Cross-Sectional Study to Define and Validate "Doctor/Pharmacy Shopping" as Outcomes Suggestive of Abuse and/or Addiction

Based on a review of the literature, the Food and Drug Administration (FDA) concluded that more data are needed regarding the serious risks of misuse, abuse, addiction, overdose, and death associated with the long-term use of extended release/long acting (ER/LA) opioid analgesics. Thus, the FDA is requiring that ER/LA opioid analgesic drug sponsors conduct post-marketing studies to assess these risks. The four observational post-marketing requirement (PMR) studies are labeled Study #2065-1, Study #2065-2, Study #2065-3, and Study #2065-4.

The objective of PMR Study #2065-4 is to define and validate "doctor/pharmacy shopping" as outcomes suggestive of misuse, abuse and/or addiction.

Study #2065-4 consists of three sub-studies, Study 4A, Study 4B, and Study 4C. In the current study (#2065 sub-study, Study 4A), the IMS® LRx database, a longitudinal pharmacy database that captures both third-party and self-pay transactions, will be linked to the PharMetrics Plus™ database, which contains provider, facility and pharmacy claims. From these databases, patients with shopping behavior will be identified and evaluated against an algorithm for abuse/addiction that will be developed and validated in PMR Study #2065-3B. Study #2065-3B will review the literature for health information suggestive of abuse/addiction in medical charts and medical claims. Then, through iterative processes that involve experts reviewing claims profiles and medical charts, text searches of medical records using natural language processing and supervised machine learning will identify and create a compilation of codes and health information with the best specificity and sensitivity to identify patients with opioid abuse/addiction. The compilation of codes and health information will be used to develop the algorithm that Study 4A will apply to define abuse/addiction. Restricting the validation of shopping behavior to only patients with diagnoses related to abuse would ignore the population with undiagnosed abuse. By supplementing ICD-9 codes with other information available on claims (e.g., addiction treatments, emergency visits) to define abuse/addiction, the measurement bias will be decreased and the estimates of the association of shopping behavior with abuse/addiction will be more valid. Categories of doctor shopping will be derived from the population distributions of the multiple variables analyzed for opioid users. The 4 categories will range from no shopping behavior (lowest level) to extensive shopping behavior (most extreme). The specifics of how the categories will be defined will be driven by the data.