Prediction of Opioid Use Disorder (OUD) or Opioid Overdose using Artificial Intelligence (AI): a Systematic Review

Background and aims: Opioid Use Disorder (OUD), defined as physical or psychological dependenceon opioids, has been a global obstacle for decades and has become a major issue withthe rise in opioid overdose deaths caused by OUD. Predicting the possibility of an overdose oridentifying people at risk of developing OUD can be a significant step toward solving this issue.Artificial Intelligence (AI) is any system that comprehends the environment and can take actionbased on the information taken to maximize the chance of achieving the goal. Artificial intelligencemodels have recently been developed to predict these cases based on data from previouspatients. Clinical data of patients with OUD and overdose history has been statistically analyzedto predict the risk of opioid use adverse events in specific populations. However, statisticalanalysis cannot predict the risk of adverse events in a precise patient due to differences betweenindividuals of the same population. To indicate this risk in an individual, AI algorithms have beenutilized in recent studies.Method: A comprehensive systematic literature search was conducted in electronic databases, includingPubMed, Scopus, Embase, and Google Scholar, up to October ۲۰۲۲. Relevant keywordssuch as “adverse opioid reaction,» «opioid overdose,» and «artificial intelligence» have beenused for this purpose. Two authors evaluated the retrieved publications independently. All studiesthat used AI models or algorithms to predict OUD or opioid overdose were included. Studies thatweren’t written in English or conference papers were excluded. Any study that met our inclusioncriteria was then critically appraised by two authors independently. Data such as AI algorithmsand databases used in studies were extracted.Results: ۲۰۷۱ relevant publications were retrieved from electronic databases. After thoroughlyexamining the titles and abstracts and removing duplicates (n=۶۲۴), ۲۷ studies remained. Fulltexts of these articles were reviewed, and ultimately ۱۱ studies were included in our review. MachineLearning (ML) was used in eight of these studies, Deep Learning (DL) in two, and NaturalLanguage Processing (NLP) in one. Most of the included articles were related to the last ۱۰ years.The risk of OUD or overdose and also the harmless dose of opioids were predicted using AI models.Some studies evaluated early detection of OUD or overdose.Conclusion: Predicting OUD and its overdose not only saves lives but can also help countries›security. Using AI to predict the risk of OUD and overdose is a new step, and standard AI algorithmsare insufficient. Changes are required for these models to be entirely suitable for this purpose.AI algorithms have shown promising performance in maintaining big data and providing analmost exact prediction of adverse outcomes of opioid use. However, the morality of OUD-specificAI interventions and the protection of personal health data has not been discussed adequately.These AI models have been used in industry and education but they haven’t emerged in medicaleras due to the insufficiency of AI models.

کلیدواژه ها:

artificial intelligence ، opioid overdose ، Opioid-Related Disorders ، Machine Learning ، Opioid