Advancing Knowledge Discovery for Postoperative Pain Management

点击翻译按钮获取中文摘要

ABSTRACT Surgery is common and appropriate postoperative pain management is critical as poor management can impair recovery and lead to adverse events, including prolonged opioid use and transition to chronic pain. Literature suggests significant disparities exist with regard to pain management and its quality-of-life impacts, particularly among vulnerable populations (e.g. depressed, obese and diabetics). However, there lacks risk stratification tools to identify individuals at high risk for these disparate pain outcomes. Although pain scores are routinely collected in electronic health records (EHRs), shared algorithms to utilize them for care improvement are limited. To advance the efficient and effective use of the abundant amount of electronic data now available, a common data model (CDM) is necessary: standardized structures, terminologies, and rules to represent EHR data. Using a CMD for postoperative pain research would facilitate timely evidence generation across multiple populations and settings, which can provide critical evidence to stakeholders and move the field away from pain treatment for the ‘average’ patient to pain treatment for an individual. In this grant, we propose an innovative approach to advance the systematic analysis of postoperative pain across populations. Our approach will leverage the Observational Medical Outcomes Partnership (OMOP) CDM to develop tools that use standardize data formats and naming conventions; OMOP has over 140 collaborating sites gloablly. We will further utilize analytical tools developed by Observational Health Data Sciences and Informatics (OHDSI) on this CDM to facilitate disseminate across the research community. Our approach will develop scalable, open source risk stratification tools for adverse pain outcomes across diverse populations. We will accomplish this work in three aims. First, we will develop clinical phenotypes to identify and extract key discriminating features necessary to assess postoperative pain using EHRs. Next, we will develop pain risk stratification models using machine learning, including deep learning, methods and tools based on phenotypes developed in Aim 1. Finally, we will validate our models externally at the VA and disseminate our work through open source libraries and public websites. This project will deliver validated risk-stratification tools derived from real world evidence to identify patients at high risk for adverse pain outcomes following surgery, which can potentially reduce prescribed opioids circulating in the community– a key to curbing the opioid epidemic.