University of Southern California Dissertations and Theses

Essays on opioid prescribing and abatement

(USC Thesis Other)

Essays on opioid prescribing and abatement

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

ESSAYS ON OPIOID PRESCRIBING AND ABATEMENT

by

Andy Nguyen

A Dissertation Presented to the
Faculty of the University of Southern California Graduate School
University of Southern California

In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(HEALTH ECONOMICS AND POLICY)

August 2018

1

DEDICATION

To my loving family:

Ba Mẹ ,
Ba Mẹ đã nâng đỡ con và hy sinh cho con rất nhiều để mới có thể thành công được ngày hôm
nay. Con sẽ luôn nhớ những sự khó khăn của Ba Mẹ trong những năm qua. Cám ơn Ba Mẹ rất
nhiều. Con sẽ không bao giờ quên những sự hy sinh, giúp đở, và yêu thương Ba Mẹ đã dành cho
con trong suốt cuộc đời con.
I love you both so much!

To Bryan,
Thank you for being there every step of the way with me, motivating me to be a better person
every day. You are an inspiration to me and have helped push me to achieve my goals.

To Ann and Mark,
Thank you for introducing me to this field and for being a beam of support throughout my life.
From paging to graduate school, your guidance in my life has been a great factor in any
achievements I may have made. You both have been great role models in my life.

To Cora and Nicholas,
You two have been such a ray of sunshine throughout this process. Thank you for the joy you
continually bring me each and every day. I can’t wait to see the great things that you two will
accomplish throughout your lives. Uncle Andy loves you very much.

To Bà Ngoại, all my cousins, aunts, and uncles,
Thank you for the love and support you have given me throughout the years. I am so proud to be
a member of this family.

To my dear friends, especially those from Ðoàn Anê Thành, Ane Thanh Lion Dance, the
United States House of Representatives Page Program Class of 2008-2009, Phi Delta Theta
Fraternity, and all those that I have made throughout the past years:
Thank you for the joy you bring into my life. I would not be the person I am today if you were
not a part of my life. Eis Aner Oudeis Aner! Excelsior!!

I love you all!

2

ACKNOWLEDGEMENTS

First off, I thank God for all his providence, who has given me everything I have today.

This dissertation would not be possible without the help of my dissertation advisor and
mentor, Dr. Jason Doctor, who has guided me throughout my tenure in graduate school. His
support, patience, and encouragement has helped me grow as an independent researcher. I am
extremely grateful for his confidence in me throughout our research work. I am forever indebted
to and thankful for the opportunity to work with Dr. Doctor.
I would also like to thank the members of my dissertation committee and dissertation
proposal committee. Dr. Michael Menchine, Dr. Rebecca Myerson, Dr. John Romley, and Dr.
Geoffrey Joyce all provided such vital feedback and insightful advice which helped me
throughout my entire dissertation process.
I am grateful for the San Diego County Medical Examiner’s Office and the San Diego
County Prescription Drug Abuse Medical Task Force, particularly Dr. Jonathan Lucas, Dr.
Roneet Lev, and Dr. Steven Campman, who have supported my research work in the San Diego
ME Office.
My learning and growth in this field would not have been possible without the amazing
faculty and staff in the Health Economics and Policy program at the University of Southern
California School of Pharmacy and Leonard D. Schaeffer Center. I would also like to thank the
Data Core who made my research possible through access to data as well as guidance when I had
coding issues. I am also very appreciative of Dr. Jeffrey McCombs, who has been there for me
the entire time since I first considered applying to the program, who helped me secure funding,
and who has been a wonderful friend and beam of support throughout graduate school.
3

Lastly, I thank all the friends I have made at USC, especially Roxanna Seyedin, Nikhil
Bhagvandas, Kinpritma Sangha, and Wendy Cheng, my companions throughout this journey
who has made graduate school so enjoyable and who provided so much advice throughout my
time at USC. I am excited to see what we will all accomplish throughout our careers and lives.
Fight On!

4

TABLE OF CONTENTS

Common Terminology 6
List of Tables 8
List of Figures 9
C hapter 1: Introduction 10
Objectives of the Dissertation 24
References 26
Chapter 2: Prescribing Patterns Leading to Prescription-Related Death 33
2.1 Background 34
2.2 Methods 36
2.3 Results 38
2.4 Discussion 40
2.5 Conclusion 42
2.6 References 43
2.7 Figures and Tables 46
Chapter 3: Opioid Prescribing Decreases After Learning of a Patient’s Fatal Overdose 6 3
3.1 Introduction 6 4
3.2 Methods 6 6
3.3 Results 70
3.4 Discussion 7 2
3.5 References 7 4
3.6 Figures and Tables 76
3.7 Supplementary Materials 84
Chapter 4: Predicting Long-Term Opioid Use in Patients Following Heart Valve Replacement: An
Instrumental Variables Approach 8 5
4.1 Background 8 7
4.2 Methods 88
4.3 Results 9 2
4.4 Discussion 9 4
4.5 Conclusion 9 6
4.6 References 9 7
4.7 Figures and Tables 99
Chapter 5 : Conclusion 1 08
5

COMMON TERMINOLOGY

Acute Pain – Pain that usually starts suddenly and has a known cause, like an injury or surgery.
It normally gets better as your body heals and lasts less than three months.
Chronic pain – Pain that lasts 3 months or more and can be caused by a disease or condition,
injury, medical treatment, inflammation, or even an unknown reason.
Days’ Supply: Number of days of supply for the prescription of a specific drug.
Drug misuse – The use of prescription drugs without a prescription or in a manner other than as
directed by a doctor, including use without a prescription of one’s own; use in greater amounts,
more often, or longer than told to take a drug; or use in any other way not directed by a doctor.
Drug abuse or addiction – Dependence on a legal or illegal drug or medication. See Opioid use
disorder.
Extended-release/long-acting (ER/LA) opioids – Slower-acting medication with a longer
duration of pain-relieving action.
High-dose prescription – prescriptions with a dosage greater than or equal to 90 MME/day
Illicit drugs – The non-medical use of a variety of drugs that are prohibited by law. These drugs
can include: amphetamine- type stimulants, marijuana/cannabis, cocaine, heroin and other
opioids, synthetic drugs, and MDMA (ecstasy).
Immediate-release opioids – Faster-acting medication with a shorter duration of pain-relieving
action.
Medication-assisted treatment (MAT) – Treatment for opioid use disorder combining the use
of medications (methadone, buprenorphine, or naltrexone) with counseling and behavioral
therapies.
Morphine milligram equivalents (MME) – The amount of milligrams of morphine an opioid
dose is equal to when prescribed.
Morphine milligram equivalent per day (MME/Day) Calculation
ME/day (Unit Strength uantity onversion F actor)/ Days Supply M = × Q × C ′
-MME conversion factor is determined by the National Drug Code.

6

Non-medical use – Taking drugs, whether obtained by prescription or otherwise, not in the way,
for the reasons, or during the time period prescribed. Or the use of prescription drugs by a person
for whom the drug was not prescribed.
Opioid use disorder – A problematic pattern of opioid use that causes significant impairment or
distress. A diagnosis is based on specific criteria such as unsuccessful efforts to cut down or
control use, or use resulting in social problems and a failure to fulfill obligations at work, school,
or home, among other criteria. Opioid use disorder has also been referred to as “ opioid abuse or
dependence ” or “ opioid addiction .”
Overdose – Injury to the body (poisoning) that happens when a drug is taken in excessive
amounts. An overdose can be fatal or nonfatal.
Physical dependence – Adaptation to a drug that produces symptoms of withdrawal when the
drug is stopped.
Prescription drug monitoring programs (PDMPs) – State-run electronic databases that track
controlled substance prescriptions. PDMPs help providers identify patients at risk of opioid
misuse, abuse and/or overdose due to overlapping prescriptions, high dosages, or co-prescribing
of opioids with benzodiazepines.

Definitions taken from the Centers for Disease Control and Prevention
Website, (available at https://www.cdc.gov/drugoverdose/opioids/terms.html)

7

LIST OF TABLES

Table 2.1. Decedent Characteristics 58
Table 2.2. First Prescription Characteristics 59
Table 2.3. Certifications of Initial Prescriber 60
Table 2.4. Drugs Initially Prescribed 61
Table 2.5. Cumulative Number of Prescribers Per Decedent on Average, by 6 months 62
Table 3.1. Blocking variables and group 78
Table 3.2. Decedent Characteristics 79
Table 3.3. Prescriber Characteristics 80
Table 3.4. Results 81
Table 4.1. Patient Characteristics 101
Table 4.2. Initial Prescription Characteristics 102
Table 4.3. Marginal Effects: Complete Results 103
Table 4.4 CDC Threshold Results, Marginal Effects 104

8

LIST OF FIGURES

Figure 2.1. Consort Diagram 46
Figure 2.2. Cumulative Growth of Connections with Prescribers, Alluvial chart by 6 month
periods 47
Figure 2.3. Connections, 22-24 Months Prior to Decedent Death 48
Figure 2.4. Connections, 19-24 Months Prior to Decedent Death 49
Figure 2.5. Connections, 16-24 Months Prior to Decedent Death 50
Figure 2.6. Connections, 13-24 Months Prior to Decedent Death 51
Figure 2.7. Connections, 10-24 Months Prior to Decedent Death 52
Figure 2.8. Connections, 7-24 Months Prior to Decedent Death 53
Figure 2.9. Connections, 4-24 Months Prior to Decedent Death 54
Figure 2.10. Connections, Two Years Prior to Decedent Death 55
Figure 2.11. New Connections by Certifying Board, Raw Counts 56
Figure 2.12. New Connections by Certifying Board, Normalized with Min-Max Scaling 57
Figure 3.1. Study timeline 76
Figure 3.2. Consort diagram 77
Figure 4.1. Patient Inclusion Timeline 99
Figure 4.2. Study Consort Diagram 100

9

CHAPTER 1
INTRODUCTION
Introduction to the Problem
Physical pain is an issue which affects all people at some point in time throughout their
lives. There are many root causes of acute or chronic physical pain in patients. Pain is associated
with all sorts of diseases as well as injury and surgeries. There are multiple ways to effectively
manage pain when it occurs. It is estimated that approximately 1.5 billion people worldwide, 100
million of which come from the United States, currently suffer from chronic pain alone (1,2) . For
both cases of acute or chronic pain, a common pain management practice in medicine is the use
of prescription or over the counter pharmaceutical drugs. In more severe cases, physicians often
prescribe prescription opioid drugs to help patients with their pain. Some pain sufferers also seek
pain reduction through illicit drug use. This study aims to look at opioids, the opioid epidemic,
the possible effects of medication assisted therapy on health outcomes, and explores possible
methods to curb prescription opioid prescribing, particularly behavioral interventions.

Background and Literature Review
While the opioid epidemic has only gained worldwide attention through the past couple
decades, the abuse of addiction to opioids has been a long standing issue. This literature review
will look at opioids often abused, prescription and illicit. It will also explore the history of the
opioid epidemic, including the current state of the crisis, as well as overdose deaths resulting
from opioids.

10

Opioids
There are a wide range of opioids which are currently available for pain therapy,
including: natural opiates, semi-synthetic opioids, and synthetic opioids. Natural opiates are
substances which can be naturally found in the opium poppy. Examples of natural opiates which
are prescribed include morphine and codeine. Semi-synthetic opioids are drugs which have been
derived from the substances found in opium. Hydrocodone, hydromorphone, oxycodone,
oxymorphone, and buprenorphine are examples of semi-synthetic opioids.
Synthetic opioids such as methadone, fentanyl, levorphanol, tramadol, and tapentadol are those
which have been manufactured to be chemically similar to opiates. Some synthetic opioid have
been created to very high potency levels. Fentanyl, is 50 times more potent than heroin and even
100 times more potent than morphine. Fentanyl comes in two forms: pharmaceutical fentanyl,
which is the form which is prescribed to patients, as well as non-pharmaceutical fentanyl, which
has been illicitly manufactured (3) . Non-pharmaceutical fentanyl is often combined with illicit
drugs such as heroin and cocaine. Hydrocodone, oxycodone, and methadone are the most
commonly prescribed opioids (4) .
Pain sufferers also seek other methods of curbing their pain through the use of illicit
drugs such as heroin. Heroin, a morphine prodrug, is chemically similar to the opioid pain
medications such as hydrocodone and oxycodone (5) . Heroin use has greatly increased in the
United States for both men and women and as well age groups and income levels. Women,
people with higher incomes, and those privately insured have seen the largest increases in heroin
use. People have also increased their use of cocaine and prescription opioids in combination to
heroin abuse. Heroin use has significantly increased in young adults between 18-25 years old,
11

doubling in the past decade. 45% of people who have used heroin also suffered from a
prescription opioid addiction (6).
One important consideration is the possibility of transition from prescription painkillers
to illicit heroin use. Prescription opioid use has been shown to be a strong risk factor for
beginning heroin use (7) . The literature finds that 4 out of 5 heroin users had first used
prescription opioids (8) . Furthermore, non-medical use of pain relievers was 19 times higher than
those did not (9) . Between 4-6% of prescription opioid abusers will eventually begin to use
heroin (6) .The increasing use of prescription opioids causes concern due to its association with
future heroin use.
The medical community has also turned to hybrid opioid antagonists and opioid
antagonists to combat opioid overdoses. Buprenorphine, a hybrid opioid antagonist, is used in
medication assisted treatment and has been shown to be an effective treatment for opioid use
disorder (6) . However, the treatment capacity of buprenorphine is far lower than the rates of
opioid abuse in most states (11) . The increasing availability of buprenorphine play a vital role in
future opioid use disorder treatment. Naloxone is a commonly prescribed opioid antagonist,
mostly used to reverse opioid overdoses. Buprenorphine and naloxone is often a combination
medication which is prescribed for opioid deterrence.
Naltrexone is another opioid antagonist that is used to help manage opioid dependence.
Naltrexone treatment typically requires for a patient to undergo a managed withdrawal period
and the treatment typically results in reduced cravings and tolerance to opioids (12) . Methadone,
a fully synthetic opioid, has also been used to treat addiction by changing how the brain and
nervous systems respond to pain.
12

Pain
Pain, as described by the International Association for the Study of Pain as an
“unpleasant sensory and emotional experience associated with actual or potential tissue damage,”
is a common symptom of medical conditions or procedures (13) . Among primary care patients,
approximately 21.5% report having persistent pain, more commonly reported by women.
Persistent pain is also associated with a four-fold likelihood of having a psychological disorder
such as anxiety and depression (14) . In the United States, 14.4 million adults reported having the
highest level of pain (15) . Pain as the fifth vital sign in medical treatment was introduced in
1996. Since then doctors included pain along with blood pressure, heart rate, respiratory rate, and
temperature rates by using an intensity scale (16) . This opened the door for more communication
between doctors and patients regarding pain to assist in treatment. However, there has been
pushback saying that including this fifth vital sign, something that could not be measured, has
resulted in excess opioid prescribing. Pain intensity has also been criticized as an assessment
measure for treatment of chronic pain because it results in the wrong goals for care and adverse
selection, where the wrong patients are the ones receiving the highest opioid regimens (17,18) .
Reported pain has played a vital role in the use of opioids.

History of the Opioid Epidemic
Painkiller use is not a novel pain therapy. There is a long history which has led to the
current opioid epidemic. With the growing need for pain therapy, prescription opioids have been
available in the United States for nearly a century, beginning in the 1930s. However, opioids
13

were not commonly prescribed until the 1990s. This was due to pharmaceutical companies
convincing the medical community that prescription painkillers would not cause addiction. Soon
after, opioid prescriptions became gradually more prevalent (19,20). In the 1980s and 1990s,
there were several studies that concluded that opioids had low addictive behavior in patients and
could be prescribed safely long-term (21,22). However, there had been no long-term longitudinal
studies regarding opioid use. The idea that extended opioid prescription is safe and had low risk
of addiction lead to the dramatic increases of prescription painkiller use. Patients were also keen
to the idea that they could be trusted to manage their own prescriptions, leading many to feel
empowered (23).
Purdue Pharma, a large manufacturer of OxyContin, led the campaign to push to
encourage widespread long-term use of opioids for chronic non-cancer pain by commissioning
more than 20,000 pain related education programs between 1996 and 2002 (24) . Some
physicians, who were spokespersons for opioid manufacturers, also published papers to convince
the medical community that addiction was so rare and unlikely to occur with extended opioid use
(25) . Purdue Pharma released OxyContin and it was approved by the Federal Drug
Administration in 1995 (26,27) . By 1996, opioid prescribing had increased greatly. In 2000,
Purdue Pharma also sponsored a book by the Joint Commission which mentioned studies that
reaffirmed the claim that there was no evidence of addiction for patients receiving prescription
painkillers (28) . The book also looked to discredit clinician’s concerns regarding the possibility
of opioid addiction.
The lack of overall information and training has also led to the increase in adverse events
regarding opioids. Physicians who have not received proper education and training have had
14

issues such as calculating the correct dosages when switching patients over from one drug to
another, correct levels of opioid prescribing, and giving opioids to patients who are at greatest
risk for abuse (29) . Other issues of consideration are drug interactions and the effects of the
drugs. The lack of information for public officials to enact policies have also lead to the increase
in prescription opioid abuse (30) . Due to this and many other factors, hydrocodone consumption
has doubled between 1999 and 2011 and oxycodone use had increased by close to 500% (31) .
Overall, painkiller use has drastically grown in the past two decades into the opioid epidemic of
today.

Current state of opioid addiction
The current opioid has been deemed by the United States Centers for Disease Control and
Prevention (CDC) as the worst drug overdose epidemic in United States history (32) . Since the
medical community being more open to long term prescribing of opioids, the amount of
painkillers sold have drastically increased. According to the Department of Justice and the CDC,
the amount sold to pharmacies, hospitals, and doctors’ offices have quadrupled from 1999 to
2010 (33,34) . However, the growth in prescription opioids is not associated with a change in the
amount of pain reported by patients (35,36) . The overall prescribing rates per 100 persons have
also decreased since 2012, after the initial increase, down about 4.9% annually from 2012
through 2016. The decrease in these rates may be due to more education, awareness of addiction,
and methods to combat opioid addiction. However, the rates for prescriptions with greater than
or equal to 30 days supply increased by 55.1%. The CDC also reports that there were nearly 215
million opioid prescriptions dispensed by retail pharmacies to nearly 62 million patients in 2016,
15

with females and people aged 65+ receiving the most in their respective demographic groups. In
the same year, the average number of opioid prescriptions was 3.5 and the average day’s supply
per prescription was 18.1 (37) .
The CDC estimates that approximately 12.5 million people aged 12 or older misused
prescription opioids in 2015 with over 2 million suffering from an opioid use disorder. Substance
abuse disorders involving prescription painkillers have also resulted in high rates of
hospitalizations and emergency department visits for opioid-related poisonings. Heroin also
attributed to thousands of visits to the hospital and emergency departments (36) .
Different demographic subgroups are also have distinct prescribing and drug use patterns.
Older age groups, from aged 40+, were more likely to receive more opioid prescriptions. Women
were also more likely than men. Non-hispanic whites were the most likely to be prescribed
compared to others (37,38) . The amount of milligram morphine equivalency (MME) also varies
among groups, with non-Hispanic Whites and higher unemployment being more associated with
being prescribed an opioid (39) . Non-medical drug use is common in youth. In a survey, 1 out of
8 high school seniors reported involvement in opioid misuse. 23% of those teens also report that
they usually combine it with marijuana use (40) .

Opioid Drug Deaths
High levels of opioid use or mixing opioids with substances such as illicit drugs or
alcohol can lead to poisoning, often also leading to death. Despite the number of prescriptions
decreasing as stated earlier, drug overdose deaths, a majority of which involving opioids,
continue to increase (41,42). According to the CDC, drug overdose deaths have tripled from
16

1999 to 2015, with the number of deaths involving opioids quadrupling since then (37,39) . In
2015 alone, there were 54,404 drug overdose deaths in the United States and 33,091 (63%)
opioid involved deaths. While methadone deaths have decreased in the past half a decade, other
illicit and synthetic opioid drug use deaths have dramatically increased. During 2015,
demographic subgroups which saw an increase included: males, those aged 25-44, and
non-hispanic whites (42). Opioids deaths have risen to a level that is higher than the number of
deaths from suicide and motor vehicle crashes, or deaths from cocaine and heroin combined.
Approximately 40% of prescription opioid deaths involve drug abuse such as patients having
multiple prescriptions, drug diversion, or doctor shopping (43) .
For synthetic and illicit drugs, the overdose death rate has also dramatically increased,
with the exception of those involving methadone. Tramadol and fentanyl related deaths increased
72.4% in one year between 2014 and 2015 (42) . Fentanyl is often found in opioid overdose
deaths. Illicit drugs also appeared on toxicology reports in more than half of deaths which
involved fentanyl (44-46) . Pharmaceutical fentanyl, particularly transdermal patch fentanyl was
rarely found in opioid involved deaths. However, a concern is that many medical examiner
offices do not test for fentanyl in their toxicology labs, meaning the number of deaths involving
fentanyl could be significantly higher (47) . Heroin involved deaths have also risen in number.
They have quadrupled since 2010, with males aged 25-44 having the highest death rate (48) .
Opioid overdose death not only affect the individual but also their friends and family.
The CDC estimates that over 115 people die each day in the United States from an opioid
overdose (49) . Over half a million Americans died between the year 2000 and 2015. In 2014
17

alone, more than 47,055 families had a drug overdose death to one of their members (42) .
Opioids negatively impact individuals and households across the United States.

Doctor Shopping
Doctor shopping refers to when a patient receives multiple controlled substances during a
short period of time, likely for a singular illness or medical incident, through multiple
prescribers. This occurs often times without the knowledge of the other doctor (50) . Studies have
shown that doctor shopping and prescription related deaths as well as hospital admission rates
are highly related, one finding that 21.4% of decedents had evidence of doctor shopping behavior
(51,52) . Another study found that 40% of deaths in 2011 involved drugs obtained through
multiple prescriptions, doctor shopping, and drug diversion (43) . A study in San Diego of all
prescription drug related deaths in the county during 2013 found that more than 50% of
prescriptions went to doctor shoppers (53) . The literature also shows that doctor shoppers obtain
much higher levels of morphine milligram equivalent pills than the those in the general
population, between three to six times the amount (54) . Doctor shopping behavior is an area of
concern with looking at risk factors for prescription related deaths.

Co-Prescribing
Benzodiazepines, a class of drugs which are used to treat anxiety, are commonly
prescribed in conjunction with opioids. However, combining the two can be unsafe. This practice
can often lead to overdose deaths. About one-third of all overdoses involving opioids also
include the consumption of a benzodiazepine. In addition, 23 percent of opioid related overdose
18

deaths involved the use of a benzodiazepine (55) . The likelihood of an opioid related death also
increases when a patient receives both prescriptions for a benzodiazepine and an opioid, with
multiple studies having results that state the increased risk as 10 times higher than patients who
only received a prescription for opioids (56,57) . Benzodiazepines have shown to have an
increasing involvement in opioid overdose related deaths (58) .
Though not advised, the practice of co-prescribing opioids and benzodiazepines is far
from rare. In a study looking at veterans, approximately 27% of patients received both opioids
and benzodiazepines during the study period (57) . Other studies have also found that around 1 in
5 patients who were prescribed an opioid was also prescribed a benzodiazepine (59) . The
combination of these two drugs has grown in prevalence and seems to have played a role in the
growing number of opioid overdose deaths currently facing the United States.

Prescription Drug Monitoring Programs
Many states have adopted what is known as a prescription drug monitoring program
(PDMP). The purpose of PDMPs are to assist the state and health care workers in tracking
controlled substances. These databases can help in the efforts to reduce drug prescribing, detect
doctor shopping, avert co-prescribing, as well as prevent drug diversion. PDMPs typically record
and store a variety of data including but not limited to: drug prescribed, quantity prescribed, days
supply, date filled, prescribing and patient identifiers, pharmacy data, as well as payment
methods. Access to the PDMP data depends primarily on the state, with the majority allowing
law enforcement, prescribers, and pharmacists the ability search and view data. To this date, all
19

50 states as well as the District of Columbia and the territory of Guam have an operational
PDMP (60-62) .
Prescription drug monitoring programs can provide valuable insight into prescribing
patterns and patient behavior which allows for a more informed response to the opioid epidemic.
The operations of PDMPs have also shown to have an impact on prescribing behaviors and
doctor shopping (62) . Several studies have shown that the utilization of the PDMP has directly
influenced what providers prescribe to patients, either by drug type of the quantity that is
prescribed to them. PDMP use has been associated with a mitigation of opioid abuse in a number
of studies with considerable evidence that PDMPs have been effective on reducing prescribing
(63-65) . A few studies did find no significant association between PDMP use and opioid
prescribing levels (66) . However, this was an analysis of older data, when access to PDMPs and
technology was much more limited than today. Overall, the literature generally finds that PDMPs
do reduce doctor shopping and curb prescription opioid abuse. A reduction in the amount of
prescriptions has also been associated with PDMPs (67) . Prescription drug monitoring programs
are a valuable resource to doctors, pharmacists, and law enforcement in the efforts to combat the
opioid epidemic.
PDMPs are a promising tool with a lot of potential which has not been tapped. Most
states, with the exception of 14 states, do not have mandatory PMDP use for prescribers. While it
is accessible, a staggeringly low number of prescribers actually use it. Some studies find that
prescribers consult a PDMP less than 25% of the time when prescribing an opioid to patients,
one reporting a figure as low as 14% (68,69) . Some reasons that were listed for lack of PDMP
use included the PDMP being time-consuming and difficulty navigating the data. However, some
20

new state policies in 42 states has now allowed for delegates to be appointed from an office’s
staff to access the data and improve the workflow for prescribers, pharmacists, or law
enforcement (70) . PDMPs generally are still in their infantile stages and have much room to
grow, opening up endless possibilities in the efforts to combat the opioid epidemic.

Controlled Substance Utilization Review and Evaluation System
California has recorded data on Schedule II prescriptions since 1939, beginning with the
California Triplicate Prescription Program (TPP). In 1997, California also created the Controlled
Substance Utilization Review and Evaluation System (CURES), a prescription drug monitoring
program which was created to work in parallel with the TPP in order to compare effectiveness
between the two. Several years later, CURES officially replaced TPP and became a permanent
fixture. In 2009, California made CURES a searchable database for prescribers. Since then the
program was revamped in 2016 and has grown as a whole, tracking all DEA schedule II, III, and
IV controlled substance prescriptions in the state of California. Pharmacists are also required to
report weekly on drugs that are dispensed in the state. The use of California’s PDMP program is
voluntary (71,72) .

Government Response
In addition to state prescription drug monitoring programs, individual states and the
United States Federal government have adopted measures in response to the ongoing opioid
epidemic. For example, 22 states adopted opioid prescribing guidelines, 12 states who had
legislation to eliminate pill mills, and 35 states which granted more naloxone prescribing (70) .
21

The adoption of opioid dosing guidelines have also shown to be effective in reducing long-acting
opioid dosage, the number of patients receiving high levels of morphine milligram equivalent
dosages, as well as the rate of opioid-related deaths particularly in injury workers (73) .
The Federal government has also laid out several plans to combat the crisis. In October
2017, President Trump declared the opioid crisis a public health emergency. The United States
Food and Drug Administration (FDA) also took several steps such as requiring safety labeling on
opioid medications. They also required warnings about the dangers of misuse and abuse (74) .
The FDA has also been more active in working with other agencies to come up with new ideas to
reverse the epidemic. The United States Department of Health and Human Services has also
created a five-point strategy which includes: better data, better research, better pain management,
better targeting of overdose reversing drugs, as well as better addiction prevention, treatment,
and recovery services (75) .

Medical School Education
There is also concern that the quality of education has played a significant role in the
amount of opioids being prescribed. A study found that the majority of doctors felt as if they did
not receive adequate training on prescription drug abuse and addiction during the tenure in
medical school (76) . Research has shown that some physicians have received less than 12 hours
of education regarding pain management during medical school (77) . The ranking of medical
school has also been linked to prescribing patterns in physicians. A study finds that when
controlling for specialty and practice location, doctors who attended higher the top medical
schools tend to prescribe lower prescriptions than their counterparts who attended lower ranked
22

programs (78) . However, states have participated in requiring more mandatory education on pain
management in medical school (79) . Further steps could be taken to improve education during
and after medical school to have more informed physicians.

Economic Burden
The opioid epidemic not only affects the lives of those involved, but also results in a
large economic burden. This is due to the need for increased health care, substance abuse
treatment, criminal justice costs, as well as loss of productivity. In 2006, it was estimated that the
total United States societal cost of non-medical prescription opioids use was 53.4 billion dollars.
By 2015, that number had grown to an estimated 78.5 billion dollars (80-82) . It was also found
that people who abused opioid are also more likely to utilize other medical services, leading to
higher healthcare costs to themselves, payers, and society (83) .

Gaps in the Literature
The opioid epidemic has not been around for long. Therefore, there are many avenues for
research that have not previously been explored. One such example is a deeper look into
prescriptions, especially those leading up to an opioid related overdose death of a patient. There
is room for more information and identification of risk factors leading up to death which can
prevent future deaths by finding critical points in a person’s prescription history. There is also a
gap in the literature to see how opioid use can affect other illnesses that a patient may have. As
shown above, opioid consumption can also lead to other medical costs. Research needs to delve
further into specific instances where opioids have an adverse effect on the other aspects of a
23

person’s health. Thirdly, the use of prescription drug monitoring programs can be used to
develop possible interventions which may lead to a reduction of opioid prescribing and levels of
opioid consumption. There are a vast number of opportunities for research in the ever growing
opioid crisis.

Objectives of the Dissertation
The urgency of the opioid epidemic is clear. As the growing number of opioid related
overdose deaths continues to climb each year, the need for more information through effective
research becomes more evident. This study seeks to fill certain gaps in the literature to better
inform the medical community and public policy officials to make better decisions in the effort
to curb opioid prescribing and opioid related overdose deaths.
Chapter two of this dissertation takes a deeper look into what happens to a patient before
they die from to an opioid related death. The paper explores CURES, California’s prescription
drug monitoring programming to look at two years worth of prescription data for patients who
have been labeled by the San Diego Medical Examiner as a prescription drug related death. It
explores prescribing and consumption patterns leading up to a prescription related death. Data
visualization techniques are also employed to visually assess patterns.
Chapter three investigates whether prescribers change their opioid prescribing levels after
learning about a patient’s death. Prescribers who were involved in a prescription-related death in
San Diego County between July 1, 2015 and June 30, 2016 were included in a randomized
control trial where a letter was sent out informing providers of a patient’s death. Prescribing data
24

before and after the letter was sent was analyzed to examine prescribing levels, prescribing
volume, and number of new starts following the letter.
Chapter four analyzes long-term opioid use in patients with heart valve replacement using
an instrumental variables approach. We use the initial provider’s overall opioid prescribing
patterns as an instrument for the initial dose of a prescription to a patient following heart valve
replacement surgery to assess likelihood of long-term opioid use. Data from the Optum
Clinformatics Data Mart between years 2007-2016 was assessed. Primary outcomes include
whether a patient had prescription opioid use at 90+ days, 180+ days, and 270+ days, 90-180
days, and 180-270 days following heart valve replacement surgery.

25

References
1. Global pain management market to reach US$60 billion by 2015, according to a new report.
Global Industry Analysts , Inc. (available at www.prweb.com/pdfdownload/8052240.pdf) (2011).
2. Institute of Medicine., Relieving Pain in America: A Blueprint for Transforming Prevention,
Care, Education, and Research . Washington, DC: The National Academies Press . (2011).

3. Centers for Disease Control and Prevention. Opioid Overdose. (available at
https://www.cdc.gov/drugoverdose/data/fentanyl.html) (2016).

4. Ossiander E., Using Textual Cause-of-Death Data to Study Drug Poisoning Deaths,
American Journal of Epidemiology , 179, Pages 884–894, (2014)

5. National Security Council., Prescription Nation 2016: Addressing America’s Drug Epidemic
(available at
http://www.nsc.org/RxDrugOverdoseDocuments/Prescription-Nation-2016-American-Drug-Epid
emic.pdf) (2016).

6. Centers for Disease Control and Prevention. . Today’s Heroin Epidemic. (available at
https://www.cdc.gov/vitalsigns/heroin/index.html) (2015).

7. Compton W.M., Jones C.M., Baldwin G.T., Relationship between nonmedical
prescription-opioid use and heroin use. The New England Journal of Medicine, 374,154-163.
(2016).

8. Jones, C.M., Heroin use and heroin use risk behaviors among nonmedical users of
prescription opioid pain relievers – United States, 2002-2004 and 2008-2010. Drug and Alcohol
Dependence, 132(1), 95-100. (2013).

9. Muhuri PK, Gfroerer JC, Davies MC., Associations of nonmedical pain reliever use and
initiation of heroin use in the United States. Substance Abuse and Mental Health Services
Administration CBHSQ Data Review. (2013).

10. Guidelines for the Psychosocially Assisted Pharmacological Treatment of Opioid
Dependence, World Health Organization. (2009).

11. Jones CM., Campopiano M., Baldwin G., McCance-Katz E., National and State Treatment
Need and Capacity for opioid Agonist Medication-Assisted Treatment, Am J Public Health
105(8): e55-63. (2015)

12. Naltrexone. Substance Abuse and Mental Health Services Administration. (available at
https://www.samhsa.gov/medication-assisted-treatment/treatment/naltrexone) (2016).

26

13. IASP Task Force on Taxonomy. Part III: Pain Terms, A Current List with Definitions and
Notes on Usage. [ed.] Merskey H. and Bogduk N. Classification of Chronic Pain, Second
Edition. Seattle : s.n., 1994, pp. 209-214.

14. Persistent Pain and Well-Being: A World Health Organization Study in Primary Care. Gureje
O., Korff MV., Simon GE., et al. 2, 1998, JAMA, Vol. 280, pp. 147-151.

15. Estimates of Pain Prevalence and Severity in Adults: United States, 2012. Nahin, RL. 8,
2015, The Journal of Pain, Vol. 16, pp. 769-780.

16. DW, Baker. The Joint Commission’s Pain Standards: Origins and Evolution. s.l. : The Joint
Commission, 2017.

17. Must we reduce pain intensity to treat chronic pain? Sullivan MD., Ballantyne JC. 1, 2016,
Journal of Pain, Vol. 157, pp. 65-69.

18. Opioid Therapy for Chronic Pain in the US: promises and perils. Sullivan MD., Howe CQ. 1,
2013, Journal of Pain, Vol. 154, pp. S94-100.

19. Morone NE., Weiner DK., Pain as the fifth vital sign: exposing the vital need for pain
education. Clin Ther., 35(11), (2013).

20. Van AZ., The promotion and marketing of oxycontin: commercial triumph, public health
tragedy. Am J Public Health, 99(2): 221-227. (2009).

21. Portenoy RK., Foley KM., Chronic use of opioid analgesics in non-malignant pain: report of
38 cases. Pain 25(2): 171–86. (1986).

22. Meldrum ML., The property of euphoria: research and the cancer patient. Opioids and Pain
Relief: A Historical Perspective, 198(196), 200–208. (2016).

23. Meldrum ML., The prescription as stigma: opioid pain relievers and the long walk to the
pharmacy counter. Prescribed: Writing, Filling, Using and Abusing the Prescription in Modern
America. 195, p. 201, (2012).

24. US GAO (Gen. Account. Off.). OxyContin abuse and diversion and efforts to address the
problem. GAO. (available at https://www.gao.gov/new.items/d04110.pdf) (2003).

25. Portenoy RK., Opioid therapy for chronic nonmalignant pain: clinicians' perspective. J. Law
Med. Ethics, 24, 296–309. (1996).

26. Timeline of Selected FDA Activities and Significant Events Addressing Opioid Misuse and
Abuse. Federal Drug Administration (available at
https://www.fda.gov/Drugs/DrugSafety/InformationbyDrugClass/ucm338566.htm) (2018).

27

27. Report of the International Narcotics Control Board for 2014., United Nations. (available at
https://www.incb.org/documents/Publications/AnnualReports/AR2014/English/AR_2014.pdf)
(2015).

28. Phillips DM. JCAHO pain management standards are unveiled.
JAMA. 284(4) : 428-429. (2000).

29. Modesto-Lowe V., Brooks D., Petry N., Methadone deaths: risk factors in pain and addicted
populations. J Gen Intern Med, 25(305) , 9. (2010).

30. Association of State and Territorial Health Officials. Prescription drug overdose: state health
agencies respond. Arlington: Association of State and Territorial Health Officials. (available at
https://www.cdc.gov/drugoverdose/pdf/rxreport_web-a.pdf) (2008).

31. Jones CM., Trends in the distribution of selected opioids by state, US, 1999–2011. Presented
at Natl. Meet. Safe States Alliance, June 6, Baltimore, MD (2013).

32. Paulozzi LJ.,The epidemiology of drug overdoses in the United States. Presented at Promis.
Leg. Responses to the Epidemic of Prescription Drug Overdoses in the U.S., Maimonides Med.
Cent. Dep. Psychiatry, Dec. 2, Grand Rounds, Brooklyn (2010).

33. US Department of Justice, Automation of Reports and Consolidated Orders System
(ARCOS). US Department of Justice, Drug Enforcement Administration (2011).

34. Paulozzi LJ., Jones CM., Mack KA., Rudd RA. Vital Signs: Overdoses of Prescription
Opioid Pain Relievers—United States, 1999—2008. MMWR, 60(43): 1487-1492., (2011).

35. Chang H., Daubresse M., Kruszewski S., et al., Prevalence and treatment of pain in
emergency departments in the United States, 2000 – 2010. Amer J of Emergency Med 32(5):
421-31. (2014).

36. Daubresse M., Chang H., Yu Y., Viswanathan S., et al., Ambulatory diagnosis and treatment
of nonmalignant pain in the United States, 2000 – 2010. Medical Care, 51(10): 870-878. (2013).
31. Mattson CL. et al., Annual Surveillance Report of Drug-Related Risks and Outcomes CDC
2017, CDC (2017).

37. Frenk SM, Porter KS, Paulozzi LJ., Prescription Opioid Analgesic Use Among Adults:
United States, 1999 – 2012., National Center for Health Statistics Data Brief, 189: 1-8. (2015).

38. Paulozzi LJ, Strickler GK, Kreiner PW, and Koris CM., Controlled substance prescribing
patterns – prescription behavior surveillance system, eight states, 2013. MMWR Surveill
Summ; 64(8): 1-14. (2015).

39. Guy et al., Vital Signs: Changes in Opioid Prescribing in the United States, 2006-2015.
MMWR 66(26), 697-704 (2017).
28

40. Teens Mix Prescription Opioids with Other Substances (2013). National Institude on Drug
Abuse. (available at
https://www.drugabuse.gov/related-topics/trends-statistics/infographics/teens-mix-prescription-o
pioids-other-substances) (2013).

41. Wide-ranging online data for epidemiologic research (WONDER). Atlanta, GA: CDC,
National Center for Health Statistics. (Available at http://wonder.cdc.gov.) (2016).

42. Rudd RA., Seth P., David F, Scholl L. Increases in drug and opioid-involved overdose deaths
— United States, 2010–2015. MMWR Morb Mortal Wkly Rep. 65(50-51) (2016).

43. Manchikanti L., Helm S., Fellows B., et al., Opioid epidemic in the United States. Pain
Physician 2012; 15:ES9-ES38. (2012).

44. Seth P., Fentanyl Law enforcement submissions and increases in synthetic opioid-Involved
overdose deaths - 27 states, 2013-2014. Morb Mortal Wkly Rep. 2016. 65(33): 837-43. (2016).

45. O’Donnell JK. et al., Deaths Involving Fentanyl, Fentanyl Analogs, and U-47700 — 10
States, July–December 2016. MMWR . 66(43) ; 1197-1202. (2017).

46. Peterson AB, Gladden RM, Delcher C, Spies E, Garcia-Williams A, Wang Y, et al.,
Increases in fentanyl-related overdose deaths - Florida and Ohio, 2013-2015. Morb Mortal Wkly
Rep. 2016. 65(33) :844-9. (2016).

47. 2015 National Drug Threat Assessment Summary. U.S. Department of Justice Drug
Enforcement Administration. (2015).

48. Heroin Overdose Data. CDC., Website(available at
https://www.cdc.gov/drugoverdose/data/heroin.html).

49. Drug Overdose, CDC.,Website (available at https://www.cdc.gov/drugoverdose/index.html).

50. Sansone, Randy A., and Lori A. Sansone., Doctor Shopping: A Phenomenon of Many
Themes. Innovations in Clinical Neuroscience 9.11-12 42–46. (2012).

51. Hall AJ, Logan JE, Toblin RL, Kaplan JA, Kraner JC, et al., Patterns of abuse among
unintentional pharmaceutical overdose fatalities. JAMA. 300(22 ): 2613-20. (2008).

52. Jena A., Goldman D., Weaver L., Karaca-Mandic P., Opioid prescribing by multiple
providers in Medicare: retrospective observational study of insurance claims. BMJ 348 (2014).

53. Lev R. et al., Who is prescribing controlled medications to patients who die of prescription
drug abuse? The American Journal of Emergency Medicine . 34(1), 30-5 (2016).

29

54. Han, Huijun et al., Increasing trends in schedule II opioid use and doctor shopping during
1999–2007 in California. Pharmacoepidemiology and drug safety 23.1 26–35. (2014).

55. Centers for Disease Control and Prevention (CDC). Multiple Cause of Death, 1999-2015.
CDC WONDER Online Database. (Website at https://wonder.cdc.gov/mcd-icd10.html).

56. Dasgupta N, Funk MJ, Proescholdbell S, Hirsch A, Ribisl KM, Marshall S., Cohort study of
the impact of high-dose opioid analgesics on overdose mortality. Pain Med Malden Mass.
17(1) :85-98. (2016).

57. Park TW, Saitz R, Ganoczy D, Ilgen MA, Bohnert ASB., Benzodiazepine prescribing
patterns and deaths from drug overdose among US veterans receiving opioid analgesics:
case-cohort study. BMJ. 350 (2015).

58. Jones CM., McAninch JK., Emergency Department Visits and Overdose Deaths From
Combined Use of Opioids and Benzodiazepines. American Journal of Preventive Medicine.
49(4) . 493-501. (2015).

59. Paulozzi LJ et al., Controlled Substance Prescribing Patterns – Prescription Behavior
Surveillance System, Eight States, 2013., MMWR 64(SS09); 1-14 (2015).

60. Using prescription drug monitoring program data to support prevention planning. Samhsa’s
Center for the Application of Prevention Technologies. (available at
https://www.samhsa.gov/capt/sites/default/files/resources/pdmp-overview.pdf).

61. Prescription drug monitoring frequently asked questions. DMP TTAC . (available at
http://www.pdmpassist.org/content/prescription-drug-monitoring-frequently-asked-questions-faq
).

62. CDC. (2017). What states need to know about PDMPs. (available at
https://www.cdc.gov/drugoverdose/pdmp/states.html).

63. Feldman L., Williams KS., Coates J., Knox M., Awareness and Utilization of a Prescription
Monitoring Program Among Physicians. Journal of Pain & Palliative Care Pharmacotherapy . 4;
313-317. (2011).

64. Green TC. et al., How Does Use of a Prescription Monitoring Program Change Medical
Practice? Pain Medicine. 13(10) ; 1314-1323. (2012).

65. Reifler LM. et al., Do Prescription Monitoring Programs Impact State Trends in Opioid
Abuse/Misuse? Pain Medicine , 13(3); 434-422. (2012).

66. Paulozzi LJ., Kilbourne EM., Prescription Drug Monitoring Programs and Death Rates from
Drug Overdose. Pain Medicine, 12(5) ; 747-754. (2011).

30

67. Worley J., Prescription Drug Monitoring Programs, a Response to Doctor Shopping:
Purpose, Effectiveness, and Directions for Future Research. Issues in Mental Health Nursing. 33,
(2012).

68. Rutkow L., Turner L., Hwang C., Alexander GC., Most primary care physicians are aware of
prescription drug monitoring programs, but many find the data difficult to access., Health
Affairs . 34(3) (2015).

69. Shatterproof (2016, March) Prescription Drug Monitoring Programs: Critical Elements of
Effective State Legislation. (available at https://secure shatterproof org/page/-/Shatterproof_WP_
FINAL pdf?_ga=1 228326110 217109100 14 61943552) (2016).

70. Network for Public Health Law, Legal Interventions to Reduce Overdose Mortality:
Naloxone Access and Overdose Good Samaritan Laws. (available at https://www networkforphl
org/_asset/ qz5pvn/network-naloxone-10-4 pdf) (2016).

71. State of California Department of Justice. Controlled substance utilization review and
evaluation system. (available at https://oag.ca.gov/cures) (2017).

72. State of California Department of Justice. CURES Prescription drug monitoring program.
(available at
https://www.deadiversion.usdoj.gov/mtgs/pharm_awareness/conf_2013/august_2013/san_diego/
small.pdf) (2013).

73. Franklin GM. et al., Bending the prescription opioid dosing and mortality curves: impact of
the Washington State opioid dosing guideline. American Journal of Industrial Medicine. 55:
325-331. (2012).

74. FDA announces enhanced warnings for immediate-release opioid pain medications related to
risks of misuse, abuse, addiction, overdose and death. (available at
https://www.fda.gov/newsevents/newsroom/pressannouncements/ucm491739.htm) (2016).

75. U.S Department of Health and Human Services. Help, Resources and Information; National
Opioids Crisis. (available at https://www.hhs.gov/opioids/) .

76. The National Center on Addiction and Substance Abuse. Under the Counter: The Diversion
and Abuse of Controlled Prescription Drugs in the US . (2005).

77. Mezei, L , & Murinson, B., Pain Education in North American Medical Schools Journal of
Pain, 1199-1208. (2011).

78. Schnell M., Currie J., Addressing the opioid epidemic: is there a role for physician
education? NBER Working Paper Series, National Bureau of Economic Research. (2017).

31

79. Federation of State Medical Boards Continuing Medical Education Board by Board
Overview. (available at http://www fsmb org/Media/Default/PDF/
FSMB/Advocacy/GRPOL_CME_Overview_ by_State pdf) (2015).

80. Curtis, F., Chao Z., Luo, F., Xu L., The economic burden of prescription opioid overdose,
abuse, and dependence in the United States, 2013. Medical Care: 54(10) : 901-906. (2016).

81. Hansen RN., Oster G., Edelsberg J., Woody GE., Sullivan SD. Economic costs of
nonmedical use of prescription opioids. Clin. J Pain. 27(3) : 194-202. (2011).

82. Birnbaum HG., White AG., Schiller M., Waldman T., Cleveland JM., Roland CL. Societal
costs of prescription opioid abuse, dependence, and misuse in the United States. Pain Med,
12(4) : 657-67. (2011).

83. Meyer R., Patel AM., Rattana SK., Quock TP., Mody SH. Prescription opioid abuse: a
literature review of the clinical and economic burden in the United States. Popul. Health Manag.,
17(6) : 372-87. (2014).

32

CHAPTER 2:
PRESCRIBING PATTERNS LEADING TO PRESCRIPTION-RELATED DEATH

Authors: Andy Nguyen 1* , Roneet Lev 2 , Jonathan Lucas 3 , Michael Menchine 4 , and Jason N.
Doctor 1

Affiliations:
1 Schaeffer Center for Health Policy and Economics, University of Southern California, Los
Angeles, CA.
2 Scripps Mercy Emergency Department, San Diego, CA.
3 San Diego County Medical Examiner's Office, San Diego, CA.
4 Department of Emergency Medicine, University of Southern California, Los Angeles, CA

Abstract
Introduction : The number of prescription-related deaths continue to climb each year. From July
1, 2015 to June 30, 2016, San Diego County had 170 prescription related deaths for which a
decedent’s Controlled Substance Utilization Review and Evaluation System report could be
accessed.
Methods: We utilize two years of California’s Prescription Drug Monitoring Program opioid
and benzodiazepine prescription data leading up to a decedent’s prescription-related death. We
assess patterns using descriptive analysis and data visualization.
Results: Patients continue to see new providers throughout the two years prior to death. On
average, decedents received prescriptions from 7.35 providers cumulatively during the 24 month
period. Prescribers on average prescribed to 1.26 decedents within our cohort.
Conclusion: Patients continue to seek new prescribers leading up to their death; notably
providers that practice Internal Medicine, Family Medicine, and Psychiatry and Neurology in the
final two months.

33

2.1 Background
In the United States, there are approximately 100 million people who suffer from chronic
pain (1) . Patients dealing with pain often turn to prescription medication to reduce their
symptoms of pain. Prescription painkillers can come in many forms: natural opiates,
semi-synthetic opioids, synthetic opioids. Those who suffer from pain may also seek other
methods to suppress their pain through the use of illicit drugs such as heroin, which is chemically
similar to semi-synthetic opioids hydrocodone and oxycodone (2) . Opioid pain therapy has
grown since the 1990s, in part due to several studies concluding that opioids had low addictive
behavior in patients and that they could be prescribed safely long-term (3,4) . By 1996, opioid
prescribing had increased greatly. Overall, lack of information for public officials to enact
policies and training for physicians in opioid prescribing has led to the increase in opioid-related
adverse events (5) .
The opioid crisis has now been deemed as the worst drug overdose epidemic in United
States history by the United States Centers for Disease Control and Prevention (CDC) (6) . The
amount of prescription drugs sold to pharmacies, hospitals, and doctors’ offices quadrupled
between 1999 and 2010 but was not associated with the amount of pain reported by patients
(7-9) . The CDC estimates that approximately 12.5 million people aged 12 or older misused
prescription opioids in 2015 with over 2 million suffering from an opioid use disorder, often
resulting in high rates of hospitalizations and emergency department visits (10,11) .
High levels of opioid use or mixing opioids with other substances can lead to poisoning
and often, fatal consequences. Drug overdose deaths, a majority involving opioids, continue to
increase (12,13) . In 2015, there were 54,404 drug overdose deaths in the United States and
34

33,091 opioid involved deaths (14) . A significant number of prescription opioid deaths are
patients with a history of drug abuse including multiple prescriptions, drug diversion, or doctor
shopping (15) . Synthetic opioids tramadol and fentanyl related deaths also increased 72.4% in
one year between 2014 and 2015. However, fentanyl is often not tested for in medical examiner
toxicology labs, meaning that the number of deaths involving fentanyl could be significantly
higher (15) . Death certificates have also been found to understate opioid involvement in deaths
(16,17) . Opioid deaths are a growing concern and further study into risk factors is required.
Benzodiazepines are also commonly prescribed in conjunction with opioids. However,
the combination of the two can be unsafe and lead to an overdose death. Approximately 23
percent of opioid related deaths also involved the use of a benzodiazepine (12,18) . A study
looking at veterans found that 27% percent of veterans received opioids and benzodiazepines and
approximately half of the deaths involving prescription drugs involved patients who were
prescribed both. Increases in benzodiazepines dose was also associated with an increased risk of
death resulting from drug overdose (19) . Benzodiazepines alone have also been associated with a
higher risk of mortality. Benzodiazepine-related deaths had a six-fold increase between 1996 and
2013 (20) . In recent years, the mortality rate has remained stable . The combination of the two
class of drugs has increased in prevalence and plays a significant role in the growing number of
opioid overdose deaths.
Patients occasionally seek treatment from multiple prescribers, at times receiving
multiple controlled substances during a short period of time from them. This often happens
where prescriptions are written without the knowledge of a patient’s other doctors (21,22) . One
study found that 21.4% of those dying of prescription related deaths had evidence of doctor
35

shopping behavior (23) . A study in San Diego centered on prescription drug-related deaths
related to all prescriptions filled found that more than 50% of prescriptions went to doctor
shoppers (24) . A study of Medicare claims also found that 34.6% of beneficiaries filled opioid
prescriptions from multiple prescribers. Doctor shopping behavior was also associated with more
hospital admissions with a positive correlation between amount of prescribers and rate of
admissions (25) . Twenty states have enacted specific laws and provision to help curb doctor
shopping behaviors (21) .
Many states have adopted what is known as a prescription drug monitoring program
(PDMP) to assist state and health care workers in tracking controlled substances. These
databases house information all on prescriptions for controlled substances in efforts to reduce
drug prescribing, detect doctor shopping, avert co-prescribing, as well as prevent drug diversion.
Prescription drug monitoring programs are a valuable resource to doctors, pharmacists, and law
enforcement in efforts to combat the opioid epidemic.
This study aims to further study prescription patterns by looking at prescription drug
monitoring program data two years prior to a decedent’s death. We examine a cohort of
decedents’ 24 month prescribing history to better understand patterns leading up to a
prescription-related death. We also use data visualization techniques to observe these prescribing
patterns.

2.2 Methods
The cohort for this study consists of prescription-related deaths in San Diego County
between July 1st, 2015, and June 30, 2016. Decedent cause of death was determined by the San
36

Diego County Medical Examiner. Medical examiner death reports include decedent’s first name,
last name, date of birth, date of death, race, gender, known addresses, as well as cause of death.
Cause of death possibilities consist of prescription drugs, over-the-counter drugs, alcohol use,
illicit drug use, and other. The cause of death marker was used to identify decedents who had
prescription drugs as a primary cause or contributing cause. After prescription-related cases were
identified, the CURES database was searched for a two year history of prescriptions. Patients
who had no history of legitimate prescriptions were eliminated from the study so that we could
observe patterns in prescribing for those who received their drugs through the usual proper
channels. An index date was created using the date of death each decedent.
California’s Controlled Substance Utilization Review and Evaluation System (CURES)
provides records of all controlled substance prescriptions, including opioids and
benzodiazepines, sold and dispensed in pharmacies throughout California. The data also provides
date filled, drug name, quantity supplied, days supply, as well as patient and provider identifiers.
Two years of opioid and benzodiazepine prescription data prior to the individual index
date was extracted from the CURES database for each decedent and kept on site at the medical
examiner’s office. Specific prescription information was used to calculate milligram morphine
equivalents (MME) using the conversion formula and conversion factors published by the CDC.
Decedent and prescriber information are used to establish a connection between a decedent and
their provider. Prescriber identifiers were used to find Board Certification information using the
California Medical Board search function.
CURES data was combined with the medical examiner reports to provide further details
for study. This dataset has significant advantages over other datasets in that it provides all
37

prescription data, whereas other datasets may only have drug information that was submitted for
claims or self-reported in a survey. The combination of medical examiner data also gives the
demographic information and especially the cause of death information, which is not available in
most other datasets.
We run and report descriptive statistics on the initial prescription in our dataset,
particularly MME per day prescribed as well as the patients relationship with that prescriber
throughout the 24-month period. The total number of prescribers a decedent receives
prescriptions from throughout the 24-month time period as well as a number of decedents a
provider prescribes to throughout that time period is assessed and reported.
We also study new connections, which we define as the initial prescription from a
prescriber to a decedent that establishes a new relationship that did not previously exist in our
dataset. We employ data visualization to optically analyze the progression of new connections
leading up to death. We also look at who is prescribing to those patients and if the patient is
prescribed an opioid, benzodiazepine, or both in the new connection. Data preparation and the
calculation of statistics is performed in EXCEL 2016. An alluvial chart is created by
RawGraphs.io. Social network visualization is performed in R 3.4.1 using the igraph package
and the Distributed Recursive Layout (26) . Tableau Desktop 2018.1 is employed for the layered
area graphs.

2.3 Results
The mean(SD) age for our decedent cohort is 48.84(13.1). Slightly more than half(57%)
of the decedents were male. The cohort was also predominantly white; 134(78.8%) were
38

identified as Caucasian. Ninety-five patients (55.9%) within the cohort were identified with
cause of death that was solely due to prescription drugs. The remaining fraction of the cohort
involved prescription drugs with another substance; either illicit, alcohol, over-the counter drugs,
or a combination. Table 1 displays the decedent characteristics.
We find that the mean(SD) MME per day a patient receives in our initial prescription is
21.5(52.8), with a median of 3.33 MME per day. Patients stay with their first prescriber for a
mean(SD) of 591.3(209.6) days with a median of 704 days. On average(SD), patients receive
50.3(31.2) prescriptions, with a median of 52, from that prescriber throughout the 24 months of
data. Table 2 shows results for the initial MME per day prescribed as well as the decedent’s
relationship with the first prescriber. 70(41.18%) of patients received a prescription from a
internal or family doctor. Prescribers with a Psychiatry and Neurology board certification was
the next common initial prescriber, writing 29(17.06%) of the first prescriptions. The first
prescriber’s certification breakdown can be found on Table 3. We also find that 68(40%) of the
initial prescriptions are Schedule III drugs, an opioid combination with acetaminophen.
62(36.5%) of the prescriptions were benzodiazepines. A full list of the drug initially prescribed
to the cohort can be found on Table 4.
At two years prior to death, the cohort of decedents had a mean(SD) of 0.87(1.04)
prescribers. At 12 months, that number rose to 3.84(4.16) prescribers. Throughout the 24 month
period, the cohort of patients received a prescription from 989 providers, averaging(SD)
7.35(7.28) prescribers throughout the two years. Table 5 reports the average growth of
prescribers per decedent leading to death. The providers on average(SD) had 1.26(.67) decedents
39

in the cohort. An alluvial chart depicting the cumulative growth of prescribers can be found on
Figure 2.
The data visualizations depicting the progression of new connections can be found on
Figures 3-10. Figure 11 shows the longitudinal raw count of new connections per prescriber and
Figure 12 shows the longitudinal data normalized after min-max scaling.

2.4 Discussion
The results depict interesting patterns leading up to the death of a patient. The initial
prescription to our decedents do not raise concern. Sixty-eight (40%) of the prescriptions are a
Schedule III analgesic that combines an opioid with acetaminophen. The MME per day
prescribed is also within the guidelines released by the CDC in 2016 (27) . The patients on
average, had a long relationship with the initial prescriber. Given that patients on average
received prescriptions from more than 7 prescribers throughout the two years, it can be inferred
that patients concurrently saw multiple providers, rather than leave their current provider to seek
a new connection with another.
The data visualization provides our most startling findings. The network graph shows the
rapid progression of new connections made throughout the two years. The graphs also show that
throughout the 117(68.8%) out of the 170 decedents were connected through receiving
prescriptions from the same prescribers. 778(78.7%) out of the 989 prescribers were connected
through those decedents who saw multiple prescribers or by prescribing to multiple decedents.
The network graphs show how quickly patients and prescribers can be connected, even when
only considering opioid and benzodiazepine prescriptions. This also suggests that while PDMP
40

data is available, prescribers may not be checking to see what other prescriptions patients are
concurrently receiving. This may be due to prescribers choosing not to check the PDMP due to
the barriers of use, such as difficulty navigating the data (28) . Further study on methods to
encourage prescriber use of PDMP data is needed.
The longitudinal new connections graph shows that within the final two months leading
up to death, there is a sharp increases in new connections between our cohort of decedents and
prescribers that practice Internal Medicine, Family Medicine, and Psychiatry and Neurology.
While clinicians practicing Psychiatry and Neurology see a sharp rise in new connections for
benzodiazepines, internal doctors and family doctors have a sharp rise in new connections where
patients are prescribed an opioid or a benzodiazepine. While there is concern within the medical
community that Emergency Medicine doctors are fueling the opioid crisis, we find that Internal
and Family Medicine practices are contributing more to this epidemic instead (29) .
This research has several limitations. Fifty(22.7%) of decedents who were classified as a
prescription-related death did not have a CURES report. This means that the patients died from a
prescription drug that was not prescribed to them. However, our study is solely focused on
patients who have gone through a prescriber. Patients with legitimate prescriptions are more
likely to be affected by a prescriber intervention. Another limitation is that data up to the two
years leading to death is used. This prevents us from capturing a patient’s entire history of opioid
consumption. Despite this, the data visualizations still give valuable insights to the rapid
progression of new connections, especially those within the final two months leading up to death.
Further study into a patient’s entire prescription history is needed.
41

2.5 Conclusion
Throughout the course of the two years leading up to death, decedents continually seek
more prescribers which lead to a connected prescribing network. In the final two months before
death, there is a sharp increase in new connections, notably between decedents with doctors
certified for Internal Medicine, Family Medicine, and Psychiatry and Neurology.

42

2.6 References
1. Institute of Medicine. Relieving Pain in America: A Blueprint for Transforming Prevention,
Care, Education, and Research. Washington, DC: The National Academies Press. (2011)

2. National Security Council., Prescription Nation 2016: Addressing America's Drug Epidemic
(available at
http://www.nsc.org/RxDrugOverdoseDocuments/Prescription-Nation-2016-American-Drug-Epid
emic.pdf) (2016).

3. Portenoy RK, Foley KM., Chronic use of opioid analgesics in non-malignant pain: report of
38 cases. Pain 25: 171-86. (1986).

4. Meldrum ML. The property of euphoria: research and the cancer patient. Opioids and Pain
Relief: A Historical Perspective. 196:200-208. (2002).

5. Modesto-Lowe V, Brooks D, Petry N., Methadone deaths: risk factors in pain and addicted
populations. J Gen Intern Med 25:305-9. (2010)

6. Paulozzi LJ.,The epidemiology of drug overdoses in the United States. Presented at Promis.
Leg. Responses to the Epidemic of Prescription Drug Overdoses in the U.S., Maimonides Med.
Cent. Dep. Psychiatry, Dec. 2, Grand Rounds, Brooklyn (2010).

7. US Department of Justice. Automation of Reports and Consolidated Orders System
(ARCOS). Springfield, VA: US Department of Justice, Drug Enforcement Administration
(DEA); 2011.

8. Paulozzi LJ, Jones CM, Mack KA, Rudd RA., Vital Signs: Overdoses of Prescription Opioid
Pain Relievers-United States, 1999-2008. MMWR 60(43) :1487-1492. (2011).

9. Chang H, Daubresse M, Kruszewski S, et al. Prevalence and treatment of pain in emergency
departments in the United States, 2000 - 2010. Amer J of Emergency Med 32(5) : 421-31.
(2014).

10. Mattson CL. et al., Annual Surveillance Report of Drug-Related Risks and Outcomes CDC
2017., CDC.

11. Daubresse M, Chang H, Yu Y, Viswanathan S, et al., Ambulatory diagnosis and treatment of
nonmalignant pain in the United States, 2000 - 2010. Medical Care 51(10) : 870-878. (2013).

12. Wide-ranging online data for epidemiologic research (WONDER). Atlanta, GA: CDC,
National Center for Health Statistics. (available at http://wonder.cdc.gov). (2016).

13. Rudd RA., Seth P., David F, Scholl L. Increases in drug and opioid-involved overdose deaths
- United States, 2010-2015. MMWR Morb Mortal Wkly Rep. 65(50-51) (2016).
43

14. O'Donnell JK. et al., Deaths Involving Fentanyl, Fentanyl Analogs, and U-47700 - 10 States,
July-December 2016. MMWR. 66(43) ; 1197-1202. (2017)

15. Manchikanti L., Helm S., Fellows B., et al., Opioid epidemic in the United States. Pain
Physician 2012; 15: ES9-ES38. (2012).

16. Seth P., Fentanyl Law enforcement submissions and increases in synthetic opioid-Involved
overdose deaths - 27 states, 2013-2014. MMWR. 2016. 65(33):837-43. (2016).

17. Ruhm CK., Drug poisoning deaths in the United States, 1999-2012: a statistical adjustment
analysis., Popul Health Metr. 14. (2016).

18. Benzodiazepines and Opioids. Website (available at
https://www.drugabuse.gov/drugs-abuse/opioids/benzodiazepines-opioids) (2017).

19. Park TW. et al., Benzodiazepine prescribing patterns and deaths from drug overdose among
US veterans receiving opioid analgesics: case-cohort study. BMJ. 350. (2015).

20. Bachhuber MA. et al., Increasing Benzodiazepine Prescriptions and Overdose Mortality in
the United States, 1996-2013. American Journal of Public Health. 106(4); 686-688. (2016)

21. Doctor Shopping Laws, CDC. (available at
https://www.cdc.gov/phlp/docs/menu-shoppinglaws.pdf).

22. Sansone, Randy A., and Lori A. Sansone., Doctor Shopping: A Phenomenon of Many
Themes. Innovations in Clinical Neuroscience 9: 42–46. (2012)

23. Hall AJ., Patterns of abuse among unintentional pharmaceutical overdose fatalities.
JAMA . 300(22) :2613-20. (2008).

24. Lev R. et al., Who is prescribing controlled medications to patients who die of prescription
drug abuse? The American Journal of Emergency Medicine . 34(1) :30-35. (2016)

25. Jena AB. et al., Opioid prescribing by multiple providers in Medicare: retrospective
observational study of insurance claims. BMJ 348:g1393 (2014)

26. Martin, S., Brown, W.M., Klavans, R., Boyack, K.W. DrL: Distributed Recursive (Graph)
Layout. SAND Reports . 2936: p. 1-10. (2008)

27. Dowell, D. et al. CDC Guideline for Prescribing Opioids for Chronic Pain — United States,
2016. MMWR Recommendations and Reports . 65(1);1–49. (2016)

44

28. Rutkow L., Turner L., Hwang C., Alexander GC., Most primary care physicians are aware of
prescription drug monitoring programs, but many find the data difficult to access., Health
Affairs . 34(3) (2015).

29. Axeen, Sarah et al. Emergency Department Contribution to the Prescription Opioid Epidemic
Annals of Emergency Medicine . 71(6). 659 - 667.e3 (2018)

45

2.7 Figures and Tables
Figure 2.1. Consort Diagram

46

Figure 2.2. Cumulative Growth of Connections with Prescribers, Alluvial chart by 6 month
periods

47

Figure 2.3. Connections, 22-24 Months Prior to Decedent Death

48

Figure 2.4. Connections, 19-24 Months Prior to Decedent Death

49

Figure 2.5. Connections, 16-24 Months Prior to Decedent Death

50

Figure 2.6. Connections, 13-24 Months Prior to Decedent Death

51

Figure 2.7. Connections, 10-24 Months Prior to Decedent Death

52

Figure 2.8. Connections, 7-24 Months Prior to Decedent Death

53

Figure 2.9. Connections, 4-24 Months Prior to Decedent Death

54

Figure 2.10. Connections, Two Years Prior to Decedent Death

55

Figure 2.11. New Connections by Certifying Board, Raw Counts

56

Figure 2.12. New Connections by Certifying Board, Normalized with Min-Max Scaling

57

Table 2.1. Decedent Characteristics

58

Table 2.2. First Prescription Characteristics

59

Table 2.3. Certifications of Initial Prescriber

60

Table 2.4. Drugs Initially Prescribed

61

Table 2.5. Cumulative Number of Prescribers Per Decedent on Average, by 6 months

62

CHAPTER 3:
OPIOID PRESCRIBING DECREASES AFTER
LEARNING OF A PATIENT’S FATAL OVERDOSE
Authors: Jason N. Doctor 1 , Andy Nguyen 1 , Roneet Lev 2 , Jonathan Lucas 3 , Henu Zhao 1 , Tara
Knight 1 , and Michael Menchine 4

Affiliations:
1 Schaeffer Center for Health Policy and Economics, University of Southern California, Los
Angeles, CA.
2 Scripps Mercy Emergency Department, San Diego, CA.
3 San Diego County Medical Examiner's Office, San Diego, CA.
4 Department of Emergency Medicine, University of Southern California, Los Angeles, CA.

Abstract:
In the current opioid epidemic, most prescription deaths occur among people with common
conditions, like back pain, where prescribing risks outweigh benefit. Psychological insights may
explain this prescribing phenomenon. People judge risk as low without available personal
experiences in memory. They are less careful when not observed. They also may falter without
an injunction from authority. We test a way to overcome these pitfalls. We randomized 861
clinicians prescribing to 170 persons with subsequent fatal overdose to either receive courtesy
notification of their patient’s death and a safe prescribing injunction from their County’s medical
examiner or to a control condition. Morphine equivalent (ME) doses filled by patients of letter
recipients compared to controls decreased -9.7% (CI 95%:- 6.2% to -13.2%; aggregate 𝚫 ME
-0.5kg (-0.3kg to -0.7kg); p < 0.001) over a 3-month period following intervention. We also
observed both fewer high dose prescriptions and new patients started on opioids among letter
recipients.

63

3.1 Introduction
The U.S. is in the grips of its worst drug crisis in its history. The crisis is driven by twin
epidemics of illicit and prescription opioid use. Both epidemics have their origins in the 1990s
when medicine undertook a campaign to eliminate pain. Without credible evidence, experts
depicted the risk of opioid addiction as low. ( 1) Today, more than a third of Americans are
prescribed opioid painkillers every year. Volumes dispensed are 3 times higher than in 1999. The
greater availability of prescription opioids has led to an alarming rise in harms. In 2015, there
were over 365,000 emergency department visits for misuse and 20,101 prescription overdose
deaths, more than have ever been recorded in U.S. history. ( 2) In addition, 1.9 million Americans
suffer from opioid addiction. In March of 2016, the Centers for Disease Control and Prevention
(CDC) issued guidelines for prescribing opioids for chronic pain that encouraged sparing and
judicious use of opioids. In an unprecedented step, the U.S. surgeon general, in August of 2016,
sent a letter to each and every physician that included a “pocket card” summarizing the CDC
guidelines. The effects of this effort are not yet known.
This opioid crisis would not have happened without social and behavioral influence.
Early on there was a call for use of behavioral strategies to expand pain care and to conduct
“behavioral research into analgesic prescribing”. ( 3) Leading pain doctors asserted that patients
have “a right to pain relief”. ( 4) This framed the decision to treat pain as response to a societal
struggle, rather than as a one where benefits must outweigh harms. A national initiative called
“Pain as ‘the 5th vital sign’” sought to numerically “code” pain within the individual. ( 5) The 5th
vital sign initiative made pain more visible in the clinic, enabled communication about pain
64

treatment and dissociated pain from its broader social and economic determinants that could
detract from a pill-centric response.
Just as social and psychological factors led to overprescription of opioids, they deserve
consideration in bringing prescribing down to safer levels. There are several such factors that
may maintain high levels of opioid prescribing. First, the U.S. healthcare system is fragmented.
There is no unified patient health record. As a result, patients who return to clinic uneventfully
for an opioid refill are disproportionately observed as compared to patients who do not return
having died of a poisoning. This biased exposure to survivors yields non-representative samples
from which to form clinical judgments. ( 6) A related concern is how people construe risk. People
rely upon knowledge that is impactful, recent and easy to retrieve from their memory when
judging probabilities and making decisions. The more easily people can call some event to mind,
the greater their belief in that event’s occurrence. The effect of this “availability” of information
in risk judgments has been documented in several experiments. ( 7–10) If personal experience
dictates that opioid harms are not available to memory, then it is likely that judged risk of harms
will be low and decisions to avoid harms will occur infrequently. Second, an individual’s
behavior improves when outside persons attend to it. ( 11) When no one is assigned to pay
attention, bad outcomes may go unnoticed in the clinic. Yet, if clinicians perceive that people are
observing their behavior, then this might lead to more careful prescribing decisions. Third,
people respond to what others approve or disapprove of, particularly when delivered by a person
in authority. ( 12) The most devastating harm an opioid can inflict is death from a poisoning. A
message communicating a patient’s death from the medical examiner—a person with authority to
investigate suspicious causes of death—may have particular weight. We hypothesize that
65

decreases in opioid prescribing may follow from an intervention that builds on the
aforementioned behavioral insights: 1) people will judge risks by scanning their memory for
personal experiences that indicate harm, 2) they will alter their behavior once they know that
they are being observed, and 3) they will respond to norms that are conferred by an authority in
the cause of death of their patient. In a randomized experiment, we test an intervention based on
these insights, evaluating the effect of a personal letter from the medical examiner notifying
clinicians of a scheduled drug death in their practice, along with guideline recommendations, and
its effect on subsequent opioid prescribing.

3.2 Methods
In this study, the sample consists of clinicians (physicians, physician assistants, dentists
and nurse practitioners) practicing in San Diego County who, within a 12 month period, had
prescribed a Schedule II-IV drug (i.e., drugs as defined by the United States Controlled
Substances Act as having an abuse potential, but also a currently accepted medical use) to a
person who suffered a Schedule II-IV drug poisoning. The sample was a countywide
heterogenous group of primary care clinicians, specialists and allied health professionals with
scheduled drug prescribing privileges. The study was a decedent-cluster randomized field
experiment; clusters of prescribers for each decedent were randomly allocated to either treatment
or control. This design avoids treatment contamination—clinicians prescribing to the same
patient sharing information about the intervention. If two or more decedents had the same
prescriber, then before randomization that prescriber was assigned to one and only one of those
decedents by random draw. Two factors served as random blocks: 1) whether the decedent
66

received a prescription from a clinician with multiple deaths (two levels), and 2) whether the
cause of death was due to opioids only, opioids in combination with benzodiazepines, or
benzodiazepines only (three levels). The first block functioned to form equivalent groups on
prescriber risk posture and the second block formed equivalent groups on preference for type of
scheduled drug prescribed in practice. Blocks were crossed to form 2 x 3 = 6 randomization
groups. The institutional review board at the University of Southern California approved all
study procedures and waived informed consent for participants under HHS regulations at 45
CFR 46.116(c), as the study was evaluating a County public service safe prescribing program.
Protected health information was not shared outside the San Diego County Medical Examiner’s
Office and its employees and volunteers working in an official capacity. The trial was registered
at clinicaltrials.gov (NCT02790476) before it began.
The Controlled Substance Utilization Review and Evaluation System (CURES) system
provided records of all opioids dispensed at California pharmacies attributable to each provider
in our sample. We converted included drugs (See list in Table S1) to milligram morphine
equivalents (MME) using MME conversions formulas published by the CDC here . For ease of
interpretation, our analysis assesses changes over time before and after the intervention in natural
log-transformed MME between treatment and control. Deaths where prescription drugs were the
primary or contributing cause were cataloged by the medical examiner between the period of
July 1st, 2015 and June 30th, 2016. As part of the county safe prescribing program, prescribers to
those deaths were identified in the CURES system. All California pharmacies and clinics that
dispense controlled substances must submit reports to CURES on a weekly basis. In consultation
with the program manager of CURES, a letter notifying CURES administrators that prescriptions
67

from these prescribers would be evaluated prospectively after the safe prescribing letters were
sent was submitted to the CURES system. First name, last name, date of birth and address
identified each decedent in the medical examiner reports and CURES data. Drug Enforcement
Agency (DEA) number, identified the prescribers in CURES data. Data from eligible clinicians
were extracted from CURES and kept on site at the medical examiner’s office for de-identified
analytic file preparation. A file stripped of patient and clinician identifiers was prepared and
released for analysis on secure servers at University of Southern California.
A letter signed by the Chief Deputy Medical Examiner of San Diego County (see letter in
S2) notified prescribers of the death in their practice. The letter outlined the annual number and
types of prescription drug deaths seen by the medical examiner, discussed the value of and way
to access the State’s prescription drug monitoring program, and discussed five CDC guideline
recommended safe prescribing strategies: 1) Avoid co-prescribing of opioids with
benzodiazepines, 2) prescribe minimal dose necessary for acute pain, 3) consideration of slow
tapers with pauses to below 50 MME per day, 4) avoid prescriptions lasting greater than
3-months for pain, and 5) prescribing naloxone in conjunction with opioids for patients taking >
50 MME per day. Letters were generated, signed and posted in the U.S. mail January 27th, 2017.
Randomization was carried out blocking on the two factors described above, resulting in
six blocks. Six decedent lists were generated from the crossed block levels. Using random.org’s
sequence generator, ( 13) true random integer sequences derived from atmospheric noise
determined decedent order in each list. For each ordered list, prescribers to decedents in the first
half were those who received the intervention.

68

Descriptive and inferential statistics were carried out in STATA (Version 14.0; Stata
Corporation, College Station, TX). The cmp command in STATA was used to compute a
difference in differences estimator within a mixed-model censored linear regression. ( 14) The
difference in differences estimator compares the average change over time in MME dispensed
for prescribers in the group receiving the letter, compared to the average change over time for
those prescribers in the control group. For ease of interpretation and to ensure normally
distributed data, we evaluate the natural log of MME. Censored regression has a continuous
component and a discrete one. The natural log MME doses estimated over days where opioids
were dispensed in the name of a prescriber represent the continuous part, and days with no
opioids filled in that prescriber’s name represent the discrete part of the model. We denote
estimation of the dependent variable as to distinguish it from uncensored log(MME)
*
︿

estimation, , that considers only > 0mg dispensing days. The analysis is represented log(MME)
︿

by:
[1] β x β x β x δ log(MME)
*

ijk
︿
=
1 1ij
+
2 2ij
+
3 3ij
+
i(k)

where , and are fixed effects coefficients on time, , intervention, ,and time by β
1
β
2
β
3
x
1
x
2

intervention interaction, , respectively. Alphabetic subscripts describe the i th prescriber, j th x
3

prescription filled and k th decedent, the nested random intercept is normally distributed δ
i(k)

with mean zero and variance, , for each i prescriber nested in (i.e., having prescribed to) σ
2
i(k)

decedent k , i ( k ). With natural log transformed data, the value 100·[ exp { } - 1] measures β
3

precisely the percentage change in MME attributable to the intervention (see Woolridge (2008),
69

pp. 212 and 410). ( 15) We evaluated data 3 months pre-intervention and 3 months
post-intervention, with post-intervention beginning one month after the letter was sent to
washout contamination by prescriptions written before receipt of the letter but filled afterwards.
We perform sensitivity analysis in regression assumptions that are described in the supplemental
materials (S3). As secondary hypotheses, we evaluate if in the letter condition there are fewer
opioid prescriptions at a high dose (≥ 50 and > 90 MME per day), and also if there fewer new
patients entering the database who were started on opioids. High dose prescriptions were
evaluated using censored regression on percent of clinician fills on high dose and chance of a fill
being a “new starts” were evaluated using logistic regression. These analyses were also
conducted within a difference and difference framework.

3.3 Results
The study timeline is provided in Figure 1, while Figure 2 describes the flow of decedent
and prescriber identification, intervention allocation, follow-up, and data analysis. The medical
examiner investigated 220 deaths in San Diego County between July 1, 2015 and June 30th,
2016 for which a Schedule II-IV prescription drug was the primary or contributing cause of
death. Of these 220 persons, 170 of them (77.3%) had one or more legitimate Schedule II-IV
drug prescriptions filled at a pharmacy in California in the 12 months prior to their death.
Observed frequencies within randomization blocking variables (cause of death and whether
decedents received prescriptions from a clinician with multiple deaths is presented in Table 1).
There were 861 prescribers to the 170 decedents with Schedule II-IV poisoning. Of these, 725
had prescribed to only one and 136 had prescribed to more than one decedent. Sample
70

characteristics are presented in Table 2. There are 16,000 DEA registrants in San Diego County
and approximately 10,000 active medical licenses to MDs and DOs. (16) This means that
795/10,000 or roughly 8% of MD or DO prescribers with a DEA number in San Diego County
had prescribed a Schedule II-IV drug to a patient who died of a Schedule II-IV drug overdose in
the 12 month period between July 1st, 2015 and June 30th, 2016.
We observed 1,279,691 prescriptions filled during the study period. From these and dates
of prescriptions dispensed we computed daily MME for 234,379 prescriber fill x days. Table 4
shows the average daily milligram ME dispensing rates per prescriber attributed to intervention
and control providers 3 months before and 1-4 months after letters were sent. In the control
group, prescribing increased pre- to post- 0.3 milligram ME (CI 95% 4.0 to -3.4 milligram ME),
whereas in the intervention group prescribing decreased -13.7 milligram ME (CI 95% -10.1 to
-17.5 milligram ME) per prescriber per day.
There was a -9.7% (CI 95%: -6.2% to -13.2%) percentage decrease in milligram ME
doses dispensed at pharmacies for prescriptions written by those clinicians in the intervention
group as compared to controls over a period beginning 1 month and ending four months after the
day the letters were mailed (p < 0.001). In addition, letter recipients were 7% (CI 95%: 2% to
11%) less likely than control prescribers to start a new patient on opioids (p < 0.01). We also
observe modest but statistically significant reductions in high dose prescribing in the intervention
group compared to control: a 3% decrease for 50 milligram ME daily doses (p < 0.05) and 4.5%
decrease for 90 milligram ME daily doses (p < 0.05) dispensed. Collectively, the 388 clinicians
in the intervention condition avoided prescribing 0.5 kg of morphine than would be expected
without the letter over the 90-day post intervention period. In clinical terms, each day clinicians
71

receiving the intervention wrote one fewer prescription of 5mg hydrocodone to be taken 4x a day
for 21 days. Sensitivity analysis that transformed MME ensured normality and a lag-correlation
analysis to ensure independence of errors each had no effect on results (See S3).

3.4 Discussion
We find a simple intervention to inform clinicians of a scheduled drug harm to their
patient resulted in fewer subsequent opioids dispensed from prescriptions. There is a strong
ethical justification for informing clinicians about harm to their patients in a factual and
non-judgmental way so that they may learn from this information and provide better care other
persons in the future. In addition, such information appears to lead to more cautious prescribing
without restricting clinician freedom to prescribe opioids through mandated prescribing limits.
Mandated limits do not consider individual patient circumstances that may arise in the course of
care. We observe modest prescribing reductions, suggesting that clinicians exercised greater
caution with opioids rather than abandoning their use.
Brief exposure to opioids in opioid-naive persons makes long-term opioid use more
likely (17) , (18) and excess prescription opioids (found in medicine cabinets or diverted) are a
source of misuse and are linked to transition to heroin. (19, 20) Correcting course in prescribing
after learning of a patient’s death from a Schedule II-IV prescription may lessen the impact of
the aforementioned harms.
While there are some limitations to the generalizability of our findings, San Diego
County is a diverse county, with a broad representation of providers and which constitutes
approximately 1% of the U.S. population. The control condition involved no increased
72

awareness of opioids or education. Yet, the study took place five months after the U.S. Surgeon
General issued guideline pocket cards to all clinicians in the U.S., including those in our control
arm, describing the CDC’s guidelines. We do not address appropriate or inappropriate
prescribing at a patient level.
Judicious prescribing represents only one of the components necessary to correct the
missteps caused by overly enthusiastic use of opioids in the clinic. Access to medication assisted
therapy, counseling, naloxone for resuscitation after overdose and efforts to address social
determinants responsible for increased opioid use all play an equally important role in ending the
crisis.
The intervention described here is scalable. Each county in the U.S. reports prescription
opioids deaths to the National Center for Health Statistics and each state maintains a vital records
death file. Each state contains a prescription drug monitoring program that tracks prescriptions to
decedents. It is feasible to “close the loop” on deaths in the clinic to encourage safe prescribing
through use of behavioral insights.

73

3.5 References
1. P. T. M. Leung, E. M. Macdonald, M. B. Stanbrook, I. A. Dhalla, D. N. Juurlink, A 1980
Letter on the Risk of Opioid Addiction. N. Engl. J. Med. 376, 2194–2195 (2017).
2. Website, (available at http://wonder.cdc.gov/ucdicd10.html ).
3. M. B. Max, Improving Outcomes of Analgesic Treatment: Is Education Enough? Ann.
Intern. Med. 113, 885 (1990).
4. F. Brennan, D. B. Carr, M. Cousins, Pain management: a fundamental human right.
Anesth. Analg. 105, 205–221 (2007).
5. D. W. Baker, History of The Joint Commission’s Pain Standards. JAMA . 317, 1117
(2017).
6. R. M. Dawes, D. Faust, P. E. Meehl, Clinical versus actuarial judgment. Science . 243,
1668–1674 (1989).
7. J. A. Cantone, R. L. Wiener, Quicker and Sooner Counterfactuals: The Availability
Heuristic and Negligence Decisionmaking. PsycEXTRA Dataset , , doi: 10.1037/e506102012-061 .
8. C. Ochoa, K. W. Jolly, H. M. Hosch, Facial Morphing and the Availability Heuristic in
Alibi Evaluations. PsycEXTRA Dataset , , doi: 10.1037/e509482008-001 .
9. A. Tversky, D. Kahneman, Availability: A heuristic for judging frequency and
probability. Cogn. Psychol. 5, 207–232 (1973).
10. W. Bruine de Bruin et al. , Promoting protection against a threat that evokes positive
affect: The case of heat waves in the United Kingdom. J. Exp. Psychol. Appl. 22, 261–271
(2016).
11. J. McCambridge, J. Witton, D. R. Elbourne, Systematic review of the Hawthorne effect:
new concepts are needed to study research participation effects. J. Clin. Epidemiol. 67, 267–277
(2014).
12. R. B. Cialdini, Crafting Normative Messages to Protect the Environment. Curr. Dir.
Psychol. Sci. 12, 105–109 (2003).
13. M. Haahr, RANDOM.ORG - Sequence Generator, (available at
https://www.random.org/sequences/ ).
14. J. F. McDonald, R. A. Moffitt, The Uses of Tobit Analysis. Rev. Econ. Stat. 62, 318
(1980).
15. J. Wooldridge, Introductory Econometrics: A Modern Approach (Cengage Learning,
74

2008).
16. R. Lev et al. , Who is prescribing controlled medications to patients who die of
prescription drug abuse? Am. J. Emerg. Med. 34, 30–35 (2016).
17. A. Shah, C. J. Hayes, B. C. Martin, Characteristics of Initial Prescription Episodes and
Likelihood of Long-Term Opioid Use - United States, 2006-2015. MMWR Morb. Mortal. Wkly.
Rep. 66, 265–269 (2017).
18. M. L. Barnett, A. R. Olenski, A. B. Jena, Opioid-Prescribing Patterns of Emergency
Physicians and Risk of Long-Term Use. N. Engl. J. Med. 376, 663–673 (2017).
19. R. A. Pollini et al. , Problematic use of prescription-type opioids prior to heroin use
among young heroin injectors. Subst. Abuse Rehabil. 2, 173–180 (2011).
20. T. J. Cicero, M. S. Ellis, H. L. Surratt, S. P. Kurtz, The Changing Face of Heroin Use in
the United States. JAMA Psychiatry . 71, 821 (2014).
22. Box, G., & Cox, D. (1964). An Analysis of Transformations. Journal of the Royal
Statistical Society. Series B (Methodological), 26(2), 211-252 (1964)..

75

3.6 Figures and Tables

Figure 3.1. Study timeline

76

Figure 3.2. Consort diagram

77

Table 3.1. Blocking variables and group
Received Rx from
prescriber with
multiple deaths
Received Rx from
prescriber with
single death
Marginal total
Benzodiazepines only

7 10 17 deaths
Benzodiazepine +
Opioids

93 22 115 deaths
Opioids only

22 16 38 deaths
Marginal Total 122 48

TOTAL = 170

78

Table 3.2. Decedent Characteristics

79

Table 3.3. Prescriber Characteristics

80

Table 3.4. Results

81

3.7 Supplementary Materials

Table S1. Included Drugs in Analysis
Included Excluded
Codeine Buprenorphine
Fentanyl tablet Butorphanol
Fentanyl patch Dihydrocodeine
Hydrocodone Fentanyl lozenge
Hydromorphone Fentanyl powder*
Methadone Fentanyl spray*
Morphine Levorphanol tartrate
Oxycodone Meperidine
Oxymorphone Opium
Tapentadol Tramadol
*Excluded because prescription was likely administered in office by a healthcare professional

List taken from CDC Commonly Prescribed Opioids
https://www.cdc.gov/mmwr/volumes/65/rr/rr6501e1.htm

82

S2. Letter to Prescribers

San Diego County Medical Examiner Office

Dear ______________(name prescriber),

This is a courtesy communication to inform you that your patient (Name, Date of Birth) died on
(date). Prescription drug overdose was either the primary cause of death or contributed to the
death.

The San Diego Medical Examiner’s office sees between 250 and 270 prescription
medication-related deaths each year. A significant proportion of deaths are due to the
combination of multiple prescription medications. Patients may obtain legitimate prescriptions
for opioids, benzodiazepines, muscle relaxants, and sleep aids from more than one prescriber.
When taken in any combination, these medications put patients at greater risk of death. We also
see many deaths that are a result of long-term therapeutic prescribing.

Controlled Substance Utilization Review and Evaluation System (CURES) helps prescribers
who are dedicated to avoiding prescribing controlled substances when they are likely to do more
harm than good. CURES contains information about whether other clinicians had prescribed
controlled substances to your patient. This type of information can help prescribers make
informed decisions and avoid duplicate or additive types of medications from being provided to
patients. We ask that you commit to prescribe safely by registering for and regularly logging in
to CURES before prescribing controlled substances. On the CURES website you may run a
report on any patient you are considering prescribing controlled substances to in order to find
their detailed prescription history. CURES data is available for only the last 12 months for
patients.

You can register for CURES at
https://cures.doj.ca.gov/registration/confirmEmailPnDRegistration.xhtml.
You can access CURES at https://cures.doj.ca.gov

The following evidence-based interventions also lower overdose death rates:

1. Avoid co-prescribing an opioid and a benzodiazepine. We found this combination in
over 50% of CURES reports and in over 20% of toxicology results of patients who died of an
overdose.
83

2. Minimize opioid prescribing for acute pain . According to the Centers for Disease
Control and Prevention (CDC), clinicians should avoid opioids, and when necessary, start with
the lowest effective dose of immediate-release opioids. Three days or less will often be
sufficient. Opioids should not be considered first-line or routine therapy for chronic pain.[1]
3. Taper opioids to safer doses. The CDC recommends that for patients already on
long-term opioid high dose opioid therapy, taper to a dose that is lower than 50 milligrams of
morphine equivalent and that slow opioid tapers as well as pauses in the taper may be needed for
long-term users .[2]
4. Avoid “the 90-day cliff.” We found that nearly 70% of patients who died were
prescribed the same medication for 3 consecutive months. The CDC recommends opioids
should be discontinued if benefits do not outweigh risks (if realistic goals for pain and function
have not been met).[3]
5. The CDC recommends prescribing naloxone to patients on higher than 50 milligrams
of morphine equivalents daily.[4]

We are aware of the challenges in balancing the potential harm and benefit of controlled
medication prescribing for your patients. Therefore, please visit:

http://sandiegosafeprescribing.org/

and click the link named “Did you get a letter from the Medical Examiner?” Here you will find
links to the CDC guidelines, local addiction referral resources, including medication-assisted
treatment, a clinical advice hotline, regimens for successful tapering and other information.

Learning of your patient’s death can be difficult. We hope that you will take this as an
opportunity to join us in preventing future deaths from drug overdose.

Sincerely,

Jonathan Lucas, MD
Chief Deputy Medical Examiner

[1] Recommendations #1, 6, CDC Guideline for Prescribing Opioids for Chronic Pain, 2016
[2] Recommendation #5, 7, CDC Guideline for Prescribing Opioids for Chronic Pain, 2016
[3] Recommendation # 2, 7, CDC Guideline for Prescribing Opioids for Chronic Pain, 2016
[4] Recommendation #8, CDC Guideline for Prescribing Opioids for Chronic Pain, 2016

84

CHAPTER 4:
PREDICTING LONG-TERM OPIOID USE IN PATIENTS FOLLOWING HEART
VALVE REPLACEMENT: AN INSTRUMENTAL VARIABLES APPROACH

Authors: Andy Nguyen 1 , Michael Menchine 2 , and Jason Doctor 1

Affiliations :
1 Schaeffer Center for Health Policy and Economics, University of Southern California, Los
Angeles, Ca.
2 Department of Emergency Medicine, University of Southern California, Los Angeles, Ca.

Abstract
Introduction: Several studies have analyzed the association between initial prescription opioids
and long-term opioid use. However, there may be confounding variables which can explain the
correlation.
Method: This study aims to employ an instrumental variables approach using the initial
provider’s overall opioid prescribing patterns as an instrument for the initial dose of a
prescription to a patient following heart valve replacement surgery to assess likelihood of
long-term opioid use. Data from the Optum Clinformatics Data Mart between years 2007-2016
was assessed. Primary outcomes include whether a patient had prescription opioid use at 90+
days, 180+ days, and 270+ days, 90-180 days, and 180-270 days following heart valve
replacement surgery.
Results: A 1 log-distance increase in logged morphine milligram equivalents (MME) per day for
a patient’s initial prescription results in a 11.28 percentage point increase in likelihood of
long-term opioid use for 90-180 days (CI 95% 6.3-16.2) and 13.8 percentage point increase
180-270 days (CI 95% 9.5-18.1) following heart valve replacement. A 1 log-distance increase in
85

logged total MME for a patient’s initial prescription results in a 13.3 percentage point (CI 95%
7.5-19) increase in likelihood of persistent painkiller use for both 90-180 days and a 16.1
percentage point increase in likelihood 180-270 days following heart valve replacement surgery.
Conclusions: Higher levels of initial opioid dose following heart valve replacement surgery
causes a higher likelihood of long-term opioid use.

86

4.1 Background
Opioid analgesics are commonly prescribed to patients following surgery (1) . Despite the
intention of acute pain relief, several studies have found that opioid use can persist following
major and minor surgery (2-7) . Approximately 5.9% of opioid-naïve patients have persistent
opioid use 90-180 days following a minor surgical procedure. The incidence is around 3-6.5%
for patients with major procedures (4, 6) . Post-surgical pain relief may contribute to long-term
opioid use, which is currently a major health issue in the United States.
There are several risk factors associated with long-term opioid in opioid-naïve patients.
One risk factor is the level of morphine milligram equivalents (MME) per day a patient is
initially prescribed, particularly if the dose is 90+ MME (8) . Gender and age were found to be
significant risk factors of long-term opioid use (3) . Patients initiated with tramadol or a
long-acting opioid were also associated with a higher incidence of long-term opioid use.
Furthermore, patients with 400+ total MME in their first 30 days following surgery are 3 times
more likely to be chronic opioid users (8) .
The current literature has not addressed potential endogeneity issues. There may exist
unobservable confounding variables. One such example may be a patient’s ability to cope with
pain. Despite being opioid-naïve, they may request a higher dose initial prescription to help
relieve pain. This may also result in long-term opioid use for future instances of pain. A patient’s
ability to cope with pain is an unobservable factor. Therefore, simple regression techniques
would render biased results.
Heart valve replacement surgery is a unique situation because it deals with an ailment
that is not directly related to pain. Patients do not seek a cardiothoracic surgeon for their level of
87

opioids prescribed, but rather for their abilities and reputation with such surgeries. However, the
amount of MME a patient receives is directly related to the provider’s propensity to prescribe
opioids while avoiding any self-selection bias. Therefore, we hope to take advantage of this
situation to employ an instrumental variables model using the patient’s initial provider’s history
as a high-intensity or low-intensity opioid prescriber as an instrument to assess the relationship
between initial opioid levels and long-term use. Due to the patient’s criteria in choosing a
surgeon and care facility, we avoid the potential self-selection bias issue with regards to opioid
levels.
We aim to address the potential endogeneity issue to further study the relationship
between initial opioid prescription levels and the likelihood of a patient having long-term opioid
use following heart valve replacement surgery. We hypothesize that initial opioid levels will
have a causal effect on long-term opioid use after controlling for unobservable confounding
variables in these patients.

4.2 Methods
The Optum Clinformatics Data Mart is utilized for this study. The Optum database is a
rich claims-based data source due the wealth of information available as well as nature of its
size, following 80 million lives globally. The data spans across all therapeutic areas and includes
patients commercially insured as well as those with Medicare Advantage. Longitudinal inpatient,
outpatient, and pharmaceutical data from the year 2007 through 2016 was accessed and
analyzed. Medical claims contain information on patient diagnosis, procedures and services
88

performed. Prescription claims contain drug name, quantity prescribed, days supply, as well as
prescriber identification and payment information.
Patients with heart valve replacement surgery were identified using the Healthcare
Common Procedure Coding System (HCPCS), Current Procedural Technology (CPT) codes, as
well as ICD-9 and ICD-10 procedure codes. Patients with transcatheter aortic heart valve
replacement (TAVR), transcatheter aortic heart valve implantation (TAVI), aortic, mitral,
tricuspid, or pulmonary valve replacements were considered for the analysis. A complete list of
surgeries and their respective procedural codes are included in Supplementary Table 4.1 . An
index date was created to denote the day of the heart valve replacement surgery.
In order to be included in the cohort for analysis, patients must have eligibility during the
time that they receive the heart valve replacement, meaning that the patients are insured through
a plan that uses Optum’s health technology services. In addition, patients needed to have
continuous eligibility 90 days prior to and 365 days after the index date. We did not differentiate
between insurance plan type. Patients were included in the cohort as long if they did not have a
gap in coverage.
Only opioid-naïve patients are considered for this analysis to isolate the effect of the
initial opioid prescription following index surgery. This prevents any residual opioid
consumption from previous prescriptions from influencing the results. Patients are considered
opioid-naïve if they have no prescriptions for any opioid 90 days prior to the index date. Even if
a patient receives a prescription at 91 days prior to index surgery, there is a minimum period of
time when a patient will have no access to prescription opioids. This eliminates patients that are
89

dependent on opioids and require constant pain therapy from the cohort for analysis. A complete
individual timeline for inclusion in the study is shown in Figure 4.1.
We considered a patient’s initial prescription as a prescription written within 7 days
following the heart valve replacement surgery (7) . We also compared the prescription fill date to
discharge to ensure that it was surgery related. A de-identified Drug Enforcement Agency (DEA)
number is used to identify prescribers in the dataset. Several patients did not have a DEA number
associated with their initial prescription and were excluded from the analysis. Prescriber
identification is required for the instrument in this study.
Opioid prescriptions were converted to milligram morphine equivalents for ease of
comparison using the drug name, drug strength, quantity supplied, day’s supply, as well as
conversion factors and formula provided by the CDC on Supplementary Table 4.2. We consider
both the logged MME per day as well as the logged total MME prescribed in the initial
prescription.
We compiled a list of initial prescribers from the initial prescriptions and accessed all
opioid prescriptions written by each prescriber in the Optum claims database. An average MME
per day for all prescriptions written by each individual provider was calculated and used to
classify them as a high-intensity or low-intensity prescriber relative to their peers using a median
split.
Following initial prescription, patients were followed for 365 days after the index date
and flagged for any future opioid fills. Future opioid prescriptions were only considered if they
were written by a prescriber other than the initial provider to prevent the instrument from being
directly correlated with any future prescriptions.
90

We observe several different time points after a patient’s index date to assess long-term
opioid use. This allows for a more granular view of any effect of initial prescription dose and
how long it may persist. The main outcomes of this analysis are whether patients had an opioid
prescription in the following time frames: 90+, 180+, and 270+ days after index surgery as well
as between 90-180 and 180-270 days after index surgery.

Statistical Analysis

A naïve probit model would result in biased estimates due to confounding variables. In
light of this, the following instrumental variables probit model is employed to estimate the effect
of initial prescription dose on long-term opioid use in the reduced form:
(1) Log(η ) Y
*
i
= β
i
+ Xγ
1
+ u

i

(2) og(η ) γ v L
i
= X
2
+
i

is the binary dependent variable representing the i th observation and whether the patient has Y
*
i

a future opioid prescription during one of the defined time frames. X is a matrix that represents
the exogenous variables: the patient age, gender, and the instrument, whether the prescription
was written by a high-intensity or low-intensity prescriber. denotes the treatment og(η ) L
i

variable (logged initial prescription MME dosage) either as the MME per day or the total MME
prescribed. , and are the coefficients of their respective variables and u i as well as v i are β , γ
1
γ
2

the error terms. In the instances where the endogeneity issue with is not present, og(η ) L
i

Equation 1 is sufficient. A Wald-test is employed to assess whether the standard probit model or
instrumental variables bivariate probit model is appropriate. Stata MP Version 14.0 (StataCorp,
College Station, TX) was used for all the data preparation and analyses.
91

4.3 Results
Descriptive Statistics
1,222 patients met the inclusion criteria from the 36,255 patients with heart valve
replacement surgery in the dataset. The number of patients excluded during each data preparation
step is shown in Figure 4.2, the study consort diagram. Since the Optum data only includes
patient year of birth, patient age is estimated by taking the difference between year of index and
the patient’s year of birth. Approximately half (49.92%) the patients were age 65+ with a
mean(SD) age of 63.08(13.99). The study cohort is predominantly male (66.2%). A more
detailed patient characteristic table can be found on Table 4.1.
The treatment variable for the main analysis is the initial prescription that is written for
patients after heart valve replacement surgery. There were 1241 prescriptions written for
individual drugs, with 19 patients who received two different drugs on the same initial
prescription date. 1,101 (88.7%) received a Schedule III or IV combination narcotic analgesic of
ibuprofen or acetaminophen with an opioid. There were 140 (11.3%) Schedule II opioid
prescriptions. Of the Schedule II drugs, 75 (53.6%) prescriptions were oxycodone and 43
(30.7%) fills were for tramadol. The mean(SD) MME per day prescribed to patients on the initial
fill is 80.3(79.33). This is clinically similar to a prescription for 8 pills of 10 mg Hydrocodone
per day. The mean(SD) total MME for the initial prescription was 459.15(461.48). A detailed
breakdown for the characteristics of the initial prescription can be found in Table 4.2.

92

Main Results
Both initial MME per day and total MME as the treatment variable are predictors for
long-term opioid use. We find that higher levels of initial dosage increase the likelihood of future
opioid consumption. A 1 long-distance increase in log MME per day for the initial prescription
results in a .1128 (CI 95% 0.063-0.162) increase in probability, or a 11.28 percentage point
increase in likelihood, of opioid use 90-180 days days after index surgery. The marginal effect
increases for 180-270 days after index surgery, at 13.8 percentage point increase in likelihood.
When considering the logged total prescription MME for both outcomes of opioid use 90-180
and 180-270 days after index surgery, the marginal effects from a 1 log-distance increase
resulted in a 13.3 percentage point, and 16.1 percentage point increase in likelihood,
respectively. While the Wald-test for endogeneity is statistically significant between 90 and 270
days, the Wald-test for regressions 270+ days after surgery, -0.038 (CI 95% -0.218-0.143), ρ =
fails to reject exogeneity. Marginal effects for opioid-use for 270+ days after index surger are
also statistically insignificant. The marginal effects for the instrumental variables probit are
reported for the outcomes: 90+ days, 180+ days, 90-180 days, and 180-270 days after index
surgery. The marginal effects from the naïve probit model are reported for the 270+ day
outcome. Both the logged MME per day and logged total MME model an F-statistic which
indicates a strong instrument (> 10): 691.283 and 440.168, respectively. The full results and
marginal effects are reported on Table 4.3.
A secondary analysis also looked at the Center for Disease Control’s recommended
thresholds, 50 MME/day and 90 MME/day as thresholds. We analyzed them as separate
segments and looked at likelihood of long-term opioid use whether a patient fell within the
93

0-49.9 MME/day, 50-89.9 MME/day, or 90+ MME/day category, particularly look at the
marginal effect of going up to the next category. If the initial prescription was between 50-89.9
MME/day, the patient was 9.2 percentage points more likely to have long-term opioid use
90-180 days after index surgery compared to those under 50 MME/day. If a patient had 90+
MME/day, they were 12.2 percentage points more likely for opioid use 90-180 days after index
heart replacement surgery than if they had received 50-89.9 MME/day. Full results are on Table
4.4.

4.4 Discussion
There is a significant increase in the likelihood of long-term opioid use with higher levels
of initial opioid dosage, either in log MME per day or log total prescription MME. While other
studies have not addressed endogeneity, this study provides an original look at the causal
relationship between the two. One concern raised with the results of this study is that while
physicians are treating an underlying heart condition, a higher opioid prescription may put a
patient, who at the time of index surgery is opioid-naïve, at more risk of long-term opioid use.
Our results suggest that while an opioid prescription may be written to help with a
patient’s initial pain, the initial dosage may increase the likelihood of a host of downstream
effects resulting from long-term opioid use. An area of concern is the possibility of addiction.
While addiction to opioids only occurs in a small percentage of patients with exposure to
opioids, addiction can develop after months of use (9) . The results of this study report an
increased likelihood of long-term use which in turn, can have the consequence of addiction to
pain relievers.
94

Another particular issue that may arise with opioid consumption and addiction is the
possibility of transitioning to heroin use. Approximately 4-6% of all prescription opioid users
eventually begin to use heroin (10) . This is especially concerning for those with heart valve
replacements because intravenous drug use is associated with higher risks of infective
endocarditis, a relationship that increases the likelihood of a second surgery (11,12) . Our
findings suggest that further caution is needed when considering the amount of pain relievers to
treat a patient’s initial discomfort.
While other studies in the literature have looked at whether patients continue opioids at
90+ days after surgery, this research provides a more granular and extended view by looking at
different time points within the first year after surgery (2,4,6) . The results found are more
conservative than the two-month window suggested by the International Association for the
Study of Pain (13) . The results suggest that the effects of initial prescription MME levels are
consistent between 90 and 270 and increase in likelihood of long-term opioid use reduces at
270+ days. However, this effect is still positive and statistically significant. Further research into
future opioid use at further time points will be important to gain more insight on the persistence
of the effects.
There are several limitations to our study. One limitation of the study is the use of claims
data. Data may be incomplete and patients may also not claim all their prescription opioids.
However, the results would suggest a conservative estimate of the marginal effects of initial
prescription MME levels. Further study is needed with complete data such as data from
prescription drug monitoring programs or electronic health records. Overall, our study provides a
robust analysis on the effects of the initial prescription and long-term opioid use.
95

4.5 Conclusion
The results of this study suggest that higher amounts of MME prescribed in the initial
prescription cause a higher likelihood of long-term opioid use. Both levels of MME per day and
total MME are significant predictors of future prescription painkiller consumption. This study
contributes to the literature by addressing endogeneity issues to further explore the downstream
effects of initial prescribing following heart valve replacement surgery.

Funding
This work was supported in part by a grant from Edwards Lifesciences.

96

4.6 References

1. Garimella, V., Cellini, C. Postoperative Pain Control. Clin Colon Rectal Surgery . 26(3):
191-196. doi: 10.1055/s-0033-1351138 (2013)

2. Goesling J. et al. Trends and Predictors of Opioid Uise Following Total Knee and Total
Hip Arthroplasty June, 157(6): 1259-1265. doi:10.1097/j.pain.0000000000000516.
(2016)

3. Sun, E. et al. Incidence of and Risk Factors for Chronic Opioid Use Among
Opioid-Naive Patients in the Postoperative Period. JAMA Intern Med. 176(9):1286-1293.
doi:10.1001/jamainternmed.2016.3298 (2016).

4. Clarke, H. et al. Rates and risk factors for prolonged opioid use after major surgery:
population based cohort study. BMJ. 348: g1251. doi: 10.1136/bmj.g1251.(2014).

5. Brat G.,Postsurgical prescriptions for opioid naive patients and associate with opioid
overdose and misuse: retrospective cohort study. BMJ . 360:j5790 doi:
https://doi.org/10.1136/bmj.j5790 (2018).

6. Brummett, C. et al. New Persistent Opioid Use After Minor and Major Surgical
Procedures in US Adults. JAMA Surg . 152(6): e170504.doi:10.1001/jamasurg.2017.0504
(2017).

7. Alam, G. et al. Long-term analgesic use after low-risk surgery: a retrospective cohort
study. Arch Intern Med. 12;172(5):425-30. doi: 10.1001/archinternmed.2011.1827.
(2012).

8. Shah, A., Hayes, C., Martin B., Factors Influencing Long-Term Opioid Use Among
Opioid Naive Patients: An Examination of Initial Prescription Characteristics and Pain
Etiologies J of Pain 18(11): 1374-1383. DOI: https://doi.org/10.1016/j.jpain.2017.06.010
(2017).

9. Volkow, N., McLellan, T. Opioid Abuse in Chronic Pain - Misconceptions and
Mitigation Strategies. N Engl J Med . 374:1253-1263 DOI: 10.1056/NEJMra1507771
(2016).

97

10. Jones, C.M., Heroin use and heroin use risk behaviors among nonmedical users of
prescription opioid pain relievers – United States, 2002-2004 and 2008-2010. Drug and
Alcohol Dependence, 132(1), 95-100. doi: 10.1016/j.drugalcdep.2013.01.007. (2013).

11. Ashley, E., Niebauer, J. Cardiology Explained. Remedica. (2004).

12. Olaison, P. Current best practices and guidelines indications for surgical intervention in
infective endocarditis. Infect Dis Clin North Am. 16(2):453-75. DOI:
https://doi.org/10.1016/S0891-5520(01)00006-X (2002).

13. Macrae W. Chronic postsurgical pain. Epidemiology of pain . IASP Press, 125-42. (1999)

98

4.7 Figures and Tables

Figure 4.1. Patient Inclusion Timeline

99

Figure 4.2. Study Consort Diagram

100

Table 4.1. Patient Characteristics

101

Table 4.2. Initial Prescription Characteristics

102

Table 4.3. Marginal Effects: Complete Results

103

Table 4.4 CDC Threshold Results, Marginal Effects

104

Supplementary Table 4.1. HCPCS/ICD Codes for Heart Valve Replacement Surgery

HCPCS
33361
Transcatheter aortic valve replacement (TAVR/TAVI) with prosthetic valve; percutaneous femoral
artery approach
33362
Transcatheter aortic valve replacement (TAVR/TAVI) with prosthetic valve; open femoral artery
approach
33363
Transcatheter aortic valve replacement (TAVR/TAVI) with prosthetic valve; open axillary artery
approach
33364
Transcatheter aortic valve replacement (TAVR/TAVI) with prosthetic valve; open iliac artery
approach
33365
Transcatheter aortic valve replacement (TAVR/TAVI) with prosthetic valve; trans- aortic approach
(e.g., median sternotomy, mediastinotomy)
33366
Transcatheter aortic valve replacement (TAVR/TAVI) with prosthetic valve; trans- apical exposure
(e.g., left thoracotomy)
33367
Transcatheter aortic valve replacement (TAVR/ TAVI) with prosthetic valve; cardio-pulmonary
bypass support with percutaneous peripheral arterial and venous cannulation (e.g., femoral vessels)
(list separately in addition to code for primary procedure)
33368
Transcatheter aortic valve replacement (TAVR/ TAVI) with prosthetic valve; car-diopulmonary
bypass support with open peripheral arterial and venous cannulation (e.g., femoral, iliac, axillary
vessels) (list separately in addition to code for primary procedure)
33369
Transcatheter aortic valve replacement (TAVR/TAVI) with prosthetic valve; cardiopulmonary
bypass support with central arterial and venous cannulation (e.g., aorta, right atrium, pulmonary
artery) (list separately in addition to code for primary procedure)
33405
Replacement, aortic valve, with cardiopulmonary bypass; with prosthetic valve other than homograft
or stentless valve
33406 Replacement, aortic valve, with cardiopulmonary bypass; with allograft valve (freehand)
33410 Replacement, aortic valve, with cardiopulmonary bypass; with stentless tissue valve
33411 Replacement, aortic valve; with aortic annulus enlargement, noncoronary sinus
33412 Replacement, aortic valve; with transventricular aortic annulus enlargement (Konno procedure)
33413
Replacement, aortic valve; by translocation of autologous pulmonary valve with allograft
replacement of pulmonary valve (Ross procedure)
33430 Replacement, mitral valve, with cardiopulmonary bypass
33465 Replacement, tricuspid valve, with cardiopulmonary bypass
105

33475 Replacement, pulmonary valve
ICD Procedure Codes
3520 Replacement of unspecified heart valve
3521 open/other rep aortic valve tissue
3522 open/other rep aortic valve
3523 Opn/oth rep mtrl vlv-tis
3524 Opn/oth rep mitral valve
3525 Opn/oth rep pulm vlv-tis
3526 Opn/oth repl pul valve
3527 Opn/oth rep tcspd vlv-ts
3528 Opn/oth repl tcspd valve
3505 Endovas repl aortc valve
3506 Trnsapcl rep aortc valve
3507 Endovas repl pulm valve
3508 Trnsapcl repl pulm valve
3509 Endovas repl uns hrt vlv

106

Supplementary Table 4.2. MME Conversion Factors and Equation

Image and factors taken from:
https://www.cms.gov/Medicare/Prescription-Drug-Coverage/PrescriptionDrugCovContra/Downl
oads/Opioid-Morphine-EQ-Conversion-Factors-Aug-2017.pdf

Eq. 1 otal MME Drug Strength uantity onversion F actor T =
*
Q
*
C
Eq. 2 ME per day (Drug Strength uantity onversion F actor)/Days Supply M =
*
Q
*
C

107

CHAPTER 5
CONCLUSION
The United States currently faces a nuanced public health issue, the growing opioid
epidemic. The growing addiction in patients and number of opioid-related deaths is a cause for
concern and needs to be addressed. This dissertation seeks to answer some vital questions
regarding opioid prescribing and provide critical evidence that may be pivotal to developing new
policies to combat the crisis.
The first study looked at opioid prescribing patterns two years leading up to
prescription-related death. While there is not much known about those decedents before their
death, the data visualizations from the first study of this dissertation provide valuable insights
into the longitudinal growth of patient-prescriber relationships, notably the sharp increase in new
connections between the cohort and doctors practicing Internal Medicine, Family Medicine, and
Psychiatry and Neurology in the final two months before death. These findings can particularly
useful in the development of prescription drug monitoring programs as well as education for
medical students, especially those seeking those board certifications. CURES currently faces a
lag between prescription fills and a data update on the website. Quick turnaround or a unified
accessible electronic health records database may help inform providers of a patient’s complete
prescription history to identify risk behaviors, particular patterns similar to our cohort of
decedents.
This dissertation also addresses to possible downstream effects of exposure to drugs. The
findings of the third study highlight the need for judicious prescribing, even in opioid-naïve
patients. While addressing patient pain following heart valve replacement surgery, minor
108

increases in opioid prescribing levels on a single prescription can have an effect on whether a
patient continues to use prescription opioids long-term. The results in this study can be important
evidence to providers when considering an initial opioid prescription.
The second study takes another perspective on opioid prescribing by employing behavioral
economic techniques to reduce MME levels. The research suggests that providing prescribers a
feedback loop to inform them of the possible consequences from past prescribing can influence
future prescribing. Due to the complex nature of the opioid epidemic, it is quite unlikely that one
method can address the entire issue and provide a complete solution. However, the findings of
this study is one piece of the puzzle toward addressing the crisis.
Future research on opioid prescribing and abatement can build on the findings reported
by this dissertation. As technology improves and more data becomes available surrounding
opioid prescribing and death, researchers, health care providers, will be able to employ it in
efforts to reduce opioid prescribing and consumption. This dissertation has contributed to the
understanding of the opioid epidemic and provides evidence that can be utilized to combat the
public health crisis.

109

Abstract (if available)

Abstract The urgency of the opioid epidemic is clear. As the growing number of opioid related overdose deaths continues to climb each year, the need for more information through effective research becomes more evident. This study seeks to fill certain gaps in the literature to better inform the medical community and public policy officials to make better decisions in the effort to curb opioid prescribing and opioid related overdose deaths. ❧ Chapter two of this dissertation takes a deeper look into what happens to a patient before they die from to an opioid related death. The paper explores CURES, California’s prescription drug monitoring programming to look at two years worth of prescription data for patients who have been labeled by the San Diego Medical Examiner as a prescription drug related death. It explores prescribing and consumption patterns leading up to a prescription related death. Data visualization techniques are also employed to visually assess patterns. ❧ Chapter three analyzes long-term opioid use in patients with heart valve replacement using an instrumental variables approach. We use the initial provider’s overall opioid prescribing patterns as an instrument for the initial dose of a prescription to a patient following heart valve replacement surgery to assess likelihood of long-term opioid use. Data from the Optum Clinformatics Data Mart between years 2007-2016 was assessed. Primary outcomes include whether a patient had prescription opioid use at 90+ days, 180+ days, and 270+ days, 90-180 days, and 180-270 days following heart valve replacement surgery. ❧ Chapter four investigates whether prescribers change their opioid prescribing levels after learning about a patient’s death. Prescribers who were involved in a prescription-related death in San Diego County between July 1, 2015 and June 30, 2016 were included in a randomized control trial where a letter was sent out informing providers of a patient’s death. Prescribing data before and after the letter was sent was analyzed to examine prescribing levels, prescribing volume, and number of new starts following the letter.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Characterization of health outcomes in patients with hemophilia A and B: Findings from psychometric and health economic analyses

PDF

Three essays on estimating the effects of government programs and policies on health care among disadvantaged population

PDF

The impact of healthcare interventions using electronic health records: an evaluation within an integrated healthcare system

PDF

Physician behavior and low-value care

PDF

Perioperative ketamine as a strategy to decrease opioid use in anesthesia care for opioid dependent patients with chronic pain: an extensive literature review with practice recommendations

PDF

Essays in opioid use and abuse

PDF

Perioperative ketamine as a strategy to decrease opioid use in anesthesia care for opioid dependent patients with chronic pain: an extensive literature review with practice recommendations

PDF

Addressing federal pain research priorities: drug policy, pain mechanisms, and integrative treatment

PDF

Economic, clinical, and behavioral outcomes from medical and pharmaceutical treatments

PDF

Three essays in health economics

PDF

The value of novel antihyperlipidemic treatments in the U.S. healthcare system: Reducing the burden of cardiovascular diseases and filling the gap of low adherence in statins

PDF

Documenting the impact of under-treatment for chronic disorders

PDF

Current state of perioperative medication management of methadone, buprenorphine, and naltrexone

PDF

Outcomes of antibiotic use among children with acute respiratory tract infections

PDF

Opioid withdrawal symptoms, opioid use, and injection risk behaviors among people who inject drugs (PWID)

PDF

Impact of pharmacy-based transitional care on healthcare utilization and costs

PDF

Sportiod and teen athletes

PDF

Current state of perioperative medication management of methadone, buprenorphine and naltrexone

PDF

Essays in health economics and provider behavior

PDF

A multi-system evaluation of medication for opioid use disorder for people who use opioids

Asset Metadata

Creator Nguyen, Andy (author)

Core Title Essays on opioid prescribing and abatement

School School of Pharmacy

Degree Doctor of Philosophy

Degree Program Health Economics

Publication Date 07/29/2019

Defense Date 06/08/2018

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag OAI-PMH Harvest,opioids,prescribingp pain

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Doctor, Jason (committee chair), Menchine, Michael (committee member), Myerson, Rebecca (committee member), Romley, John (committee member)

Creator Email andyn@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-40130

Unique identifier UC11668780

Identifier etd-NguyenAndy-6564.pdf (filename),usctheses-c89-40130 (legacy record id)

Legacy Identifier etd-NguyenAndy-6564.pdf

Dmrecord 40130

Document Type Dissertation

Format application/pdf (imt)

Rights Nguyen, Andy

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA