Characterization and analysis of high incidence of potentially unsafe prescribing of some Extended-Release opioid analgesics using Natural Language Processing of Electronic Health Records clinical notes

CERSI Collaborators: Molly Jeffery, PhD (Mayo Clinic) (PI), William Becker, MD (Yale), W. Michael Hooten, MD (Mayo Clinic), Hongfang Liu, PhD (Mayo Clinic)

FDA Collaborators: Catherine Dormitzer, PhD, Cynthia Kornegay, PhD, Judy Racoosin, MD, MPH

CERSI Subcontractors: Optum Labs- Aylin Altan, PhD, Christopher Hane, Vijay Nori, Louis Rumanes, Darshak Sanghavi, MD, Jane Sullivan

Project Start Date: August 2017
Project End Date: April 2020

Regulatory Science Challenge

As part of an effort to improve how opioid prescribing is assessed in Risk Evaluation and Mitigation Strategy (REMS) programs, FDA considered measuring how often an opioid analgesic that required a certain amount of prior exposure to lower dose opioid analgesics (i.e., opioid tolerance) was prescribed to patients that did not have that prior exposure (opioid non-tolerance). Both the initial study conducted by FDA using Medicare data, and a subsequent investigation using data from the Sentinel network, suggested a high incidence of potentially unsafe prescribing behavior of prescribing opioid analgesics at doses requiring tolerance to opioid non-tolerant patients. Because these studies relied solely on medical billing data, FDA desired further replication and validation of this finding in other datasets.

Project Description and Goals

First, using the OptumLabs medical billing data, this study sought to determine the frequency of opioid non-tolerant subjects who were inappropriately prescribed doses of any opioid analgesic labeled for opioid-tolerant patients only. Second, this study sought to provide insight into prescribing behaviors by using natural language processing to explore free text fields within the patients’ electronic health records among those identified as being non-opioid tolerant but who are receiving a higher dosage opioid prescription. This exploration provided insights into whether patients were correctly identified as not being opioid tolerant and whether physicians may have offered reasons for prescribing opioids at higher dosages to opioid non-tolerant patients.

The project team found that more than half of patients initiating these medications did not have evidence of prior opioid tolerance, suggesting they were at increased risk of opioid-related harms, including fatal overdose. Data from EHRs did not contribute substantial additional evidence of opioid tolerance beyond the data found in prescription claims. Little evidence was found in EHR notes to suggest the reason the clinician prescribed the medication to opioid non-tolerant patients.

Accomplishments

Results were presented by Dr. Molly Jeffery at the FDA Joint Drug Safety and Risk Management Advisory Committee and Anesthetic and Analgesic Drug Products Advisory Committee meeting on August 3, 2018 during a presentation titled “Characterization of Potentially Unsafe Prescribing of Opioid Analgesics Requiring Prior Opioid Tolerance (OTOs)”. These findings contributed to regulatory action to modify the TIRF REMS.

Project catalyzed future research (B9a).

Publications

Jeffery MM, Chaisson CE, Hane C, Rumanes L, Tucker J, Hang L, McCoy R, Chen CL, Bicket MC, Hooten WM, Larochelle M, Becker WC, Kornegay C, Racoosin JA, Sanghavi D. Assessment of potentially inappropriate prescribing of opioid analgesics requiring prior opioid tolerance. JAMA Netw Open. 2020;3(4):e202875. doi:10.1001/jamanetworkopen.2020.2875