ABTRACT Musculoskeletal pain is a highly prevalent, costly, and burdensome condition commonly managed with opioids. Opioids do not often adequately address pain and have high risk for misuse and addiction. As a result, clinical practice guidelines strongly recommend non-pharmacologic treatments for musculoskeletal pain. Physical therapists are one of the most common providers of non-pharmacologic pain care in the US. However up to 43% of people receiving non-pharmacologic treatments delivered by physical therapists continue to seek care for their pain condition beyond the physical therapy episode, signaling a sub-optimal response to non- pharmacologic treatment. Additional treatments received often are higher risk, more invasive, and not recommended by current practice guidelines (e.g., opioids, surgery). The overarching aim of this project is to use real-world data to better understand risk factors for poor response to non-pharmacologic care delivered by physical therapists. This knowledge will enhance individual treatment decision making (e.g.., when to consider coordinated multimodal care) and provide direction on how to re-design non-pharmacologic care models to better address risk for poor response. In Aim 1, we will identify baseline factors associated with an unfavorable change in pain-related disability (a common precursor to escalation of care) among patients undergoing treatment by a physical therapist. Using the ATI Outcomes Registry Dataset, one of the largest outcomes datasets specific to physical therapy, we will derive common trajectories of self-reported disability following initiation of care (Sub-Aim 1a) and develop multivariable models to identify risk factors for an unfavorable disability change trajectory (Sub-Aim 1b). Aim 2 will use a separate, large claims-based dataset comprised of industrial union workers to identify baseline factors associated with escalation of care (i.e., subsequent use of imaging, opioids, injection, and surgery) following initiation of treatment by a physical therapist for musculoskeletal pain (Sub-Aim 2a). Sub-Aim 2b will use a subset of the claims-based dataset that is also included in the ATI Outcomes Registry to determine the incremental change in probability of care escalation associated with each one-point improvement in disability score. This information could help clinicians set disability improvement targets for a meaningful reduction in risk for escalating care. In exploratory analyses, we will assess how the mass disruption of healthcare delivery related to COVID-19 impacts these questions. We will perform stratified analyses for each Aim to 1) determine whether disability trajectories and factors associated with those trajectories have changed following the onset of COVID-19, and 2) establish whether receiving physical therapy prior to or after onset of COVID-19 is associated with differences in risk for poor outcomes. Collectively, this work aims to improve the delivery of non-pharmacolo...