Trajectories of non-pharmacologic and opioid health services for pain management in association with military readiness and health status outcomes: SUPIC renewal

Brandeis University, Waltham MA

Investigators

Linked publications & trials

Abstract

PROJECT SUMMARY Parent Grant Description This application is to supplement our work funded by the National Center for Complementary and Integrative Health (5R01 AT008404) that is examining suboptimal and evidence-based pain management in service members served by the Military Health System (MHS) and veterans served in the Veterans Health Administration (VHA). This innovative study contains data on over 1.66 million service members in all branches who have two years of observational data in the MHS between federal fiscal years 2017-2021. Chronic pain contributes to high rates of disability and healthcare burden in the United States. The Institute of Medicine estimated the annual cost burden is $635 billion from healthcare costs and lost productivity. The MHS and the VHA are not immune to the broad challenges of pain management witnessed in the civilian healthcare system. Service members experience frequent injuries, physical stressors, and comorbid psychological conditions; all of which jeopardize military readiness and require careful consideration in pain management approaches. Service members who have functional limitations are medically or administratively separated from active duty service, leaving them to continue their care in the VHA. Spurred by the significant personal and economic costs of inadequate pain management and over-reliance on opioids, the US healthcare system, MHS, and VHA joined together to publish the National Pain Strategy and updated clinical practice guidelines promote complementary, integrative, and other non-pharmacologic therapies (NPT) for pain management. The SUPIC project has been investigating changes in the MHS and VHA through observational studies with this overall goal: to advance knowledge on the military readiness and health status outcomes associated with different care trajectories that utilize nondrug pain management strategies. The Specific Aims of the SUPIC parent NCCIH grant are to: Aim 1: a) Characterize trajectories of nonpharmacological therapies utilization for pain management; b) examine relationships between NPT utilization (e.g., specific modalities, utilization trajectories) and multidimensional outcomes (e.g., military readiness, opiate utilization, health status); and c) study both trajectories and outcomes in discrete high-risk subgroups of patients with discrete comorbid mental health (posttraumatic stress disorder, depression) and pain conditions (e.g., back pain, other musculoskeletal, headache). Aim 2: Identify facility- and provider-level factors (full-time equivalent availability for health occupations, propensity to refer to NPT or prescribe opioids) that explain variation in NPT use and opioid prescribing between MHS facilities. Aim 3: a) Describe the demographic, clinical, and treatment history characteristics of individuals in the existing SUPIC cohort with chronic pain in the MHS who do and do not transition to VHA care; b) identify associations between NPT receipt in the MHS and long-term health outcomes in the VHA (main effects); and c) identify selected moderators and mediators of these associations. Goals of the Supplement The SUPIC project was developed to characterize pain management strategies utilized in the MHS and understand the association of optimal and suboptimal pain management strategies with downstream outcomes among US service members. However, one challenge to observational studies such as SUPIC is “missing variable bias” related to important contextual factors not available in these data. Specifically, information on community features representing social determinants of health (SDOH) such as sociodemographic, housing, delivery system characteristics, unemployment, and income are lacking. These structural and community features influence one another, as well as individuals within communities, but are absent in most observational healthcare studies. The goal of this supplement is to extend the SUPIC project’s current analysis by augmenting our data with social indicators of health and make the data ready for future machine-learning (ML) applications by forming clusters based on SDOH and appending ML-ready indicators to be explored in future applications using ML approaches. The specific aims of this supplement are to: Aim 1. Create ML-ready datasets of community-level social determinant of health (SDOH) measures, suitable for geographically linking with other data resources such as claims data. We will acquire, subset, and clean the Agency for Healthcare Research and Quality (AHRQ) Social Determinants of Health Data and enhance its usefulness for ML analyses through two steps. First, using agglomerative hierarchical clustering we will create summary SDOH taxonomy measures for each of the three levels of available geography (county, ZIP, census tract). These measures, suitable for attaching to more granular data sources, will provide low-dimensional summaries of community SDOH characteristics. Second, we will generate power terms and interactions among the SDOH features up to order of magnitude 3, as well as interactions with the cluster indicators to facilitate downstream ML analyses. Aim 2. Analyze, interpret and document the clusters created under Aim 1 to promote transparency and ensure their utility for downstream researchers. To characterize the composition of the clusters we will perform descriptive analyses of the features within each cluster and compare these features across clusters. In addition, we will estimate a multinomial logit model of cluster membership to identify the features that are most associated with each of the clusters. To enhance the transparency and usefulness of the data, we will create a handbook documenting the specifics of the cluster analyses, profiling the clusters, naming and interpreting them, and characterizing their spatial distribution. Aim 3. Demonstrate the feasibility and value of linking the community-level SDOH database with claims data by linking the SDOH and SUPIC databases. We will geographically link the SDOH data to SUPIC claims data and illustrate the application of a ML estimation algorithm that draws on data across ecological levels. Among military members who are being treated for pain management, we will conduct an exploratory analysis of the risk of an adverse outcome (defined as primary diagnosis of opioid use disorder, emergency department diagnosis of self-harm event, or opioid/other drug overdose) using Lasso regression on one-half the SUPIC data. We will then use causal forests on the reserved half of the data to obtain a causal estimate of the effect of one of our ML predictors on adverse outcomes. The algorithms used, and results from, this application to claims data will be described in the ML-ready public use file data documentation. Aim 4. Deliver the ML-ready database and our extensive documentation to the Inter-university Consortium for Political and Social Research (ICPSR) data repository for public use. We will prepare and submit the enhanced SDOH dataset, along with documentation, to the ICPSR data repository. The database will provide county-, zip code-, and tract-level AHRQ SDOH features, and the cluster indicators from our analyses. Documentation will characterize the algorithms to create the SDOH clusters and describe and interpret the cluster indicators. Algorithms for application to claims data will be described. This unrestricted public data set will be ready to support future machine learning analyses with multiple populations in the US, in addition to future ML grants with the SUPIC military database. With a strong, established, and expandable research database of MHS data and an experienced team of investigators and ML experts, this supplement is poised to transform the SUPIC database into a ML-ready resource, combined with ML-ready social determinants of health indicators, which can be used to extend much needed work on pain management that can directly inform MHS leaders and facilitate informed clinical and organizational changes among other institutions as well.