10102_Influence Of Medicare Formulary Restrictions On Evidence-Based Prescribing Practices

Yale University
Yale University
EliScholar – A Digital Platform for Scholarly Publishing at Yale
EliScholar – A Digital Platform for Scholarly Publishing at Yale
Yale Medicine Thesis Digital Library
School of Medicine
January 2019
Influence Of Medicare Formulary Restrictions On Evidence-Based
Influence Of Medicare Formulary Restrictions On Evidence-Based
Prescribing Practices
Prescribing Practices
Aishwarya Vijay
Follow this and additional works at: https://elischolar.library.yale.edu/ymtdl
Recommended Citation
Recommended Citation
Vijay, Aishwarya, “Influence Of Medicare Formulary Restrictions On Evidence-Based Prescribing Practices”
(2019). Yale Medicine Thesis Digital Library. 3961.
https://elischolar.library.yale.edu/ymtdl/3961
This Open Access Thesis is brought to you for free and open access by the School of Medicine at EliScholar – A
Digital Platform for Scholarly Publishing at Yale. It has been accepted for inclusion in Yale Medicine Thesis Digital
Library by an authorized administrator of EliScholar – A Digital Platform for Scholarly Publishing at Yale. For more
information, please contact elischolar@yale.edu.
1

Influence of Medicare Formulary Restrictions on Evidence-Based
Prescribing Practices

A Thesis Submitted to the Yale University School of Medicine in Partial
Fulfillment of the Requirements for the Degree of Doctor of Medicine

By
Aishwarya Vijay
2020

2

Abstract
Controlling the cost of prescription drugs is integral to improving health outcomes, and
patient access and adherence to treatment. While prescription drugs can often provide essential
therapeutic benefit, previous studies have suggested that inappropriate prescription drug use is a
principal cause of adverse drug events as well as abuse and diversion of drugs. Thus, balancing
the benefits and harms to promote appropriate prescription drug use is an essential component of
healthcare delivery in the United States. There are multiple ways appropriate prescription drug
use is promoted. Black-box warnings and drug labeling controlled by the FDA as well as
guidelines released by the CDC, such as the 2013 guidelines released during the opioid epidemic,
aim to promote appropriate prescription at a population level. At a patient-level, drug formularies
have multiple strategies in place to promote safe and cost-effective prescribing of individual
medications.
The Center for Medicare & Medicaid Services (CMS) makes use of prescription drug
formularies that are used for the coverage of around 17% of the US population. These
formularies have uniformly adopted utilization management strategies, such as quantity limits,
prior authorization, and step therapy, in order to promote safe, evidence-based and cost-effective
prescribing. These strategies are in place to impact drug prescription rates as well as to
incentivize use of biological or therapeutically interchangeable generics over brand-name drugs.
Thus far, the implementation of utilization management strategies for commonly prescribed
drugs has not been thoroughly studied.
This study presents three main analyses conducted and published in the peer reviewed
literature during my time in medical school. The first characterized the change in opioid
prescription versus non-opioid analgesics in both the outpatient and emergency room setting in
3

the context of the 2013 CDC guidelines encouraging prescription on non-opioid analgesic
alternatives. We found that overall rates of pain medication prescribing were high and that opioid
pain medication prescription increased in the outpatient setting only, whereas non-opioid pain
medication prescribing increased in both the outpatient and ED settings, an area that has not been
previously reported or well-investigated.
The second study characterized how Medicare formulary restrictions were applied to
opioid “potentiators”, which are commonly used in conjunction with opioids and increase
patients’ risk of adverse events. We found that from 2013-2017, Medicare prescription drug plan
formularies had relatively unchanged rates of benzodiazepine, non-benzodiazepine sedative-
hypnotic, and gabapentinoid coverage with small increases in use of quantity limits, and that
more than a quarter of formularies provided unrestrictive coverage of these potentially unsafe
opioid potentiators in 2017.
The third and final study herein presents a more global analysis of whether Medicare
used formulary restrictions to promote prescription of therapeutically interchangeable generics
over the top 100-grossing brand-name drugs in light of the 2020 CMS plans for an indication-
based formulary design. We showed that a substantial portion of CMS formularies provided
similarly restrictive coverage of brand-name drugs and their therapeutically interchangeable
generics, including the same tier placement or utilization management, thereby missing
opportunities to incentivize prescribing of less costly generics.
Overall, the results of this comprehensive study on safe and cost-effective drug
prescription showed that while current formulary design includes opportunities to reduce costly
and potentially unsafe prescribing, the impact of these tools is sub-optimal. These results
highlight the need for both physician and patient education on the utility of the formulary
4

restriction strategies. On a larger scale, it suggests that these strategies alone may not be
sufficient to reduce over-prescription of potentially unsafe drugs like opioid potentiators, or to
incentivize prescription of cost-saving generics over brand-name drugs. The Center for
Medicare & Medicaid Services (CMS) has proposed an indication-based formulary design
starting in 2020, allowing Medicare Advantage and Part D prescription drug plans to cover drugs
only for select indications, which could increase formulary negotiating power and secure more
competitive pricing. This might be the change needed in order to ensure continued patient access
to affordable and safe prescription drugs.

Acknowledgements
This thesis is the product of 3.5 years of research conducted as a medical student at Yale. When I
first started at Yale Medical School, I had just finished a global health project in Malaysia
working with marginalized prisoner and transgender populations. While there, I came to realize
that societal and financial barriers to care are important factors in determining health outcomes,
sometimes far more than the actual clinical encounter. I wanted to conduct research that effected
change at a larger, more structural level, research that would affect patients across different ages,
gender, ethnicity and ideology. This body of work addresses issues in prescribing practices at a
nationwide, policy level in order to promote cost-effective, evidence-based access to drugs for all
patients. I hope it will have a part in changing the way we prescribe medicine as a medical
community. I also aim to use it to inform my own clinical practice during and after residency and
fellowship training.
5

I would like to thank Dr. Joseph Ross for his role as both my thesis advisor and mentor during
the entirety of medical school. I am grateful for your time, advice and thoughts over these past
four years. I would also like to thank Dr. Sanket Dhruva, who has acted not only a research
mentor and collaborator but also as an invaluable source of career and life advice. I am eternally
grateful for the generous funding the Yale School of Medicine has provided through the years to
enable me to devote my time to this work. Finally, I would like to thank my parents, my sister,
Varsha, and my fiancé, Steen, for their constant support of my pursuits.
Publication Note
All research presented in this thesis was conducted during my time as a Yale medical student.
The analyses presented have been published as cited below, in addition to a publication on off-
label prescription that has not been presented in this thesis.
Vijay A, Gupta R, Liu P, Dhruva SS, Shah ND, Ross JS. Medicare Formulary Coverage of
Brand-Name Drugs and Therapeutically Interchangeable Generics. Journal of general internal
medicine. 2019 Oct 17:1-3.
Vijay A, Ross JS, Shah ND, Jeffery MM, Dhruva SS. Medicare Formulary Coverage and
Restrictions for Opioid Potentiators from 2013 to 2017. Journal of general internal medicine.
2019 Apr 15;34(4):518-20.
Vijay A, Rhee TG, Ross JS. US prescribing trends of fentanyl, opioids, and other pain
medications in outpatient and emergency department visits from 2006 to 2015. Preventive
medicine. 2019 Jun 1;123:123-9.
Vijay A, Becker JE, Ross JS. Patterns and predictors of off-label prescription of psychiatric
drugs. PloS one. 2018 Jul 19;13(7):e0198363.
6

1. INTRODUCTION
1.1 Safe and Cost-Effective Prescribing
Access to safe, cost-effective prescription drugs is integral to increasing patient
adherence, improving patient health outcomes and ultimately decreasing all-cause medical costs
1, 2. Previous studies have suggested that inappropriate prescription drug use is a principal cause
of adverse drug events (ADEs), which in turn can lead to additional physician visits,
hospitalizations, injury, deterioration of body functioning, and death 3. Inappropriate prescription
drug use on the patient side can also lead to addiction, diversion and overdose deaths 4. Thus,
balancing harms and benefits of prescription drug use by incentivizing appropriate prescription is
paramount in ensuring positive health outcomes across a broad range of patient populations.

At a population level, safe drug prescribing is controlled by the Food and Drug
administration (FDA) through labeling and black box warnings, as well as through CDC
guidelines 5. Cost-effective drug prescription can be promoted in part through the incentivization
of generic drugs over brand-name equivalents 6. At a patient level, there have been various
strategies adopted: requiring communication between pharmacist and physician at time of
dispensation, requiring prescription drug monitoring programs to be in place for high-risk
medications, and utilization management strategies incorporated within drug formulary policies
4, 7, 8.
Utilization management strategies, in theory, act to control costs of expensive branded
drugs as well as prevent over-prescription of potentially unsafe drugs. These strategies include
tiering of formularies (drugs are divided into “tiers,” with the first tier typically representing
generics at the lowest level of patient cost-sharing, and a higher tier requiring higher patient cost-
7

sharing), prior authorization (requiring physicians to obtain approval from the health plan before
prescription for coverage) and quantity limits (limiting the amount of drug a patient can receive
over a given amount of time) 9. A case study of opioid coverage among a private insurer showed
that implementing these restriction strategies lead to a 15% decrease in opioid prescribing,
suggesting that these methods can be used for their intended effect 10. Another study on
rosiglitazone, which has a black box warning on increased risk of myocardial ischemia, showed
that there was reduced rosiglitazone prescribing associated with Medicaid plans that
implemented formulary restrictions compared with plans without formulary restrictions, although
overall, these restrictions were underutilized 11.

1.2 Controlling Prescription with Restriction Strategies – Effective or Not?
Studying the impact of formulary management on drug prescription is a new and emerging field,
still understudied. Previous studies have often focused on a specific therapeutic drug class, from
anti-thrombotics to antihyperglycemic agents, or specific FDA labeling, such as black box
warnings. By and large, the results of these studies show a) that many drugs of concern remain
relatively unrestricted, b) that the restrictions had little impact on how providers managed
treatment regimens, and c) that for many drugs, brand-name and generics are treated very
similarly. All of this taken together suggests sub-optimal utilization or relative ineffectiveness of
the formulary management strategies despite pilot studies. Furthermore, even in cases where
formulary restrictions were shown to decrease prescription of targeted drugs, there was less
consensus on whether this actually affected patient costs and health outcomes 12. Table 1 shows
results from these past studies.
8

Table 1. Past Studies Examining Impact of Formulary Management Strategies on Drug
Prescription
Author
Title
Year
Therapeutic
Area Studied
Main Finding
Liang et. al
Medicare
formulary
coverage for
top-selling
biologics
2009
Top 20
Biologics from
2006-2009
–
Cost-sharing
and utilization
management of
top-selling
biologics
increased from
2006-2009,
thus decreasing
access
Samuels et. al.
Medicare
Formulary
Coverage
Restrictions for
Prescription
Opioids, 2006
to 2015
2017
Short and long-
acting opioids
(except
methadone)
–
Increasing use
of quantity
limits and, to a
lesser extent,
prior
authorization
on opioid
medications
from 2006-
2015
–
Overall, high
rates of
unrestrictive
coverage
persisted for
many opioids,
especially at
high doses,
Dhruva et. al.
Association
between FDA
black box
warnings and
Medicare
formulary
coverage
changes
2017
Nine drugs that
received black-
box warning
from 2013-
2017
–
Medicare
formularies
became more
restrictive for
half of the
drugs
–
A substantial
proportion of
formularies
remained
unrestrictive
Shaw et al.
Coverage of
Novel
2018
144 novel
therapeutic
–
Most novel
agents were
9

Therapeutic
Agents by
Medicare
Prescription
Drug Plans
Following FDA
Approval
agents
approved by
the FD between
2006-2012
covered, but
access was
often restricted
through prior
authorization
or step therapy
and was
dependent on
plan choice
Alghamdi et.
al.
Analysis of
formulary
coverage and
cost of biologic
disease
modifying anti‐
rheumatic drugs
in Medicare Part
D
2018
Biologic
DMARDS
–
Majority of
formularies
placed
restrictions on
the utilization
of biologic
DMARDs.
–
Biologic
DMARDs were
increasingly
placed in
higher
specialty tiers
that required
high cost‐
sharing
payments.
Roberto et. al.
Impact of
Formulary
Restrictions on
Medication
Intensification
in Diabetes
Treatment
2018
Second-Line
Anti-
hyperglycemics
–
Formulary
restrictions had
no statistically
significant
impact on
selection of and
days’ supply of
second-line
anti-
hyperglycemics
Dayoub et. al.
Evolution of
Medicare
Formulary
Coverage
Changes for
Antithrombotic
Therapies After
Guideline
Updates
2019
Anti-
thrombotics
(DOACs and
warfarin)
–
Formularies are
providing
increased
restrictiveness
(higher tiering)
with increasing
DOAC
coverage
10

1.3 Opioid Prescriptions – A Changing Landscape
The United States is currently facing an opioid epidemic, which began in the mid-1990s
with increased pharmaceutical marketing, as well as promotion by both hospital accrediting
bodies and official medical societies 13. Emergency department (ED) visits for opioid overdoses
rose 30% across the country from July 2016 through September 2017 14, 15. Opioid-related deaths
were five times higher in 2016 than 1999 15. In response, the Centers for Disease Control and
Prevention (CDC) issued guidelines in 2013 encouraging the replacement of opioid medications
with non-opioid alternatives to treat chronic pain 16. Despite such efforts, opioid-related harms
have been rising nationwide.
While a study of nationwide opioid prescriptions from 2002-2013 suggested that opioid
prescriptions began to decline prior to the 2013 CDC guideline announcement 17, the response to
these guidelines has not been very well studied. Samuels et al. demonstrated that prescription of
opioids through CMS formularies remained relatively unrestricted, especially at high doses and
for the particular medications that have higher rates of overdose deaths 18. Partly as a result of
these findings combined with the 2013 CDC guidelines, Medicare recently proposed formulary
changes to restrict opioid availability based on maximum daily dosage and initial fill quantity.19
In order to fully understand the impact of these findings and characterize the relationship
between formulary restriction and nationwide prescription rates, it was necessary to examine the
nationwide changes in opioid prescribing rates versus non-opioid analgesic prescriptions after
the 2013 CDC guidelines were announced, especially in an outpatient setting where formulary
restrictions are quite relevant to patient access to medication. The objective of the first study was
11

thus to investigate and compare current prescribing rates of opioid medications, including
fentanyl, and of non-opioid medications in the outpatient and emergency department settings
using a nationally-representative sample.

1.4 Opioid potentiators – a new epidemic
A currently under-recognized but important concern concurrent to the opioid epidemic is
the over-prescription of “opioid potentiator” drug classes: benzodiazepines, non-benzodiazepine
sedative-hypnotics, and gabapentinoids.20, 21 These drugs have risks when used on their own;
benzodiazepines, in particular, have the second highest overdose death rate after opioids.22 In
addition, they increase the risk of an adverse event when taken with opioids. Figure 1, taken
from the CDC, graphically shows the role of opioid and opioid potentiator co-prescription within
the umbrella of the opioid epidemic.
Figure 1. Opioid Overdose Deaths Involving Benzodiazepines (source: CDC,
Multiple Cause of Death 2009-2015).
12

Unfortunately, it appears there has been little effort to decrease prescribing of these
potentially dangerous drugs. Between 1996 and 2013, the number of adults who filled a
benzodiazepine prescription increased by 67%, and the quantity of benzodiazepines obtained
more than tripled .23 While the MMA excluded benzodiazepines in 2006 because of multiple
reported adverse effects in the elderly, they eventually gained coverage in 2014 under Part D for
any medically accepted indication 24, 25. A recent study indicates that subsequent to a 2016 CDC
guideline release recommending avoidance of concurrent opioid and benzodiazepines use, the
intensity of benzodiazepine prescription has not reduced and the rate of co-prescribing only
decreased by a small amount 26.
Overuse of the non-benzodiazepine sedative hypnotics is associated with increased
mortality and adverse outcomes such as fractures, falls and cognitive impairment.27 Nonetheless,
more than 50% of patients within hospitals may receive these medications, which are sometimes
13

continued after discharge.28 Gabapentinoids have also seen a surge in prescribing in recent years
for a broad range of pain diagnoses. In a recent study of a Medicaid managed care population,
95% of gabapentin prescribing was for off-label indications.29
Despite evidence that these medications are being increasingly prescribed and can have
devastating effects, especially in combination with opioids, examination of how Medicare
controls coverage of opioid potentiators had not been previously characterized. The aim of the
second study was to characterize Medicare formulary coverage and restriction of
benzodiazepines, non-benzodiazepine sedative-hypnotics, and gabapentinoids from 2013-2017.
1.5 Therapeutic Exchange – Incentivizing Generic over Brand Drug Prescription to Reduce
Patient Costs
U.S. prescription drug sales, excluding physician administered drugs, accounted for
nearly 10% of total healthcare spending in 2017 30. Given that generic drugs are generally less
expensive than brand-name drugs for patients, and that these lower out-of-pocket costs can
improve patient adherence, preferential prescription of generic drugs over brand-name is one
important target in improving health outcomes 1. While generic substitution is critical to
curtailing prescription drug spending, a previous study has shown that 72% of current
formularies favor pricier, branded drugs over bioequivalent generics in at least one therapeutic
area 31.
It is apparent that the incentivization of generic prescribing through formulary restriction
is not uniform across drug classes. The issue is further complicated in that not all brand-name
drugs have an approved bioequivalent generic. However, for many drugs, therapeutically
interchangeable generics are available, offering potential cost savings if substituted. Therapeutic
14

interchangeables are drugs within the same class, with similar clinical effect and safety profile,
but with a different chemical composition of the drug of interest 32- for brand-name drugs
without an approved generic bioequivalent, a TE can usually be substituted. In fact, one study
estimated that between 2010 and 2012, $73 billion could have been saved by TE substitution for
the most commonly prescribed medication classes 33.
The Center for Medicare & Medicaid Services (CMS) has proposed an indication-based
formulary design starting in 2020 34, allowing Medicare Advantage and Part D prescription drug
plans to cover brand-name drugs only for select indications. This could potentially increase
formulary negotiating power and secure more competitive pricing. The indication-based
formulary design also defines a role of the therapeutic interchangeable, as the formulary must
ensure coverage with a therapeutic interchangeable of any indication not covered by the
corresponding brand-name drug. With the new formulary design in the horizon, the third study
aimed to understand if and how 2016 Medicare prescription drug plan formularies incentivize
selection of brand-name drugs without bioequivalent generics compared to their corresponding
therapeutically interchangeable generic drugs through tier placement and utilization management
strategies.
2. STATEMENT OF PURPOSE
The purpose of this thesis is to describe three published studies that systematically
characterize the relationship between CMS formulary regulations and a) safe and evidence-based
prescribing, using opioids and opioid potentiators as a case study, b) cost-effective prescribing
using therapeutic exchanges across a broad, nationally representative drug sample.
15

Study 1: U.S. Prescribing Trends of Opioids, Fentanyl and Other Pain Medications in
Outpatient and Emergency Department Visits from 2006-2015:
Examination of national opioid versus non-opioid analgesic prescription rates before and after
release of CDC guidelines encouraging prescription of non-opioid analgesics.
Study 2: Medicare Formulary Coverage and Restrictions for Opioid Potentiators from
2013-2017:
Characterization of CMS formulary coverage, including utilization management strategies, of
opioid potentiators such as benzodiazepines, non-benzodiazepine sedative hypnotics and
gabapentinoids.
Study 3: Medicare Formulary Coverage of Brand-Name Drugs with Available FDA-
Approved Therapeutically Interchangeable Generics
Characterization of how 2016 Medicare prescription drug plan formularies incentivize selection
of top 100-grossing brand-name drugs without bioequivalent generics compared to their
corresponding therapeutically interchangeable generic drugs through tier placement and
utilization management strategies.
Medicare files provide a broad and impactful perspective on key components of health
care in the United States. Medicare is the largest national insurer, accounting for 29% of United
States’ total prescription drug spending and covering 17% of the nation’s patient population.
Thus, it has a strong impact on nationwide drug demand. In fact, Medicare coverage policies
often drive private insurance coverage decisions 9. Finally, Medicare primarily provides
prescription drug coverage to an older adult population (>65 yo) vulnerable because of the need
for more medications combined with limited or fixed incomes 3. Therefore, findings on the
16

impact of Medicare formulary restrictions on prescription drug policy are fairly nationally
representative and especially impactful regarding safe and affordable access to prescription
drugs.
3. STUDY 1 – METHODS AND RESULTS
3.1 U.S. Prescribing Trends of Opioids, Fentanyl and Other Pain Medications in
Outpatient and Emergency Department Visits from 2006-2015
Data Source
We used 2006-2015 data from the National Ambulatory Medical Care Survey (NAMCS)
and National Hospital Ambulatory Medical Care Survey (NHAMCS), which provide nationally
representative samples of office-based outpatient visits and emergency department visits,
respectively (https://www.cdc.gov/nchs/ahcd/ahcd_questionnaires.htm). NAMCS and NHAMCS
both sample non-federally employed physicians who are primarily engaged in direct patient care
– the sampling design utilizes a stratified two-stage sample, with physicians selected in the first
stage and visits in the second stage. The data provide an analytic base that serves as an important
tracking tool on ambulatory and emergency care utilization regarding national trends, medication
use, and practice patterns in the US. Samples included 390,538 visits in NAMCS and 305,570
visits in NHAMCS.
Drug Sample
To characterize pain medication prescribing, we examined the first eight medications
listed for all outpatient and ED visits, ensuring consistency across all survey years. We
constructed three indicator variables using generic names of medications: fentanyl products (i.e.,
fentanyl and droperidol-fentanyl), all opioid products other than fentanyl (including analogues),
and all other non-opioid pain medications. Opioid products other than fentanyl consisted of the
17

following medications: codeine, meperidine, methadone, alfentanil, hydromorphone, morphine,
oxycodone, pentazocine, propoxyphene, sufentanil, opium, levorphanol, oxymorphone,
butorphanol, nalbuphine, buprenorphine, hydrocodone, dihydrocodeine, remifentanil, tapentadol,
and their combined products. Other non-opioid pain medications are nonsteroidal anti-
inflammatory drugs (NSAIDs), non-analgesics, and other drugs (i.e., acetaminophen, aspirin,
diclofenac, ibuprofen, indomethacin, ketoprofen, ketorolac, naproxen, phenylbutazone,
piroxicam, tolmetin, tramadol, gabapentin, and pregabalin).
Demographics
We included a number of patient demographic and clinical covariates provided during
visits. Demographic variables included: age (<19, 19-44, 45-64, or ≥65), gender, race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, or other), primary source of payment (private, Medicare, Medicaid, or other). Medicare is a federal program that provides health coverage for US adults over the age of 65, and Medicaid is a state and federal program that provides health coverage for low-income individuals and families. Clinical variables included visit diagnosis and physician specialty. Both NAMCS and NHAMCS collect up to three visit diagnoses for each sampled visit using the International Classification of Diseases, 9th edition, Clinical Modification (ICD-9-CM) diagnostic codes. We categorized visit diagnosis into three groups: cancer-related pain diagnoses, non-cancer related pain diagnoses, and no pain-related diagnosis. For physician specialty, we distinguished between generalists (i.e., general/family practice, internal medicine, pediatrics, and obstetrics and gynecology) vs. other in NAMCS. In NHAMCS, we distinguished clinical specialty by clinical degree (i.e., MD vs. other). We also reported number of visits in the past 12 months (0, 1-2, 3-5, or ≥6), number of chronic conditions (0-1 or ≥2), and number of concomitant medications (0-5 or ≥6) prescribed in NAMCS datasets. 18 Statistical Analysis We determined the proportion of visits for which any pain medication was prescribed and examined associations with selected characteristics (e.g., age, sex, race/ethnicity, clinical comorbidities, concomitant medication use, and physician specialty), using Bonferroni-adjusted bivariate analyses. Next, we determined the proportion of visits for which any pain medication was prescribed across survey years, overall and for each pain medication class, also stratifying overall analyses by selected patient and visit characteristics. We used Chi-Square analysis to compare rates in 2006-2007 and 2014-2015. All analyses were conducted using Stata MP/6-Core version 15.1 (College Station, TX), accounting for the complex survey design and sampling weights. 3.2 U.S. Prescribing Trends of Opioids, Fentanyl and Other Pain Medications in Outpatient and Emergency Department Visits from 2006-2015 Selected characteristics of the study subjects Between 2006 and 2015, 66,987 (17.4%) of 390,538 office-based outpatient visits (nationally-representative of 961 million visits) and 134,953 (45.0%) of 305,570 ED visits (nationally-representative of 130 million visits) listed a pain medication prescription (Table 2). 56.3% of office-based outpatient visits were to primary care physicians, and of these visits, 18.3% involved a prescription for a pain medication. Among office-based outpatient visits, pain medication prescription was highest among patients aged 45-64, non-Hispanic Black patients, patients with Medicare coverage, patients receiving care from primary care physicians, and patients receiving care for a pain-related diagnosis (all p-values < 0.001). Among ED visits, pain medication prescription was highest among patients aged 19-44, males, Hispanic patients, 19 patients with private insurance, patients receiving care from MDs, and patients receiving care for a pain-related diagnosis (all p-values < 0.001). Table 2. Selected characteristics (weighted %) of visits in which pain medications were prescribed, 2006-2015 NAMCS and NHAMCS. NAMCS NHAMCS Total (column %) Pain medication prescriptio n (row %) P- value† Total (column %) Pain medication prescriptio n (row %) P- value† Sample size Unweighted sample 390,538 66,987 305,570 134,953 Weighted visits 961,261,3 67 167,349,60 6 130,155,3 21 58,568,338 Age <19 18.9 9.6 <0.001 24.1 40.0 <0.001 19-44 24.1 17.3 39.0 51.9 45-64 29.7 21.8 21.7 47.8 ≥65 27.3 18.1 15.2 31.1 Gender Female 58.5 17.6 0.045 54.9 43.6 <0.001 Male 41.5 17.1 45.1 46.1 Race/ethnicity Non-Hispanic White 71.8 17.5 <0.001 59.7 44.9 0.001 Non-Hispanic Black 10.3 19.0 22.5 44.8 Hispanic 12.5 16.9 14.6 46.3 Othera) 5.3 14.1 3.2 43.3 Primary source of payment Private 53.7 15.3 <0.001 32.6 47.8 <0.001 Medicare 25.9 20.1 18.7 35.5 Medicaid 12.5 17.7 28.8 45.3 Other 7.9 21.6 19.9 50.0 Physician specialty Generalistsb) 56.3 18.3 <0.001 - - - Otherc) 43.7 16.3 - - Clinician specialty MDs - - - 90.1 45.8 <0.001 Otherd) - - 9.9 39.6 Repeat of visits in the past 12 months 0 visit 6.9 12.4 <0.001 - - - 1-2 visits 36.4 15.7 - - 3-5 visits 31.2 18.3 - - 6+ visits 25.4 21.2 - - Chronic conditionse) - - - <2 68.2 14.7 <0.001 ≥2 31.8 23.8 Concomitant medications prescribed - - - <6 83.9 13.0 <0.001 ≥6 16.1 37.7 Visit diagnosis Cancer-relatedf) 4.7 14.9 <0.001 0.6 46.5 <0.001 20 Other pain-relatedg) 5.0 43.8 13.0 60.6 No indication 90.3 16.1 86.5 42.7 Note: † compares proportion differences by any pain prescription using a weight-corrected, Bonferroni-adjusted chi-squared statistic. a) includes Asians, American Indian/Alaska Natives (AIANs), Native Hawaiian or Other Pacific Islanders (NHOPI), or 2+ reported racial/ethnic groups; b) includes general/family practice, internal medicine, pediatrics, and obstetrics and gynecology; c) includes psychiatry, general surgery, orthopedic surgery, cardiovascular diseases, dermatology, urology, neurology, ophthalmology, otolaryngology, and others; d) includes physician assistants (PAs) and nurse practitioners (NPs); e) was based 14 chronic conditions (yes/no) collected by the NAMCS (e.g., arthritis, congestive heart failure, and diabetes); f) was based on ICD-9-CM diagnostic codes 140-239, 338.3X; and g) was based on ICD-9-CM codes 338.XX, 350.1X-350.2X, 354.4X, 355.71, 379.91. 388.7X, 719.4X, 724.1X-724.2X, 729.1X, 780.96, 786.5X, 789.XX. National prescribing trends of opioids and other pain medications The proportion of all outpatient visits in which any pain medication was prescribed increased significantly from 15.0% in 2006-2007 to 20.5% in 2014-2015 (p<0.001). Among ED visits, the proportion did not change significantly, ranging from 44.2% in 2006-2007 to 44.5% in 2014-2015 (p=0.72) (Table 3). Non-opioid pain medication prescription increased in both settings, from 9.2% to 12.6% (p<0.001) in the outpatient setting and from 26.3% to 29.2% (p=0.001) in the ED setting in 2006-2007 and 2014-2015, respectively. Table 3. Pain medication prescribing trends, 2006-2015 NAMCS and NHAMCS. Years (%) 2006- 2007 vs. 2014- 2015, P-value 200 6- 200 7 200 8- 200 9 201 0- 201 1 201 2- 201 3 201 4- 201 5 NAMCS Visits in which any pain medication prescribed 15. 0% 16. 4% 17.4 % 18.0 % 20.5 % <0.001 Visits in which any pain medication from the specific class prescribed Opioid and combined products† 5.9 % 6.8 % 7.1 % 7.6 % 8.1 % <0.001 Non-analgesics, NSAIDs, tylenol, tramadol, and non-opioid combined products‡ 10. 4% 11. 3% 12.1 % 12.8 % 14.9 % <0.001 NHAMCS 21 Visits in which any pain medication prescribed 44. 2% 45. 6% 46.8 % 44.0 % 44.5 % 0.719 Visits in which any pain medication from the specific class prescribed Opioid and combined products† 25. 1% 25. 7% 27.0 % 24.2 % 21.9 % 0.001 Non-analgesics, NSAIDs, tylenol, tramadol, and non-opioid combined products‡ 26. 4% 27. 7% 28.2 % 27.2 % 29.6 % <0.001 Note: †codeine, meperidine, methadone, alfentanil, hydromorphone, morphine, oxycodone, pentazocine, propoxyphene, sufentanil, opium, levorphanol, oxymorphone, butorphanol, nalbuphine, buprenorphine, hydrocodone, dihydrocodeine, remifentanil, tapentadol, and their combined products; ‡includes gabapentin and pregabalin for non-analgesics, and NSAIDs include acetaminophen, aspirin, diclofenac, ibuprofen, indomethacin, ketoprofen, ketorolac,naproxen, phenylbutazone, piroxicam, tolmetin, and tramadol. Factors of prescribing any pain medication There were several patient factors predictive of higher rates of prescribing of any pain medication among both outpatient and ED visits (Table 4). Among outpatient visits, pain medication prescription was highest among visits by patients aged 45-64 years, increasing significantly over time to 25.6% in 2014-2015 (p<0.001), and among visits by patients with Medicare, increasing significantly over time to 24.2% in 2014-2015 (p<0.001). In contrast, among ED visits, pain medication prescription was lowest among visits by patients with Medicare insurance, but increased significantly over time to 36.4% in 2014-2015 (p=0.003). Table 4. Stratified analysis of pain medication prescribing trends by key patient and visit characteristics, 2006-2015 NAMCS and NHAMCS. Years (%) 2006-2007 vs. 2014- 2015, P-value 2006- 2007 2008- 2009 2010- 2011 2012- 2013 2014- 2015 NAMCS Visit diagnosis Cancer-related* 11.9% 14.3% 16.0% 15.2% 16.4% 0.037 Other pain-related† 43.4% 43.6% 43.3% 44.2% 44.1% 0.846 No indication 13.9% 15.1% 16.0% 16.5% 19.1% <0.001 Physician specialty Generalist‡ 16.3% 17.7% 17.8% 18.4% 21.6% <0.001 Other§ 13.1% 14.4% 16.8% 17.6% 19.2% <0.001 Age <19 9.4% 10.1% 10.1% 8.4% 10.0% 0.504 19-44 15.7% 15.5% 17.6% 18.7% 19.5% 0.001 22 45-64 18.8% 20.3% 21.7% 22.5% 25.6% <0.001 ≥65 14.6% 17.0% 18.2% 19.0% 21.4% <0.001 Gender Female 15.0% 16.9% 17.6% 18.1% 20.6% <0.001 Male 14.9% 15.5% 17.1% 18.0% 20.2% <0.001 Race/ethnicity Non-Hispanic White 15.1% 16.7% 17.5% 17.9% 20.8% <0.001 Non-Hispanic Black 15.6% 16.7% 19.8% 19.6% 22.9% <0.001 Hispanic 14.5% 15.1% 16.6% 18.9% 19.5% 0.003 Other|| 13.4% 14.1% 12.7% 15.0% 15.3% 0.331 Primary source of payment Private 13.5% 14.8% 15.2% 15.7% 17.8% <0.001 Medicare 15.8% 19.1% 19.8% 20.9% 24.2% <0.001 Medicaid 16.6% 15.9% 19.0% 17.3% 19.5% 0.055 Other 19.1% 19.5% 22.6% 23.9% 22.8% 0.168 NHAMCS Visit diagnosis Cancer-related* 41.0% 46.2% 46.8% 47.8% 49.3% 0.074 Other pain-related† 57.9% 62.2% 63.2% 60.0% 59.3% 0.262 No indication 42.5% 43.3% 44.3% 41.3% 41.9% 0.548 Clinician specialty MDs 45.4% 46.8% 47.5% 44.7% 44.8% 0.580 Other¶ 35.5% 36.5% 43.5% 37.7% 42.5% <0.001 Age <19 39.6% 41.9% 41.2% 39.0% 38.2% 0.350 19-44 51.4% 52.9% 54.2% 50.0% 51.1% 0.785 45-64 45.5% 47.5% 49.9% 47.4% 48.6% 0.024 ≥65 30.1% 30.1% 31.9% 31.1% 32.1% 0.118 Gender Female 45.2% 46.7% 48.0% 44.9% 45.8% 0.570 Male 42.9% 44.2% 45.4% 42.6% 42.9% 0.955 Race/ethnicity Non-Hispanic White 44.6% 45.3% 46.5% 43.6% 44.2% 0.646 Non-Hispanic Black 43.1% 46.1% 46.8% 43.7% 44.1% 0.561 Hispanic 44.2% 46.4% 48.5% 45.8% 46.2% 0.157 Other|| 42.4% 43.5% 45.6% 40.3% 44.6% 0.395 Primary source of payment Private 48.2% 48.2% 49.5% 46.1% 46.8% 0.253 Medicare 32.6% 35.1% 37.0% 35.3% 36.4% 0.003 Medicaid 43.0% 46.8% 46.7% 44.7% 45.3% 0.067 Other 48.5% 51.0% 52.1% 49.3% 48.3% 0.839 Note: *was based on ICD-9-CM diagnostic codes 140-239, 338.3X; †was based on ICD-9-CM codes 338.XX, 350.1X-350.2X, 354.4X, 355.71, 379.91, 388.7X, 719.4X, 724.1X-724.2X, 729.1X, 780.96, 786.5X, 789.XX; ‡includes general/family practice, internal medicine, pediatrics, and obstetrics and gynecology; §includes psychiatry, general surgery, orthopedic surgery, cardiovascular diseases, dermatology, urology, neurology, ophthalmology, otolaryngology, and others; ||includes Asians, American Indian/Alaska Natives (AIANs), Native Hawaiian or Other Pacific Islanders (NHOPI), or 2+ reported racial/ethnic groups; and ¶includes physician assistants (PAs) and nurse practitioners (NPs). 23 4. STUDY 2 – METHODS AND RESULTS 4.1 Medicare Formulary Coverage and Restrictions for Opioid Potentiators from 2013- 2017 Data Source We used 2013, 2015, and 2017 Medicare Prescription Drug Formulary Files, which include data on all Medicare Advantage and Stand-alone Part D formularies. This data was gathered from the CMS Prescription Drug Plan Formulary and Pharmacy Network Files. The following variables for each plan were collected for each opioid potentiator: coverage, prior authorization, specialty tier, quantity limit amount, and step therapy. Drug Sample We identified all benzodiazepines, non-benzodiazepine sedative-hypnotics, and gabapentinoids available in oral formulations. Benzodiazepines studied included alprazolam, chlordiazepoxide, clonazepam, clorazepate, diazepam, estazolam, flurazepam, lorazepam, oxazepam, quazepam, temazepam, and triazolam. Non-benzodiazepine sedative-hypnotics included doxepin, zaleplon, zolpidem. Gapapentinoids included gabapentin, gabapentin enacarbil, and pregabalin. We characterized median formulary coverage of the lowest dose of the generic version of each drug, or the brand-name version when generics were unavailable. We focused on generics since they are used more commonly than the bioequivalent brand-name version, and on the lowest dose because higher doses can be created from lower doses. We excluded two brand- name drugs (Rozerem and Lunesta) that became available as a generic between 2013 and 2017, as generic availability impacts brand-name formulary coverage. 24 Statistical Analysis For each drug in each year, we determined the proportion of formularies not providing coverage; providing restrictive coverage using one or more utilization management strategy (quantity limit, prior authorization, or step therapy); or providing unrestrictive coverage (no utilization management). We summarized median coverage across all drugs in all three years. Analyses were conducted in R Studio version 3.2.3. 4.2 Medicare Formulary Coverage and Restrictions for Opioid Potentiators from 2013- 2017 Formulary Restrictions on Opioid Potentiators There were 12 benzodiazepines, 3 non-benzodiazepine sedative-hypnotics and 3 gabapentinoids eligible for study. The median proportion of formularies not providing coverage across all drugs was 21.8% (interquartile range [IQR], 0.3-64.8%) in 2013, 14.4% (IQR, 0.0- 66.3%) in 2015, and 17.6% (IQR, 0.0-68.7%) in 2017 (Table 5). The median proportion of formularies providing restrictive coverage was 63.3% (IQR, 49.6-69.4%) in 2013, 70.1% (IQR, 65.8-81.2%) in 2015, and 66.8% (IQR, 54.4-77.9%) in 2017, with the largest growth in use of quantity limits, a smaller increase in prior authorization, and infrequent use of step therapy. The median proportion of formularies providing unrestrictive coverage in the 3 years was 33.3% (IQR, 27.1-43.7%), 27.0% (IQR, 16.3-32.2%), and 27.9% (IQR, 18.0-41.6%), respectively. In 2017, 47.9% of formularies provided unrestrictive coverage of at least 1 benzodiazepine, 39.9% of at least 1 non-benzodiazepine sedative-hypnotic, and 67.2% of at least 1 gabapentinoid. Table 5. Median Medicare Prescription Drug Plan Formulary Coverage and Use of Utilization Management Strategies for Benzodiazepines, Non-benzodiazepine Sedative-hypnotics, and Gabapentinoidsa, 2013-2017 Median Formulary Coverage (Interquartile Range), %