i Final Contract Report Approaches to Drug Overdose Prevention Analytical Tool (ADOPT): Evaluating Cost and Health Impacts of a Medicaid Patient Review & Restriction Program Prepared for Office of the Associate Director for Policy Policy Research, Analysis, and Development Office Centers for Disease Control and Prevention Grant No: 1U58CD001370-01 Prepared by Joy Melnikow, MD, MPH Zhuo Yang, MS Meghan Soulsby, MPH Dominique Ritley, MPH Kenneth Kizer, MD, MPH University of California, Davis Center for Healthcare Policy and Research December 2012 The views in this report are those of the authors. No official endorsement by the Center for Disease Control and Prevention is intended or should be inferred.
160
Embed
Approaches to Drug Overdose Prevention Analytical Tool (ADOPT ... Opioid... · patient review and restriction programs (PRR), sometimes referred to as patient “lock-in” programs.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
i
Final Contract Report
Approaches to Drug Overdose Prevention
Analytical Tool (ADOPT):
Evaluating Cost and Health Impacts of a
Medicaid Patient Review & Restriction
Program
Prepared for
Office of the Associate Director for Policy
Policy Research, Analysis, and Development Office
Centers for Disease Control and Prevention
Grant No: 1U58CD001370-01
Prepared by
Joy Melnikow, MD, MPH
Zhuo Yang, MS
Meghan Soulsby, MPH
Dominique Ritley, MPH
Kenneth Kizer, MD, MPH
University of California, Davis
Center for Healthcare Policy and Research
December 2012
The views in this report are those of the authors. No official endorsement by the Center for Disease Control and Prevention is
intended or should be inferred.
ii
Acknowledgments
The authors of this report greatly appreciate the contributions of our content expert, Dr. Barth
Wilsey, UC Davis, as well as CDC staff who gave of their time and expertise to assure correct
assumptions and clean data informed the micro-simulation model.
i
Introduction
Many assert that effective policies aimed at preventing the increasing abuse of prescription
opioids among Medicaid beneficiaries could save tens of millions of dollars by substantially
reducing associated mortality, morbidity and associated healthcare costs. However, the lack of
cost-effectiveness data gives rise to the need for evidence-based investigations. This report
presents the design and findings of a micro-simulation model that evaluates how certain policies
might reduce the misuse/abuse of prescription opioids in the Medicaid population, thus reducing
associated, preventable health care costs and outcomes. Use of illegal opioids (e.g., heroin),
prescriber fraud, and opioid diversion fall outside the scope of this project.
We developed this model in response to a CDC request to examine the effectiveness of Medicaid
patient review and restriction programs (PRR), sometimes referred to as patient “lock-in”
programs. The Approaches to Drug Overdose Prevention Analytical Tool (ADOPT) is an
evidence-based tool to help inform policy decisions regarding prescription drug overdose
prevention policies. This micro-simulation model simulates the prescription opioid behavior of
an adult Medicaid enrollee cohort to explore the impact and the cost-effectiveness of such
programs. By applying various PRR policies to the simulated prescription opioid behavior, users
can assess the cost and health impacts of the policies. ADOPT supports interactive features that
allow users to customize the population demographics and policy details, and performs a "what-
if" analysis to project the outcomes of a specified policy within that population. Although
ADOPT has the potential to analyze and compare different approaches to opioid overdose/abuse
prevention (such as prescriber/patient education or monitoring strategies), the current version
focuses on the Medicaid patient review and restriction (PRR) program. The model was informed
by an analysis of a MarketScan® Medicaid dataset and a literature review.
Report Summary
This report is divided into three primary sections: Parts One and Two, the literature review and
MarketScan® data analysis; and Part Three, which contains information from the two previous
sections that are used to inform the design of the model. Each section is written to stand alone;
however, readers are encouraged to read the report in its entirety to understand the context
surrounding the model.
Part One: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – A Review of the Literature
This literature review provides supporting material for the investigation into the health and
economic burden of prescription opioid misuse and abuse and state-level policies that may
reduce or eliminate these burdens, especially within the Medicaid population. This review
provides the necessary context to evaluate two state-level policies -- patient review and
restriction programs and prescription drug monitoring programs -- recommended by the White
House and CDC. Some data from this literature review are used to inform the cost-effectiveness
model developed for the CDC to examine the effectiveness of patient review and restriction
programs. The review includes a summary of the prevalence of prescription opioid misuse and
abuse, sources of opioids, and prescribing patterns at the state level and in the Medicaid
ii
population. It also includes studies about the health outcomes related to prescription opioid
misuse and abuse, the health care and societal costs attributable to misuse and abuse, and policy
options that may eliminate or curtail such misuse and abuse.
Part Two: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – MarketScan® Data Analysis
Part Two presents an analysis of MarketScan® Medicaid data using the 90-day exposure window
and the episode-based regression models to estimate the relationship between the risk of
overdose and daily opioid dose. The results from this analysis describe the characteristics of the
patient population, their prescription opioid use, and their risk of overdose events. It also
characterizes the pharmacy shopping behavior among the long term opioid users and their rates
of overdose. These data points were used to calibrate the micro-simulation model presented in
Part Three.
Part Three: Evidence-Based Tools for Promoting Health Policy and Disease Prevention -
Prescription Opioid Overdose
Part Three introduces The Approaches to Drug Overdose Prevention Analytical Tool (ADOPT)
This is an Excel-based micro-simulation model that simulates the patterns of Medicaid enrollees’
prescription opioid use in order to evaluate associated health outcomes and costs under different
restriction policies.* It compares the counterfactual scenarios of implementing a prescription
drug overdose/abuse prevention policy versus the absence of such a policy, and evaluates the
cost and health impact of the policy. The model’s interactive features allow users to customize
the population demographics and policy details, and perform a "what-if" analysis to project the
outcomes of the specified policy within that population. Although ADOPT has the potential to
analyze and compare different approaches to drug overdose/abuse prevention (such as
prescriber/patient education or monitoring strategies), the current version focuses on the
Medicaid patient review and restriction (PRR) program.
Key Findings
Key findings from the MarketScan® data analysis:
Higher doses of opioids are associated with an increased risk of overdose in the
Medicaid population, whereas the type of opioid drug, after adjusting for dose and
other risk factors, shows little effect on overdose risk.
Medicaid opioid users who exhibited pharmacy shopping (>=4 pharmacies in any 90
days) have higher (1.8-fold) risk of overdose than those who did not, even after
adjusting for dose and other risk factors.
Medicaid opioid users who had overlapping prescriptions (same drug type with more
than 25% overlapping supply days) have about 3-fold increase in overdose risk.
Overlapping prescriptions could be a meaningful indicator for the PRR program to
identify high risk patients.
Key findings from the ADOPT model:
* The model was informed by an analysis of the MarketScan® Medicaid dataset and literature review; these analyses
are presented in Sections 1 and 2, respectively, of this report.
iii
ADOPT can resemble the individual patterns of opioid use in the Medicaid
population to a satisfactory extent, though some details, such as modeling of drug
type may need further revision and calibration.
Based on our exploratory analysis, the less selective PRR program criteria, show a
greater overall reduction in prescription opioid use and overdose prevention, however
they have a small effect on the average prescription reduction and overdose
prevention per program enrollee. Conversely, more selective criteria are less
effective but more efficient in targeting the high-risk users. These programs are more
likely to identify those actually misusing or abusing prescription opioids, but they
require a large population pool to justify the investment on this program.
More state-specific input values are needed to conduct more relevant analysis.
Report Assumptions and Limitations
This model relies on a number of assumptions including:
Outcome probabilities, patterns of prescription opioid use, and drug pricing derived from
MarketScan® data are generalizable to individual state Medicaid programs
The fixed PRR program cost is $300,000 annually, and the variable cost is $200 per
program enrollee (representing the additional labor and material expenditures that
increase as the program caseload increases)†
PRR enrollees remain enrolled in Medicaid and the PRR program for the duration of the
policy period
Enrollees consumed prescriptions “as prescribed” – does not consider opioid diversion
Overdose risk is based on acquisition of prescription opioids, not use of illegal opioids
Characteristics of a subsequent episode of opioid use are correlated with those of a
previous episode
All PRR program enrollees’ overlapping prescriptions (i.e., two prescriptions of the same
drug type, one of which had a supply for 5 days or longer, overlapped by 25% or more of
the days prescribed) are eliminated in the scenario of having the PRR program
All PRR program enrollees’ prescriptions that contribute to an aggregate daily dose more
than 80mg morphine equivalent will be reduced to an aggregate daily dose of 80mg
morphine equivalent in the scenario of having the PRR program
A complete list of assumptions can be found in Part 3 of this report.
Additionally, this model has several limitations including:
1. Geographic variation: Although MarketScan® data comes from multiple states (12
states in 2012), it may not be representative of the national data. It is possible that in
certain states, the Medicaid opioid users behave differently than the MarketScan®
population – in which case the analysis may not be accurate.
2. Baseline scenario: Under-estimated prevalence of opioid abuse/misuse: ADOPT uses
the MarketScan® Medicaid dataset to simulate the scenario of not having a PRR
† Cost estimations are based on estimates from Oklahoma and Washington state PRR programs, per content expert
discussions with the CDC; Jones, C.M., Roy, K. Email correspondence, August 2012
iv
program, then identifies the subjects who meet the program enrollment criteria and
calculates the health and financial impact if the PRR was established. However, it is
possible that some states already had a PRR program when the MarketScan® data were
collected, in which case the prevalence of opioid abuse or misuse (including drug
shopping) would be under-estimated. This may cause an undervaluation of a PRR
program in the analysis.
3. Prescriber information is imputed: Many PRR programs use the number of opioid
prescribers as an eligibility criterion; however, the MarketScan® data do not contain
prescriber information. Therefore the model uses the reported correlation between
numbers of pharmacies and prescribers from the Massachusetts’ PRR program database.
It is possible that this correlation may not reflect the experience of the MarketScan®
population.
4. Incomplete representation of PRR criteria: The current version of ADOPT can only
analyze some of the criteria that may be used in a PRR program, but in practice, PRR
programs often use other criteria such as emergency department use, number of office
visits, or history of substance abuse. If data become available, they could be incorporated
into the model.
5. Uncertainty in estimation of overdose risk: ADOPT uses the hazard ratios for opioid
overdose that are derived from the MarketScan® inpatient and outpatient datasets.
However, overdose rates may be higher than observed because patients may have expired
before entering the hospital. In addition, overdose events were identified by using the
diagnostic codes. Misclassification of diagnostic codes may cause under-estimation or
over-estimation of the overdose risk.
6. Uncertainty of PRR program costs: The model uses a fixed program cost of $300,000
annually and a variable cost of $200/program enrollee, however, these costs may not
reflect the actual cost incurred by states operating PRR programs and they do not include
the start-up costs for states newly implementing programs.
7. Assumptions about effects of PRR programs may be inaccurate: For example,
assumptions that these programs reduce dosage or overlapping prescriptions.
Despite these limitations, the ADOPT model demonstrates the potential to simulate individual
prescription consumption behavior with satisfactory similarity to real prescription consumption
behavior based on calibration with MarketScan® data. Using the current model structure and
interface, it is possible to add new functions if and when future data becomes available.
Ultimately, the strength of the ADOPT is its ability to be customized with state-specific data,
which will produce more timely, accurate, and relevant conclusions than those reached using the
MarketScan® data. Policy makers now have the opportunity to introduce valid, evidence-based
information into their decision making process about state-specific patient review and restriction
programs to ensure that the most cost-effective policies target those enrollees who will benefit
the most.
v
Approaches to Drug Overdose Prevention
Analytical Tool (ADOPT):
Evaluating Cost and Health Impacts of a
Medicaid Patient Review & Restriction
Program
Part 1
Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – A Review of the Literature
Part 1 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Sources of Misused and Abused Opioids ............................................................................................ 1.12 Doctor and Pharmacy Shopping .......................................................................................................... 1.13
Risk Factors Associated with Prescription Opioid Misuse and Abuse .................................................... 1.22 Sociodemographic Risk Factors ....................................................................................................... 1.22 Behavioral Risk Factors ................................................................................................................... 1.23 Pain Level & Comorbidities ............................................................................................................ 1.24 Mental Health ................................................................................................................................... 1.24 Opioid Dose and Supply .................................................................................................................. 1.25
OUTCOMES RELATED TO OPIOID MISUSE AND ABUSE ............................................................ 1.26 Health Care Utilization ........................................................................................................................ 1.26
Characteristics Associated with Prescription Opioid-Related Healthcare Utilization ..................... 1.27 Health Outcomes: Opioid-Related Comorbidities ............................................................................... 1.28 Health Outcomes: Opioid-Related Mortality ....................................................................................... 1.31
Characteristics Associated with Opioid-Related Mortality .............................................................. 1.31 State-Specific Opioid-Related Mortality ......................................................................................... 1.33 Opioid-Related Mortality: Opioid Type and Dosing Patterns ......................................................... 1.33 Opioid-Related Mortality and Doctor/Pharmacy Shopping ............................................................. 1.36
Outcomes: Health Care Costs .............................................................................................................. 1.39 Health Care Costs and Doctor Shopping ......................................................................................... 1.42 State-Specific Health Care Costs ..................................................................................................... 1.43
Literature Review Sources ................................................................................................................... 1.58 Databases of Peer-Reviewed Literature ........................................................................................... 1.58 National Data Sources ...................................................................................................................... 1.58 Federally Maintained Sources of Grey Literature ............................................................................ 1.58 State Maintained Sources of Grey Literature ................................................................................... 1.58 Nonprofit Organizations .................................................................................................................. 1.58
Literature Review Search Terms.......................................................................................................... 1.59 BIBLIOGRAPHY .................................................................................................................................... 1.60
Part 1 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – A Review of the Literature
1.3
List of Tables
Table 1-1. Selected Policies and Programs to Reduce Prescription Drug Misuse and Abuse ................... 1.6
Table 1-2: Defining Prescription Drug Use Patterns ................................................................................. 1.8
and 2007 ................................................................................................................................................... 1.14
(MED) and morphine equivalent dose per day (MED/d) are used repeatedly in studies32-38
to
Nonmedical Use: Patients incorrectly using a
medication for which they had a prescription AND/OR use without a prescription belonging to the user
Misuse: Patients incorrectly
using a medication for which they had a
prescription
•High potential for abuse, no currently accepted medical use in the US. Includes illicit drugs such as heroin and marijuana.
Schedule I
•High potential for abuse leading to severe psychological or physical dependence. Includes prescription drugs such as oxycodone, as well as drugs such as cocaine.
Schedule II
•Less potential for abuse relative to Schedule II, leading to moderate or low physical dependence or high psychological dependence. Includes prescription drugs such as hydrocodone
Schedule III
•Low potential for abuse relative to Schedule III drugs. Includes prescription drugs such as diazepam.
Schedule IV
•Low potential for abuse relative to Schedule IV drugs. Includes limited quantities of narcotics (ex: containing no more than 200 milligrams of codeine per 100 milliliters or per 100 grams).
Butalbital + codeine (with or without aspirin/acetaminophen/ ibuprofen) 0.15
Dihydrocodeine (with or without aspirin/acetaminophen/ibuprofen) 0.25
Pentazocine (with or without aspirin/acetaminophen/ibuprofen) 0.37
Buprenorphinea
25.0-40.0
Butorphanol 7.00
Schedule II Short-Acting*
Morphine sulfate 1.00
Codeine sulfate 0.15
Oxycodone (with or without aspirin/acetaminophen/ibuprofen) 1.50
Hydromorphone 4.00
Meperidine hydrochloride 0.10
Oxymorphone 3.00
Fentanyl citrate transmucosalb 0.125
Tapendatol short actingc
not established
Schedule II Long-Acting*
Morphine sulfate sustained release 1.00
Fentanyl transdermald 2.40
Levorphanol tartrate 11.0
Oxycodone HCL control release 1.50
Methadone 3.00
Oxymorphone extended releasec
3.00
Hydromorphone extended releasec
5.00
Tapentadol extended releasec
not established
Source: Von Korff et al (2008); FDA Blueprint for Prescriber Education for Extended-Release and Long-Acting
Opioid Analgesics (2012)
Note: The majority of these conversation factors are based on Von Korff’s CONSORT (CONsortium to Study
Opioid Risks and Therapeutics) study. Opioids delivered by pill, capsule, liquid, transdermal patch, and
transmucosal administration were included in the data, but opioids formulated for administration by injection or
suppository were not included.
*Prescription opioids are classified as short- or long-acting based on their duration. Short-acting opoids result in a
more rapid increase and decrease in blood serum levels, where as long-acting opioids release gradually into the
bloodstream or have a long half-life for prolonged activity.39
aBuprenorphine is typically used for opioid detoxification and maintenance
40
bTransmucosal fentanyl conversion to morphine equivalents assumes 50% bioavailability of transmucosal fentanyl
and 100 micrograms transmucosal fentanyl is equivalent to 12.5 to 15 mg of oral morphine. cData for oxymorphone, hydromorphone and tapentadol obtained from FDA Blueprint for Prescriber Education for
Extended-Release and Long-Acting Opioid Analgesics
Part 1 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – A Review of the Literature
1.11
dTransdermal fentanyl conversion to morphine equivalents is based on the assumption that one patch delivers the
dispensed micrograms per hour over a 24 hour day and remains in place for 3 days.
Diversion. Diversion, as defined by the Uniform Controlled Substances Act, is the “transfer of a
controlled substance from a lawful to an unlawful channel of distribution of use.”41
Diversion
can occur in multiple forms, including theft, forgery, and illegal purchase, either from drug
dealers or illegal Internet pharmacies. An increasingly common form of diversion in some areas
of the United States are “pill mills,” which is a term used to describe a provider (physician, clinic
or pharmacy) that is inappropriately prescribing and/or dispensing prescription drugs.42
This
report focuses on two additional forms of diversion – doctor shopping and pharmacy shopping –
defined as using multiple physicians and/or multiple pharmacies to obtain prescription drugs
inappropriately.
BACKGROUND: PRESCRIPTION OPIOID MISUSE & ABUSE
Prevalence of Prescription Opioid Misuse, Abuse and Dependence
Over the past two decades, the prevalence of self-reported, nonmedical use and misuse of
prescription opioids has increased in the United States.43-47
According to the 2010 National
Survey on Drug Use and Health (NSDUH)†, the prevalence of past-year nonmedical prescription
opioid use was 4.8% (approximately 12.2 million individuals) and 2.0% (5.1 million individuals)
reported past-month nonmedical use.27
In 2010, nearly as many individuals admitted initiating
nonmedical opioid use within the past 12 months (2.0 million) as those initiating use of
recreational marijuana (2.4 million), the most commonly abused illicit drug.11
Since 2002, the
NSDUH has found that lifetime, past-year, and past-month nonmedical opioid use has remained
relatively stable (Table 1-4).27
Table 1-4. Prevalence of Nonmedical Prescription Opioid Use, 2002-2010
Year Lifetime
N (%)
Past Year
N (%)
Past Month
N (%)
2002 29, 611 (12.6) 10,992 (4.7) 4,377 (1.9)
2003 31,207 (13.1) 11,671 (4.9) 4,693 (2.0)
2004 31,768 (13.2) 11,256 (4.7) 4,404 (1.8)
2005 32,692 (13.4) 11,815 (4.9) 4,658 (1.9)
2006 33,472 (13.6) 12,649 (5.1) 5,220 (2.1)
2007 33,060 (13.3) 12,466 (5.0) 5,174 (2.1)
2008 34,861 (14.0) 11,885 (4.8) 4,747 (1.9)
2009 35,046 (13.9) 12,405 (4.9) 5,257 (2.1)
2010 34,776 (13.7) 12,213 (4.8) 5,100 (2.0)
Source: SAMHSA, Office of Applied Studies, National Survey on Drug Use and Health, 1999-2009
Note: Prevalence among individuals ages 12 and older. Numbers are in thousands.
Multiple iterations of the NSDUH,46
as well as surveys of patients hospitalized for opioid
withdrawal48
and enrolled into methadone maintenance programs49
have found that the most
commonly used opioids were compounds containing hydrocodone and oxycodone. Of those
reporting past-year nonmedical opioid use in the 2002-2004 NSDUH, Becker et al found that the
† Administered by the Substance Abuse and Mental Health Services Administration, this survey is based on a target
of 67,500 face-to-face interviews with a representative sample of civilian, non-institutionalized individuals ages 12
and older.
Part 1 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – A Review of the Literature
1.12
majority reported use of only one opioid (39%) or 2-4 different opioids (45%), while 16%
reported use of more than four different opioids.50
Results from a 1998-2006 telephone survey‡
found higher rates of concomitant non-opioid medications among “regular” opioid users, defined
as individuals “using at least 5 days per week for at least four continuous weeks.” Approximately
21% of regular opioid users reported taking 10 or more medications, compared to 16% of non-
regular users and 4.5% of non-opioid users. Among regular opioid users, 18% had used opioids
for more than 5 years, 47% used opioids for more than 2 years, and 21% had used opioids for
less than six months.51
Jones analyzed the 2010 NSDUH and found that approximately one
million individuals reported chronic nonmedical use (defined as use for 200 days or more),
which is a 75% increase from 2002-2003.52
As previously mentioned, the NSDUH has found
that the prevalence of nonmedical opioid use has remained relatively stable for the past decade,27
yet opioid-related deaths have increased significantly.13
Jones hypothesizes that the increase in
chronic nonmedical use may be one contributing factor to the increased mortality.52
State-Specific Prevalence Rates
From 2008-2009, the prevalence of nonmedical opioid use ranged from 3.6% (Nebraska) to 8.1%
(Oklahoma), compared to a national prevalence of 4.8%.1 In 2008, nearly 21% of respondents to
Utah’s BRFSS§ survey reported using at least one prescription opioid within the past year; of
those respondents, 3.2% reported nonmedical use of that prescription, either more frequently or
in higher doses than directed.53
Medicaid-Specific Prevalence Rates
Using 2002-2003 data from the Medicaid Analytic eXtract (MAX) database, McAdam-Marx et
al estimated the prevalence of prescription opioid abuse in the Medicaid population at 87 per
10,000 population and found the majority of Medicaid abusers (59%) lived in the Eastern region
of the United States.18,54
Sources of Misused and Abused Opioids
Individuals misusing and abusing prescription opioids acquire their drugs from a variety of
sources.11,55
The 2010 NSDUH found that 17% of users obtained their opioids through a
prescription from one doctor, whereas 83% obtained their prescription opioids from other
sources (Figure 1-3).11,26,56
In comparison, the 2006 NSDUH found that 11% of respondents
obtained their opioids through a prescription from one doctor.46
Rosenblum et al surveyed nearly
5,700 individuals entering into methadone treatment facilities across the country and found
similar results – 28% of individuals reported their most frequent source of opioids was through a
doctor’s prescription.49
Studies have found that the source of abused opioids varies by gender and age. Back et al
analyzed results from the 2006 NSDUH and found that men were more likely to obtain
prescription opioids from family/friends (either purchased or for free) or from a drug dealer or
‡ Administered by the Sloane Epidemiology Center of Boston University, this is a telephone survey administered
from February 1998 through September 2006 on prescription and non-prescription medication use during the
previous 7 days . The sample consists of 19,150 randomly identified civilian, non-institutionalized individuals ages
18 and older. § The Behavioral Risk Factor and Surveillance System (BRFSS) survey is a cross-sectional telephone survey of
adults ages 18 years and older conducted by state health departments, with support from the CDC.
Part 1 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – A Review of the Literature
1.13
stranger, while women were more likely to take prescription opioids from family/friends without
asking.46
Green et al found that women abusing hydrocodone/ acetaminophen primarily obtained the drug
from within their social network, such as friends or family members (44.6%), followed by a
dealer (37.5%), or their own
prescription (28.8%), whereas
males abusing this opioid
obtained it primary from a
dealer (45.2%), followed by
within their social network
(40.6%), or their own
prescriptions (25.4%).57
Cicero et al surveyed opioid
users entering substance abuse
treatment facilities and found
that approximately 90% of
males and females under age
20 acquired their prescription
opioids from a dealer, while
less than half of individuals
over age 51 did so. Cicero et
al also found that males were
more likely than females to
acquire their opioids from a
dealer, (OR=1.64), whereas
females were more likely than males to use a doctor’s prescription to obtain opioids
(OR=1.71).58
Doctor and Pharmacy Shopping
“Doctor shopping” and “pharmacy shopping” are two methods of diverting prescription opioids
that contribute to nonmedical opioid use, misuse and abuse. These terms refer to visiting multiple
providers (“doctor shopping”) or pharmacies (“pharmacy shopping”) to obtain medically
unnecessary prescription opioids. As shown in Table 1-5, doctor and pharmacy shopping have
been defined using a variety of cut-off points for classifying a patient as having potential
controlled substance misuse or mismanagement that would warrant further evaluation.16,17,59-65
Published thresholds vary by number of providers or pharmacies seen by a single patient to
obtain any opioid over a given time period. However, as the numbers of providers or pharmacies
are not direct measures of misuse alone, such information should be used in conjunction with
prescription patterns to identify potential misuse and determine if intervention is needed.
Table 1-5. Thresholds for Defining Doctor and Pharmacy Shopping
Citation Number of Opioid Prescribers And/
Or
Number of Pharmacies
Dispensing Opioids
Parente et al (2004) >6 in one year OR >3 in a year
Hall et al (2008) >5 during the year before death
Katz, Panas et al (2010) >1 - >10 over a 12-month period AND >1 - >10 over a 12-month period
White et al (2009) > 2 over a 3-month period OR
Obtained
free from a
friend or
relative
55%
Prescribed
by one
doctor
17%
Bought from
a friend or
relative
12%
Took from a
friend or
relative
without
asking
5%
Got from
drug dealer
or stranger
4%
Other source
7%
Source: Centers for Disease Control and Prevention. Policy Impact:
Prescription Painkiller Overdose. November 2011
Figure 1-3. Sources of Prescription Opioids, 2010
Part 1 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – A Review of the Literature
1.14
Citation Number of Opioid Prescribers And/
Or
Number of Pharmacies
Dispensing Opioids
> 3 over a 12-month period
U.S. GAO (2009) >6 over two year period
U.S. GAO (2011) > 5 in one year
Wilsey et al (2010) >2 within one month AND >2 within one month
Wilsey et al (2011) 2-5 in one year
Peirce et al (2012) >4 in the 6 months before death OR >4 in the 6 months before death
There may be justifiable reasons that patients use multiple providers. Wilsey et al recognized the
potential for patients to either (1) substitute clinicians, (2) obtain medications from a practitioner
covering for the patient's customary provider, or (3) receive treatment from another practitioner
that could be entirely appropriate (dentist, emergency room doctor, etc.). Assuming this is
accurate, they defined the occurrence of a multiple provider episode as occurring when an
individual received a prescription for the same medication from two or more practitioners filled
by two or more pharmacies within a 30-day period. This practice negates the traditional
“gatekeeper” role of a single pharmacist who would know if the patient was obtaining a
medication in a justifiable manner (e.g., seeing a physician on-call for the patient’s customary
doctor). Using these criteria, they found that opioid prescriptions (12.8%) were most frequently
involved in multiple provider episodes, followed by benzodiazepines (4.2%), stimulants (1.4%),
and anorectics (0.9%), respectively. The greatest associations with multiple provider episodes
were simultaneously receiving prescriptions for different controlled substances (polypharmacy of
controlled substances).64
A second study by this group attempted to find a threshold for
identifying patients who used multiple providers. Using data from the California prescription
drug monitoring program, this study found that patients who used two to five providers to obtain
opioids did not differ consequentially in terms of their demographics and prescription utilization
characteristics from patients who used only one provider during a one-year period. This was
consistent with the proposition that many patients who use up to five prescribers in a one-year
period might have justifiable reasons for doing so.63
A 2009 United States Government Accountability Office (GAO) report on Medicaid fraud and
abuse of controlled substances reviewed claims during Fiscal Years (FY) 2006 and 2007 in five
states - California, Illinois, New York, North Carolina, and Texas. The GAO defined doctor
shopping as seeing six or more different prescribers to obtain prescriptions for the same type of
controlled substance during FY 2006 and 2007, and found that nearly 65,000 Medicaid recipients
met these criteria. Table 1-6 shows the number of beneficiaries receiving prescription opioids
from multiple prescribers.16
Table 1-6. Medicaid Beneficiaries Receiving Prescription Opioids from Multiple Prescribers,
FY 2006 and 2007
Prescription Opioid Number of Prescribers in Selected States
6-10 11-15 16-20 21-50 51+ Total
Fentanyl 777 41 6 1 0 825
Hydrocodone 31,364 3,518 723 340 9 35,954
Hydromorphone 590 67 14 11 0 682
Methadone 824 76 9 2 0 911
Morphine 810 50 8 1 0 869
Oxycodone 5,349 435 73 18 0 5,875
Part 1 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – A Review of the Literature
1.15
Prescription Opioid Number of Prescribers in Selected States
6-10 11-15 16-20 21-50 51+ Total
Total Prescription Opioids 39,714 4,187 833 373 9 45,146
Total Controlled Substances* 64,239 5,066 926 396 9 70,636
Source: United States Government Accountability Office (2009)
Note: The numbers in the columns do not represent unique beneficiaries. There are 64,920 total unique beneficiaries
*Additional substances included amphetamine derivatives, benzodiazepine, methylphenidate, and non-
benzodiazepine sleep aids
In 2011, the GAO subsequently released a report examining the prevalence of doctor shopping in
the Medicare population. The GAO reviewed 2008 Medicare Part D claims for controlled
substances from five states - California, Georgia, Maryland, Massachusetts, and Texas (only
California and Texas were also included in the Medicaid analysis). This analysis defined doctor
shopping as seeing five or more different prescribers from January through December 2008 to
obtain prescriptions for the same type of controlled substance. Using this definition, the GAO
found that approximately 170,000 Medicare recipients met this criterion. Table 1-7 shows the
number of beneficiaries receiving prescription opioids from multiple prescribers. In both GAO
analyses, hydrocodone and oxycodone were the prescription opioids most often received from
multiple prescribers; in the Medicare analysis, these two drugs were involved in more than 80%
of the doctor shopping events.17
Table 1-7. Medicare Beneficiaries Obtaining Prescription Opioids from Multiple Prescribers,
2008
Prescription Opioid Number of Prescribers in Selected States
5-10 11-15 16-20 21-50 51+ Total
Codeine with Acetaminophen 1,500 21 4 0 0 1,525
Fentanyl 5,043 24 8 2 0 5,077
Hydrocodone 92,801 3,553 700 335 5 97,394
Hydromorphone 2,453 77 13 8 0 2,551
Meperidine 149 8 0 0 0 157
Methadone 3,414 9 0 0 0 3,423
Morphine 6,354 33 4 0 0 6,391
Oxycodone 54,183 1,974 440 235 5 56,837
Tramadol 4,364 134 33 14 0 4,527
Total Prescription Opioids 170,261 5,833 1,202 594 10 177,882
Total Controlled Substances* 181,823 5,927 1,214 600 10 189,574
Source: United States Government Accountability Office (2011)
Note: The numbers in the columns do not represent unique beneficiaries. There are 170,029 unique beneficiaries
*Additional substances included amphetamine derivatives, benzodiazepine, carisoprodol, methylphenidate, and
non-benzodiazepine sleep aids
Epidemiologists have also explored the relationship between shopping behavior and drug-related
death. Using 2005-2007 data from West Virginia’s prescription drug monitoring program and
drug-related death data compiled in their state forensic database, Peirce et al analyzed trends in
doctor and pharmacy shopping among living and deceased individuals. They found a
significantly greater proportion of deceased subjects were doctor shoppers and pharmacy
shoppers (25% and 17%) than living subjects (4% and 1%). As depicted in Figure 1-4, these
researchers reported that 55% of pharmacy shoppers also met criteria for doctor shoppers,
whereas only 20% of doctor shoppers met criteria for pharmacy shopping.65
Thus, as the authors
Part 1 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – A Review of the Literature
1.16
point out, there is clearly a relationship between doctor and pharmacy shopping, although one
can occur without the other.
Figure 1-4. Dual Doctor and Pharmacy Shoppers in West Virginia, 2005-2007
Based on data from Massachusetts’ prescription drug monitoring database, Katz et al analyzed
individual’s use of multiple prescribers and pharmacies for Schedule II opioids in 2006. As
displayed in Table 1-8, the majority of patients (76.9%) used one prescriber and one pharmacy
for their opioid prescriptions, nearly 10% used two prescribers and one pharmacy, and 3% used
two prescribers and two pharmacies. The authors found that, relative to the overall sample size
(562,591), the number of patients using high numbers of prescribers or pharmacies was small –
only 1.5% of the sample used 5 or more prescribers and less than 0.5% used 5 or more
pharmacies. However, the authors did find that individuals with more prescribers were also more
likely to use more pharmacies (see Table 1-9). For example, fewer than 1% of individuals with
one prescriber used four or more pharmacies, whereas nearly 70% of individuals with ten or
more prescribers used four or more pharmacies.61
To define shopping behavior, the above studies relied on counting the number of prescribers or
number of pharmacies a subject uses during a specified period, but they did not distinguish
successive prescribers from concomitant prescribers. Cepeda et al examined overlapping
prescriptions, defined as at least 1 day of overlapping prescriptions written by two or more
different prescribers at any time during an 18-month period. Overlapping prescriptions are not
unique to opioids and, thus, provide a useful comparison to medications not likely to be misused.
Cepeda et al determined that having two or more overlapping prescriptions written by different
prescribers and filled at three or more pharmacies differentiated opioids from non-controlled
substances (i.e., diuretics) and constituted shopping behavior.66
In a subsequent study, Cepeda et
al estimated the prevalence of doctor shopping by analyzing prescription drug claims from a
large database that includes 65% of all retail prescriptions in the country. They found that only a
very small proportion of individuals met this criterion (Table 1-10). Of the 25,161,024
individuals in the dataset who received one opioid prescription during the study period, only
0.30% (75,215) met criteria for doctor shopping. Even among those identified as doctor
shoppers, very few had significantly high utilization; only 11% used >5 prescribers and 6.7%
used > 6 pharmacies. Cepeda et al also found variation in the schedule and formulation most
often dispensed to doctor shoppers compared to non-doctor shoppers (Figure 1-5).
20%
80%
Doctor Shopping Both Doctor & Pharmacy
Shopper
55% 45%
Pharmacy Shopping Both Pharmacy &
Doctor Shopper
Part 1 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and Prevention Policies – A Review of the Literature
1.17
Table 1-8. Distribution of Patients by Number of Prescribers and Pharmacies, 2006 #
Taken From: Katz et al (2010). Usefulness of prescription monitoring programs for surveillance – analysis of Schedule II opioid prescription data in
Massachusetts, 1996-2006. Figure 2A.
Table 1-9. Distribution of Patients Using Multiple Prescribers and Pharmacies, 2006 # Pharmacies Used
# Providers Used n 1+ 2+ 4+ 6+ 8+ 10+
1 443,956 100.00% 2.5% 0.12% 0.01% 0.00% 0.00%
2 75,191 100.0% 27.0% 1.12% 0.10% 0.01% 0.00%
3 24,919 100.0% 41.2% 3.74% 0.44% 0.07% 0.02%
4 9,980 100.0% 51.2% 7.57% 0.98% 0.13% 0.03%
5 4,274 100.0% 61.2% 14.09% 2.32% 0.28% 0.05%
6 1,887 100.0% 68.0% 20.83% 5.14% 1.17% 0.21%
7 1,025 100.0% 78.0% 30.34% 9.85% 1.27% 0.10%
8 543 100.0% 81.6% 34.44% 13.08% 3.87% 0.37%
9 296 100.0% 84.5% 47.30% 19.26% 7.09% 1.35%
10+ 520 100.0% 91.2% 69.23% 42.88% 24.62% 11.92%
Total Population 562,591 100.00% 9.38% 0.90% 0.17% 0.05% 0.02%
Source: Katz et al (2010)
Part 1 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and Prevention Policies – A Review of the Literature
1.18
Table 1-10. Distribution of Patients Using Multiple Prescribers and Pharmacies, 2008
# of Prescribers Number of Pharmacies
3 4 5 6 7+ Total (n)
2 43.74% 3.32% 0.75% 0.19% 0.11% 48.10% (36,178)
3 24.46% 4.67% 1.70% 0.54% 0.26% 31.63% (23,790)
4 3.29% 3.21% 1.67% 0.69% 0.40% 9.26% (6,967)
5 0.77% 1.48% 1.18% 0.63% 0.41% 4.46% (3,357)
6+ 0.41% 1.23% 1.47% 1.19% 2.25% 6.55% (4,923)
Total
(n)
72.67%
(54,658)
13.91%
(10,460)
6.75%
(5,080)
3.24%
(2,439)
.34%
(2,578)
100%
(75,215)
Taken from: Cepeda et al (2012). Opioid Shopping Behavior: How Often, How Soon, Which Drugs, and What Payment Method? Table 2.
Note: Total population was 25,161,024; of that, 72,215 (0.30% met criteria for doctor shopping
Figure 1-5. Type and Formulation of Prescription Opioids Dispensed to Doctor Shoppers Compared to Non-Doctor Shoppers
Source: Cepeda et al (2012). ). Opioid Shopping Behavior: How Often, How Soon, Which Drugs, and What Payment Method?
Note: Other category includes Schedule IV opioids and unscheduled opioids. IR=Immediate Release; ER=Extended Release; Combo=Combination products (i.e., those containing
acetaminophen, or NSAIDs)
0.0%
5.0%
10.0%
15.0%
20.0%
25.0%
30.0%
35.0%
40.0%
45.0%
50.0%
II Only III Only Other II & III II &
Other
II, III, &
Other
III &
Other
Per
cen
t D
isp
ense
d
Schedule Type
Non-Doctor Shopper Doctor Shopper
0.0%
10.0%
20.0%
30.0%
40.0%
50.0%
60.0%
70.0%
80.0%
90.0%
IR ER Combo IR & ER IR &
Combo
ER &
Combo
IR, ER,
&
Combo
Per
cen
t D
istr
iub
tio
n
Formulation Type
Non-Doctor Shopper Doctor Shopper
Part 1 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – A Review of the Literature
1.19
Doctor shoppers were more likely to be written prescriptions containing Schedule II opioids and
extended release formulations, whereas non-doctors shoppers were more likely to be written
prescriptions containing Schedule III opioids and immediate release or combination formulations
(i.e., opioid and acetaminophen, etc).67
The variable definitions for thresholds (i.e., cut-off points) for doctor or pharmacy shopping
await corroboration via interviews with patients who use multiple physicians and/or pharmacies
to obtain prescription opioids. If investigators could differentiate justifiable from illicit shopping
behavior, the sensitivity and specificity of the various cut-off points could be determined. Similar
studies have been devised involving the unlawful channeling of regulated pharmaceuticals from
legal sources to the illicit marketplace by interviewing club drug users, street-based illicit drug
users, methadone maintenance patients, and HIV-positive individuals who abuse and/or divert
drugs.49,68-70
As in the aforementioned instance, these studies would require anonymity and
confidentiality.68
Opioid Prescribing Patterns
Over the past three decades, physicians began more aggressive management of chronic non-
cancer pain (CNCP), which contributed to the increase in duration of opioids use 51,71-75
despite
limited evidence of the efficacy of opioids for the treatment of this type of pain.8,76-81
A
comparison of National Ambulatory Medical Care Survey (NAMCS) results from 1980 and 2000
revealed that although the number of visits for musculoskeletal pain remained constant during
that time, the prevalence of opioids prescribed at both acute pain visits and chronic pain visits
increased (8% to 11% and 8% to 16%, respectively). Additionally, the use of stronger opioids
(such as morphine) during chronic pain visits more than tripled (from 2% to 9%), which
translates to an additional 4.6 million visits in which strong opioids were prescribed.82
Dorn,
Meek and Shah saw a similar trend analyzing NAMCS and National Hospital Ambulatory
Medical Care Survey (NHAMCS) data for chronic abdominal pain-related outpatient visits. They
found that the number of visits for this condition decreased by approximately 18% from 14.8
million in 1997-1999 to 12.2 million in 2006-2008, yet the prevalence of visits in which an
opioid was prescribed increased nearly 107% over the same period (5.9% to 12.2%).83
In 2001,
NAMCS data showed that opioids were prescribed in 63 primary care visits per 1,000 total visits,
compared to 41 per 1,000 total visits in 1992. Physicians reported that a pain-related diagnosis
(back pain, acute musculoskeletal pain, and headache) was the primary diagnosis in nearly two-
thirds of visits resulting in an opioid prescription.84
A review of 2009 pharmacy dispensing data
(representing pharmacies dispensing over half of all prescriptions in the U.S) found that a large
proportion of opioid prescriptions were prescribed for patients between the ages of 40-59 years
old (45.7%, or 36.4 million). Additionally, over half of all opioid prescriptions in this dataset
(56.4%) were dispensed to patients who had previously filled a prescription for an opioid in the
past 30 days.85
From 2000-2009, Kenan et al found that the number of opioid prescriptions per 100 individuals
increased 35.3%, from 61.9 to 83.7. Additionally, the average prescription size (expressed as
morphine milligram equivalent [MME] per day multiplied by the prescription duration) of both
hydrocodone and oxycodone prescriptions increased nearly 70%, from 170MME to 288MEE and
923MME to 1566MME, respectively.86
In 2005, nearly 100 million prescriptions were written
for hydrocodone, making it the most commonly prescribed drug in the United States. In
Part 1 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – A Review of the Literature
1.20
comparison, the second and third most common prescriptions, atorvastatin (a cholesterol
lowering medication) and amoxicillin, had approximately 63 million and 52 million prescriptions
written, respectively.22,87
In 2009, the number of opioid prescriptions dispensed rose to 257
million (a 48% increase since 2000).25
The CDC estimated that in 2010, enough opioids were
sold to provide a typical dose of hydrocodone (5mg every 4 hours) to every American adult for
one month.1
State-Specific Opioid Prescribing Patterns
Over the past two decades, state data show increases in the number patients receiving a
prescription for opioids and the number of opioid prescriptions written and dispensed, especially
for Schedule II opioids. According to a nationwide analysis of 9 million prescription drug claims
from 2000, 64.2 per 1,000 total claims were for opioids. Although some states had a prescription
opioid claim rate below 20 per 1,000 total claims (California, Texas, Illinois, Michigan, and New
York), some states had rates over 100 per 1,000 total claims (Alaska, Arizona, Delaware,
Maryland, Massachusetts, New Hampshire, South Carolina, and Tennessee). States with long-
standing prescription drug monitoring programs (PDMP’s) had among the lowest rates.47
From 1996-2002, Franklin et al found that the number of Washington State employees with
prescriptions for Schedule II opioids increased 2.5 times. Additionally, as a proportion of all
opioid prescriptions, Schedule II opioid prescriptions increased from 19.3% to 27.2% during that
period. There was also a 55% increase in the average daily morphine equivalent dose of long-
acting opioids (from 88mgMED/day to 132mgMED/day) during that period. Since the current
CDC43
and Washington State opioid dosing guidelines recommend that physicians refer patients
for a pain management consultation for prescriptions over 120mg morphine equivalent dose per
day, Franklin et al opined that it is conceivable that the average daily dose would not have
reached or exceeded this threshold if the guideline had been enacted during this time.33
Dembe
et al reviewed Ohio workers compensation data from 2008-2009 and found that nearly 10% of
claimants had prescriptions for opioids exceeding 120mg.88
Studies of state prescribing data also found that a small number of prescribers are responsible for
prescribing the majority of opioids. Using California workers compensation claims data for
prescription drugs filled from 2005-2009, Swedlow et al analyzed physician prescribing patterns
for Schedule II opioids among this population. They found that the top 1% of prescribers
(approximately 93 physicians) accounted for one-third of the total Schedule II opioid
prescriptions and slightly more than 40% of the total milligrams morphine equivalent (MME)
prescribed. The top 10% of prescribers (approximately 917 physicians) accounted for almost
21% of the total Schedule II opioid prescriptions and nearly 87% of the total MME prescribed.
From 2005-2009, the top 10% of prescribers had an average of 17.5 claims in which they
prescribed Schedule II opioids (compared to 3.5 claims in the group overall), totaling nearly
750,000 MME (compared to slightly over 87,000 MME in the group overall).89
Blumenschein et
al found a similar trend among users of the Kentucky All Schedule Prescription Electronic
Reporting Program (KASPER). Analyzing 2005-2009 KASPER data revealed that the top 10%
of prescribers were responsible for the vast majority of all prescriptions for controlled
substances.90
Part 1 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – A Review of the Literature
1.21
Medicaid-Specific Prescribing Patterns
Analyses of Medicaid prescription drug claims have revealed increases in the number of these
beneficiaries receiving a prescription for opioids and the number of opioid prescriptions written
and dispensed. From 1996-2002, the Medical Expenditures Panel Survey (MEPS) found a 22.5%
increase in the number of Medicaid enrollees using prescription opioids.14
Compared to the
MEPS data analysis, Brixner et al found a larger increase in the proportion of Medicaid
recipients receiving prescription opioids using CMS (Centers for Medicare and Medicaid
Services) data. From 1998-2003, they found an 83% increase in Medicaid recipients with these
prescriptions, from 15 million to more than 27.5 million.15
Zerzan and colleagues found an even
larger increase among fee-for-service Medicaid enrollees from 1996-2002. They analyzed CMS
prescription drug claims data for 49 state fee-for-service Medicaid programs and found a 309%
increase in the number of opioid prescriptions dispensed between 1996 and 2002, compared to
170% for non-opioid prescription drugs. During that same period, oxycodone and methadone
prescriptions increased 1,615% and 790%, respectively.91
Zerzan et al also measured variation in
opioid dispensing rates by state; they used “defined daily dose”**
(DDD) per 1,000 Medicaid
recipients per day (DDD/1000/d). In 1996, dispensing rates varied from 6.9 - 44.1 DDD/1000/d
and increased to 7.1-164.97 DDD/1000/d in 2002. From 1996-1997, 8% of state Medicaid
programs (four states) were able to maintain or decrease their dispensing rates, while two-thirds
of states at least doubled their rates. In 2002, there was a 23-fold difference between the states
with the highest and lowest overall opioid dispensing rates.91
Using data from the Trends and Risks of Opioid Use for Pain, (TROUP††
) study, Edlund et al
found the 4% of users in the 95th
– 99th
percentile in the Medicaid population consumed
approximately 26% of total opioids and the top 1% (99th
– 100th
percentile) consumed 20% of
total opioids by milligrams morphine equivalents (MME). In comparison, the 4% and 1% of the
commercially insured population consumed 27% and 43% of total opioids, respectively.93
Braden et al found different results when comparing any chronic opioid use in these two
populations. In their analysis, TROUP data revealed that Medicaid recipients were twice as
likely to have any opioid use compared to commercial enrollees, and four times as likely to have
greater than a 90 day-supply. The authors speculate that this may be partially attributable to a
greater comorbidity and disability burden in this population.94
In another analysis of TROUP data, Sullivan et al found that the proportion of the Arkansas
Medicaid population with greater than 180-days supply of prescription opioids grew from 9.5%
to 16.0% from 2000-2005 (a 68.5% increase), compared with an increase from 2.1% to 3.2% in
the commercially insured group (a 49.9% increase). Surprisingly, the prevalence of individuals
with prescriptions for doses greater than 120mg morphine equivalent dose (MED) per day did
not vary by insurance type, nor did the mean cumulative opioid dose received within a calendar
year. During this period, the Medicaid group had a larger increase in the cumulative yearly dose
per user for short-acting Schedule II opioids than the commercially insured group (191.2% vs.
**
“Defined Daily Dose” is a conversion factor established by the World Health Organization’s Collaborating Centre
(WHOCC) for Drug Statistics Methodology and is another method of standardizing drug dose. The WHOCC defines
DDD as “the assumed average maintenance dose per day for a drug used for its main indication in adults.91,92
††
The Trends and Risks of Opioid use for Pain (TROUP) study was conducted from 2000-2005. The study
compared trends and risks of opioid use, misuse and abuse in two populations – a national commercially insured
population (HealthCore Blue Cross and Blue Shield) and the Arkansas Medicaid population.
Part 1 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – A Review of the Literature
1.22
95.5%, respectively), while the commercially insured group had a larger increase in the
cumulative yearly dose per uses of long-acting Schedule II opioids (54.0% vs. 38.3% in the
Medicaid group).34
An analysis of Michigan Medicaid recipients receiving care from a large,
rural family medicine group found that the average number of opioid prescribers per patient
within a six-month period was 3.7 (ranging from 2-10 per patient). Patients using non-opioid
analgesics had 3.2 fewer prescriptions per 6 months and were less likely to have 6 or more
prescriptions (OR=0.24, 95% CI=0.08-0.73) than those on opioids alone. This analysis also
found that the average number of opioid prescriptions per patient within the six-month period
averaged 8.4 (ranging from 3-28 per patient), with 64% of patients having more than six
prescriptions.95
Risk Factors Associated with Prescription Opioid Misuse and Abuse
Studies have identified a variety of factors associated with an individual’s risk for opioid misuse,
abuse or dependence 46,47,96-99
including (but not limited to) male gender, 37,46,60
simultaneous use
of another illicit substance or prescription drug abuse,37,50,100-104
individuals reporting severe
pain,37,73
comorbid conditions,9,18,105
and daily opioid dose.96,99,106
Sociodemographic Risk Factors
National surveys and studies report that numerous sociodemographic factors are associated with
increased risk for prescription opioid misuse and abuse, including gender,46,107,108
age, 46,50,99,101,106
race/ethnicity,107
employment status,50
income,50,107
and geographic location.107,109
Gender
Multiple iterations of the NSDUH found that lifetime, past year and past month utilization
among both males and females increased from 2002-2010, however, prevalence rates for females
remain below those of males (see Table 1-11).27
Table 1-11. Prevalence of Nonmedical Prescription Opioid Use by Gender, 2002 & 2010 Lifetime (%) Past Year (%) Past Month (%)
Males Females Males Females Males Females
2002 14.3 11.0 5.2 4.2 2.0 1.7
2010 15.7 11.8 5.6 4.0 2.3 1.7
Source: SAMHSA, Office of Applied Studies, National Survey on Drug Use and Health, 2002-2009 Note: Prevalence among individuals ages 12 and older
Based on TROUP‡‡
data, Thielke et al found the prevalence of long-term opioid use (defined as
90 days or more of prescribed opioids within a calendar year) increased among males and
females in both the commercially insured population and the Arkansas Medicaid population. The
largest increase among males in both groups occurred in the 45-64 year old group. Females ages
45 and older saw the largest increase in the commercial population, compared to females ages 65
and older in the Medicaid group.108
In analyses of data from the TROUP§§
study, Edlund et al
found that heavy utilization (defined as “individuals in the top 5% of total opioid use”) was
associated with male gender in both the commercially insured and Medicaid populations.93
‡‡
The Trends and Risks of Opioid Use for Pain (TROUP) study was conducted from 2000-2005. The study
compared trends and risks of opioid use, misuse and abuse in two populations – a national commercially insured
population (HealthCore Blue Cross and Blue Shield) and the Arkansas Medicaid population.
Part 1 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – A Review of the Literature
1.23
Race/Ethnicity
According to the 2010 NSDUH, Native Americans/Alaska Natives had the highest prevalence of
nonmedical opioid use, followed by Whites. Native Hawaiians/Pacific Islanders had the second
lowest lifetime and past-year prevalence rates (behind Asians), but their past-month prevalence
was higher, second only Native American/Alaska Natives (see Table 1-12).27
Table 1-12. Prevalence of Nonmedical Opioid Use by Race/Ethnicity, 2010 Lifetime
(n=34,776,000)
Past Year
(n=12,213,000)
Past Month
(n=5,100,000)
Overall 13.7 4.8 2.0
White 15.2 5.1 2.2
Black or African American 10.6 3.6 1.6
Hispanic or Latino 11.3 4.8 2.0
Asian 6.3 3.2 0.7
Native American or Alaska Native 19.3 8.8 4.0
Native Hawaiian or Other Pacific Islander 6.4 3.4 2.4
Source: SAMHSA, Office of Applied Studies, National Survey on Drug Use and Health, 2010
Geographic Location
The 2010 NSDUH found a higher prevalence of nonmedical opioid use among respondents
residing in the western region of the country, and among those from small metro counties (Table
1-13).27
Cicero et al used data from the Researched Abuse, Diversion and Addiction Related
Surveillance (RADARS) System and found that the majority of prescription opioid abuse occurs
outside of large metropolitan areas, and that certain regions have significantly higher rates of
abuse, such as the rural North East, upper Northwest, and Appalachia.110
White et al estimated
that the majority of privately insured abusers resided in the Southern region of the United States
(52%).9,54
Table 1-13. Prevalence of Nonmedical Opioid Use by Geographic Location, 2010 Lifetime
(n=34,776,000)
Past Year
(n=12,213,000)
Past Month
(n=5,100,000)
Overall 13.7 4.8 2.0
Region
Northeast 12.2 4.3 1.9
Midwest 13.1 4.7 2.0
South 13.3 4.4 1.9
West 16.4 6.1 2.3
County Type
Large Metro (>1 million population) 13.8 4.8 1.9
Small Metro (20,000 -999,999 population) 14.3 5.0 2.1
Nonmetro (0-19,999 population) 12.5 4.4 2.1
Source: SAMHSA, Office of Applied Studies, National Survey on Drug Use and Health, 2010
Behavioral Risk Factors
Multiple studies and surveys have identified non-opioid substance abuse as a risk factor for
prescription opioid misuse and abuse.46,50,93,99,103
The NSDUH repeatedly identifies illicit
substance and alcohol abuse as a risk factor for past-year nonmedical opioid use and abuse.50,103
Among males, other illicit drug abuse or dependence (such as cocaine, heroin, hallucinogens, or
inhalants) and alcohol abuse or dependence are significantly associated with past-year opioid
Part 1 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – A Review of the Literature
1.24
misuse.46,103
Men are more likely to misuse opioids as the result of legal and behavioral
issues.111
Significant predictors of opioid misuse among females include serious psychological
distress and cigarette use, as well as emotional issues and affective distress.46,103,111
The 2006
NSDUH found that individuals who reported past-year non-medical use of tranquilizers or
sedatives were 16 times more likely to misuse prescription opioids.46
Pain Level & Comorbidities§§
Research by Toblin et al and Bohnert et al found that individuals reporting severe pain were at
higher risk for opioid misuse and abuse.37,73
Edlund et al found that heavy utilization (defined as
“individuals in the top 5% of total opioid use”) in the TROUP***
study was associated with
headache (commercially insured population only), back pain (both populations), and arthritis
(Medicaid population only).93
Edlund et al also found that in the commercially insured and
Medicaid populations, the likelihood heavy utilization increased with the number of chronic non-
cancer pain diagnoses.93,99
Mental Health
Studies consistently report that mental health diagnoses are associated with increased
nonmedical use of prescription opioids, as well as increased risk for abuse.46,50,99,108,112
A review
of 2002-2004 NSDUH identified mental health diagnoses, including depressive symptoms, panic
symptoms, and social phobic/agoraphobic symptoms as factors associated with past-year
nonmedical prescription opioid use and abuse.50
The 2006 NSDUH found that one in four
prescription opioid abusers reported a history of serious psychological distress, and that
compared with men, rates of distress were higher among women (14.5% vs. 11.2%,
respectively).46
One study found that mental health disorders are 50-100% more common among Medicaid
recipients, compared to the overall population and another found that 29% of Medicaid
beneficiaries across six states had both a mental health condition and a history of substance
abuse.18
Using data from the TROUP study, Thielke et al found the prevalence of long-term
opioid use was higher among individuals with a mood disorder, regardless of age, gender, or
insurance group. Although the prevalence of long-term users without a mood disorder was
similar among both populations, the prevalence of long-term users with a mood disorder was
significantly higher in the Medicaid population. Similarly, the prevalence of long-term use was
higher among individuals in each population with an anxiety disorder compared to those without
the disorder, but the prevalence of long-term use with an anxiety disorder was higher in the
Medicaid population compared to commercially insured population.108
Also using TROUP data,
Edlund et al found that persons with mental health or mood disorders were more likely to be
heavy utilizers as well (defined as “individuals in the top 5% of total opioid use”). The likelihood
of being a heavy user in the commercially insured population increased with the number of
substance abuse diagnoses.93,99
§§
As comorbidities are both a risk factor for, and outcome of, nonmedical prescription opioid use, misuse and abuse,
relevant research will be discussed in both the risk factor and outcomes sections. ***
The Trends and Risks of Opioid use for Pain (TROUP) study was conducted from 2000-2005. The study
compared trends and risks of opioid use, misuse and abuse in two populations – a national commercially insured
population (HealthCore Blue Cross and Blue Shield) and the Arkansas Medicaid population.
Part 1 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – A Review of the Literature
1.25
Opioid Dose and Supply
The literature identifies daily dose and days’ supply of prescription opioids as additional risk
factors for misuse and abuse. Compared to patients receiving low daily opioid doses, Dunn et al
found that patients prescribed high daily opioid doses (defined as >100mg morphine equivalent
dose per day, or MED/d) were more likely to be male (48% vs. 40%) and current smokers (40%
vs. 28%). Patients prescribed high daily doses were also more likely to have been previously
treated for depression (32% vs. 26%) and/or substance abuse (14% vs. 5%).36
An analysis of
opioid abuse within the South Central Veterans’ Affair Health Care Network found that patients
prescribed opioids continuously for 211 days were more likely to develop abuse or dependence
compared with patients prescribed opioids continuously for 91-120 days.96
Analyses of the
TROUP†††
study have also found that risk for prescription opioid misuse and abuse is associated
with an individual’s daily opioid dose and supply (see Table 1-14). In both the commercially
insured and Medicaid sample, Sullivan et al found that patients with a daily dose greater than
120mg MED/d were at a significantly increased risk of opioid misuse and abuse/dependence,
compared with patients with daily doses below 120mg MED/d.106
Edlund et al found that higher
average daily dose and greater number of day’s supply of prescribed opioids were associated
with opioid abuse in the commercially insured sample, but only higher average dose was
associated with abuse in the Medicaid sample.99
Table 1-14. Odds of Opioid Misuse and Abuse by Average Daily Dose and Days Supply Possible Misuse
a Probable Misuse
a Abuse/Dependence
b
Commercially
Insured (OR)
Arkansas
Medicaid
(OR)
Commercially
Insured (OR)
Arkansas
Medicaid
(OR)
Commerically
Insured (OR)
Arkansas
Medicaid
(OR)
Daily Dose
<Median mg/day 1.00 1.00 1.00 1.00 1.00 1.00
Median-120 mg/day 1.65 1.21 2.68 1.80 1.48 1.11
>120 mg/day 2.37 2.02 6.70 4.69 2.19 1.70
Days Supply
91-160 days NR NR NR NR 1.00 1.00
161-185 days NR NR NR NR 1.48 1.01
>185 days NR NR NR NR 1.79 1.18
Source: aSullivan et al (2010);
bEdlund et al (Nov 2010)
Note: Median daily dose was 32mg and 35 mg morphine equivalent in the commercially insured and Arkansas
Medicaid samples, respectively. NR=not reported.
Summary
Over the past two decades, as physicians managed pain more aggressively and prescribed
stronger opioids more frequently and at higher doses, studies and surveys at the state and
national level have documented an increase in the prevalence of nonmedical use, misuse, and
abuse of these drugs, particularly products containing hydrocodone, oxycodone and methadone.
Risk factors identified for prescription opioid misuse and abuse include (but are not limited to)
demographic factors (gender, race/ethnicity, etc), non-opioid substance use, and comorbid
mental health disorders. State- and national-level surveys show increased numbers of patients
who receive high doses of prescription opioids (in excess of 100mgMED/d), and who are chronic
users (continuous use for longer than 90 days). Analyses of state- and Medicaid/Medicare data
Part 1 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – A Review of the Literature
1.26
found that relatively few patients obtain their prescription opioids through doctor or pharmacy
shopping; however, the literature suggests that this minority of patients may be at higher risk of
overdose and incur increased healthcare costs and we will address this literature in the next
section.
OUTCOMES RELATED TO OPIOID MISUSE AND ABUSE
As extensively detailed in the previous section, the past two decades have seen significant
increases in the number of opioids prescribed and dispensed, as well as the prevalence of
prescription opioid nonmedical use, misuse, and abuse. These increases correlate with increases
in negative outcomes, including prescription opioid-related healthcare utilization and death, as
well as health care and societal costs.
Health Care Utilization
The increase in opioid prescriptions and the prevalence of nonmedical opioid use and abuse has
been associated with increased health care utilization (such as ED visits, hospitalizations and
physician visits) related to these drugs. Using nationwide claims data for approximately 2 million
employer-insured individuals from 1998-2002, White et al compared opioid abusers†††
to
nonabusers had significantly higher prevalence rates for a number of specific comorbidities.
These comorbidities included non-opioid poisoning, hepatitis (A, B, or C), psychiatric illnesses,
and pancreatitis, which were approximately 78-, 36-, 9-, and 21-times higher (P<0.01) among
opioid abusers, respectively. The authors concluded that the high costs associated with care of
opioid abusers were driven primarily by high prevalence rates of these costly comorbidites and
high utilization rates of medical services and prescription drugs. With regard to medical services, 97% of prescription opioid abusers had at least one outpatient physician visit, 67.8% had at least
one hospital inpatient stay, 45.5% had at least one outpatient mental health visit, and 12.6% had
at least one inpatient mental health stay (versus 71.5%, 5.5%, 4.1% and 0.2% of nonabusers,
respectively)..9
In 2010, the Drug Abuse Warning Network (DAWN) found that ED visits related to nonmedical
prescription opioid use occurred at a rate of 137.4 visits per 100,000 population, accounting for
nearly one-third of the 1,345,645 total ED visits involving misuse or abuse of prescription
drugs.12
As seen in Table 1-15, ED visits related to opioids such as oxycodone, hydrocodone
and methadone have increased substantially from 2004-2010.10
A review of the literature by
Webster et al found that while methadone was associated with 30% of all overdose-related ED
visits, when adjusted for the number of outpatient prescriptions, methadone-related ED visits
were 23 times higher than visits for hydrocodone and six times higher than visits for
oxycodone.113
Table 1-15. ED Visits for Prescription Opioids Compared to Illicit Drugs, 2010
Type of Drug-Related ED Visit Number of
Visits Rate
a Percent Change
2004-2010
†††
Patients were identified as abusers if they had at least one claim with an ICD-9 code related to prescription
opioid abuse from 1998-2002 (304.0, 304.7, 305.5, and 965, but excluding 965.01). A group of matched controls without an opioid abuse diagnosis served as the comparison group.
Part 1 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – A Review of the Literature
1.27
Type of Drug-Related ED Visit Number of
Visits Rate
a Percent Change
2004-2010
Total Drug-Related ED Visitsb 4,916,328 1589.0 94%
Prescription Opioids 425,247 137.4 156%
Oxycodone 182,748 59.1 255%
Hydrocodone 115,739 37.4 149%
Methadone 76,237 24.6 86%
Illicit Drugs 1,171.024 378.5 NR
Cocaine 488,101 157.8 NR
Heroin 224,706 72.6 NR
Marijuana 461,028 149.0 22%
Methamphetamine 94,929 30.7 48%
Source: Drug Abuse Warning Network (2010)
NR = Not Reported aRate is per 100,000 population
bIncludes all drugs (illicit drugs, alcohol and prescription drugs) and all causes (suicide attempts, abuse, adverse
drug reactions, etc)
Characteristics Associated with Prescription Opioid-Related Healthcare Utilization
As seen in Table 1-16, opioid-related ED visits vary by numerous demographic factors. In 2010,
ED visits related to all prescription opioids were higher among males and individuals ages 45-54.
In addition to having the highest number of opioid-related visits, the rate of increase from 2004-
2010 was highest among individuals ages 45-54. Oxycodone-related ED visits among females
were increasing at a similar rate as males, but hydrocodone-related ED visits among males are
increasing at a much higher rate than among females.10
Table 1-16. Opioid-Related ED Visits by Gender and Race/Ethnicity, 2010 Opiates (Total) Oxycodone Hydrocodone
# of visits Rate
a %
Changeb
# of
visits Rate
a %
Changeb
# of
visits Rate
a %
Changeb
Gender
Male 229,107 150.6 171% 104,028 68.4 257% 55,846 36.7 180%
Neuropathic Pain NR NR 9.8% 7.6% 3.2% 0.8% 2.8% 0.7%
Source: aWhite et al (2005). Privately insured population based on administrative claims data for approximately 2 million insured members from 16 large
employers. bMcAdam-Marx et al (2010). Medicaid population based on data from the Medicaid Analytic eXtract (MAX) from the Centers for Medicare and Medicaid
Services (CMS). cWhite et al (2011). Privately insured population based on administrative claims from 40 self-insured Fortune 500 companies. Florida Medicaid population based
on administrative claims for all Medicaid-eligible beneficiaries in the state.
Note: Controls were randomly selected, demographically matched individuals. Abusers were patients with least one claim with an ICD-9 code related to prescription opioid abuse during the study period (304.0, 304.7, 305.5, and 965.0 (excluding 965.01) NR = Not Reported
Part 1 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – A Review of the Literature
1.31
psychiatric disorders, substance abuse, and medical comorbidities will require combined research
and policy-making efforts to establish a knowledge base to inform risk-reduction and effective
use of evidence based treatment.
Health Outcomes: Opioid-Related Mortality
As seen in Table 1-18, prescription opioid overdose deaths have increased over 250% over the
past decade.13
Longitudinal studies have found that nearly 100 opioid-related overdose deaths
occur each day in the United States, which is greater than deaths attributable to heroin and
cocaine combined.26
In 2008, national mortality data shows that prescription opioid overdose
deaths account for over 40% of all drug overdose deaths. Among prescription opioid overdose
deaths, methadone-related deaths account for one-third.13
In 2008, the CDC reported that
prescription opioid overdose deaths occurred at a rate of nearly 4.8 deaths per 100,000
population;115
in comparison, the methadone-related overdose death rate was approximately 1.5
overdose deaths per 100,000 in the same year.116
Prescription opioid overdose has now surpassed
firearms and motor vehicle accidents as the leading cause of unintentional injury or death among
35-54 year olds, and, behind motor vehicle accidents, the second leading cause overall.1,43,105,117
Opioid Dosage Level (mg morphine equivelant per day)
Source: (i) Dunn et al (2010). Risk of opioid-related overdose death or definite/probable opioid-related non-fatal overdose
(ii) Dunn et al (2010). Risk of opioid-related overdose death or serious non-fatal event (serious=requiring hospitalization, unconsciousness, respiratory failure)
(iii) Gomes et al (2011). Risk of opioid-related death expressed as adjusted odds ratio*
(iv) Bohnert et al (2011). Risk of opioid-related death among patients with a chronic pain diagnosis
(v) Bohnert et al (2011). Risk of opioid-related death among patients with an acute pain diagnosis
(vi) Bohnert et al (2011). Risk of opioid-related death among patients with a diagnosed substance use disorder
Note: All risks are expressed as hazard ratios unless otherwise. MED/d = morphine equivilant dose per day
Figure 1-6. Relationship between Opioid Dosage Level and Fatal/Non-Fatal Overdose Risk
Part 1 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – A Review of the Literature
1.38
with three or more unique controlled substances dispensed within the past 6 month, compared to
7% of controls. Approximately 29% of deaths involved individuals who only had one controlled
substance dispensed in the past 6 months, compared to three-quarters of the control group.65
In Ohio, there were 1,047 unintentional poisoning deaths in 2008, and prescription opioids were
involved in 37% of these (approximately 837 deaths). According to the state Department of
Health, Violence and Injury Prevention Program, 16% of unintentional overdose deaths in Ohio
involved individuals with a history of doctor shopping, occurring most frequently among females
ages 25-44 (the Department does not break out doctor/pharmacy shoppers by opioid-related
deaths only). One-quarter of all unintentional poisoning deaths involved individuals who
obtained opioids through diversion, with higher rates among males’ ages 15-34 and 65 and older,
as well as females ages 15-24. Methadone was diverted more frequently than other opioids, and
71% of methadone deaths involved diverted methadone.133
Methadone used for pain treatment
and methadone used for opioid substance abuse treatment are not distinguished in Ohio overdose
data. Some experts have conjectured that methadone deaths are more likely to result from
methadone’s use as a pain medication. The underlying rationale is that the timing of the
increased deaths coincided with the increase in use of methadone as a prescription analgesic and
the increased dangers of methadone to opioid naïve patients versus long-term users of opioids in
treatment.134
Gomes et al examined the prevalence of doctor and pharmacy shopping among residents of
Ontario, Canada who received opioids through a publicly funded prescription drug coverage
program and suffered an opioid-related overdose death from 1997-2006 . Of the 593 opioid-
related deaths among individuals receiving opioids for non-cancer pain, they found that
approximately 10% of decedents obtained prescriptions for opioids from four of more prescribers
and 10% filled their opioid prescriptions at four or more pharmacies in the six months before
death (compared to 5.7% and 4.3% of controls, respectively). Nearly 3.5% of decedents obtained
their prescriptions from six or more prescribers and 2.6% filled their opioid prescriptions at six
or more pharmacies during the same period (compared to 1.6% and 0.7% of controls,
respectively).38
As a result of recent research on the relationship between mortality and opioid dose and number
of prescribers, the CDC has identified a subset of high risk patients. This subset consists of the
10% of opioid users seeking care from multiple doctors (i.e., doctor shoppers) and are prescribed
high doses (defined as >100mgMED/day) and account for 40% of all opioid-related overdoses
(see Figure 1-7).35
Part 1 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – A Review of the Literature
1.39
Outcomes: Health Care Costs
Health care related to the nonmedical use of opioids has been estimated to cost insurers (both
private and public) approximately $72.5 billion annually.1 Banthin & Miller used data from the
Medical Expenditures Panel Survey (MEPS) to examine trends in prescription opioid utilization
and costs between 1996/1997 and 2001/2002. From 1996/1997 to 2001/2002, Medicaid
expenditures for prescription opioids increased 153% from nearly $257 million to $650 million,
whereas expenditures for all prescription drugs increased 104% during that time, from $11.6
billion to $23.7 billion.14
Prescription opioid expenditures per user increased by 107%, from $75
to $155, but the number of recipients using prescription opioids only increased 22.5% (compared
to increases of 106.6% and 11.5% for all prescription drugs).14
Using data from the U.S. Centers
for Medicare and Medicaid (CMS), Brixner et al also analyzed the trends in prescription opioid
utilization and costs from 1998-2003. The authors found that during that time, Medicaid
expenditures for opioids increased nearly 300%, from $311 million to approximately $1.2
billion, accounting for 4% of total Medicaid prescription drug expenditures.15
In two different studies, White et al compared the costs incurred by patients with a diagnosis of
prescription opioid abuse compared to controls (individuals without this diagnosis) in two
privately insured samples (1998-2002 and 2003-2007) and a Florida Medicaid sample (2003-
2007) (see Figure 1-8). Data for the 1998-2002 privately insured sample came from claims data
for approximately 2 million members from 16 large, nationwide employers; data for the 2003-
2007
10%
40% 10%
40% 80%
20%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
High Risk: Patients seeing multiple doctors for high dose opioids
Low Risk: Patients seeing one doctor for high dose opioids
Low Risk: Patients seeing one doctor for low dose opioids
Taken from: CDC Grand Rounds: Prescription Drug Overdoses - a U.S. Epidemic. Figure 3
Opioid Users Opioid-Related
Overdoses
Figure 1-7. Percentage of Opioid Users and Overdoses, by Risk
Group
Part 1 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – A Review of the Literature
1.40
Part 1 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and Prevention Policies – A Review of the Literature
Data Source ............................................................................................................................................ 2.4
Episode-Based Model ........................................................................................................................ 2.6
Model Settings ....................................................................................................................................... 2.6
Subgroup Analysis: The Role of Pharmacy Shopping in Overdose Events .......................................... 2.8
Peak Number of Pharmacies .............................................................................................................. 2.8
Study Population Characteristics ........................................................................................................... 2.9
Characteristics of Prescription Opioid Use .......................................................................................... 2.10
Opioid Prescriptions by Drug Type ................................................................................................. 2.10
Predominant Opioid Prescriptions among Long-Term Episodes of Opioid Use ............................. 2.11
Supply days of opioid prescriptions by drug type ............................................................................ 2.11
Characteristics of Overdose Events ..................................................................................................... 2.12
Number of Overdose Events ............................................................................................................ 2.12
Overdose Events and Estimated Costs by Encounter Type ............................................................. 2.12
Overdose Rates by Patient Characteristics ....................................................................................... 2.13
Overdoses by Type of Prescription Opioid Use ............................................................................... 2.16
Relationship between Overdose Risk and Prescribed Dose: Results of the 90-Day Exposure Window
Model ............................................................................................................................................... 2.16
Pharmacy Shopping among Long-Term Opioid Users ........................................................................ 2.17
Number of Patients with an Overdose Event, by Peak Number of Pharmacies ............................... 2.17
Comparison between Different Pharmacy Shopping Criteria .......................................................... 2.18
Combined Criteria: Peak Number of Pharmacies and Overlapping Prescriptions .......................... 2.19
Relationship between Potential Pharmacy Shopping and Overdose Risk ....................................... 2.21
Table 2-7. Overdoses by Type of Opioid Use ......................................................................................... 2.16
Table 2-8. Overdose Rates and Hazard Ratios by Dose Level and Predominant Drug Type .................. 2.16
Table 2-9. Comparison of Different Pharmacy Shopping Criteria in Medicaid MarketScan® Dataset . 2.19
Table 2-10. Comparison of Different Pharmacy Shopping Characteristics by Demographics, Overdose
Events and Opioid Consumption Patters ................................................................................................. 2.19
Table 2-11. Hazard Ratios of Overdose, Including Indicators for Pharmacy Shoppinga and Overlapping
Prescriptions in Medicaid MarketScan® Dataset, 2008-2010 ................................................................ 2.21
List of Figures
Figure 2-1. Regression Model Schematic: Prescriptions, Exposure Windows, and Overdose Events in
Models 1 and 2 ........................................................................................................................................... 2.7
Figure 2-2. Percentage of Supply Days for Prescription Opioids ............................................................ 2.11
Figure 2-3.Overdoses among Long-Term Users by Peak Number of Pharmacies Visited ...................... 2.18
Part 2 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – MarketScan™ Data Analysis
2.4
INTRODUCTION
To respond to the CDC’s request for a model to examine the effectiveness of Medicaid patient
review and restriction (PRR) programs, we developed a micro-simulation model of an adult
Medicaid enrollee cohort to explore the impact and the cost-effectiveness of such programs. The
model was informed by an analysis of the MarketScan® Medicaid sample, in conjunction with
the literature review presented separately.
METHODS
Data Source
This study uses MarketScan® data, a commercially available administrative claims dataset that
includes information on demographics (age, race and gender), Medicaid enrollment duration,
diagnosis, and health care utilization (i.e., prescription drugs, hospital and emergency department
visits). The study population for this analysis consisted of Medicaid beneficiaries who received
at least one opioid analgesic prescription for non-cancer pain between January 2008 and
December 2010. We excluded individuals:
With less than 24-months continuous Medicaid enrollment;
Younger than age 12 years at the start of continuous enrollment;
With history of cancer diagnosis (ICD-9 CM neoplasms 140-293.2, excluding 173.X, 210-
239 and 232);
Residing in any long-term care facilities;
Who filled any opioid prescription in the first 3 months of the continuous enrollment period
(this is necessary to exclude subjects whose time-to-event estimation is subject to left
truncation).
We identified 427,411 Medicaid beneficiaries in the MarketScan® data during the 24-month
period who met the inclusion criteria.
Definitions
Episode of Opioid Use. We defined an “episode of opioid use” as commencing with the
dispensing date of an opioid prescription with no previous prescription in the dataset, or having a
gap longer than 31 days from the end run-out date of a previous opioid prescription. “Episode
duration” is defined as the number of days from the first fill date to the end date of the last opioid
prescription with no prescription gaps exceeding 31 days after the previous refill.
Long-term Episode of Opioid Use. An episode is defined as “long-term” if the duration is longer
than 90 days with 3 or more prescriptions dispensed in that time.
Pharmacy Shopping. Pharmacy shopping defined as visiting multiple pharmacies to obtain
medically unnecessary prescription opioids and contributes to nonmedical opioid use, misuse and
abuse. Pharmacy shopping has been defined in the literature using a variety of cut-off points for
classifying a patient as having potential controlled substance misuse or mismanagement that
would warrant further evaluation. Published thresholds vary by number of pharmacies seen by a
single patient to obtain any opioid over a given time period. 1-5
Peak Number of Pharmacies. Within a long-term episode of use, we defined the “peak number
of pharmacies” visited as the maximum number of unique pharmacies IDs that appeared in
Part 2 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – MarketScan™ Data Analysis
2.5
opioid prescription claims during any 90 days in that episode. The peak number of pharmacies
visited may be a more accurate indicator of prescription opioid consumption patterns than the
total number of pharmacies visited for the entire episode, which is affected by the episode length.
In other words, long-term continuous opioid use over many months may have multiple
pharmacies due to a change in residence or a pharmacy switch, but if multiple pharmacies are
used in a shorter (90-day) period, this is more likely to represent opioid misuse or abuse.
Morphine Equivalent Dose and Average Daily Dose. Consistent with previous studies,6-10
we
compared the effects of multiple types of opioid drugs using a drug conversion method known as
the “morphine equivalent dose.” The morphine equivalent dose (MED) is calculated by
multiplying the strength of the opioid prescription by the quantity and by a drug-specific
conversion factor (expressed in milligrams morphine equivalent, or MME). The majority of these
conversation factors are based on Von Korff’s CONSORT (CONsortium to Study Opioid Risks
and Therapeutics) study.10
For details on the drug-specific conversion factors used, please refer
to the Appendix. The total MED is calculated by adding MEDs for all opioid prescriptions within
an episode. The average daily dose is the total MED divided by episode duration. The average
milligrams morphine equivalent daily dose (mgMED/d) is categorized into 4 levels: 0-
<20mgMED/d; 20-<50mgMED/d; 50-<100mg/d; and 100mg/d or more.
Overlapping Prescriptions. Overlapping prescriptions was defined as two prescriptions of the
same drug type that overlapped by 25% or more of the days prescribed and the former of the two
prescriptions had a supply time of 5 days or longer. The origin of the 25% cutoff point is from
the clinical opinions of an expert panel in which early opioid refills were defined as patients who
filled opioid prescriptions when 25 percent or more of an existing prescription should have
remained available.11
We restricted it to the same opioid category because patients could have
legitimate concomitant use of two or more different types of opioids. We required the
prescription dispensed earlier than the other have at least 5 days of supply, because the 25%
cutoff point was too sensitive for prescriptions with short supply days – a refill on the same date
as the run-off day of a previous fill with less than 5-day supply would be mistakenly considered
as overlapping prescription.
Opioid Overdose Events. Opioid overdose events were identified using inpatient and outpatient
claims data for the study population. We defined “definite cases of overdose” as claims with
ICD-9 codes indicating opioid-related poisoning (965.0, 965.00, 965.02 and 965.09) or
accidental poisoning (E935.1 and E935.2). We defined “probable cases of overdose” as claims
with ICD-9 codes indicating adverse effects of opioid use (E935.1 and E935.2) plus at least one
ICD-9 code indicating overdose-related symptoms on the same day (see Appendix for the full
list). We included both definite and probable cases in the analysis. We excluded suicidal
poisoning by opioid drugs (E950.0), poisoning undetermined whether accidentally or
purposefully inflicted (E980.0), and opioid drug dependence (304.X and 305.X). We grouped
inpatient and outpatient claims into overdose encounters and classified the encounters into 3
types: hospitalizations, ED visits, and outpatient visits (see Appendix for detailed rules for
grouping and classification). If an individual had multiple overdose encounters during his/her
continuous Medicaid enrollment period, only the earliest one (i.e. initial overdose) was counted.
Part 2 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – MarketScan™ Data Analysis
2.6
STATISTICAL ANALYSIS
Regression Models
We constructed two regression models -- the 90-day exposure window model and the episode-
based model -- to estimate the relationship between the risk of overdose and daily opioid dose.
Each regression model explored a different question.
90-Day Exposure Window
The 90-day exposure window model includes all initial overdoses regardless of the drug source
(prescribed to patient or obtained through diversion), thereby providing a more accurate
estimation of the population at risk of overdose from any prescription opioid. The exposure
window model is used in other studies7,9
to examine the population overdose risk, thus providing
cross-validation with our analytical results.
Episode-Based Model
Administrative data cannot capture the real pattern of opioid use, which includes illicit use
supported through diversion. Preliminary analysis of our study population indicated some
evidence of drug diversion contributing to overdose episodes.* Therefore, we designed the
episode-based model to examine overdoses that happened within an episode of opioid use, but
not those that happened when no documented prescription opioid was in use.
Model Settings
The episode-based model treats each episode as a separate observation period. In each episode, a
patient was exposed to opioids at one of four average daily MED levels (0-20mg/d, 20-50 mg/d,
50-100 mg/d, and >100 mg/d) for the whole episode. We used categorical rather than continuous
variables to describe average daily MED because a continuous variable model could be subject
to bias caused by patients obtaining opioids with extremely high dose. The time at risk for
overdose lasts until the end of an episode, or the day of the first overdose (if any) that occurred
within the episode, or the censoring date.†Gaps between episodes were not included in
calculating the time at risk. Exposure windows were defined as 90 days prior to an initial
overdose (including the event date) for each overdose patient, and 90 days past the fill date of the
first prescription for all other patients remaining at risk for overdose at the time of that patient’s
event.
Error! Reference source not found. depicts how the exposure windows and the time at risk are
sed in each model. Subject 1’s prescriptions are clustered into two episodes of opioid use, and
therefore Subject 1 has two observation periods in the episode-based model (model 1) but only
one in the 90-day exposure window model (model 2). Subject 3’s overdose, which occurred with
no legitimate prescription on record, is not included in the episode-based model. Subject 4 has
three exposure windows in the 90-day exposure model, because she was at risk for overdose
when three overdose events (including hers) occurred.
* We conducted a preliminary analysis in STATA which shows that some overdose happens when there was no
opioid prescription in use, which suggests opioid from other sources (probably through diversion). † The censoring date is used if the prescription lasts beyond the end of the enrollment period
Part 2 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and Prevention Policies – MarketScan™ Data Analysis
2.7
Figure 2-1. Regression Model Schematic: Prescriptions, Exposure Windows, and Overdose Events in Models 1 and 2
Subject 4 in model 1
Subject 4 in model 2
Overdose event
CensoredSupply of opioid prescription
Exposure window
30 daysSubject 1’s prescription history
Subject 1 in model 1
Subject 1 in model 2
Subject 2 in model 1
Subject 2 in model 2
Subject 3 in model 1
Subject 3 in model 2
Subject 2’s prescription history
Subject 3’s prescription history
Subject 4’s prescription history
Fill date of first opioid prescription
End of episode of opioid use
Part 2 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – MarketScan™ Data Analysis
2.8
Both models adopt Cox proportional hazards regression analysis12
to estimate the risk of
overdose as a function of average daily MED levels. In the episode-based model, a patient’s
multiple observation periods (i.e. multiple episodes) are clustered into one group. In the 90-day
exposure window model, the average daily MED level is treated as a time-varying covariate. The
regression analysis for each model is adjusted for demographic variables including gender, age,
and race as well as clinical variables including history of depression diagnosis, history of alcohol
abuse, and concurrent use of sedative/hypnotics.
Subgroup Analysis: The Role of Pharmacy Shopping in Overdose Events
Patient review and restriction programs rely on identifying opioid users at risk for misuse or
abuse based on various criteria, including (but not limited to) the number of pharmacies used, the
number of physicians providing opioid prescriptions, and the number of emergency department
visits. The MarketScan® database contained pharmacy but not physician identifiers, so we
focused on the role of use of multiple pharmacies on risk of overdose events.
Peak Number of Pharmacies
We restricted the study population to long-term opioid users because patient review and
restriction programs are not applicable to short-term opioid use, as the program is unlikely to use
prescription history in the past to regulate future prescription behavior if the period of opioid use
is short (i.e., less than 90 days based on our definition). The study population was classified by
the peak number of different pharmacies visited during a specified timeframe. The number of
pharmacies was calculated by counting the number of unique, de-identified pharmacy IDs from
the prescription claims database within a specified timeframe. The peak number of pharmacies
was the highest number of pharmacies visited for that patient. The peak number was classified
into 1, 2, 3, 4, and 5 or more pharmacies and three time periods were used - 90 days, 180 days
and 1 year. We calculated the number and percentage of patients having an opioid-related
overdose event(s) in each subgroup classified by the peak number of pharmacies based on each
of the three time periods.
Pharmacy Shopping Criteria
Based on the number of pharmacies and the timeframe, we created six different definitions for
pharmacy shopping:
1. Obtaining prescriptions from 3 or more pharmacies over a 1 year period
2. Obtaining prescriptions from 4 or more pharmacies over a 1 year period
3. Obtaining prescriptions from 3 or more pharmacies over a 180-day period
4. Obtaining prescriptions from 4 or more pharmacies over a 180-day period
5. Obtaining prescriptions from 3 or more pharmacies over a 90-day period
6. Obtaining prescriptions from 4 or more pharmacies over a 90-day period
We calculated how many long-term users met each definition of pharmacy shopping, and how
many pharmacy shoppers had one or more opioid-related overdose event during the follow-up
period. We used opioid-related overdose as a surrogate measure of opioid misuse and abuse and
calculated the diagnostic odds ratio (DOR) for each definition. Typically, DOR is a measure of
the effectiveness of a diagnostic test. DOR is defined as the ratio of the odds of testing positive if
the subject has a disease relative to the odds of testing positive if the subject does not have the
disease.13
We replaced “test” with the criterion for pharmacy shopping and “disease” with
Part 2 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – MarketScan™ Data Analysis
2.9
opioid-related overdose, thereby using DOR to assess the efficacy of each definition to identify
opioid users at high risk of overdose, and, presumably, also at high risk of abuse and misuse.
Characteristics of Prescription Fill Behaviors
We combined the two items - pharmacy shopping definition with the highest DOR (the indicator
for overlapping prescriptions) – to categorize the study population into four subgroups:
1. Patients without pharmacy shopping behavior and without overlapping prescriptions
2. Patients without pharmacy shopping behavior but with overlapping prescriptions
3. Patients with pharmacy shopping behavior but without overlapping prescription
4. Patients with both pharmacy shopping behavior and overlapping prescriptions
We calculated the demographic characteristics, overdose risk, and prescription fill patterns (in
terms of prescription frequency, dose level, and drug type by DEA classification) of each
subgroup.
Hazard Ratios for Overdose Due to Pharmacy Shopping and Overlapping Prescriptions
We examined whether pharmacy shopping behavior and overlapping prescriptions were
associated with an increased risk of opioid-related overdose. The 90-day exposure model, which
was modified to include two additional indicators for the history of pharmacy shopping and
having overlapping prescriptions, was used for this purpose. The hazard ratios of pharmacy
shopping and overlapping prescriptions were calculated.
RESULTS
Study Population Characteristics
Among the 427,411 Medicaid patients included in the study population, 69.2% were female,
68.3% were 18 years of age and older and 54.4% were white (see. The prevalence of diagnosed
depression and alcohol abuse was 5.0% and 1.2%, respectively. Long-term users‡ accounted for
21.1% of the study population (90,010 individuals). The majority of long-term users (51.5%)
were over 45 years of age, whereas only 1.9% were between ages 12 and18 years. Long-term
users had a significantly higher prevalence of depression (10.6%) and alcohol abuse (3.0%) than
the overall study population.
Table 2-1.Baseline Characteristics of the Study Population Total
(n=427,411)
Long Term Users
(n=90,010) P Value
Mean Months Enrolled 28.9 31.3 <0.001
Female 69.2 70.3 <0.001
Age
12 – 17 31.7 1.9
<0.001 18 – 29 26.0 14.6
30- 44 20.9 32.0
45 and over 21.4 51.5
Race
White 54.4 67.7 <0.001
‡ Long-term opioid users were defined as those who have at least one episode of opioid use longer than 90 days
with at least 3 prescriptions.
Part 2 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – MarketScan™ Data Analysis
2.10
Total
(n=427,411)
Long Term Users
(n=90,010) P Value
Black 37.1 24.7
Hispanic 1.9 1.0
Other 6.6 6.6
Depression diagnosis,% 5.0 10.6 <0.001
Alcohol abuse, % 1.2 3.0 <0.001
Note: Values are expressed as percentages
Characteristics of Prescription Opioid Use
Opioid Prescriptions by Drug Type
Error! Reference source not found. lists the number of each opioid drug prescribed during the
tudy period. Most prescriptions (79.2%) were dispensed to long-term users. Hydrocodone, a
schedule III opioid, is by far the most commonly prescribed opioid in both the overall study
population (45.7%) and the subset of long-term users (44.6%). The most commonly prescribed
short-acting§ Schedule II opioid is oxycodone (15.1% and 14.9% in the overall population and
long-term users, respectively). Prescriptions for long-acting§ Schedule II drugs are higher among
long-term users (11.5%) than in the rest of the study population (0.5%, data not shown).
Table 2-2. Commonly Prescribed Opioids in a Sample of the U.S. Adult Medicaid Population,
Part 2 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – MarketScan™ Data Analysis
2.17
Hazard ratios (HR) for dosage levels were adjusted for gender, age, race, history of depression,
history of alcohol abuse, and concurrent use of sedative/hypnotics. In both the episode-based
model and the 90-day exposure model, higher dosage level was associated with a significantly
increased risk of overdose. Patients who received an average daily dose of 100mg/d or higher
had almost a 5-fold increase in overdose risk (4.76 [95% CI: 3.83 to 5.91] in the episode-based
model and 4.89 [95% CI: 3.67 to 6.52] in the 90-day exposure model, compared with the group
of patients who received the lowest dosage level (1 to <20 mg/d). Compared with the same group
(1 to <20 mg/d), patients who did not receive any recent prescription had hazard ratios of 0.31
(95% CI: 0.23 to 0.44) in the 90-day exposure model (note that no opioid use is not included in
the episode-based model).
The episode-based model also included the most frequently prescribed opioid type in an episode
(i.e., predominant opioid type) as covariates. Compared with the episodes where hydrocodone
was the most frequently prescribed opioid, episodes with a different predominant drug type did
not show any statistically significant difference in terms of adjusted overdose risk, except for
morphine in extended-release format which showed a 35% increase (HR: 1.35, 95% CI: 1.06 to
1.72) in overdose risk. Overdoses involving codeine had the lowest unadjusted overdose rate of
5.64 per 1,000 person-years (95% CI: 3.88 to 7.92 per 1,000 person-years), whereas its adjusted
risk was very close (HR: 1.02 [95% CI: 0.71 to 1.46]) to that of the reference drug, hydrocodone.
Pharmacy Shopping among Long-Term Opioid Users
The study population in this subgroup analysis is restricted to the 90,010 long-term users.**
The
main analysis found that approximately one in five opioid users met criteria for long-term use
(21.1% of all opioid users). Inclusion of short-term users would substantially bias the
comparison between shoppers and non-shoppers.
Number of Patients with an Overdose Event, by Peak Number of Pharmacies
Over a 1-year period, the numbers of patients using 3, 4, and 5 or more pharmacies were 15,901
(16.8%), 9,766 (10.8%), and 19,409 (20.6%), respectively. That means nearly half (49.2%) of
long-term users visited three or more different pharmacies in a 1-year period. If the time span is
narrowed to any 180 consecutive days, the numbers of patients using 3, 4, and 5 or more
pharmacies changed to 16,806 (18.7%), 9,012 (10.0%), and 12,339 (13.7%), respectively. After
further restricting the time setting to any 90 consecutive days, the corresponding numbers and
percentages fell to 15,647 (17.3%), 7,564 (8.4%), and 5,519 (6.1%), respectively. That means
only 14.5% of the long-term users used 4 or more pharmacies during any 90 consecutive days,
compared with 31.2% over a 1-year period.
In all three time periods (90-days, 180-days, and 1 year), the percentage of patients having
opioid-related overdoses in each category monotonically increases as the peak number of
pharmacies increases. However, the most dramatic increase is seen among patients using 4 and 5
or more pharmacies in the 90-day period. We found that 3.4% of patients who ever visited 4
pharmacies in any 90 consecutive days and 5.4% of patients who ever visited 5 or more
pharmacies in any 90 consecutive days had at least 1 overdose event in the follow-up time,
compared with only 0.4% to 1.5% among those visiting 1 to 3 pharmacies in any 90 consecutive
**
Those who had at least 1 episode of opioid use for 90 days or longer with at least 3 opioid prescriptions dispensed
in that episode.
Part 2 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – MarketScan™ Data Analysis
2.18
days. In the 180-day and 1-year timeframes, the difference between the high and low numbers of
pharmacies visited was not as significant.
Figure 2-3 presents the distribution of the number of patients (bars, left y axis) and the
percentage of overdose (dots, right y axis) in each category based on the peak number of
pharmacies. Three sets of bars and dots represent three different time spans (1 year, 180-days
and 90-days) in which the numbers of pharmacies were counted.
Figure 2-3.Overdoses among Long-Term Users by Peak Number of Pharmacies Visited
Comparison between Different Pharmacy Shopping Criteria
A comparison of the six different definitions of pharmacy shopping is listed in Table 2-9.
Corresponding to the information provided in Figure 2-3, 49.2% (44,266 of 90,010) of sample
Medicaid recipients are categorized as pharmacy shoppers when using the definition of 3 or more
pharmacies during a one-year period; whereas only 14.5% (13,083) are eligible when using the
definition of 4 or more pharmacies over 90 consecutive days. This shows that the more
restrictive criteria is less sensitive in identifying overdose cases – less than half (47.0%) of
overdosed subjects are included in the pharmacy shopping group defined as “≥4 pharmacies in
any 90 consecutive days,” compared with nearly 70% (69.9%) when defined as “≥3 pharmacies
0%
1%
2%
3%
4%
5%
6%
0
5,000
10,000
15,000
20,000
25,000
30,000
35,000
1 2 3 4 ≥ 5
Percentage o
f Overd
ose
(in
dicated
in d
ots)
Nu
mb
er o
f Lo
ng-
term
Use
rs
(i
nd
icat
ed in
bar
s)
Peak Number of Pharmacies Visited
# of pharmacies in 1 year setting # of pharmacies in 180-day setting # of pharmacies in 90-day setting % of overdose for 1-year setting % of overdose for 180-day setting % of overdose for 90-day setting
Part 2 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – MarketScan™ Data Analysis
2.19
in a year.” To quantitatively compare the different criteria, the diagnostic odds ratio (DOR)††
is
calculated for each criterion, a higher value of which is indicative of better test performance. The
criterion of “≥4 pharmacies in any 90 consecutive days” has the highest value of 5.23.
Table 2-9. Comparison of Different Pharmacy Shopping Criteria in Medicaid MarketScan®
Dataset Pharmacy Shopping
Criteria
Eligible
Recipients
Overdose
Events
% Overdose among
Eligible Recipients
% of Total
Overdose DOR
1-year setting,
≥ 3 pharmacies 44,266 825 1.86% 69.86% 2.38
1-year setting,
≥ 4 pharmacies 29,175 620 2.13% 52.50% 2.30
180-day setting,
≥ 3 pharmacies 38,157 818 2.14% 69.26% 3.05
180-day setting,
≥ 4 pharmacies 21,351 599 2.81% 50.72% 2.73
90-day setting,
≥ 3 pharmacies 28,730 788 2.74% 66.72% 4.28
90-day setting,
≥ 4 pharmacies 13,083 555 4.24% 46.99% 5.23
Combined Criteria: Peak Number of Pharmacies and Overlapping Prescriptions
To identify the group of prescription opioid users at the highest risk for an overdose event (and
perhaps those most likely to benefit from a patient review and restriction program), we combined
our results from prescription utilization patterns and peak pharmacy use. As the criterion of “4
or more pharmacies in any 90 consecutive days” had the highest DOR, we used it to define high
risk pharmacy shopping behavior. In addition to this definition, we used an indicator for
overlapping opioid prescriptions in order to further distinguish high risk patterns. Table 2-10
shows that, among the 90,010 long-term users, 6,024 (6.7%) had both pharmacy shopping
behavior and overlapping prescriptions; 3,885 (4.3%) did not exhibit pharmacy shopping
behavior, but did have overlapping prescriptions; 7,059 (7.8%) did have shopping behavior but
no overlapping prescriptions; the rest (81.1%) had neither shopping behavior nor overlapping
prescriptions.
Table 2-10. Comparison of Different Pharmacy Shopping Characteristics by Demographics,
Overdose Events and Opioid Consumption Patters
No shopping,
No overlapping RX
No shopping,
Overlapping RX
Shopping,
No overlapping RX
Shopping,
Overlapping RX
Demographics
Number 73,042 3,885 7,059 6,024
Mean age 44.1 45.6 38.0 42.0
††
DOR is a measure of the effectiveness of a diagnostic test. DOR is defined as the ratio of the odds of testing
positive if the subject has a disease relative to the odds of testing positive if the subject does not have the disease.13
We replaced ‘test’ with criterion for pharmacy shopping and ‘disease’ with opioid-related overdose, thereby using
DOR to assess the efficacy of each definition to identify opioid users at high risk of overdose, and, presumably, also
at high risk of abuse and misuse.
Part 2 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Butalbital + codeine (with or without aspirin/acetaminophen/
ibuprofen)
0.15
Dihydrocodeine (with or without
aspirin/acetaminophen/ibuprofen)
0.25
Pentazocine (with or without aspirin/acetaminophen/ibuprofen) 0.37
Buprenorphinea
25.0-40.0
Butorphanol 7.00
Schedule II Short-Acting
Morphine sulfate 1.00
Codeine sulfate 0.15
Oxycodone (with or without aspirin/acetaminophen/ibuprofen) 1.50
Hydromorphone 4.00
Meperidine hydrochloride 0.10
Oxymorphone 3.00
Fentanyl citrate transmucosalb 0.125
Tapendatol short actingc
not established
Schedule II Long-Acting
Morphine sulfate sustained release 1.00
Fentanyl transdermald 2.40
Levorphanol tartrate 11.0
Oxycodone HCL control release 1.50
Methadone 3.00
Oxymorphone extended releasec
3.00
Hydromorphone extended releasec
5.00
Tapentadol extended releasec
not established
Sources: Von Korff et al (2008); FDA Blueprint for Prescriber Education for Extended-Release and Long-Acting
Opioid Analgesics (2012)
Note: The majority of these conversation factors are based on Von Korff’s CONSORT (CONsortium to Study
Opioid Risks and Therapeutics) study. Opioids delivered by pill, capsule, liquid, transdermal patch, and
transmucosal administration were included in the data, but opioids formulated for administration by injection or
suppository were not included. aBuprenorphine is typically used for opioid detoxification and maintenance
16
bTransmucosal fentanyl conversion to morphine equivalents assumes 50% bioavailability of transmucosal fentanyl
and 100 micrograms transmucosal fentanyl is equivalent to 12.5 to 15 mg of oral morphine. cData for oxymorphone, hydromorphone and tapentadol obtained from FDA Blueprint for Prescriber Education for
Extended-Release and Long-Acting Opioid Analgesics
dTransdermal fentanyl conversion to morphine equivalents is based on the assumption that one patch delivers the
dispensed micrograms per hour over a 24 hour day and remains in place for 3 days.
Part 2 of 3: Prevalence of Prescription Opioid Misuse and Abuse, Related Outcomes, and
Prevention Policies – MarketScan™ Data Analysis
2.24
ICD-9 Codes Indicating Overdose-Related Symptoms
276.4 Mixed acid–base balance disorder
292.1 Drug-induced psychotic disorders (including 292.11 and 292.12)
Data Sources .......................................................................................................................................... 3.4
Definitions for the MarketScan®
Data Analysis ..................................................................................... 3.6
Overview: The Micro-Simulation Process and Simulation Process ...................................................... 3.7
MODEL CALIBRATION ....................................................................................................................... 3.11
Number of Prescriptions ...................................................................................................................... 3.12
Days’ Supply, Dose Level, and Generic Drug Type ............................................................................ 3.13
Distribution of Number of Pharmacies Visited .................................................................................... 3.14
Demographic Characteristics and Opioid Use Patterns of Potential Pharmacy Shoppers ................... 3.14
Model Limitations ................................................................................................................................ 3.33
APPENDIX I ........................................................................................................................................... 3.35
The Simulation Process ........................................................................................................................ 3.35
Step 1: Simulate the Basic Individual Profile .................................................................................. 3.35
Step 2: Simulate Predominant Drug Type in An Episode of Drug Use ........................................... 3.35
Step 16: Check Individual Eligibility for the Patient Review and Restriction Program .................. 3.46
Step 17: Summarize the Cost and Health Outcomes of the Simulated Cohort ................................ 3.46
APPENDIX II .......................................................................................................................................... 3.47
Table 3-13. Annual Health Impact of the PRR Program in a Population of 10,000 Long-Term Users under
Different Eligibility Scenarios* ............................................................................................................... 3.29
Table 3-14. Cost Analysis of the PRR Program under Different Eligibility Scenarios ........................... 3.32
Table 3-15. Most Frequently Used Opioid Types in Market Scan Data .................................................. 3.36
Table 3-16. Hazard Ratios for Prescription Opioid Overdose ................................................................. 3.45
Table 3-17. Distribution of Overdose and Cost Estimates ....................................................................... 3.45
List of Figures
Figure 3-1. Overview of ADOPT Micro-Simulation Model ...................................................................................... 3.4
Figure 3-2. Overview of the Simulation Process ........................................................................................................ 3.9
Figure 3-3. Simulated Individual’s Prescription History from ADOPT Model for Comparison with Table 3-8 Data
Figure 3-4. ADOPT Output: Program Summary ...................................................................................................... 3.25
Figure 3-5. Example of Random Sampling of Predominant Drug Type .................................................................. 3.37
Figure 3-6. Subsequent Episodes of Opioid Use ...................................................................................................... 3.38
Figure 3-7. Example of Opioid Type Distribution Table, for Predominant Drug Type of Hydrocodone and Episode
Length between 180- and 364-Days ......................................................................................................................... 3.40
3.4
INTRODUCTION
The Approaches to Drug Overdose Prevention Analytical Tool (ADOPT) is an evidence-based
tool created by UC Davis to help inform policy decisions regarding policies to prevent
prescription opioid misuse/abuse and consequent adverse health outcomes. Specifically, it is an
Excel-based, micro-simulation model that simulates patterns of prescription opioid use by
Medicaid recipients to evaluate associated health outcomes and costs.* It compares the
counterfactual scenarios of implementing a prescription drug misuse/abuse prevention policy
versus the absence of such a policy, and evaluates the cost and health impact of the policy. The
model’s interactive features allow users to customize the population demographics and policy
details, and perform a "what-if" analysis to project the outcomes of the specified policy within
that population (see Figure 3-1). Although ADOPT has the potential to analyze and compare
different approaches to drug misuse/abuse prevention (such as prescriber/patient education or
monitoring strategies), the current version focuses on the Medicaid patient review and restriction
(PRR) program sometimes referred to as a “lock-in” program).
This section of the report (Part 3) explains the design, calibration, cost estimation, and basic
operation of the model. It also provides an example of the projected program policy results
(using MarketScan® Medicaid data to inform the ADOPT model) on prescription opioid use,
related over-dose events, and their cost impact. The strength of this model is its ability to be
customized to state-specific data, as these findings will lead to more valid conclusions for state-
specific populations and policies than the example provided here.
Figure 3-1. Overview of ADOPT Micro-Simulation Model
User-defined patient population
characteristics
User-defined Policy Intervention
Individual prescription history, without policy
intervention
Individual prescription history, with policy
intervention
Simulate
Apply the policy to simulated cohort
Opioid and medical costs, health outcomes , without
policy intervention
project
Opioid and medical costs, health outcomes , with
policy intervention
project
CompareIncremental cost and
health outcome of the policy intervention
METHODS
Data Sources
The ADOPT model allows users to specify the values of some of input parameters, while others
are not user-modifiable. The default values of the modifiable parameters and the values of non-
modifiable parameters come from multiple sources, including the analysis of MarketScan®
* The model was informed by an analysis of the MarketScan® Medicaid dataset and a literature review; these
analyses are presented in Parts 1 and 2, respectively, of this report.
3.5
Medicaid data, the literature (for prescriber information), content experts (for PRR program
cost),† and government documents (for PRR program eligibility criteria) (see Table 3-1).
Table 3-1.ADOPT Input Parameters: Data Sources and Modifiability of Input Category Input Parameter Type Source User Modifiable?
Number of Above Conditions Required for Eligibility Yes
* Risk factors include depression diagnosis, alcohol use, and concurrent sedative/hypnotic drug use.
† Cost estimations are based on estimates from Oklahoma and Washington state PRR programs, per
content expert discussions with the CDC; Jones, C.M., Roy, K. Email correspondence, August 2012
The MarketScan® Medicaid data are a commercially available administrative claims dataset that
include information on demographics (age, race and gender), Medicaid enrollment duration,
diagnosis, and health care utilization (i.e., prescription drugs, hospital and emergency department
visits). It contains approximately 7 million Medicaid enrollees from multiple states (the number
of states varies by year; in 2012 there were 12 states). The study population for this analysis
consisted of Medicaid beneficiaries who received at least one opioid analgesic prescription for
non-cancer pain between January 2008 and December 2010. We excluded individuals:
with less than 24 months continuous Medicaid enrollment;
† Cost estimations are based on estimates from Oklahoma and Washington state PRR programs, per content expert
discussions with the CDC; Jones, C.M., Roy, K. Email correspondence, August 2012
3.6
younger than age 12 years at the start of continuous enrollment;
with history of cancer diagnosis (ICD-9 CM neoplasms 140-293.2, excluding 173.X, 210-
239 and 232);
residing in any long-term care facilities;
having any opioid prescription filled in the first 3 months of the continuous enrollment
period. This last criterion enabled us to accurately identify the period of continuous opioid
use.
We identified 427,411 Medicaid beneficiaries in the MarketScan® data during the 24 month
period who met the inclusion criteria.
Definitions for the MarketScan®
Data Analysis
Episode of Opioid Use. An “episode of opioid use” commenced with the dispensing date of an
opioid prescription with no previous opioid prescription in the dataset, or having a gap longer
than 31 days from the end run-out date of a previous opioid prescription. “Episode duration” was
the number of days from the first fill date to the end date of the last opioid prescription with no
prescription gaps exceeding 31 days after the previous refill.
Long-term Episode of Opioid Use. An episode was “long-term” if the duration is longer than 90
days with 3 or more prescriptions dispensed, concurrently or in succession, in that time.
Pharmacy Shopping. Pharmacy shopping was defined as visiting multiple pharmacies to obtain
prescription opioids, which contribute to medically unnecessary opioid use, misuse and abuse.
Published thresholds (for identifying misuse/abuse) vary by number of pharmacies visited by a
single patient to obtain any opioid over a given time period.1-5
Peak Number of Pharmacies. Within a long-term episode of opioid use, we defined the “peak
number of pharmacies” visited as the maximum number of unique pharmacy IDs that appeared
in opioid prescription claims during any 90 days in that episode. The peak number of
pharmacies visited may be a more accurate indicator of prescription opioid consumption patterns
than the total number of pharmacies visited for the entire episode, which is affected by the
episode length. In other words, long-term, continuous opioid use over many months may include
multiple pharmacies due to a change in residence or a pharmacy switch, but multiple pharmacies
used in a shorter (90-day) period are more likely to represent opioid misuse or abuse.
Morphine Equivalent Dose and Average Daily Dose. Consistent with previous studies,6-10
we
compared the effects of multiple types of opioids using a drug conversion method known as the
“morphine equivalent dose.” The morphine equivalent dose (MED) is calculated by multiplying
the strength of the opioid prescription by the quantity and by a drug-specific conversion factor
(expressed in milligrams morphine equivalent, or MME). The majority of these conversion
factors are based on Von Korff’s CONSORT (CONsortium to Study Opioid Risks and
Therapeutics) study.10
The total MED was calculated by adding MEDs for all opioid
prescriptions within an episode. The average daily dose was the total MED divided by episode
duration. The average daily dose was categorized into 4 levels: 0-<20mg/d; 20-<50mg/d; 50-
<100mg/d; and 100mg/d or more.
Overlapping Prescriptions. Overlapping prescriptions were two prescriptions of the same opioid
type, one of which had a supply for 5 days or longer, that overlapped by 25% or more of the days
3.7
prescribed. The 25% cutoff point originated from the clinical opinions of an expert panel in
which early opioid refills were defined as patients who filled opioid prescriptions when 25
percent or more of an existing prescription should have remained available.11
We restricted it to
the same opioid category because patients could have legitimate concomitant use of two or more
different types of opioids. We required the earliest prescription dispensed to have at least 5 days
of supply, because the 25% cutoff point was too sensitive for prescriptions with short supply
days – a refill on the same date as the run-out day of a previous fill with less than 5-day supply
would be mistakenly considered as overlapping prescription.
Opioid Overdose Events. “Opioid overdose events” were based on inpatient and outpatient
claims data for the study population. We defined “definite cases of overdose” as claims with
ICD-9 codes indicating opioid-related poisoning (965.0, 965.00, 965.02 and 965.09) or
accidental poisoning (E935.1 and E935.2). We defined “probable cases of overdose” as claims
with ICD-9 codes indicating adverse effects of opioid use (E935.1 and E935.2) plus at least one
ICD-9 code indicating overdose-related symptoms on the same day (see Appendix II for full
list). We included both definite and probable cases in the analysis. We excluded suicidal
poisoning by opioid drugs (E950.0), poisoning undetermined whether accidentally or
purposefully inflicted (E980.0), and opioid drug dependence (304.X and 305.X). We grouped
inpatient and outpatient claims into overdose encounters and classified the encounters into 3
types: hospitalizations, ED visits, and outpatient visits (see Appendix II for detailed rules for
grouping and classification). If an individual had multiple overdose encounters, only the earliest
one (i.e. initial overdose) was counted.
Overview: The Micro-Simulation Process and Simulation Process
We chose a micro-simulation model to study a Medicaid patient review and restriction (PRR)
program. Although micro-simulation models are complex and time consuming to build and to
run, they accommodate heterogeneity better than cohort models. For example, in some disease-
focused models, the individual heterogeneity may not be important; the disease prevalence,
incidence and mortality rates are based on population-level data (or age/gender stratified) and
using a micro-simulation may not add enough additional information to justify the additional
effort and cost required. In the case of evaluating prescription drug abuse prevention policies, it
is critical to account for the individual heterogeneity. Individuals differ in the types of drugs
used, dosage, length of drug use, number of pharmacies/prescribers used to obtain drugs, and so
on. Whether an individual meets the criteria of the preventive program depends on his/her
personal behavior. A cohort model cannot evaluate the cost and efficacy of a policy driven by
individual behavior. For this reason, we built the ADOPT as a micro-simulation model.
Micro-simulation is a valuable tool used to project the impact of a policy. Micro-simulation
modeling is often rigid in its evaluation of a pre-determined range of policy options in a specific
population (often at a national level) with a number of fixed assumptions; the conclusion applied
to another policy context may not be generalizable. To make our analysis more flexible, timely,
and relevant to state-specific concerns, we designed the model interface to allow users to
customize the analysis. The model is Excel-based and can be used by any computer running
Excel 97 or more recent versions.
For each simulated Medicaid enrollee, the model starts with simulating the basic profile
including age, gender, race, and overdose-related risk factors (such as depression diagnosis,
3.8
alcohol use, and concurrent sedative/hypnotic drug use), based on user-defined distributions of
population characteristics (See Figure 3-2, step 1). With the profile information, ADOPT then
simulates the characteristics of each episode of opioid use for each enrollee. This includes the
predominant drug type (i.e., the most frequently prescribed drug in an episode [step 2]), the
3.9
Figure 3-2. Overview of the Simulation Process
-daily dose levelandIn a specified length of time -number of pharmacies -number of prescribers-number of RX
-age-gender-race-risk factors
Simulate individual profile (step 1)
-predominant drug type (step 2)-Length of episode (step 3)-whether have concurrent drug use (step 4)-whether have overlapping RX (step 5)
Simulate information about initial episode of opioid use
-opioid drug type of each RX (step 7)-more details: strength, master form, supply days, etc (step 8)-RX date relative to the start date of the episode (step 9)-Pharmacy ID of each RX (step 10)-Prescriber ID of each RX (step 11)
Simulate prescriptions obtained with in an episode
Simulate information about subsequent episodes (step 6)
- opioid drug use in subsequent episodes is correlated with opioid drug use in the previous episode
Repeat the process for subsequent episode(s) (step 12)
Assign absolute dates to all RX (step 13)
Summarize RX history (step 14)
Simulate opioid overdose events( step 15)
Check subject’s eligibility for PRR program (step 16)
Repeat the process for the entire cohort (step 17)
3.10
length of episode (step 3), and whether the subject has concurrent prescriptions (step 4) and/or
overlapping prescriptions (step 5) in each of the episodes.
The ADOPT assumes the characteristics of a subsequent episode are correlated with those of a
previous episode (step 6). For example, an enrollee who frequently uses hydrocodone in the
previous episode is likely to use hydrocodone more than any other opioid drugs in the following
episode. After characteristics of an episode are created, the ADOPT simulates a list of
prescriptions and adds details of each prescription through multiple steps (including drug names
in step 7, prescription details [strength, quantity, supply days, generic name, and reimbursement]
in step 8, dispensing date relative to the start date of the episode in step 9, pharmacy IDs in step
10, and prescriber IDs in step 11). This process (steps 2-11) is repeated for all episodes in step
12. After all prescriptions in all episodes of opioid use are simulated for a hypothetical enrollee,
ADOPT assigns calendar dates (i.e., converting a relative date such as “6 days after the
beginning date of the second episode” to an absolute date such as “Jan-10-2010”) to all
prescriptions and sorts them into a chronological order (step 13). The daily dose and the
maximum number of pharmacies and prescribers that a subject obtains prescriptions from, as
well as the maximum number of prescriptions in a specific time frame (e.g., any 90 days or any
60 days) are calculated (step 14). The risk of overdose events is based on the daily dose (step
15). The numbers of pharmacies, prescribers, and prescriptions are then compared with the PRR
program eligibility criteria in order to determine whether the hypothetical enrollee is eligible for
program enrollment (step 16). If eligible, the enrollee’s prescription history and the risk of
having an overdose event under the PRR program will be calculated (step 16). The ADOPT
repeats the above process (step 1-16) for the entire cohort and calculates the aggregated cost and
number of overdose events (step 17).
The simulation process used by the ADOPT model requires a series of steps, which are outlined
below and presented in greater detail in Appendix I.
1. Simulate the basic individual profile
2. Simulate the predominant drug type
3. Simulate the episode duration
4. Simulate the use of concurrent opioid use
5. Simulate the use of overlapping prescriptions
6. Simulate the subsequent episodes of opioid use
7. Simulate the drug type for each prescription in an episode of use
* The model first simulates the opioid use if there is no patient review and restriction program in place, then simulates how the opioid use is changed by the program. The items
under this sub-title show the opioid use of the potential eligible patients in the scenario of no PRR program.
† Eliminated overlapping prescriptions=overlapping with a previous prescription of the same drug type for 25% or more of the total supply days of the previous prescription
‡ Prescriptions with reduced strength or quantity are those contributing to an aggregate daily dose over 80mgmorphine equivalent on any day
3.28
The ADOPT model compares the prescription opioid use by the PRR eligible cohort in the case
of having a PRR program versus the absence of a program. Without the PRR program, the
eligible cohort has more per-person-prescriptions than the remaining, non-eligible population
under all scenarios, especially Scenarios #1 and #3. Moreover, the eligible cohort has more
frequent use of long-acting Schedule II opioids (17.6% [95% CI: 14.7%-20.5%] under Scenario
#1, for example) than the remaining, non-eligible population (5.4% [95% CI: 4.7%-6.1%]).
Expectedly, the eligible cohort also has higher numbers of monthly prescriptions and average
doses than the remaining, non-eligible population. In general, more stringent, selective eligibility
criteria (such as in Scenarios #1 and # 3) yield a smaller pool of heavier opioid users (i.e. more
monthly prescriptions and higher average dose) than less selective eligibility criteria (such as
Scenarios #4 and #5).
With the PRR program, many overlapping prescriptions are eliminated, and the prescriptions
with originally excessive doses are now reduced in strength or quantity. Table 3-12 shows the
proportions of total eliminated prescriptions under Scenarios 1 through 5 as 44.5%, 17.2%,
48.8%, 14.0%, and 16.2%, respectively. The corresponding percentages for prescriptions with
reduced strength or drug quantity are 11.5%, 13.2%, 13.8%, 13.8%, and 16.0%. These values
show that the percentages of prescriptions with reduced strength or drug quantity are comparable
between all scenarios, whereas Scenarios #1 and #3 have significantly higher percentages of
eliminated prescriptions, likely because they target heavier opioid users than other scenarios. The
number of monthly prescriptions is reduced under all scenarios. The highest reduction is seen
under Scenario #3, from 3.13 (95% CI: 2.64-3.62) prescriptions per month to 1.46 (95% CI:
1.15-1.77), followed by Scenario #1 where the number is reduced from 2.79 (95% CI: 2.63-2.95)
to 1.33 (95% CI: 1.14-1.52). The reduction under Scenarios #2, #4, and #5 is relatively small.
The average dose is reduced to a comparable level (ranging from 44.5 to 50.2 g morphine
equivalent per day) across all scenarios, with Scenarios #1 and #3 showing a relatively greater
reduction in dose.
Impact on Opioid Overdose-Related Events
Table 3-13 shows baseline estimates of annual opioid-overdose-related outpatient visits, ED
visits, hospitalization, and deaths and compares these outcomes with the estimated reduction in
each event according to different PRR program criteria. In general, the less selective scenarios
result in a greater reduction in these events. However, if the absolute reduction is divided by the
number of program enrollees, the more selective scenarios have a greater per-person reduction,
indicating that the more selective scenarios better identify the patients at highest risk of an
overdose event.
3.29
Table 3-13. Annual Health Impact of the PRR Program in a Population of 10,000 Long-Term Users under Different Eligibility
Cost Analysis of Different Patient Review and Restriction Scenarios
This cost analysis of a patient review and restriction (PRR) program is considered conservative
because the model does not consider additional cost-savings achieved through reductions in
office or emergency department visits to obtain opioid prescriptions or from “external” effects
such as reduced drug diversion and, consequently, reduced overdose risk for others. At this stage,
we lack data about the cost and frequency of office/emergency department visits for obtaining
excessive opioid prescriptions. To our best knowledge, there has been no evidence regarding the
extent to which patient review and restriction programs can help to control access to prescription
opioids through diversion. Therefore, this cost analysis only considers cost-saving achieved
through reduced prescription opioid use and reduced overdose-related medical costs attributable
to the patient holding the prescription.
We included administrative costs for the program, which will vary by state, by year and by
program caseload. Our assumed program costs were derived from email communication with
content experts.**
The emails gave an estimated labor cost of $700,000 for the Washington state
PRR program in current year with a caseload of 3,800. In the ADOPT model, we assumed a
fixed annual program cost of $300,000, as well as assigning a variable cost of $200 per program
enrollee to represent the additional labor and material expenditures that increases as the program
caseload increases. Using such setting, the modeled program cost for Washington state is about
$1 million ($300,000+3,800*$200=$1,060,000), which is close to the real program cost if other
cost components (besides labor cost) were taken into account.
The results of the cost analysis are shown in Table 3-14. Without a PRR program, the estimated
total prescription opioid costs and the overdose-related medical costs are approximately $4.59
million (Standard Error: ±0.24) and $0.91 million (SE: ±0.11) per year per 10,000 long-term
users, respectively. The major cost savings of implementing the PRR program are attributable to
reduced opioid expenditure. For example, under Scenario #5, the cost savings for opioids and
overdose-related medical services are around $0.73 million (SE: ±0.17) and $0.45 million (SE:
±0.64) per 10,000 long-term users annually. In a state with 10,000 long-term users, a PRR
program implemented under Scenarios #2, #4 or #5 saves money, whereas programs under
Scenarios #1 and #3 cost more than not implementing the program.
Although Scenarios #2, #4 and #5 have the highest overall savings, their average savings per
enrollee are much lower than those achieved in Scenarios #1 and #3. On average, each program
enrollee under Scenarios #1 and #3 saves $1,395 (SE: ±211) and $2,251 (SE: ±352), compared
with less than $500 under Scenarios #2, #4 and #5. Based on average savings, we calculated the
number of enrollees needed for the PRR program to break even (i.e., beyond which, the program
starts to yield positive total savings). Under Scenarios #1 and #3, the program needs to enroll 251
and 146 patients, respectively; under Scenarios #2, #4, and #5, the program must enroll more
than 1,000 patients.
We also calculated the total number of prescription opioid users needed for the PRR program to
break even. This number is important because PRR program staff can compare this threshold
with the actual number of Medicaid opioid users in the state to project whether the program has a
**
Jones, C.M., Roy, K. Email correspondence, August 2012
3.31
large enough pool of users to have a positive financial impact. To determine the minimal number
of Medicaid prescription opioid users needed to break even requires the following information:
1) how many the PRR program enrollees are needed to break even 2) what proportion of opioid users are long-term users 3) how many users per 10,000 long-term users meet the PRR program criteria
3.32
††
The calculation is based on the assumption that the PRR program has annual fixed cost of $300,000 and variable cost of $200 per enrollee per year on top of it. ‡‡
numbers in parentheses with “±” are standard error; numbers in parentheses in a range format such as “234-281” are the estimated range based on the upper and lower bounds of