While there is good evidence for the role of biological, psychological, and social factors in the etiology and prognosis of back pain, the synthesis of the 3 in research and clinical practice is suboptimal. This precludes a personalized approach to cLPB treatment that would support improved clinical outcomes. The primary objective of this research project is to address the critical need for new diagnostic and prognostic markers and associated patient classification protocols for cLBP treatment. To achieve our objectives, we propose three aims to prioritize and/or validate novel instruments that assess critical domains of the biopsychosocial model, validate patient-centered outcome measures, and investigate their clinical utility using the UCSF REACH cLBP Clinical and Digital Cohorts.
In Aim 1, we propose to validate common data elements (CDEs) that characterize important phenotypic traits in cLBP patients. These data elements will be aligned with domains of the biopsychosocial model (Aim 1a, bio- behavioral;
Aim 1 b, pathophysiological;
and Aim 1 c, functional/biomechanical). Before CDE's are introduced into the REACH clinical cohort, they will be prioritized into three categories by measures of reproducibility, diagnostic accuracy, and clinical validity: basic, supplemental, and emerging. Through this work, we will validate an imaging suite that researchers can use to study the spine pathologies in clinical cohorts, and clinicians can use to improve their care of cLBP patients.
In Aim 2, we will define personalized outcome measures that constitute a clinically meaningful treatment effect for individual patients. These measures will be derived from the Patient-Reported Outcomes Measurement Information System (PROMIS), and will objectively determine 'what is acceptable' to the patient.
In Aim 3 we will analyze phenotypic traits, using a combination of traditional data analyses and deep learning methods, to define clinically useful cLBP phenotypes. In both Aims 2 and 3, we will utilize both traditional statistical approaches and complex machine learning techniques. If we show that our machine learning models outperform the clinicians (who are currently inundated with data), these tools can prove to be beneficial clinical decision support systems in the setting of patient-centric treatment planning. Throughout, we plan dynamic interactions with the BACPAC consortium. BACPAC/REACH collaborations will enhance our abilities to successfully attain our ultimate goal of developing algorithms for personalized cLBP treatments that lead to improved clinical outcomes.
