Improving the quality of pain management is a high priority for the Veterans Health Administration (VHA). VHA has published policy guidance that establishes an innovative stepped care model of pain management (SCM-PM) as the single standard of pain care. Other than pharmacological and procedure based interventions in which specific, easily retrievable codes are used to document care in the VHA's electronic health record (EHR), it is difficult to capture the broader array of CHA or key aspects of integrated care.
The overall objective of the project is to develop a method of assessing quality, state-of-the-art pain care (referred to as Pain Care Quality in this application) in an integrated healthcare system. Aim 1. To identify and quantify empirically-derived, key dimensions of Pain Care Quality (as described above) in veterans with musculoskeletal pain. This will involve the development of NLP and ML tools to extract pain management data from structured fields (coded) and unstructured fields (clinician text notes) in the EHR. Aim 2. To assess factors associated with Pain Care Quality in a nationally representative sample of veterans with musculoskeletal pain. Aim 3. To determine whether VHA facilities that have adopted the SCM-PM provide higher quality pain care.
The proposed project extends prior research by our investigator team by using Natural Language Processing (NLP) and Machine Learning (ML) to automate a previously validated approach to identify and quantify key dimensions of Pain Care Quality, namely assessment, especially functional assessment, integrated treatment plans, reassessment (outcomes), and patient education from the EHR. We will search the Musculoskeletal Diagnoses Cohort (MSD), (CRE 12-012) a database of nearly 6 million veterans, to identify a representative sample of veterans with MSD. Three separate annotation tasks will be done in support of this study. The first annotation task will support the creation of a ML model to help select progress notes that are rich in information about pain management for subsequent annotation efforts. The second and third annotation tasks will provide specific examples of text to develop and refine the NLP system. Once this automated solution is validated, we intend to apply it to a national sample to test important questions about Pain Care Quality among veterans with comorbid mental health conditions, access to CHA, and the SCM-PM.
the annotation process is complete and the analysis is ongoing.
This innovative solution to identifying key dimensions of healthcare has potential applicability to improving the management of other complex health problems for which existing quality of care indicators and metrics are limited. This is a high priority area for the VHA.
External Links for this Project
- Luther SL. Pain Care Quality and Integrated and Complementary Health Approaches. Spotlight on Pain Management [Cyberseminar]. Health Services Research and Development. 2016 Sep 6. [view]