Prescription Drug Monitoring Programs: A Policy with Limited Impact on the Opioid Painkiller Epidemic Courtney R. Yarbrough Ph.D. Candidate University of Georgia Department of Public Administration and Policy Supported by a grant from the Robert Wood Johnson Foundation’s Public Health Law Research program (#72227)
Research Question • What is the effect of prescription drug monitoring programs (PDMP) on prescribing for opioid and nonopioid analgesics through the Medicare Part D program? • Difference-in-differences estimation • Physician-level prescribing, 2010-2013 • Looking at days supply of prescriptions for opioids and nonopioid pain relievers, oxycodone, hydrocodone, and DEA Schedules II-IV
Findings • No significant changes in overall opioid prescribing • 6.3% decrease in days supply prescribed per physician for oxycodone • Evidence of substitution toward Schedule IV opioids and nonopioid analgesics • State statutes that explicitly do not require physician access neutralize the effects of PDMPs.
Rates of prescription painkiller sales, deaths and substance abuse treatment admissions (1999-2010) Source: Centers for Disease Control and Prevention MOTIVATION: THE OPIOID EPIDEMIC
The Opioid Epidemic • In 2012, U.S. patients filled 259 million prescriptions for opioid painkillers, enough to medicate every American adult for a month. • The U.S. consumes 80% of opioid painkillers in the world (99% of hydrocodone). • 1.9 million Americans have an opioid painkiller substance abuse disorder. More than 4 million use the drugs non-medically. • In 2014, there were almost 19,000 deaths related to opioid painkiller overdose.
Source: New York Times, Jan. 16, 2016
Prescription Drug Monitoring Programs • Forty-nine states have now enacted PDMPs as a primary response to prescription painkiller abuse. • Online databases collect dispensing data from pharmacies about prescriptions filled for controlled substances. • Physicians can consult the PDMP to see if a patient has multiple, overlapping prescriptions. • PDMPs help uncover doctor shopping behavior by providing physicians with a tool to verify a patient’s drug-seeking behavior.
Prescription Drug Monitoring Programs • They vary state-to-state in operational details. • Unsolicited reports (41) • Reporting frequency (daily - 21) • Registration requirements (21) • Statutes explicitly not requiring access (16) • Mandatory access (15)
PDMP Literature • Few studies have systematically studied the effects of PDMPs. • The literature presents conflicting results on PDMP effectiveness. • Most focus on ecological measures of outcomes such as opioid-related deaths or treatment admissions at the state-level. • Few contend with endogeneity of policy adoption.
Contribution • Observes individual-level responses to PDMPs by the policies’ intended targets— physicians • Examines possible switching between opioid and nonopioid pain treatments • Measures effects on the most commonly abused opioids — oxycodone and hydrocodone • Uses difference-in-differences estimator to help control endogeneity of policy adoption
Prescription Drug Monitoring Programs • For this study, I consider a state to have a PDMP in time t if: 1. Dispensers are required to report. 2. Physicians have access. 3. The database is available online. • I use a proportional value of PDMP if the program was implemented in time t . • I also measure if a state has a statute explicitly not requiring physician PDMP access.
Online PDMP Implementation Pre-2010: 29 States
Online PDMP Implementation Pre-2010: 29 States 2010: MA (Dec.)
Online PDMP Implementation Pre-2010: 29 States 2010: MA (Dec.) 2011: FL, KS*, OR*
Online PDMP Implementation Pre-2010: 29 States 2012: AK*, DE, MT, NJ*, RI, SD*, TX, WA 2010: MA (Dec.) 2011: FL, KS*, OR*
Online PDMP Implementation Pre-2010: 29 States 2012: AK*, DE, MT, NJ*, RI, SD*, TX, WA 2010: MA (Dec.) 2013: AR, GA*, WI*, WY* 2011: FL, KS*, OR*
Online PDMP Implementation Pre-2010: 29 States 2012: AK*, DE, MT, NJ*, RI, SD*, TX, WA 2010: MA (Dec.) 2013: AR, GA*, WI*, WY* 2011: FL, KS*, OR* Control: MD, MO, NE, NH, PA
Data – Dependent Variables • ProPublica Prescriber Checkup database (2010-2012) and Centers for Medicare and Medicaid Services (2013) • Number of prescriptions filled through Medicare Part D at the drug-provider-year level • All providers included with at least 50 Part D fills per year • Drugs suppressed if < 10 • Aggregated according to drug categories (from Medicare Formulary Reference File) to form 7 DVs
Data – Dependent Variables • Logged days supply of a physicians prescribing that is for: 1. Opioid painkillers 2. Nonopioid painkillers 3. Hydrocodone 4. Oxycodone 5. Schedule II Opioids (including oxycodone) 6. Schedule III Opioids (including hydrocodone) 7. Schedule IV Opioids (e.g., tramadol)
Data – Independent Variables • State-Level • PDMP (proportional [0,1]) • “No Required Access” Statute (proportional [0,1]) • County-Level • Part D Enrollment • Per Capita Medicare Costs • Percent of Population White, Black, Hispanic, Asian, and Other • Median Income • HHI of Physician Prescribing • Provider-Level • Provider Sex • Medical Specialty Dummies • State Fixed Effects • Year Fixed Effects
Empirical Model • OLS models with state and year fixed effects • n = 789,569 at the physician-year level • Excluding the 29 states with PDMPs prior to 2011 and including state and year fixed effects creates a difference-in-differences framework.
Pre-trend Analysis
Results PDMP Statute PDMP
Conclusion: A Limited Effect for PDMPs • PDMPs do not appear to decrease physician prescribing of opioid painkillers overall. • They have a small but targeted effect with respect to the high-profile drug oxycodone. • Back-of-the-envelope calculation shows a decrease of ~104 days supply per doctor. • Statutes explicitly not requiring physician use of a PDMP have the effect of reversing these reductions.
Conclusion: A Limited Effect for PDMPs • Hydrocodone prescribing seems to remain unchanged, despite also being heavily abused. • Small substitution effects from Schedule II to Schedule IV drugs and nonopioids analgesics might prevent some adverse effects of opioid use. • PDMPs have in recent years shown only limited success in reducing opioid prescribing, suggesting that they need to be strengthened and/or additional policy tools are required.
Next Steps • Exploit other variation in PDMP characteristics to understand what works (e.g., mandates and registration requirements) • Measure the effect of PDMPs for the prescribing outliers • Analyze the relationship between PDMPs and individual pain management
Limitations • Studies using claims data outside Medicare may arrive at different results. • DID models control for unobservable time-invariant sources of endogeneity; however, time-variation sources may persist. • Other policy changes related to opioid abuse prevention are not included (e.g., Pill Mill legislation). • If many patients are crossing state lines to non- PDMP states, the models may overestimate size of the effect.
Thank you COURTNEY R. YARBROUGH UNIVERSITY OF GEORGIA cryarb@uga.edu
Medicare Part D Data • Beneficiaries include age-eligible (65+ YO) and disability- eligible (from SSDI) individuals. • 1/3 of all beneficiaries had ≥ 1 opioid prescription. • 25% of observations in data were from claims by disability- eligible patients. 44% of disabled beneficiaries had ≥ 1 opioid prescription; 23% were chronic users. 1/3 have a musculoskeletal diagnosis (e.g., back pain). • MedPAC found evidence for 170,000 cases of doctor shopping in 2008 claims. • Inpatient hospital stays increased 10.6% annually among Medicare patients 1993-2012.
Source: Social Security Administration Credit: Lam Thuy Vo/NPR
Opioids: A Gateway Drug • Heroin poisoning deaths have tripled since 2010 (10,574 in 2014). • Both opioid painkillers and heroin are opiates and operate through similar channels on the brain, producing comparable euphoria. • 80% of new heroin users are previous abusers of opioid painkillers. • Users report transitioning to heroin because the drug is much less expensive and more accessible than prescription opioids. • Suspicions that opioid-abuse policies have driven to rise in heroin use appear to unsubstantiated (Compton, Jones & Baldwin, 2016).
PDMP Literature • Simeone and Holland (2006) find a decrease in per capita supply of opioids but no change in treatment admission. • Reifler et al. (2012) find slower growth in opioid overdoses and treatment admission in PDMP states from 2003 to 2009. • Paulozzi, Kilbourne & Desai (2011) find insignificant effects of PDMPs on overdose mortality or opioid consumption rates but find evidence of switching between Schedule II and Schedule III opioids. • Radakrishnan (2014) finds decreased abuse of oxycodone on the intensive margin and fewer SUD treatment admissions but no effect for deaths, overall opioid abuse, or heroin abuse. • Rutkow et al. (2015) observe significant but modest decreases in opioid sales in Florida after the state’s implementation of a PDMP and Pill Mill regulations.
Prescriber Checkup Dataset
Results
Recommend
More recommend
