at the time of study enrollment, including clinical movement and range of motion assessments, and patient reported outcome (PRO) measures. MSKI data will be collected monthly for up to 1-year following initial assessment. For Specific Aim #3, Service members (n=780) who are receiving conservative treatment for a musculoskeletal injury of the low back or lower extremity will be prospectively enrolled. Service members will receive treatment for MSKIs at the discretion of the Service members' healthcare providers. Participants will undergo repeat ["initial"]; 4 weeks post-initial assessment, or at RTD/A clearance, if prior to 4 weeks; and 12 weeks post-initial assessment, or at RTD/A clearance, if prior to 12 weeks) clinical movement and range of motion assessments. PRO and MSKI data will be collected monthly for up to 1-year following the initial assessment. For Specific Aim #5, Service Members who are assigned to a unit at Fort Liberty, NC (n=560) and active duty Service members receiving rehabilitation care for a MSKI at any rehabilitation clinic on Fort Liberty,NC (n=560) will be enrolled. Participants will complete our optimized pre- and post-MSKI assessments identified in Specific Aim #4. All procedures and time points will be identical to those previously described for Specific Aims #2 and #3.

For Specific Aims #2 and #3, participants will complete: 1) Movement assessments utilizing semi-automated kinematic (markerless motion capture system) and kinetic (instrumented walkway) measurements for jump-landing, triple hop, single leg squat, double leg squat, and gait; 2) Range of motion measurements including hip extension, hip abduction, knee flexion, knee extension, and dorsiflexion; 3) PRO measures including NIH PROMIS and Tampa Scale of Kinesiophobia (TSK-11),and 4) MSKI Tracking.

To address Specific Aim #2 and #4, the study team will use a statistical approach to select and evaluate candidate assessments for the prediction of MSKIs. Optimal assessments will be determined based on lowest rates of misclassification of MSKI prediction. This analysis will use a machine learning algorithm (hierarchical cluster analysis) to define groups of highly correlated response variables among the comprehensive set of clinical assessments. The study team will describe and compare variables obtained from individual assessments by the presence versus absence of MSKI for up to one year follow-up, using t-tests or a nonparametric alternative. Hierarchical cluster analysis (Ward's method) will be used to next identify clusters of correlated variables that minimize within-cluster variance in the dataset. Tree diagrams and cluster means will be used to interpret clusters in relation to individual variables and functional domains described in the literature, including performance and general health measures and self- assessment of function. Sensitivity analyses will evaluate the robustness of assigned groupings to the choice of clustering method, using alternative algorithms to calculate similarity between clusters (e.g., complete linkage or group average methods).

Cluster analysis results will be used to derive reduced, candidate assessments from potential combinations of the individual variables found characterized by the identified clusters. The candidate assessment variables will be evaluated as predictors in binary classification models of MSKI through up to one year follow-up. Misclassification rates will be reported as the percent of participants where the outcome was incorrectly predicted after thresholding using clinically relevant sensitivities and specificities. Cross-validation will be used to minimize bias in estimated probabilities. The study team will select assessments that combine to have the lowest misclassification rates and shortest completion times to create the optimal MSKI risk assessment (Specific Aim #4). Completion times will be estimated based on observed times to administer each individual item. Analyses will be repeated restricting outcomes to the most common individual MSKI types (versus no MSKI), to evaluate for qualitative differences in optimal assessments by injury type.

To address Specific Aims #3 and #4, the study team will use a statistical approach to select and evaluate candidate assessments for the prediction of MSKI outcomes (Specific Aim #3). Optimal assessments will be determined based on lowest rates of misclassification of outcomes and shortest assessment times. This analysis will use a machine learning algorithm (hierarchical cluster analysis) to define groups of highly correlated response variables among the comprehensive set of clinical assessments. The study team will use hierarchical cluster analysis (Ward's method) to identify clusters of observations that minimize within-cluster variance at the baseline visit. Tree diagrams and cluster means will be used to interpret clusters in relation to individual variables and functional domains described in the literature. Sensitivity analyses will evaluate the robustness of assigned groupings to the choice of clustering method, using alternative algorithms to calculate similarity between clusters (e.g., complete linkage or group average methods). Analyses will be repeated within subgroups of MSKI type to evaluate for differences in identified clusters across injury types.

Cluster analysis results will be used to derive reduced, candidate assessments from potential combinations of the individual variables characterized by the identified clusters. The candidate assessment variables will be evaluated as predictors in binary classification models (e.g., logistic or Lasso regression) of undesired patient outcomes. Misclassification rates will be reported as the percent of patients where the outcome was incorrectly predicted after thresholding using clinically relevant sensitivities and specificities. Cross-validation will be used to minimize bias in estimated probabilities. The study team will select assessments that combine to have the lowest misclassification rates and shortest completion times to create the optimal post-MSKI assessment (Specific Aim #4). Completion times will be estimated based on observed times to administer each individual item. Analyses will be repeated for back pain versus less frequent MSKI types, and for the 12-week follow-up visit, to evaluate for differences in optimal assessments by MSKI type and time post-MSKI. Because primary analyses will be based on baseline assessments and outcomes ascertained by internet-based questionnaires to minimize attrition, no imputation of missing data is planned.

To address Specific Aim #5, the study team will evaluate the ability of our optimal assessments (Specific Aim #4) to predict MSKIs and undesired outcomes in validation cohorts over six month follow-up. Misclassification rates for unadjusted and full models will be described using receiver operating characteristic curves (ROC) and 95% confidence intervals. Analyses will be repeated separately for back pain and lower extremity MSKIs using the optimal assessments identified for these injury types.