Predicting AOD Relapse and Treatment Completion from Social Media Use

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): Public health research and practice have not yet taken advantage of emerging changes in communication media even though methods and tools to analyze social media interactions have been developed and successfully used in marketing and business to better target prospective customers, tailor products, and predict behavior. This research will adapt these tools and take important steps in advancing science in ways that have potential to be used to improve public health. These adaptations will allow processing and meaningful interpretation of large volumes of social media data generated by individuals in Alcohol or Drug Abuse (AOD) treatment and allow us to address 3 aims. 1) It will allow use of this data source for identifying social media content that might be used to identify individuals who are at high risk for substance use relapse and treatment dropout; 2) It will provide a description of the frequency and patterns of AOD patients' public dialogue on social media with respect to topics such as alcohol and drug information and use, as well as treatment information; and 3) It will help to identify the best social media platforms to reach individuals i AOD treatment. Research staff will recruit 1,000 patients entering drug-free outpatient AOD treatment from 4 community based substance abuse treatment programs (a total of 11 sites). Participants will complete an intake battery and survey of their social media use, report weekly on their alcohol and drug use, give permission to extract treatment entry and discharge data from their clinic records, and to extract data from their Facebook and Twitter accounts. To address the first aim, social media data will be analyzed using Differential Language Analysis (DLA), an open-vocabulary technique that does not rely on pre-conceived theories regarding reasons for relapse and treatment dropout, but allows the data itself to drive an inclusive exploration of language. It finds words, phrases, and topics and presents them using word clouds, but unlike most word clouds, which scale words by their frequency, DLA scales words according to the strength of the relationship between the word or phrase and the variable tested. This open-vocabulary approach has excellent potential to reveal new insights to aid our understanding of risk factors, attitudes, and behaviors associated with relapse and treatment dropout. Eventually this information could be used to generate algorithms in the development of social media applications that would provide additional support for individuals when they are at risk for relapse and treatment dropout, or provide deserved acknowledgement for efforts when patients are fully engaged in treatment. Identifying factors that adversely affect treatment retention and sustained recovery is imperative. Less than 45% of the patients who enter treatment complete it and relapse rates have been reported as high as 92% at 12 months, with most relapsing within 3 months. Identifying social interactions that predict treatment dropout or substance use and automatically sending messages to intervene before that happens could improve and extend the lives of the 22.2 million drug-dependent individuals in the US.