TY - JOUR
T1 - Using Prescription Drug Monitoring Program Data to Assess Likelihood of Incident Long-Term Opioid Use
T2 - a Statewide Cohort Study
AU - Henry, Stephen G.
AU - Stewart, Susan L.
AU - Murphy, Eryn
AU - Tseregounis, Iraklis Erik
AU - Crawford, Andrew J.
AU - Shev, Aaron B.
AU - Gasper, James J.
AU - Tancredi, Daniel J.
AU - Cerdá, Magdalena
AU - Marshall, Brandon D.L.
AU - Wintemute, Garen J.
N1 - Funding Information:
This study was supported by the National Institutes of Health grant R01DA044282.
Publisher Copyright:
© 2021, Society of General Internal Medicine.
PY - 2021
Y1 - 2021
N2 - Background: Limiting the incidence of opioid-naïve patients who transition to long-term opioid use (i.e., continual use for > 90 days) is a key strategy for reducing opioid-related harms. Objective: To identify variables constructed from data routinely collected by prescription drug monitoring programs that are associated with opioid-naïve patients’ likelihood of transitioning to long-term use after an initial opioid prescription. Design: Statewide cohort study using prescription drug monitoring program data Participants: All opioid-naïve patients in California (no opioid prescriptions within the prior 2 years) age ≥ 12 years prescribed an initial oral opioid analgesic from 2010 to 2017. Methods and Main Measures: Multiple logistic regression models using variables constructed from prescription drug monitoring program data through the day of each patient’s initial opioid prescription, and, alternatively, data available up to 30 and 60 days after the initial prescription were constructed to identify probability of transition to long-term use. Model fit was determined by the area under the receiver operating characteristic curve (C-statistic). Key Results: Among 30,569,125 episodes of patients receiving new opioid prescriptions, 1,809,750 (5.9%) resulted in long-term use. Variables with the highest adjusted odds ratios included concurrent benzodiazepine use, ≥ 2 unique prescribers, and receipt of non-pill, non-liquid formulations. C-statistics for the day 0, day 30, and day 60 models were 0.81, 0.88, and 0.94, respectively. Models assessing opioid dose using the number of pills prescribed had greater discriminative capacity than those using milligram morphine equivalents. Conclusions: Data routinely collected by prescription drug monitoring programs can be used to identify patients who are likely to develop long-term use. Guidelines for new opioid prescriptions based on pill counts may be simpler and more clinically useful than guidelines based on days’ supply or milligram morphine equivalents.
AB - Background: Limiting the incidence of opioid-naïve patients who transition to long-term opioid use (i.e., continual use for > 90 days) is a key strategy for reducing opioid-related harms. Objective: To identify variables constructed from data routinely collected by prescription drug monitoring programs that are associated with opioid-naïve patients’ likelihood of transitioning to long-term use after an initial opioid prescription. Design: Statewide cohort study using prescription drug monitoring program data Participants: All opioid-naïve patients in California (no opioid prescriptions within the prior 2 years) age ≥ 12 years prescribed an initial oral opioid analgesic from 2010 to 2017. Methods and Main Measures: Multiple logistic regression models using variables constructed from prescription drug monitoring program data through the day of each patient’s initial opioid prescription, and, alternatively, data available up to 30 and 60 days after the initial prescription were constructed to identify probability of transition to long-term use. Model fit was determined by the area under the receiver operating characteristic curve (C-statistic). Key Results: Among 30,569,125 episodes of patients receiving new opioid prescriptions, 1,809,750 (5.9%) resulted in long-term use. Variables with the highest adjusted odds ratios included concurrent benzodiazepine use, ≥ 2 unique prescribers, and receipt of non-pill, non-liquid formulations. C-statistics for the day 0, day 30, and day 60 models were 0.81, 0.88, and 0.94, respectively. Models assessing opioid dose using the number of pills prescribed had greater discriminative capacity than those using milligram morphine equivalents. Conclusions: Data routinely collected by prescription drug monitoring programs can be used to identify patients who are likely to develop long-term use. Guidelines for new opioid prescriptions based on pill counts may be simpler and more clinically useful than guidelines based on days’ supply or milligram morphine equivalents.
KW - health policy
KW - long-term opioid use
KW - opioid analgesics
KW - pain
KW - prescription drug monitoring programs
UR - http://www.scopus.com/inward/record.url?scp=85103131093&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85103131093&partnerID=8YFLogxK
U2 - 10.1007/s11606-020-06555-x
DO - 10.1007/s11606-020-06555-x
M3 - Article
C2 - 33742304
AN - SCOPUS:85103131093
JO - Journal of General Internal Medicine
JF - Journal of General Internal Medicine
SN - 0884-8734
ER -