Finite mixture modeling of change point processes to discover opioid prescribing patterns: A case study of automated reports and consolidated ordering system data

Abstract

Abstract One key challenge of evaluating the effectiveness of opioid-related policies is to determine whether policy changes are associated with opioid prescription and usage. We explored this issue by employing a finite mixture of change point processes applied to US state-level opioid prescription data from 2006 to 2014. We identified clusters of states based on patterns of opioid prescription dosage per capita. The produced partitionings demonstrate some degree of geographic proximity when examined on the US map. Among the four drugs examined - methadone, buprenorphine, oxycodone, and hydrocodone – change points are detected for all drug classes except hydrocodone.