A methodology using unsupervised machine learning analyzed 11 million tweets filtered for commonly abused prescription opioid drugs Analyses identified 2.3 million tweets with content relevant to nonmedical use of prescription medications/drugs (NMUPD) Twitter content was associated with a high degree of discussion (approximately 80%) about polydrug abuse involving multiple types of substances The methodology can filter large volumes of twitter data with minimal human intervention to identify macro NMUPD themes and trends