We created an operational definition of possible opioid shopping in US commercial health insurance data and examined its correlates.
The population consisted of 264,204 treatment courses in persons with a fill for an opioid or diuretic prescription in 2012 and a second within 18 months. We examined counts of prescribers and pharmacies and the numbers of fills and overlaps for ability to discriminate courses of opioids from diuretics, which were a negative control. The most discriminatory measure, indicating possible shopping behavior, was cross-tabulated against other prescriptions filled and diagnoses as found in insurance claims. The associations between claims characteristics and shopping behavior were assessed in a logistic regression.
A definition that classified possible “moderate” or “extensive” shopping when a person obtained drug through at least 3 practices and at least 3 pharmacies over 18 months was highly discriminatory between opioid and diuretic treatment. Overlaps between fills and number of fills did not improve the discrimination. Data from insurance claims strongly predicted moderate-to-extensive levels of possible shopping (c=0.82). Prominent among 20 significant predictors were: state of residence; amount of opioid dispensed; self-payment; use of nonspecialist prescribers; high use of anxiolytics, hypnotics, psychostimulants, and antipsychotics; and use of both immediate release and extended-release or long-acting opioids.
The use of ≥3 prescribing practices and ≥3 dispensing pharmacies over 18 months sharply discriminated courses of opioid treatment from courses of diuretics. This pattern of fills was additionally associated with the numbers of nonspecialist and self-paid fills, the total morphine milligram equivalents dispensed, and heavier use of drugs for anxiety, sleep, attention, and psychosis.
Supplemental Digital Content is available in the text.
*WHISCON, Newton, MA
†Janssen Research and Development, Titusville, NJ
‡IMS Health, Plymouth Meeting, PA
Supported by the Opioid PMR Consortium, which is comprised of companies that hold NDAs of extended-release and long-acting analgesics, working in response to collective post-marketing requirements from the US Food and Drug Administration (http://www.fda.gov/Drugs/DrugSafety/InformationbyDrugClass/ucm363722.htm). The NDA holders include Allergan Sales, LLC (Parsippany, NJ), Collegium Pharmaceuticals, Inc. (Canton, MA), Daiichi Sankyo, Inc. (Edison, NJ), Depomed, Inc. (Newark, CA), Endo Pharmaceuticals, Inc. (Malvern, PA), Janssen Pharmaceuticals, Inc. (Titusville, NJ), Mallinckrodt, Inc. (Hazelwood, MO), Pernix Therapeutics Holdings, Inc. (Morristown, NJ), Pfizer, Inc. (NY, NY), Purdue Pharma, L. P. (Stamford, CT), and West-Ward Pharmaceuticals Corp. (Columbus, OH). A.M.W. and L.B.W. are employees of WHISCON. M.S.C. is an employee of Janssen Research and Development. D.B. and Y.Y. are employees of IMS Health. The study was designed in collaboration between OPC members and the investigators with input from FDA. The investigators maintained intellectual freedom in terms of publishing results. The data on which this analysis is based were prepared by IMS Health, using IMS Health's proprietary information. The IMS team that prepared the study files consisted of the IMS coauthors and Stephanie Roy, Heather Preston, Michel Denarie, and Goksu Dogan. Dr. Peter Kreiner of Brandeis University made available to the study his personally collected data on the characteristics of state prescription drug monitoring programs.
This study was registered with ClinicalTrials.gov as NCT02668549 on January 20, 2016.
Reprints: Alexander M. Walker, MD, DrPH, World Health Information Science Consultants, 275 Grove Street, Newton, MA 02466 (e-mail: firstname.lastname@example.org).
Received July 29, 2016
Received in revised form November 29, 2016
Accepted January 24, 2017