Enhancing Identification of Opioid-involved Health Outcomes Using National Hospital Care Survey Data

Abstract

Purpose This report documents the development of the 2016 National Hospital Care Survey (NHCS) Enhanced Opioid Identification Algorithm, an algorithm that can be used to identify opioid-involved and opioid overdose hospital encounters. Additionally, the algorithm can be used to identify opioids and opioid antagonists that can be used to reverse opioid overdose (naloxone) and to treat opioid use disorder (naltrexone). Methods The Enhanced Opioid Identification Algorithm improves the methodology for identifying opioids in hospital records using natural language processing (NLP), including machine learning techniques, and medical codes captured in the 2016 NHCS. Before the development of the Enhanced Opioid Identification Algorithm, opioid-involved hospital encounters were identified solely by coded diagnosis fields. Diagnosis codes provide limited information about context in the hospital encounters and can miss opioid-involved encounters that are embedded in free text data, like hospital clinical notes. Results In the 2016 NHCS data, the enhanced algorithm identified 1,370,827 encounters involving the use of opioids and selected opioid antagonists. Approximately 20% of those encounters were identified exclusively by the NLP algorithm.