Algorithms to Identify Nonmedical Opioid Use

Abstract

The rise in nonmedical opioid overdoses over the last two decades necessitates improved detection technologies. Manual opioid screening exams can exhibit excellent sensitivity for identifying the risk of opioid misuse but can be time-consuming. Algorithms can help doctors identify at-risk people. In the past, electronic health record (EHR)-based neural networks outperformed Drug Abuse Manual Screenings in sparse studies; however, recent data shows that it may perform as well or less than manual screenings. Herein, a discussion of several different manual screenings and recommendations is contained, along with suggestions for practice. A multi-algorithm approach using EHR yielded strong predictive values of opioid use disorder (OUD) over a large sample size. A POR (Proove Opiate Risk) algorithm provided a high sensitivity for categorizing the risk of opioid abuse within a small sample size. All established screening methods and algorithms reflected high sensitivity and positive predictive values. Neural networks based on EHR also showed significant effectiveness when corroborated with Drug Abuse Manual Screenings. This review highlights the potential of algorithms for reducing provider costs and improving the quality of care by identifying nonmedical opioid use (NMOU) and OUD. These tools can be combined with traditional clinical interviewing, and neural networks can be further refined while expanding EHR.

Keywords: Algorithm; Drug abuse; Drug abuse prediction; Drug addiction; Opioid abuse.