IMPACT Abstract Chronic pain impacts 50 million U.S. adults, severely interferes with the work and life of over 25 million, and costs $635 billion annually for medical treatment and resultant loss of productivity. While some non- pharmacological complementary pain management methods, such as Mindfulness-Based Stress Reduction (MBSR), are effective at reducing the pain of some patients, others do not respond. Clinicians lack the tools to accurately and reliably predict which patients will respond to complementary treatments. Racially and ethnically diverse populations are also underrepresented in both research and practice of complementary interventions despite increased risk for chronic pain and related adverse outcomes. In response to RFA-NS-22-050 (UG3/UH3), IMPACT – Integrative Mindfulness-Based Predictive Approach for Chronic low back pain Treatment proposes using machine learning methods (a subfield of AI) to identify biopsychosocial predictive and monitoring markers of the ’ response to MBSR for chronic low back pain (cLBP). This research will target a diverse, high risk population suffering from cLBP (total n=350). Comprehensive biopsychosocial data (locomotor activity, sleep, circadian rhythms, heart rate variability, depression, anxiety, pain outcomes, and social support) will be collected from diverse patients treated with MBSR for cLBP. Aim 1 (UG3) will involve the initiation of a clinical trial of MBSR for cLBP (n=50) and ML modeling with longitudinal biopsychosocial data and related clinical trial datasets to identify candidate predictive and monitoring markers of the response to MBSR for cLBP prior to expanding the trial in the UH3 phase. Milestones for transition from the UG3 phase (Aim 1) to the larger clinical trial of the UH3 phase (Aims 2+3) will include: (1) finalized data collection and primary analysis protocols for the clinical trial of MBSR for cLBP, (2) success with passive data collection procedures and experimentation with ML model training and testing for the identification of predictive and monitoring biopsychosocial markers of the response to MBSR for cLBP, and (3) preliminary validation of candidate ML-based biopsychosocial predictive and monitoring markers of the response to MBSR for cLBP using statistical and cross-validation methods. Aim 2 (UH3) will expand the clinical trial initiated in Aim 1 to collect biopsychosocial data from a larger sample (n=300). Aim 3 (UH3) will involve ML modeling with data collected in Aim 2 to identify and validate accurate biopsychosocial predictive and monitoring markers of the response to MBSR for cLBP. To complete our aims, clinician scientists from Boston University, University of Massachusetts Chan Medical School, and Cambridge Health Alliance with extensive expertise in successfully recruiting and engaging diverse populations in clinical trials of mindfulness interventions for pain will collaborate with biomedical, data scientists and machine learning researchers from Worcester Polytechnic Institute. This proposed project will ultimately enhance clinical decision- making and targeted treatment of cLBP.

IMPACT Narrative Chronic pain impacts 50 million US adults and over-reliance on pain opioids has caused an epidemic in which 29% of chronic pain patients abuse opioids, 6% transition to heroin, and 130 people die from opioid overdose each day. While some non-pharmacological complementary pain management methods, such as Mindfulness Based Stress Reduction (MBSR), are effective at reducing the pain of some patients, others do not respond, and clinicians are unable to accurately predict which patients will respond. We propose using machine learning methods to identify biological, psychological and social markers that can be used to accurately predict the response of diverse individuals with chronic low back pain (cLBP) to MBSR in order to enhance clinician decision-making and the targeted treatment of chronic pain.