Identifying individuals amenable to drug recovery interventions through computational analysis of addiction content in social media

摘要

Drug abuse and addiction is a growing epidemic at the forefront of public health. Within this remit, the illicit use of opioid analgesics alone has emerged as one of the fastest growing forms of drug abuse in the U.S. and the death rate from this epidemic are drawing comparison to the US AIDS epidemic. Traditional methods of epidemiology based on explicit reporting of indicator-based data from patient records or data collected through surveys is often found wanting in modeling and designing effective interventions for addiction. In addition to the non-real time nature of the aforementioned methods, this is also due to a number of reasons including the continual penetration of novel biological/chemical entities into the abuse-cycle, the complex etiology of addiction which includes among others social factors, and the multistage nature of the addiction process. The recent advent of social media presents an intriguing information resource that is free from some of the above deficiencies and may be leveraged to model the addiction process and offer perspectives that are unavailable through traditional methods of epidemiology. In this paper, we use addiction related social media content to design a computational epidemiological approach for predicting a user's propensity for seeking drug recovery interventions. Solving this problem is crucial for designing effective interventions, identifying cohorts who would be most amenable to recovery, and planning resource allocations. Our method characterizes the evolving language of drug use, identifies the interactions that influence a drug user's actions, and using machine learning techniques predicts the extent to which a user is likely to participate in addiction recovery communities. Experimental assessments on real-world data from the social media platforms Reddit and Twitter indicate the proposed method can identify users who are amenable to addiction recovery intervention with high precision, recall, and F1 values.