University of Southern California Dissertations and Theses

Impact of medication persistence on clinical and health care cost outcomes in patients with major depressive disorder using retrospective claims data

(USC Thesis Other)

Impact of medication persistence on clinical and health care cost outcomes in patients with major depressive disorder using retrospective claims data

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content i

IMPACT OF MEDICATION PERSISTENCE ON CLINICAL AND HEALTH CARE COST
OUTCOMES IN PATIENTS WITH MAJOR DEPRESSIVE DISORDER USING
RETROSPECTIVE CLAIMS DATA

by

Roxanna Seyedin

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(HEALTH ECONOMICS)

May 2022

Copyright 2022 Roxanna Seyedin

ii

Dedication

To those everywhere committed to the enhancement of human dignity.

iii

Acknowledgments

To Dr. Jeff McCombs, for his immense and persistent patience, compassion, and
guidance throughout this time.

iv

Table of Contents

Dedication ........................................................................................................................ii
Acknowledgments ........................................................................................................... iii
List of Tables ...................................................................................................................vi
Abstract .......................................................................................................................... vii
Chapter 1: Introduction .................................................................................................... 1
Background ................................................................................................................. 1
Clinical Background ..................................................................................................... 3
Drug Therapy for Major Depressive Disorder .............................................................. 7
Outcomes of Drug Therapy in Patients with Major Depressive Disorder ................... 14
Use of Claims Databases in Outcomes Research ..................................................... 15
Research Objectives .................................................................................................. 17
Chapter 2: Research Design and Methods ................................................................... 19
Data Source ............................................................................................................... 19
Study Population and Included Therapies (Objectives 1 to 3) ................................... 19
Data Structure ........................................................................................................... 20
Outcome Measures ................................................................................................... 21
Statistical Methods and Software............................................................................... 22
Chapter 3: Descriptive Statistics ................................................................................... 23
Demographic Characteristics ..................................................................................... 23
Baseline Characteristics by Episode.......................................................................... 24
Baseline Healthcare Costs by Episode ...................................................................... 28
Patient Outcomes by Episode ................................................................................... 28
Chapter 4: Analysis of Initial Treatment Episodes ......................................................... 31
Introduction ................................................................................................................ 31
Methods and Results: Factors that Influence Duration of Initial Therapy ................... 31
Methods and Results: Impact of Treatment Duration................................................. 33
Discussion ................................................................................................................. 36
Chapter 5: Analysis of Second Treatment Episodes ..................................................... 38
Introduction ................................................................................................................ 38
Methods and Results, Factors that Influence Duration of Second
Treatment Episodes ................................................................................................... 38
Methods and Results, Impact of Treatment Duration................................................. 40
Discussion ................................................................................................................. 42

v

Chapter 6: Analysis of Generic/Branded Antidepressant Therapy, Initial
Treatment Episodes ...................................................................................................... 44
Introduction ................................................................................................................ 44
Methods and Results, Other Factors that Influence Treatment Duration ................... 44
Discussion ................................................................................................................. 46
Chapter 7: Conclusions and Policy Implications ............................................................ 48
References .................................................................................................................... 50

vi
List of Tables
TABLE PAGE
1.1 ICD-9/10-CM Diagnostic Codes for MDD………………………………………………..4
1.2 Drug therapy for MDD……………………………………………………………………...8
3.1 Baseline Characteristics for Patients with MDD ..…………………………………….23
3.2 Baseline Characteristics for Patients with MDD by Episode Type…………………..25
3.3 Baseline Comorbidities for Patients with MDD by Episode Type…………………….26
3.4 Baseline (Other) Drug Therapy for Patients with MDD by Episode Type…………...27
3.5 Baseline Prior Use ($USD over 12 months) [SD] for Patients with MDD
by Episode Type……………………………………………………………………………….28
3.6 Drug Therapy Outcomes for MDD Patients by Episode Type………………………..29
3.7 Post-Treatment 1-Year Healthcare Costs ($USD) [SD] for Patients with
MDD by Episode Type………………………………………………………………………..30
4.1 OLS and Cox Model, Initial Episode Duration: N=9,090 observations……………...32
4.2 Cox Model, Duration on Time to Hospitalization/ED Visit: N=9,090
observations ..................................................................................................................34
4.3 OLS of MDD Drug Therapy Duration on Healthcare Costs: N=9,090
observations……………………………………………………………………………………35
4.4 GLM Regression of MDD Drug Therapy Duration on Healthcare Costs:
N=9,090 observations.……………………………………………………………...………...36
5.1 OLS and Cox Model, Second Episode Duration: N=7,744 observations….………..39
5.2 Cox Model, Duration on Time to Hospitalization/ED Visit: N=7,744
observations…………………………………………………………………………………....41
5.3 OLS of MDD Drug Therapy Duration on Healthcare Costs: N=7,744
observations…………………………………………………………………………………....42
5.4 GLM Regression of MDD Drug Therapy Duration on Healthcare Costs:
N=7,744 observations………………………………………………………...……………….42
6.1 OLS and Cox Model, Initial Episode Duration, Other Factors: N=9,090
observations..………………………………………………………………………………..…45
vii

Abstract

Depression is a common illness, affecting 280 million people worldwide.
i
It is associated with mood fluctuations or loss of interest in pleasurable activities of
everyday life. It is recurrent in nature, making it a long-term condition, which poses
clinical challenges that may lead to increased health resource utilization, and
furthermore, a burden to our economy. Furthermore, the majority of patients discontinue
treatment prior to the recommended treatment length to achieve symptom remission.

The overarching goal of this dissertation is to document treatment patterns of
antidepressant (AD) therapy and to examine the effect on clinical and economic
outcomes. The dissertation comprises of three objectives to explore these issues:
1) examining the factors associated with persistence with initial treatment and the
impact of persistence on hospitalization risks and costs, with special attention to the
therapeutic class of initial drug therapy, 2) determining the factors associated with
persistence with second treatment episodes and the impact on hospitalization risks and
costs, with a particular interest in the type of treatment episode and factors from a
patients’ initial treatment, and 3) to document the potential impact of generic/branded
therapy and examine medication- and disease-related factors associated with
persistence in initial therapy. These objectives will be addressed using de-identified
claims from Optum Clinformatics
®
DataMart Database.

____________________________

i
World Health Organization. Depression: Let’s Talk. 2021; http://www.who.int/news-room/fact-
sheets/detail/depression. Accessed February 24, 2022.
1

Chapter 1: Introduction

Background

Globally, major depressive disorder (MDD) continues to be a leading cause of
disability.
1
Nearly 10% of adults in the United States (US) experience at least one major
depressive episode, with individuals 18 to 25 years of age at highest risk.
2
Depression
is associated with high suicidality, and is reported to be the second leading cause of
death for pre-teens and young adults.
3
It is also correlated to pain and other chronic
health conditions such as cardiovascular disease, dementia, diabetes, among others.
4-6
Initiating treatment for MDD often includes antidepressant medications in
combination with psychotherapy. The adjunctive use of antidepressants may also be
needed to achieve wellness in patients with bipolar depression.
7
The American
Psychiatric Association (APA) guidelines recommend that patients receive a minimum
of 6-12 weeks of uninterrupted treatment for an incident depressive episode; this should
be followed by a phase ranging from 4-9 months where the medication is continued in
order to prevent relapse.
8
However, proper treatment with antidepressants is limited by
factors that affect the ability of the patient to remain persistent with therapy namely,
treatment-related adverse effects (nausea, etc.) and a delay in the onset of symptomatic
improvement to name a few.
9
Furthermore, remission rates for initial antidepressant
monotherapy is about one-third of all patients, as documented in the National Institute of
Mental Health’s Sequenced Treatment Alternatives to Relieve Depression (STAR*D)
trial.
10
Among the patients who achieved remission from this trial, roughly 90% had at
least one residual depressive symptom, such as anhedonia, fatigue, and sleep issues.
11

2

These clinical challenges create further concern when the total economic burden of
MDD is an estimated $211 billion per year and increasing, representing a 21.5%
increase from $174 billion per year in 2005.
12

These findings confirm that medication persistence is crucial for patients
suffering from MDD; avoiding gaps in treatment decreases the likelihood of relapse up
to 70%.
13
However, medication persistence over a sufficient period of treatment among
patients with MDD ranges from 25 to 50%.
14,15
This underscores the importance of
developing better data on the barriers to persistence in real-world practice for patients
with MDD. These data include information on the clinical outcomes achieved by patients
who change drug therapies over time and the effectiveness of alternative therapies in
these patients.
The goal of this dissertation is to investigate the factors that influence persistence
with MDD drug therapy and to document the impact of persistence with
antidepressants/antipsychotics on MDD patients across the patient’s initial treatment
attempt and each subsequent episode of drug therapy. We use paid claims data from a
large commercial insurer in the US. Specifically, persistence and the clinical and
economic outcomes of antidepressants/antipsychotics are evaluated based on the type
of treatment episode (initial, switching, restarting, augmentation) controlling for factors
that may affect the outcomes of treatment. The analysis of follow-up drug therapy
initiated by patients during the data period addresses gaps in the current literature,
which typically focus on initial treatment attempts and ignored the long-term treatment
patterns of patients with MDD.

3

Clinical Background

Definition of Disease

Clinical depression (also known as major-depressive disorder) is a mood
disorder typically characterized by multiple episodes in which symptoms affect how one
feels, thinks, and handles daily activities.
16
The pathophysiology of MDD is not clearly
understood. Current evidence suggests that the availability of neurotransmitter
receptors and cell signaling is the biological source of MDD. Disturbance of the central
nervous system serotonin (5-HT) activity and other neurotransmitters (dopamine,
norepinephrine, glutamate, etc.) is most widely noted.
17

There are also now compelling arguments for genetic factors making individuals
more predisposed to depression by an increased level of helplessness to stressful life
situations. A variation in the allele that encodes the serotonin transporter (5-HTT) has
been shown to put individuals with this allele variant at higher risk of developing
depression as a result of traumatic life events. The Tph1 and Tph2 genes have also
been shown to play a role in increasing suicidal risk by means of disrupting the
production of melatonin, which regulates circadian rhythm.
18

Classification of Major Depressive Disorder

According to the Diagnostic and Statistical Manual of Mental Disorders, Fifth
Edition (DSM-5), a patient must have at least five of the following depressive symptoms
present during the same two-week period to qualify for having a major depressive
episode: 1) feeling sad or dysphoric; 2) loss of interest in activities once enjoyed; 3)
changes in appetite leading to weight loss or gain unrelated to dieting; 4) disturbed
4

sleep; 5) increase in purposeless physical activity or slowed movements and speech;
6) lack of energy; 7) feeling worthless or having unnecessary guilty; 8) diminished ability
to focus or decreased decision-making; and 9) thoughts of death or suicide.
Additionally, at least one of the five symptoms must include dysphoria or anhedonia.
The criterion symptoms above (excluding weight change and suicidal thoughts) must be
present nearly every day causing functional impairment for the patient to be diagnosed
with MDD.
19

The current study will identify these patients based on International Classification
of Diseases, Ninth/Tenth Revision, Clinical Modification (ICD-9/10-CM) diagnostic
codes listed in Table 1.1 below. The severity and psychotic specifiers are only assigned
when a patient meets the full criteria for a major depressive episode.
19

Table 1.1 ICD-9/10-CM Diagnostic Codes for MDD
Severity/course specifier Single episode Recurrent episode
Mild 296.21 (F32.0) 296.31 (F33.0)
Moderate 296.22 (F32.1) 296.32 (F33.1)
Severe 296.23 (F32.2) 296.33 (F33.2)
With psychotic features 296.24 (F32.3) 296.34 (F33.3)
In partial remission 296.25 (F32.4) 296.35 (F33.41)
In full remission 296.26 (F32.5) 296.36 (F33.42)
Unspecified 296.20 (F32.6) 296.37 (F33.9)
Abbreviations: ICD-9/10-CM, International Classification of Diseases, Ninth/Tenth Revision,
Clinical Modification; MDD, major depressive disorder.

Other Common Types of Depression
Several types of depressive disorders can be effectively treated with
antidepressants, each with a defining characteristic, varying cause and clinical
5

presentation. Dysthymia, also known as persistent depressive disorder, is a type of
recurring low-level depression where the individual has depressed mood and two or
three more additional depressive symptoms that lasts at least one year in
children/adolescents and at least two years in adults.
Postpartum depression is characterized by a heterogeneous group of depressive
symptoms at the onset of pregnancy or after giving birth. Mothers may experience
persistent lethargy and sadness, with more severe cases leading to postpartum
psychosis. Seasonal affective disorder (SAD) is another form of depression that
surfaces annually during the short daylight of the winter months. It is characterized by
feelings of guiltiness, irritability, and sadness. Patients also tend to exhibit a significant
increase in appetite, which results in weight gain. SAD is MDD, with seasonal patterns.
Those with bipolar disorder, once known as manic-depressive disease, experience
depressive episodes coupled with periods of euphoria or irritability.
18

Epidemiology and Course of Major Depressive Disorder
In a national survey of US adults, the one-year prevalence rate of MDD was
10.4% and the lifetime prevalence rate was nearly doubled. The lifetime prevalence
rates were significantly lower among men than women (14.7% versus 26.1%). Seventy
percent (70%) of respondents who reported being diagnosed with MDD also reported
lifetime treatment for MDD; 13% of the same sample reported attempting suicide during
their worst episode. The average age of first MDD treatment for respondents with
lifetime prevalent MDD was 32 years old; the average length of time between the onset
of symptoms and initiating treatment was 47 months.
20

6

Single episodes of MDD are infrequent, and the recurrent course of this disorder
gives cause for why it consistently holds the highest burden of any disorder among
wealthy countries.
21
MDD patients will experience, on average, about five episodes over
their lifetime.
22
Several population studies report high recurrence rates in MDD. Rates
range from 40 to 75% in patients who had symptom improvement following an initial
depressive episode.
23-26

The recurring nature of MDD episodes has resulted in significant gaps in our
understanding of the effectiveness of antidepressant treatment over time and the costs
associated with recurring episodes. Achieving persistence and adherence are measures
used to assess the effectiveness of these therapies. Persistence is defined as the time
measured between treatment initiation and discontinuation.
27
It should not be confused
with adherence which is generally measured using the medication possession ratio
(MPR) > 80% (dividing the total days of medication supplied by 365 days) or the
proportion of days covered (PDC) > 80% (number of days with medication available
divided by 365 days).
28
Measures of adherence are more appropriately applied to
disease states in which the amount of drug therapy consumed is more relevant to
patient outcomes (e.g., statin therapy or treatments for hypertension). Measures of both
persistence and adherence have been estimated using claims data.
The current study creates a unit of observation for all episodes of antidepressant
drug therapy initiated by patients. A new episode of treatment is identified each time a
patient initiates therapy using a drug not previously used, including their initial treatment
episode, or reinitiates therapy after a break in therapy of at least 15 days. A relevant
7

study showed that four out of five patients with a 14-day gap in medication supply
(during the initial 90 days of treatment) were at risk for discontinuing therapy.
29

Drug Therapy for Major Depressive Disorder

The treatment of depression can be described in three stages: the acute phase
of six weeks to three months, followed by the continuation phase of four to nine months,
and lastly, the maintenance phase (one year post treatment initiation). Patients whose
depressive symptoms return during either the acute or continuation phase are defined
as relapsing within the same depressive episode. Depressive symptoms returning after
one year are characteristic of a new episode and defined as a recurrence. Validated
rating tools such as the Hamilton Depression Scale (HAM-D) or the Patient Health
Questionnaire-9 (PHQ-9) are commonly used to assess if remission has been achieved.
For example, remission on the HAM-D is a score less than 7; on the PHQ-9, an
individual has achieved remission if they score less than 5. It is also recommended for
patients to be persistent for four to nine months in order to prevent relapse.
8
A plethora of treatment options can be used for MDD. These options include
psychotherapy, namely cognitive-behavioral therapy (CBT), pharmacotherapy, and
other somatic therapies. However, the scope of this dissertation is limited to
pharmacotherapy.

Overview of Current Pharmacotherapy
For moderately severe and severe depression, the following second-generation
antidepressants (SGAs) are considered as first-line treatment: selective serotonin
8

reuptake inhibitors (SSRIs), serotonin and norepinephrine reuptake inhibitors (SNRIs),
mirtazapine (a tetracyclic antidepressant), and bupropion (an atypical
antidepressant).
30,31

First-generation antidepressants (tricyclic antidepressants (TCAs) and
monoamine oxidase inhibitors (MAOIs)) are not frequently prescribed, as SGAs having
fewer and less severe adverse effects and are of similar efficacy.
31
Toxicity in overdose
is also a major reason as to why MAOIs are not used in initial treatment. The most
commonly prescribed drug therapies are listed (by class) in Table 1.2.

Table 1.2 Drug Therapy for MDD
Therapeutic Class Drugs
SSRIs citalopram, escitalopram, fluoxetine, fluvoxamine,
paroxetine, sertraline,
SNRIs venlafaxine, duloxetine, desvenlafaxine, milnacipran,
levomilnacipran
TCAs amitriptyline, desipramine, clomipramine, nortriptyline,
imipramine, protriptyline, trimipramine, doxepin
MAOIs phenelzine, tranylcypromine
SMSs vilazodone, vortioxetine
Other ADs mirtazapine, bupropion
SARIs trazodone, nefazodone
AAPs olanzapine, aripiprazole, brexpiprazole, quetiapine,
risperidone
Augmentation strategies lithium, T3, buspirone
Abbreviations: MDD, major depressive disorder; SSRIs, selective serotonin reuptake inhibitors;
SNRIs, serotonin and norepinephrine reuptake inhibitors; TCAs, tricyclic antidepressants;
MAOIs, monoamine oxidase inhibitors; SMSs, serotonin modulators and stimulators; AD,
antidepressants; SARIs, serotonin antagonist and reuptake inhibitor; AAPs, atypical
antipsychotics; T3, triiodothyronine.

A provider’s treatment selection for a patient depends on past treatment
response, potential adverse side effects, medical history, age, concurrent medication
use, and cost.
32
One approach to help understand the complexity of drug treatment
9

patterns among MDD patients is to focus on episodes of treatment. This allows for
observations to be compared across various treatment groups on a per-episode basis.
The current study will define different types of episodes based on the duration of gaps in
therapy and overlap across related treatment regimens.

Initiating, Restarting, and Switching Pharmacotherapy
All patients in the analytic data file for this dissertation have an initial treatment
attempt, which is the first observed antidepressant treatment preceded by 12 months of
data. When patients reinitiate the same drug therapy after 15 days or more of a break
in therapy, the episode is defined as a restarting episode. A patient may also switch to a
different medication to optimize their treatment outcomes.
A meta-analysis focused on studies where patients were switching from an SSRI
found that 1) inadequate response to an SSRI does not necessarily translate to
inadequate response to the entire class of SSRIs; 2) switching from an SSRI to a TCA
may be a viable treatment strategy to nonresponse; and 3) switching across classes of
antidepressants with a different mechanism of action is also recommended; it should be
noted that one strategy is not superior over another.
33-35

The data set defines a switching episode when a patient changes medication
while still on therapy or within 15 days of terminating their previous therapy and must
discontinue the use of all previous medication within 60 days. For patients that change
regimens after discontinuing for 15 days or more, the episode will be identified as a
delayed switching episode. The advantages of switching therapies are that this strategy
results in fewer drug interactions and a higher likelihood of adherence.
36

10

Augmentation Pharmacotherapy
Current treatment guidelines recommend antidepressant monotherapy as first-
line treatment for patients preferring antidepressants to CBT for newly diagnosed
depression. However, evidence has shown that even if a patient experiences a partial
(yet significant) response to one agent, the likelihood of remission can increase with
augmentation (apart from combining two SSRIs, which may result in an increased the
risk of serotonin syndrome).
32
This study will define an augment episode when a patient
adds on a second medication while continuing their original medication(s) for more than
60 days. An augment episode should be distinguished from combination therapy, where
two medications are taken concurrently and initiated on the same day.
Augmenting with an atypical antipsychotic (AAP) has also been recommended as
a potential treatment option. In low doses, aripiprazole or brexpiprazole have proven
efficacy in patients taking SSRIs and are approved by the Food and Drug Administration
(FDA) for use in combination with antidepressants for MDD.
37
Olanzapine and
quetiapine extended-release are also approved for use as add-ons in treatment-
refractory patients.
8

The most common and efficacious augmentation agents are atypical
antipsychotics (AAPs), followed by lithium, triiodothyronine (T3), and vitamin D. In the
aforementioned STAR*D trial, patients who augmented citalopram with either
bupropion-SR or buspirone experienced a 30% remission rate.
10
Augmentation with
olanzapine, quetiapine, aripiprazole, and risperidone were found to be significantly more
effective than placebo.
36
Disadvantages of augmenting with AAPs include a substantial
risk of adverse events, including weight gain, akathisia (restlessness), and sedation.
11

Until the initiation of AAPs, lithium was the well-established medication for
antidepressant augmentation.
36
In a meta-analysis of augmentation studies, including
lithium, a three-fold difference in efficacy was found when lithium was compared to
other augmenting agents.
38
The use of thyroid hormones as supplementation to
antidepressants is based on evidence showing a link between MDD and thyroid
function. Of these, T3 is one of the most widely studied augmentation medications that
has proven response in patients without thyroid hormone abnormalities who have also
not experienced favorable outcomes with a TCA or an SSRI. However, a majority of the
T3 augmentation studies do not include a control group, have inadequate follow-up and
small sample sizes, or lack proper monitoring for hyperthyroidism or thyrotoxicosis.

Combination Pharmacotherapy
Combination therapy or ‘polypharmacy’ is defined here as the initiation of
treatment using two antidepressants with a different mechanism of action. Various types
of combinations have been reported; for example, starting bupropion to sertraline when
initiating treatment, or adding mirtazapine to fluoxetine as initial therapy.
32
In a three
year study where patients were given a combination of mirtazapine and venlafaxine,
after six months of treatment, 75% of those still receiving the same treatment, and 56%
of the original group of patients showed a significant response.
39

In another trial, the same combination was prescribed to patients who did not
achieve symptom relief with their initial treatment attempt. Over an eight-week period,
82% of patients exhibited response, yet only half of the patients experienced good
tolerability.
40
The disadvantages of combination therapy are that in this approach, the
12

patient is likely to experience adverse effects, there is reduced persistence, and an
increased likelihood for drug interactions to occur.
41

Limitations of Current Pharmacotherapy
The current array of treatment options for depression are not optimal. Despite the
introduction of several novel therapy classes, only three out of four patients with
depression respond to treatment with antidepressants (50% reduction in HAM-D or
other rating tool). Those who do not respond are either not adherent or are resistant to
available treatment options.
41
The latter group of patients falls into a subtype of
depression that has yet to be uniformly defined in the literature, namely treatment-
resistant depression. To date, no single class of antidepressants has distinguished itself
as superior to others; there remains a gap in evidence by the lack of data from direct
head-to-head comparisons across these alternative classes of these medications.
The current study addresses this issue through the analysis of retrospective claims
data, which provides insight into the treatment outcomes achieved under multiple
treatment situations in a real-world setting, such as switching or augmenting therapies.
The measurement of clinical improvement in MDD symptoms is not feasible
using paid claims data. There are no clinical markers of the disease, such as clinical
laboratory tests that track treatment efficacy in most disease states. Patient-reported
outcomes play a significant role in measuring clinical response in MDD, and these data
are not recorded on paid claims. For patients who do in fact respond to treatment, it
takes about six to twelve weeks before their prescribed antidepressant will provide full
symptom remission; the goal of treatment is symptom remission, otherwise relapse is
13

more likely.
41
Results from four clinical trials, all comparing the same two
antidepressants for MDD showed that 80% of patients who experienced an onset of
response after two weeks continued to respond after eight. Of the patients who did not
experience an onset of improvement at week two, only 43% reported symptom
resolution at week eight.
33
These data give rise to the notion that, in order to have
advantages over existing antidepressant treatment, novel agents should have a more
rapid onset of action.

Long-term Treatment with Antidepressants
Due to the cyclical nature of MDD, many patients will experience a recurrence.
Improved persistence during the continuation and maintenance phases of treatment,
even after experiencing symptom remission, may reduce the risk of a reoccurrence.
The Canadian Network for Mood and Anxiety Treatments (CANMAT) Clinical Guidelines
recommend patients to maintain antidepressant use for a minimum of six to nine
months; for patients with specific risk factors (recurrent episodes or significant
comorbidities), therapy is suggested to be maintained for at least two years.
42
Early
medication discontinuation is very common in treating depressive disorders. Studies
report that roughly 25% to 42% of MDD patients stop treatment after or before one
month, and 72% discontinue by three months.
43
The importance of treatment
persistence is underscored by the fact that it is strongly related to negative treatment
consequences. Early discontinuation of therapy increases the risk of suicide, and early
discontinuation often exacerbates the chronic nature of the disease.
44
14

Outcomes of Drug Therapy in Patients with Major Depressive Disorder

Risk of Suicide and Emergency Department Visits
MDD is also the most common psychiatric diagnosis associated with suicide,
attributing to 50% of all suicides.
12
The American Association of Suicidology (AAS)
reported that being diagnosed with MDD increases the risk of suicide 20-fold as
compared to undiagnosed patients or other not suffering from depression.
45
Nearly 60%
to 70% of patients who suffer from acute depression experience suicidal ideas; the
incidence of suicide among MDD patients ranges from 10% to 15%.
46
In 2010, suicide-
related costs contributed 5% of the total growing economic burden posed by MDD,
specifically due to the increased use of emergency department (ED) visits for suicide-
related injuries.
12,47

The literature suggests that pharmacological treatment with antidepressants
and/or mood stabilizers may be a successful approach to prevent suicidal behavior.
A meta-analysis comparing suicide completion rates in individuals with bipolar disorder
showed lithium therapy resulted in a five-fold reduction in suicide attempts and
completed suicides.
48

Hospitalization
Apart from treatment utilization, the economic burden of MDD also comes in the
form of costs associated with inpatient hospitalization. It is well-documented that
depression is associated with longer hospital stays, particularly for older adults with late-
life depression.
49
Approximately one in ten patients with MDD are hospitalized at some
point in their life.
50
In a retrospective analysis using data from the Premier Perspective
15

Hospital Database, 5.2% of patients with MDD that were hospitalized had a least one
MDD-related readmission within 30 days of discharge. Results of the study found that
for an average MDD hospitalization of six days, the average hospital charge per stay
was $6,713.
50
Among the reasons for MDD hospitalization, suicide attempt and ideation
are largely reported as the key factors.
51
Treatment persistence may also be a probable
explanation for a decrease in hospitalizations.
MDD constitutes a substantial economic burden on society resulting from its high
prevalence, poor treatment outcomes, and subsequent high utilization of healthcare
resources.
52
A study found that 33% to 69% of medication-related admissions were due
to poor medication persistence, resulting in an annual cost of $100 billion.
53

Unsurprisingly, another meta-analysis highlighted that patients who were non-adherent
had a higher risk of relapse and recurrence, which translated to higher hospitalization
rates.
52
Conversely, adherent patients may use more outpatient services, such as
physician visits and visits to mental health providers.
Understanding the nature of MDD treatment patterns may help identify the
unmet needs of this vulnerable patient population, with the goal of improving treatment
options and clinical outcomes in addition to identifying areas for potential cost savings to
our healthcare system.

Use of Claims Databases in Outcomes Research
The analysis of retrospective claims data has been widely used in outcomes
research. The use of a large-scale database allows for greater generalizability as
compared to research that focuses solely on randomized clinical trials with relatively
16

small sample sizes, short time periods over which outcomes are measured, and their
inherent inability to measure treatment compliance, which is stressed during the trial.
Insurance claims data are well-suited for the analysis of cost and utilization, as they
include all direct medical costs incurred by an insurer or employer. However, due to the
fact that claims data is not inherently designed for outcomes research, it is necessary to
recognize the issues which will be tied to a claim database analysis to reduce potential
bias that may occur.
Treatment-Selection Bias

When using retrospective claims data to model the effect of alternative
treatments on outcomes, factors influencing treatment selection must be addressed. In
the case of observational data, it is not possible to randomly assign patients to
treatments of interest, and patients often self-select into treatment. The major factors
that contribute to treatment selection and patient outcomes may be unobservable;
determining the treatment effect utilizing claims data creates a potential bias that should
be resolved by applying multivariate statistical methods that adjust and include the
impact of observable factors on treatment selection and outcomes.
When discerning the treatment effect of various therapies for MDD, the treatment
outcomes may be a function of various factors (i.e., comorbid conditions, illness
severity, prior treatment attempts) that also relate to and render the selection of a
certain treatment. In this case, it is erroneous to use a simple linear model to estimate
the treatment effect, as the estimated results will contain heterogeneity bias. This type
of bias occurs when one does not account for unobserved, systematic differences
between the individual units. These unobserved, individual-specific effects (ci) that are
17

fixed over time influence outcomes and must be controlled for. Explicitly, β cannot be
consistently estimated if the unobserved heterogeneity, ci is correlated with other
explanatory variables, xit : Cov (xj,c) ≠ 0.
yit = xit β + ci + uit
When unobserved heterogeneity exists, the linear estimate of β will be inconsistent
because one of the conditions for the ordinary least squares (OLS) estimate to be an
unbiased and efficient estimate are not met: Cov (xj,uj) = 0 is violated (xj is correlated
with the idiosyncratic error term), as the composite error term now includes c

and uit..
The first defense to avoid treatment selection bias is to utilize all the available
data on treatment history to document factors that may influence treatment selection
and outcomes. Much of the existing research often ignores episodes of treatment after
the first observed episode; sometimes, this may be because the data set is cross-
sectional in nature. Cross-sectional data offers a snapshot of individuals at the same
point in time. Longitudinal data allows for the analyses of data on individuals over a
period to better measure the baseline characteristics of the individual each time they
initiate treatment, thus effectively reducing the impact of observed heterogeneity. When
analyzing outcomes in drug therapy, the use of cross-sectional data is particularly
concerning because it fails to account for information that is provided in later episodes,
which may significantly impact treatment outcomes.

Research Objectives
The first objective of this research is to document the factors that determine the
duration of therapy for patients’ initial treatment episodes, including the initial drug
18

therapy that is selected. Next, we examine the impact of initial treatment duration
(namely, time to all-cause discontinuation (TTAD), on patient outcomes such as time to
hospitalization and healthcare costs. These analyses are repeated for the second
treatment episode, while considering how the duration of therapy achieved during the
initial treatment episode and type of second episode impact treatment duration and
patient outcomes achieved in the second episode. Lastly, we compare the impact of
generic versus branded antidepressant on persistence, examining what disease- and
medication-related factors are associated with persistence in therapy on initial therapy.

19

Chapter 2: Research Design and Methods
Data Source
This research uses retrospective paid claims data from the 1% Optum
Clinformatics
®
DataMart, a large US health insurance provider that insures patients
covered under both commercial and Medicare Advantage plans. The data includes
information on all paid claims for services delivered between January 2007 and
December 2019. The database includes information on both medical (inpatient,
outpatient, ambulatory, ED, and prescription drug claims, as well as demographics,
enrollment eligibility, and mortality. These data are utilized to derive comprehensive
diagnostic and drug utilization profiles that reflect the patient’s health status over time.

Study Population and Included Therapies (Objectives 1 to 3)
The data is used to identify MDD patients based on two criteria: 1) at least one
claim with a diagnosis of MDD (ICD-9-CM codes 296.2x, 296.3x or ICD-10-CM codes
F32.x, F33.x) according to the DSM-5 criteria
19
and 2) at least one filled prescription for
an antidepressant (AD) or an antipsychotic (AP). The date of any first AD/AP medication
fill was identified as the index date of the patient’s initial treatment attempt. Once the
patient’s index date was identified, all of the subsequent episodes were identified and
summarized for analysis. Once created, each initial treatment episode was then
required to have 1) no claims with an AD/AP in the prior 12 months and 2) to be
continuously eligible for 12 months prior and 2 years following their index date. The
requirement of this washout period provides the necessary pre- and post-treatment
information for analyzing treatment choice and outcomes for each episode type.
20

Patients were excluded if any of the following diagnoses were present during the
pre-index period: substance-induced mood disorder, bipolar disorder (296.4x, 296.5x or
F31.x), cyclothymic disorder (301.13 or F34.0), dementia (290.xx or F03.90),
schizophrenia (295.xx or F20.x), intellectual disability (317.x-319.x or F70-F73, F79),
autism spectrum disorder (299.00 or F84.0), Parkinson disease (332.x or G20, G21),
senility without mention of psychosis (797.xx or R41.81), and other cerebral
degenerations (331.xx or G31.xx). Patients with these disorders may use the drugs
from the therapeutic classes under study and are excluded to control for potential
disease-related confounding.

FDA approved drug therapies for MDD are included in the analysis and further
categorized by therapeutic class (SSRIs, SNRIs, serotonin modulators and stimulators
(SMSs), TCAs, MAOIs, AAPs, and other antidepressants). The therapeutic class of the
patient’s initial therapy is included as a dummy variable to account for differential
treatment effects by drug class.

Data Structure

A 12-year longitudinal dataset was created and includes all episodes of MDD
drug therapy initiated by patients during the data period. The unit of analysis is an
episode of treatment, which is indicated by the patient’s initial treatment attempt or a
change in the patient’s initial drug therapy (whether a patient changed medications or
was no longer on active therapy). Five types of episodes were defined:
1. First Observed Episode: the ‘first’ episode recorded in the available dataset
2. Restart Episodes: if therapy is reinitiated with the same medication used in
their prior episode after a break in therapy of at least 15 days
21

3. Switching Episodes: a) if a patient changed medications within 15 days of
terminating their previous therapy, and b) if a patient does not terminate their
previous therapy and switches to a new drug therapy; they must discontinue
the use of all previous medications within 60 days

4. Delayed Switching Episodes: if a patient changed medications after a break in
therapy in excess of 15 days

5. Augmentation Episodes: if a patient adds on an additional medication while
continuing their original medication(s) for 60 days or more

The three objectives will focus on the first (and second) observed episodes, restart,
switching, delayed switching, and augmentation episodes.

Outcome Measures

The sequence of analysis that is shared across all objectives include 1) an
evaluation of what impacts persistence with MDD therapy and 2) how persistence
effects patient outcomes. The analyses incorporate two types of patient outcome
measures: 1) clinical events (drug therapy utilization (measured by TTAD) and
healthcare resource utilization and 2) direct healthcare costs. These outcomes are
shared across all three objectives, and hence, utilize the same statistical methods.

Drug Therapy Utilization Patterns
The duration of MDD medication therapy is evaluated. The days of uninterrupted
drug therapy after treatment initiation is calculated. Uninterrupted therapy requires the
patient to not have a gap greater than 15 days between the end of the days-of-supply
and the subsequent prescription refill. If a patient has an oversupply due to an early refill
of their previous fill, the days of oversupply are accounted for and added to the total
duration. Additionally, the likelihood of switching, delayed switching, or augmenting with
22

other MDD therapies after therapy initiation is assessed and compared by therapeutic
class.

Healthcare Utilization and Cost Outcomes
The impact of MDD drug therapy on health care utilization is evaluated by the
time to a clinical event, namely, inpatient psychiatric-related hospitalizations, non-
psychiatric-related hospitalizations, and ED visits. The economic outcomes are
measured as average total direct healthcare costs, analyzed 1 year following treatment
initiation. Net costs include prescription drug, outpatient, and inpatient expenditures.

Statistical Methods and Software
Descriptive statistics comparing MDD patients’ baseline characteristics and
outcomes by (initial, restart, switch, augment) episodes of treatment will use a t-
test/Pearson’s chi-squared test to evaluate group differences. Multiple logistic
regression analyses will be carried out to assess the likelihood of medication
persistence among these patients. The association between MDD drug treatment
history and time-to-discontinuation will be estimated using Cox proportional hazards
regressions. The impact of treatment duration on healthcare costs will be estimated
using generalized linear model (GLM) procedures. STATA version 14 SE (STATA Corp.
College Station, TX) was utilized for processing the data and the estimation of the
econometric models.

23

Chapter 3: Descriptive Statistics

Demographic Characteristics

A total of 52,228 treatment episode observations, representing 9,090 patients
with MDD, met the inclusion criteria for the study. Each patient was identified as having
at least one claim of an AD or AP prescription between January 2007 and December
2019. Table 3.1 provides a summary of demographic characteristics of the study
sample. Descriptive statistics and univariate tests of baseline characteristics by episode
type are summarized in Tables 3.2-3.7.

Table 3.1 Baseline Characteristics for Patients with MDD
Abbreviations: MDD, major depressive disorder; k, thousand; POS, point-of-service;
HMO, health maintenance organization; PPO, preferred provider organization; EPO, exclusive
provider organization.

Race % Education %
white 75.1 unknown 0.3
unknown 2.1 less high school 0.4
hispanic 11.5 high diploma 25.1
black 9.0 less college 55.4
asian 2.2 college or more 18.8
Income Region
unknown 14.9 south atlantic 25.3
< 40k 21.0 east northcentral 13.1
40k - 49k 7.1 west southcentral 16.2
50k-59k 7.3 mountain 10.6
60k-74k 8.9 west northcentral 10.8
75k-99k 13.0 pacific 13.0
100k+ 27.9 mid atlantic 4.5
Policy east south central 3.3
pos 48.7 new england 3.2
indemnity & other 21.9 unknown -
hmo 11.6 Insurance
ppo 14.6 commercial 71.0
epo 3.2 medicare 29.0
24

Of the 9,090 initial patient episodes, 65.7% were female, and the mean age was
53 years old. The ethnic makeup of the study population includes 75.1% White, 11.5%
Hispanic, 9.0% Black, 2.2% Asian, and 2.1% from other races. The majority of the study
population resides in the South Atlantic region (25.3%), followed by the West
Southcentral (16.2%) region, and East Northcentral (13.1%) region, which corresponds
with the geographic regions served by Optum.

Baseline Characteristics by Episode

The data set contains 9,090 initial, 26,046 restart, 6,170 switch, 9,262 delayed
switch, and 2,033 augment episodes, totaling to 52,601 observations. Episodes in which
two drug therapies were filled on the exact same index date were initially counted as
two separate observations of polytherapy (95, 186, and 92 of which were seen in the
restart, switch, and delayed switch episodes, respectively). After deducting the
polytherapy observations from the 52,601 total observations, a sample of 52,228
observations remain for analysis. The majority of initial episodes were monotherapy
episodes initiated with an SSRI (64.5%). Augment episodes were further defined by
whether the drug therapy was the second medication added within the episode or the
initial medication used (or being augmented).
There are significant differences in the characteristics of patients by episode
type. Patients initiating restart, switch, and delayed switch episodes tend to be younger
than patients augmenting therapy. Not surprisingly, patients switching, or augmenting
therapy tend to use other antidepressants more frequently than in other episode types.
Significant differences in use (for every therapeutic class) for switch, delayed switch,
25

and augment episodes as compared to initial episodes exist; chi-square tests were used
to evaluate these group differences. SSRIs were the most commonly prescribed
treatment across all episode types.

Table 3.2 Baseline Characteristics for Patients with MDD by Episode Type

Table 3.3 displays a list of covariates that were included in all multivariate
analyses to control for treatment selection bias. A higher proportion of comorbid
diagnoses (in the previous year) is seen among patients who augmented therapy.
Initial
n=9,090
Restart
n=26,046
Switch
n=6,170
Delayed
Switch
n=9,262
Augment
n=2,033
Mean age, yrs
(SD)
52.8 (21.5) 53.9 (20.7) 50.2 (21.7) 52.0 (21.3) 57.6 (22.2)
Age category, yrs (n, %)
< 18 354 (3.9) 663 (2.5) 227 (3.7) 240 (2.6) 67 (3.3)
18-34 1,781 (19.6) 4,561 (17.5) 1,575 (25.5) 2,019 (21.8) 315 (15.5)
35-44 1,287 (14.2) 3,636 (14.0) 826 (13.4) 1,337 (14.4) 206 (10.1)
45-54 1,348 (14.8) 4,311 (16.6) 874 (14.2) 1,451 (15.7) 277 (13.6)
55-64 1,267 (13.9) 4,290 (16.5) 894 (14.5) 1,365 (14.7) 281 (13.8)
> 65 3,053 (33.6) 8,585 (33.0) 1,774 (28.8) 2,850 (30.8) 887 (43.6)
Commercial (%) 71.0 75.3 75.1 74.6 66.5
Female (%) 67.9 68.3 66.0 68.2 68.4
Initial Drug Therapeutic Class (n, %)
SSRI 5,865 (64.5) 15,673 (60.2) 2,209 (35.8)* 4,086(44.1)** 551 (27.1)*
SNRI 812 (8.9) 2,598 (10.0) 669 (10.8)* 1,225(13.2)** 172(8.5)*
Other AD 1,025(11.3) 3,519 (13.5) 1,156 (18.7)* 1,550(16.7)** 425 (20.9)*
SMS 548 (6.0) 1,869 (7.2) 890 (14.4)* 990 (10.7)** 368 (18.1)*
AAP 302 (3.3) 1,280 (4.9) 960 (15.6)* 876 (9.5)** 387 (19.0)*
TCA 533 (5.9) 1,101 (4.2) 283 (4.6)* 532 (5.7)** 129 (6.3)*
MAOI 5 (0.04) 6 (0.02) 3 (0.05)* 3 (0.03)** 1 (0.05)*
Polytherapy (n, %) 95 (0.04) 286 (4.6) 92 (1.0) -
*p < 0.0001 and **p < 0.005 for significant differences as compared to initial treatment episode;
Abbreviations: MDD, major depressive disorder; yrs, years; SD, standard deviation; SSRI,
selective serotonin reuptake inhibitor; SNRI, serotonin and norepinephrine reuptake inhibitor;
AD, antidepressant; SMS, serotonin modulator and stimulator; AAP, atypical antipsychotic;
TCA, tricyclic antidepressant; MAOI, monoamine oxidase inhibitor.
26

Table 3.4 includes additional drug treatment history covariates that were also included
in all multivariate analyses to control for treatment selection bias. The highest use of
pain medications (opiates, amphetamines, sedatives, and hypnotics) is observed for
patients who augment therapy.

Table 3.3 Baseline Comorbidities for Patients with MDD by Episode Type
Initial
n=9,090
Restart
n=26,046
Switch
n=6,170
Delayed
Switch
n=9,262
Augment
n=2,033
Prior Dx (prior 12 months) (%)
infections 14.7 11.6 13.6 14.0 15.1
neoplasms 20.6 19.9 18.2 20.1 21.8
endocrine 53.0 52.1 49.4 51.1 58.1
diabetes 12.2 13.0 11.3 11.9 13.7
chronic non-cancer pain 43.0 43.5 45.0 46.9 49.4
hyperlipidemia 34.0 34.6 31.1 32.9 39.0
overweight/obese 10.2 9.3 8.8 9.1 8.5
blood 14.8 13.0 12.9 13.4 16.5
mental/behavioral
disorders
74.0 70.8 75.6 71.6 73.5
neurocog disorders 4.3 3.0 4.5 3.7 8.0
schizophrenia 0.4 0.5 1.2 0.7 1.4
anxiety & ocd 31.6 30.1 35.9 32.7 32.4
bipolar disorder 6.0 4.8 8.8 6.3 7.5
smoke 9.1 9.4 10.4 10.2 10.0
alcohol disorder 1.5 1.2 1.9 1.5 1.9
drug disorder 1.8 1.3 2.6 2.1 2.2
nervous system 50.2 47.0 49.3 50.3 56.2
circulatory 46.8 44.9 42.9 44.2 54.3
hypertension 38.6 37.0 34.8 36.0 44.9
chf 3.4 3.4 2.7 2.9 4.1
stroke 3.1 3.1 3.4 2.9 5.1
ami 0.9 0.8 0.6 0.6 1.1
ihd 4.2 2.7 2.5 2.2 4.8
pvd 4.3 3.3 3.7 3.7 4.7
respiratory 48.5 47.6 47.3 50.3 46.8
digestive 33.4 31.1 31.8 34.0 35.2
27

Abbreviations: MDD, major depressive disorder; Dx, diagnosis; neurocog, neurocognitive;
OCD, obsessive-compulsive disorder; CHF, congestive heart failure; AMI, acute myocardial
infarction; IHD, ischemic heart disease; PVD, peripheral vascular disease; CKD, chronic kidney
disease.

Table 3.4 Baseline (Other) Drug Therapy for Patients with MDD by Episode Type
*Pain includes opiates, amphetamines, sedatives, hypnotics; Abbreviations: MDD, major
depressive disorder; Rx, drug.

Initial
n=9,090
Restart
n=26,046
Switch
n=6,170
Delayed
Switch
n=9,262
Augment
n=2,033
genitourinary 41.3 40.5 41.1 42.4 42.1
ckd 5.6 4.7 4.2 4.3 6.7
complication of pregnancy 2.8 2.5 2.1 2.9 2.1
skin & subcutaneous 30.5 30.5 29.9 29.8 31.2
musculoskeletal system 56.1 55.6 53.5 56.6 60.1
congenital anomalies 3.7 3.9 4.8 4.1 4.2
perinatal 0.2 0.2 0.4 0.3 0.3
symptoms of ill-defined
conditions
75.6 74.9 74.7 77.2 78.9
injury 28.3 27.3 29.4 29.3 32.8
Initial
n=9,090
Restart
n=26,046
Switch
n=6,170
Delayed
Switch
n=9,262
Augment
n=2,033
Other Rx (prior 12 months) (%)
antihistamines 7.0 8.1 9.5 9.6 8.2
antiviral 57.5 55.3 53.3 57.9 50.6
anti-neoplastics 2.1 1.8 1.2 1.9 2.6
autonomic 28.1 27.3 27.5 29.6 30.6
anti-thrombotic 7.4 5.9 5.5 5.7 8.5
anti-hypertensive 43.9 42.2 38.9 40.4 46.2
anti-inflammatory 31.8 30.1 28.3 32.2 29.3
gastrointestinal 26.6 23.7 24.4 26.5 29.4
anti-diabetic 44.0 43.2 42.6 44.2 43.9
anti-infection 23.2 22.2 21.9 23.7 23.0
muscle relaxants 2.7 3.0 2.8 3.1 3.1
pain* 62.1 62.1 63.0 66.1 68.1
therapeutic agents 8.2 8.1 6.3 7.7 8.9
28

Baseline Healthcare Costs by Episode

Table 3.5 presents the baseline healthcare expenditures (total and by type of
service). In the baseline period, patients with restart episodes had lower total health
expenditures than patients initiating a switch, delayed switch, or augment episode in the
12 months prior to episode initiation. Augment patients nearly double their inpatient,
outpatient, and prescription expenditures as compared to those patients who restart
therapy.

Table 3.5 Baseline Prior Use ($USD over 12 months) [SD] for Patients with MDD by
Episode Type
*Summation of Rx, inpatient, and outpatient expenditures; Abbreviations: SD, standard
deviation, MDD, major depressive disorder; Rx, drug.

Patient Outcomes by Episode

Table 3.6 provides the unadjusted descriptive statistics for patient drug therapy
outcomes by episode type. Persistence with drug therapy differs significantly by episode
type, notably, restart and augment episodes as compared to initial treatment episodes.
Patients restarting therapy had a significantly lower duration of therapy as compared to
all other episode types (p<0.001), whereas those who were on an augmenting or
augmented drug significantly increased their duration of therapy (p<0.001).
Initial
n=9,090
Restart
n=26,046
Switch
n=6,170
Delayed
Switch
n=9,262
Augment
n=2,033
Rx expenditures
1,829
[5,375]
2,838
[7,339]
3,063
[7,065]
2,923
[7,096]
4,265
[7,864]
MDD-related Rx
expenditures
-
365
[1,067]
691
[1,633]
448
[1,257]
823
[1,795]
Outpatient
expenditures
2,171
[6,360]
2,697
[7,099]
3,770
[8,289]
3,628
[9,471]
4,391
[9,538]
Inpatient
expenditures
3,946
[15,370]
4,424
[15,948]
7,983
[20,630]
6,335
[19,213]
9,816
[23,544]
Total health
expenditures*
7,947
[19,031]
9,961
[21,364]
14,818
[25,871]
12,886
[25,071]
18,472
[28,631]
29

Patients who had an augment episode increased their persistence 1.5-fold as compared
to patients who had a restart episode (121 days vs. 82 days). In other words, patients
who are persistent enough to add on an additional drug therapy are persistent. Notably,
the only therapeutic class to increase persistence following a patients’ initial treatment
attempt (for every episode type) were SNRIs.

Table 3.6 Drug Therapy Outcomes for MDD Patients by Episode Type

Patients restarting their previous therapy or having a delayed switch exhibited the
lowest unadjusted average cost in the one-year post-treatment period (Table 3.7).
A shorter duration of therapy for patients who switch did not however result in lower
1-year post-treatment costs. Patients restarting or switch/delayed switching therapy are
consistently less costly to treat than patients augmenting therapy.
Initial
n=9,090
Restart
n=26,046
Switch
n=6,170
Delayed
Switch
n=9,262
Augment
n=2,033
Augmented
n=2,033
Duration of therapy (days), Mean [SD]
All related
therapy
93 [133] 82 [115]* 86 [121] 87 [122] 121 [143]* 297 [183]*
SSRI 97 [136] 81 [114]* 90 [124]* 90 [123]* 126 [144]* 294 [183]**
SNRI 90 [126] 92 [127]* 110 [143]* 98 [139]* 145 [164]* 322 [180]*
Other AD 95 [143] 81 [113] 84 [119] 88 [119] 131 [145] 300 [199]*
SMS 74 [119] 78 [108] 72 [107]** 72 [102]** 103 [131]* 283 [184]*
AAP 96 [130] 87 [128]* 81 [119] 89 [127] 118 [145] 290 [184]*
TCA 68 [109] 74 [106] 75 [104]* 66 [89]* 90 [122]* 300 [187]*
MAOI 112 [152] 200 [154] 53 [20] 233 [238] - 299 [171]
*p <0.001 and **p<0.01 for significant differences compared to all other episode types;
Abbreviations: MDD, major depressive disorder; SD, standard deviation; SSRI, selective
serotonin reuptake inhibitor; SNRI, serotonin and norepinephrine reuptake inhibitor; AD,
antidepressant; SMS, serotonin modulator and stimulator; AAP, atypical antipsychotic; TCA,
tricyclic antidepressant; MAOI, monoamine oxidase inhibitor.
30

Outcome data presented in this chapter reflect unadjusted results. Chapter 4 will
use multivariable statistical methods to evaluate the factors impacting duration and the
impact of duration on patient outcomes.

Table 3.7 Post-Treatment 1-Year Healthcare Costs ($USD) [SD] for Patients with MDD
by Episode Type
*Summation of Rx, inpatient, and outpatient expenditures; Abbreviations: SD, standard
deviation, MDD, major depressive disorder; Rx, drug.

Initial
n=9,090
Restart
n=26,046
Switch
n=6,170
Delayed
Switch
n=9,262
Augment
n=2,033
Rx expenditures
2,473
[5,948]
2,795
[7,700]
3,279
[7,321]
3,058
[7,336]
4,877
[9,242]
MDD-related Rx
expenditures
294
[860]
355
[1,083]
831
[1,925]
541
[1,498]
1,174
[2,241]
Outpatient
expenditures
2,840
[8,275]
2,599
[7,584]
3,908
[10,198]
3,489
[9,101]
4,638
[10,510]
Inpatient
expenditures
3,553
[14,183]
3,666
[14,467]
5,491
[17,499]
4,943
[16,531]
6,336
[18,079]
Total health
expenditures*
8,866
[19,649]
9,061
[20,526]
12,679
[24,398]
11,490
[23,035]
15,852
[25,760]
31

Chapter 4: Analysis of Initial Treatment Episodes
Introduction

The analysis of initial episodes has two basic objectives. First, we seek to
examine the factors which may impact treatment duration with special interest being
applied to the therapeutic class of the medication use to treatment. Next, we assess the
effect of treatment duration on patient outcomes, specifically, 1) the time to a clinical
event (inpatient psychiatric-related/non-psychiatric-related hospitalizations and ED
visits) and 2) healthcare expenditures.

Methods and Results: Factors that Influence Duration of Initial Therapy
The effects of the patient’s baseline characteristics on duration of initial therapy
episode were evaluated first using an OLS regression model. The primary factors of
interest were the therapeutic class of the initial drug, prior hospitalization, prior
healthcare expenditures, and other patient history variables. The results of the OLS
model were then directionally verified with a Cox proportional hazard model, which
estimated the risk of discontinuing using time to discontinuation of initial drug therapy
(in the initial episode of treatment). Only the results for a select number of relevant
independent variables are presented in Table 4.1 (and all tables that follow).
Therapeutic class of the patient’s initial therapy does indeed have an impact on
treatment duration. Patients who use an SNRI, SMS, AAP, or TCA in their initial episode
have a statistically significant shorter duration of therapy (25-33 days) as compared to
MDD patients initially treated with an SSRI. There is a statistically significant increase in
the duration of therapy with age (p<0.0001). Racial differences are significant and exist
32

as well; minorities have a shorter duration of therapy as compared to whites (p<0.01).
Education also plays a role in duration of therapy, as patients with less than a high
school diploma have a shorter duration of therapy compared to those with at least a
college degree (p=0.001).
The OLS results are confirmed by the Cox analysis of time to discontinuation.
Patients on an SNRI have nearly a 20% increased risk of discontinuing their initial drug
therapy as compared to those who were on an SSRI (hazard ratio = 1.1712, p <0.0001).
Roughly the same significant increase in the risk of discontinuation persists with SMS,
AAPs, and TCAs as compared to SSRIs. All age groups have a significantly increased
risk of discontinuing initial therapy as compared to those who are aged 65 and older,
and these differences narrow monotonically after age 18. Minorities also have a
significantly increased risk of terminating therapy. Patients with less than a high school
diploma are at greater risk of terminating therapy as compared to those with a college
degree (hazard ratio = 1.1930, p=0.002).

Table 4.1 OLS and Cox Model, Initial Episode Duration: N=9,090 observations
OLS regression model* Cox proportional hazards
model
Independent Variables OLS Coeff. p value Hazard Ratio Pr > ChiSq
therapeutic class
SSRI ref. group - ref. group -
SNRI -24.9450 < 0.0001 1.1712 < 0.0001
Other AD -5.9224 0.196 1.0475 0.097
SMS -32.5147 < 0.0001 1.2198 < 0.0001
AAP -27.8311 < 0.0001 1.1780 0.001
TCA -22.4998 < 0.0001 1.1658 < 0.0001
MAOI -8.1656 0.907 0.9381 0.849
age category
<18 -38.2527 < 0.0001 1.2616 < 0.0001
18 – 34 -33.3754 < 0.0001 1.2209 < 0.0001
33

Methods and Results: Impact of Treatment Duration

Time to Psychiatric/Non-Psychiatric Hospitalization & ED Visit

The effect of treatment duration during the initial treatment episode on patient
risk of inpatient psychiatric and non-psychiatric hospitalization, as well as ED visits is
estimated using a cox proportional hazards model (Table 4.2). Each additional month of
drug therapy duration does not have a significant impact on psychiatric/non-psychiatric
OLS regression model* Cox proportional hazards
model
Independent Variables OLS Coeff. p value Hazard Ratio Pr > ChiSq
35 – 44 -28.9964 < 0.0001 1.1916 < 0.0001
45 – 54 -29.5866 < 0.0001 1.1879 < 0.0001
55 – 64 -11.8999 0.042 1.0803 0.022
65 < ref. group - ref. group -
race
white ref. group - ref. group -
unknown 15.2096 0.167 0.9154 0.143
hispanic -15.9333 < 0.0001 1.1217 <0.0001
black -12.6829 0.007 1.0787 0.011
asian -3.9274 0.695 1.0472 0.450
education
unknown -30.3267 0.05 1.2305 0.037
less high school -23.9306 0.001 1.1930 0.002
high school diploma 1.3107 0.784 1.0095 0.751
less college 0.4869 0.901 1.0097 0.687
college plus ref. group - ref. group -
total health expenditures
12 mo. prior (per $USD
100)
0.0323 0.005 0.9998 0.001
hospitalization 12 mo.
prior
0.7729 0.874 0.9743 0.378
OOP costs (initial Rx) 0.1713 < 0.0001 0.9989 < 0.0001
female 8.6864 0.004 0.9576 0.020
*R
2
= 0.08; Abbreviations: OLS, ordinary least squares; Coeff., coefficient; Pr, probability;
ChiSq, chi-square; ref., reference; SSRI, selective serotonin reuptake inhibitor; SNRI,
serotonin and norepinephrine reuptake inhibitor; AD, antidepressant; SMS, serotonin
modulator and stimulator; AAP, atypical antipsychotic; TCA, tricyclic antidepressant; MAOI,
monoamine oxidase inhibitor; mo., month; OOP, out-of-pocket; Rx, drug.
34

hospitalization nor ED visit risk. For inpatient mental health/psychiatric facility risk,
patients that are between 55 to 64 years old have a significantly greater risk of
hospitalization than those 65 years of age and older. For ED visits and inpatient non-
psychiatric hospitalizations, all age groups below the age of 65 have a lower risk of
event occurrence. The risk of having an ED visit was found to be significantly (33%)
lower if the patient had a hospitalization in the 12 months prior (hazard ratio = 0.6724,
p<0.0001).

Table 4.2 Cox Model, Duration on Time to Hospitalization/ED Visit: N=9,090
observations
Abbreviations: ED, emergency department; mh, mental health; psych, psychiatric; IP, inpatient;
non-psych, non-psychiatric; HR, hazard ratio; Pr, probability; ChiSq, chi-square; mo., months;
ref., reference.
inpatient mh/
psych facility
ED visit IP (non-psych)
hospitalization
Independent
Variables
HR Pr>ChiSq HR Pr > ChiSq HR Pr >ChiSq
duration of
episode
(monthly)
1.0014 0.909 0.9994 0.903 1.0049 0.312
total health
expenditures
12 mo. prior
(per $USD 100)
1.0003 0.993 0.9992 0.002 0.9988 0.002
hospitalization
12 mo. prior
1.1328 0.446 0.6724 <0.0001 0.0740 <0.0001
age category
<18 0.7635 0.492 0.4257 <0.0001 0.6556 0.009
18 – 34 1.4276 0.158 0.5699 <0.0001 0.6221 <0.0001
35 – 44 1.0944 0.722 0.5252 <0.0001 0.4277 <0.0001
45 – 54 1.5155 0.067 0.5135 <0.0001 0.4548 <0.0001
55 – 64 1.7803 0.005 0.6604 <0.0001 0.6551 <0.0001
65 < ref. group - ref. group - ref. group -
race
white ref. group - ref. group - ref.group -
unknown 1.2289 0.515 0.8446 0.307 1.0619 0.685
hispanic 1.0354 0.842 1.1038 0.160 0.9310 0.333
black 1.0329 0.862 1.0076 0.922 1.0448 0.574
asian 0.3596 0.082 1.0012 0.993 0.8405 0.278
35

Healthcare Costs Over 1 Year
Multivariate linear regressions were used to investigate the impact of duration of
initial drug therapy (in months) on healthcare expenditures. Age, race, education,
gender, income, type of insurance and policy, region, as well as total health
expenditures in the baseline period, were used as control variables. We also include a
vector of disease diagnosis, patients’ prior use of medications, and hospitalization
indicator variables as additional controls.
Table 4.3 presents the results from our OLS regression models, suggesting that
the duration of therapy did not have a significant impact on total post-index health
expenditures (-USD$44.02, p=0.260) but did in fact have a significant impact on total
medical expenditures (-USD$119.68, p=0.001). As expected, each additional month of
drug therapy significantly increased medication costs ($75.66, p<0.0001).

Table 4.3 OLS of MDD Drug Therapy Duration on Healthcare Costs: N=9,090
observations
*Per month in $USD; Abbreviations: OLS, ordinary least squares; MDD, major depressive
disorder; CI, confidence interval; IP, inpatient; OP, outpatient; Rx, drug.

Next, general linear models (GLMs) were carried out to verify and refine these
results. The models were defined by a logarithmic link function with a gamma
distribution for total health, medical, and prescription expenditures. This specification is
commonly used when modeling healthcare expenditures with skewed distributions.
OLS estimates* 95 % CI p value
Duration of therapy
Total health expenditures (IP + OP + Rx) -$44.02 -$120.64, $32.60 0.260
Total medical expenditures (IP + OP) -$119.68 -$188.22, -$51.14 0.001
Total Rx expenditures $75.66 $45.05, $106.28 <0.0001
36

Results (Table 4.4) suggest that duration of therapy had a significant effect on
post-index prescription expenditures (USD$106.41, p<0.0001).

Table 4.4 GLM Regression of MDD Drug Therapy Duration on Healthcare Costs:
N=9,090 observations
*Per month in $USD; Abbreviations: GLM, generalized linear model; MDD, major depressive
disorder; CI, confidence interval; IP, inpatient; OP, outpatient; Rx, drug.

Discussion

In this retrospective claims-based analysis, we sought to document factors that
impact persistence, with interest to whether or not differences in persistence exist
across therapeutic class of initial AD treatment for patients with MDD. We then
assessed if lack of compliance with treatment (non-persistence) was linked to an
increased risk of relevant clinical events (inpatient psychiatric-related/non-psychiatric-
related hospitalizations and ED visits) and greater healthcare costs.
Our results indicate that persistence did indeed differ by therapeutic class of
initial therapy. Patients receiving SSRI therapy had the longest duration of initial
treatment. Use of other therapeutic classes versus SSRIs resulted in a shorter duration
of therapy. A recent study found that initial treatment with SNRI therapy was associated
with greater persistence over a 1-year time frame versus other therapeutic classes
(SSRIs, other ADs, TCAs, and MAOIs).
54
This study did not consider patients initiating
AAPs and SMSs separately as we have done in the current analysis. However, our
Marginal Effect* 95 % CI p value
Duration of therapy
Total health expenditures (IP + OP + Rx)
$90.91 -$7.77, $182.82 0.072
Total medical expenditures (IP + OP)
-$15.50 -$104.36, $73.36 0.732
Total Rx expenditures
$106.41 $74.29, $138.54 <0.0001
37

results similar in that the majority of patients in our sample discontinued treatment in the
first 3 months.
We also explored the role of therapeutic class in a later (second) episode of
treatment, where the number of patients on SNRI therapy was also greater. As
expected, the difference in persistence for SNRIs and SSRIs narrowed. Patients who
use an SNRI in their second episode did not have a statistically significant shorter
duration of therapy (4 days) as compared to MDD patients treated with an SSRI.
Similar to previous findings patients on TCAs or SMSs had a shorter duration of therapy
as compared to SSRIs (27 and 34 days, respectively) and their risk of discontinuing
their second drug therapy was statistically significant and increased (25% and 18%,
respectively).
54-56

The general pattern of our results is consistent with the literature in that older age
increases the likelihood of persistence and racial differences in persistence exist even
after accounting for baseline clinical and demographic factors such as income,
education, and region, among others.
57,58

38

Chapter 5: Analysis of Second Treatment Episodes

Introduction

Next, we evaluated the follow on (second) episodes of treatment. Like initial
treatment episodes, we seek to examine the factors that impact treatment duration,
however, focusing on the type of treatment episode (switch, delayed switch, augment,
or restart) as well as drug use history-related factors from patients’ initial treatment
episode, namely, the number of days the patient was off therapy and the duration of
their initial episode. Next, we consider the effect of the treatment duration of the second
episode on the following patient outcomes of interest: 1) the time to a clinical event
(inpatient psychiatric-related/non-psychiatric-related hospitalizations and ED visits) and
2) healthcare expenditures.

Methods and Results, Factors that Influence Duration of Second Treatment
Episodes

The effects of the type of second episode, days off therapy, duration of initial
episode, prior hospitalization, prior healthcare expenditures, out of pocket medication
costs (for the drug used in the second episode), and other independent variables
related to patient history demographics were assessed first using an OLS model. The
OLS model results were then validated using a Cox proportional hazards model, which
evaluated the time to discontinuing the drug used in the second episode of treatment
(Table 5.1).
Whether a patient’s second episode of treatment was a switch, delayed switch,
augment, or restart did indeed have an impact on treatment duration. Patients who
39

augment have a statistically significant longer duration of therapy (27 days) as
compared to patients who restart their initial therapy. Similar to initial episodes, there is
a statistically significant increase in the duration of therapy with age (p<0.0001). Racial
differences also carry over into the second episode of treatment, appearing strongest
for blacks as compared to whites (p=0.001).
OLS results are validated by the Cox analysis of time to discontinuation. Having
an augment episode provides a protective effect; patients that augment have a 12%
decreased risk of discontinuing treatment as compared to those that restart (hazard
ratio = 0.88, p <0.0001). All age groups have a significantly increased risk of
discontinuing initial therapy as compared to those who are aged 65 and older, and
these differences are diluted as patients approach age 65. Black (hazard ratio = 1.10,
p=0.001) and Hispanics (hazard ratio = 1.08, p=0.001) have a significantly increased
risk of terminating therapy as compared to whites.

Table 5.1 OLS and Cox Model, Second Episode Duration: N=7,744 observations
OLS regression model* Cox proportional hazards
model
Independent Variables OLS Coeff. p value Hazard Ratio Pr > ChiSq
Type of 2
nd
Episode
restart episode
ref. group - ref. group -
delayed switch episode

-1.9114 0.587 1.0324 0.119
switch episode

5.1081 0.323 0.9742 0.392
augment episode

26.5598 < 0.0001 0.8751 < 0.0001
days off therapy (mo.)

-0.4222 < 0.0001 1.0022 < 0.0001
duration, initial Rx
(mo.)

0.0116 0.972 1.0001 0.996
age category
<18
-58.4246 < 0.0001 1.4327 < 0.0001
18 – 34
-54.4794 < 0.0001 1.4131 < 0.0001
35 – 44
-44.7473 < 0.0001 1.3217 < 0.0001
45 – 54
-34.4957 < 0.0001 1.2411 < 0.0001
55 – 64
-19.5811 0.001 1.1415 < 0.0001
40

Methods and Results, Impact of Treatment Duration

Time to Psychiatric/Non-Psychiatric Hospitalization & ED Visit
The effect of treatment duration for the patient’s second treatment episode on
patient risk of inpatient psychiatric and non-psychiatric hospitalization, as well as ED
visits is estimated using a cox proportional hazards model (Table 5.2). Each additional
month of drug therapy duration does not have a significant impact on psychiatric/non-
psychiatric hospitalization nor ED visit risk. For inpatient mental health/psychiatric
facility risk, patients that are 45 to 64 years old have a significantly greater risk of
hospitalization than those 65 years of age and older. For ED visits, all age groups below
OLS regression model* Cox proportional hazards
model
Independent Variables OLS Coeff. p value Hazard Ratio Pr > ChiSq
65 < ref. group - ref. group -
race
white ref. group - ref. group -
unknown
0.4537 0.962 0.9907 0.863
hispanic
-12.6139 0.003 1.0813 0.002
black
-16.2920 0.001 1.1013 0.001
asian
-6.8753 0.500 1.0232 0.725
education
unknown
42.0059 0.279 0.8115 0.191
less high school
12.0189 0.711 0.9346 0.712
high school diploma
-9.2784 0.066 1.0581 0.048
less college
-3.9261 0.346 1.0278 0.235
college plus ref. group - ref. group -
total health expenditures 12
mo. prior (per $USD 100)
-0.0004 0.958 0.9999 0.906
hospitalization 12 mo. prior
11.6124 0.017 0.9484 0.044
OOP costs (2nd Rx)
0.0969 < 0.0001 0.9992 < 0.0001
female
-1.4839 0.655 1.0087 0.643
*R
2
= 0.10; Abbreviations: OLS, ordinary least squares; Coeff., coefficient; Pr, probability;
ChiSq, chi-square; Rx, drug; ref., reference; mo., month; OOP, out-of-pocket.
41

the age of 65 have a lower risk of event occurrence; for inpatient non-psychiatric
hospitalization risk, similar results were shown, though not significant for patients less
than 18 years of age.

Table 5.2 Cox Model, Duration on Time to Hospitalization/ED Visit: N=7,744
observations
Abbreviations: ED, emergency department; mh, mental health; psych, psychiatric; IP, inpatient;
non-psych, non-psychiatric; HR, hazard ratio; Pr, probability; ChiSq, chi-square; mo., months;
ref., reference.

Healthcare Costs Over 1 Year
The impact of duration of drug therapy (in months) on healthcare expenditures
was investigated using multivariate linear regression analysis. The control variables
used in the analysis of initial treatment episodes were again employed.
inpatient mh/
psych facility
ED visit IP (non-psych)
hospitalization
Independent
Variables
HR Pr>ChiSq HR Pr > ChiSq HR Pr >ChiSq
duration of
episode (monthly)
1.0035 0.796 0.9919 0.190 0.9941 0.359
total health
expenditures 12
mo. prior (per
$USD 100)
1.0001 0.657 0.9993 0.007 0.9989 0.007
hospitalization
12 mo. prior
1.2986 0.110 0.6718 <0.0001 0.1158 <0.0001
age category
<18 0.9738 0.949 0.4920 <0.0001 0.9444 0.770
18 – 34 1.3334 0.301 0.5543 <0.0001 0.7725 0.031
35 – 44 1.1160 0.701 0.5127 <0.0001 0.4170 <0.0001
45 – 54 1.6475 0.046 0.4611 <0.0001 0.5512 <0.0001
55 – 64 1.7584 0.012 0.5558 <0.0001 0.7673 0.007
65 < ref. group - ref. group - ref. group -
race
white ref. group - ref. group - ref.group -
unknown 1.0924 0.813 0.8835 0.536 1.1420 0.481
hispanic 0.9543 0.811 1.0860 0.348 0.9159 0.347
black 1.1194 0.578 0.8385 0.080 0.9766 0.812
asian 0.5092 0.256 1.0087 0.964 0.8853 0.581
42

Results of the OLS regression is presented in Table 5.3 suggesting that the
duration of therapy did indeed have a significant impact on total post-index medical
expenditures (-USD$95.17, p=0.033). Each additional month of drug therapy
significantly increased medication costs (USD$92.85, p<0.0001).

Table 5.3 OLS of MDD Drug Therapy Duration on Healthcare Costs: N=7,744
observations
*Per month in $USD; Abbreviations: OLS, ordinary least squares; MDD, major depressive
disorder; CI, confidence interval; IP, inpatient; OP, outpatient; Rx, drug.

Output of the GLM model with gamma family and log link function is presented in
Table 5.4. These results suggest that the duration of therapy had a significant effect on
post-index total health (USD$133.61, p =0.047) and prescription expenditures
(USD$135.57, p<0.0001).
Table 5.4 GLM Regression of MDD Drug Therapy Duration on Healthcare Costs:
N=7,744 observations
*Per month in $USD; Abbreviations: GLM, generalized linear model; MDD, major depressive
disorder; CI, confidence interval; IP, inpatient; OP, outpatient; Rx, drug.
Discussion

In this study of second episodes of treatment, we explored whether the type of
follow-on episode (switch, delayed switch, augment, or restart) and factors related to a
patients’ initial therapy had an impact on treatment duration. Similar to initial treatment
OLS estimates* 95 % CI p value
Duration of therapy
Total health expenditures (IP + OP + Rx)
-$2.32 -$96.67, $92.03 0.962
Total medical expenditures (IP + OP)
-$95.17 -$182.41, -$7.93 0.033
Total Rx expenditures
$92.85 $55.90, $129.81 <0.0001
Marginal Effect* 95 % CI p value
Duration of therapy
Total health expenditures (IP + OP + Rx)
$133.61 $1.60, $252.73 0.047
Total medical expenditures (IP + OP)
-$1.96 -$129.09, $125.18 0.976
Total Rx expenditures
$135.57 $93.53, $177.61 <0.0001
43

episodes, we also evaluated the effect of duration of treatment on hospitalizations and
healthcare costs.
Our results suggest that patients who add on a second drug therapy (augment)
have a statistically significant longer duration of therapy (27 days) and a statistically
significant decreased risk of discontinuing therapy (hazard ratio = 0.88; p <0.0001) as
compared to patients who restart their initial therapy. These findings are similar to
findings from a very recent meta-analysis which suggest that combination/augmentation
strategies may result in improved persistence for patients who have did not achieve
adequate response from their initial therapy.
59

44

Chapter 6: Analysis of Generic/Branded Antidepressant
Therapy, Initial Treatment Episodes

Introduction

Lastly, we return to our focus on initial episodes, assessing the impact of
branded vs. generic medications on treatment duration. We also focus on selected pain
related factors that may impact persistence on initial therapy. These factors include
chronic non-cancer pain disorders (fibromyalgia, diabetic neuropathy, osteoarthritis,
lower back pain, headaches, neuropathic pain) and the pain-related medications used
to treat these conditions (opioids and sedatives).

Methods and Results, Other Factors that Influence Treatment Duration

Whether the initial drug therapy was dispensed generic does indeed have an
impact on treatment duration. Patients who initiate treatment with a generic medication
have a longer duration of treatment (about 1 week longer) as compared to those who
initiate branded AD therapy (p<0.05). If the patient suffered from a chronic pain
disorder, they had a statistically significant shorter duration of initial therapy (7 days).
The setting in which patients where prescribed their initial drug therapy also had an
impact on treatment duration. Those who received their initial prescription from a
primary care or mental health provider had a shorter duration of therapy as compared to
those who were prescribed in an inpatient or outpatient facility (19 days). In these
models, the increase in duration of therapy with age held as it did in our previous model
(see Chapter 4); racial disparities persisted as well.
Results from the Cox proportional hazards model directionally confirmed our OLS
results (Table 6.1). The presence of coexisting pain disorders exhibited a significant
45

increased risk of discontinuing therapy (p=0.009). Patients who were prescribed their
drug therapy in a professional setting versus an inpatient/outpatient hospital were 10%
more likely to discontinue (hazard ratio = 1.1049, p<0.0001). All age groups have a
significantly increased risk of discontinuing initial therapy as compared to those who are
aged 65 and older, and these differences narrow monotonically after age 18. Minorities
also have a significantly increased risk of terminating therapy. Patients with less than a
high school diploma were 26% more likely to discontinue therapy as compared to those
who had a college degree or more (hazard ratio = 1.2608, p<0.0001).

Table 6.1 OLS and Cox Model, Initial Episode Duration, Other Factors: N=9,090
observations
OLS regression model* Cox proportional hazards
model
Independent Variables OLS Coeff. p value Hazard Ratio Pr > ChiSq
dispensed generic

8.3478 0.041 0.9582 0.030
combination therapy

20.0165 0.196 0.4606 0.097
location of initial rx
(professional)

-18.5076 < 0.0001 1.1049 < 0.0001
total health expenditures
12 mo. prior (per
$USD100)

0.0377 0.002 0.9997 0.002
hospitalization 12 mo.
prior

3.3795 0.479 1.0014 < 0.0001
female

3.7951 0.217 0.9785 0.231
pain rx

-0.2633 0.933 1.0090 0.621
chronic non-cancer pain
disorder

-6.9391 0.029 1.0489 0.009
alcohol disorder

-1.6753 0.895 0.9925 0.918
drug disorder

-14.1452 0.155 1.0402 0.602
age category
<18
-25.27103 0.004 1.2969 < 0.0001
18 – 34
-30.6231 < 0.0001 1.2851 < 0.0001
35 – 44
-24.7736 < 0.0001 1.2375 < 0.0001
45 – 54
-26.0329 < 0.0001 1.2251 < 0.0001
55 – 64
-9.3309 0.116 1.1058 0.003
65 < ref. group - ref. group -
46

Discussion

In this analysis, we aimed to compare persistence with branded and generic AD
therapy, with special consideration to disease- and medication-related factors that may
impact persistence; namely, patients diagnosed with chronic non-cancer pain disorders
who may be using opioids and sedatives to treat these conditions.
Patients who initiate treatment with generic therapy had a significantly longer
duration of initial therapy than patients on branded therapy. The fact that branded
antidepressants are more costly than generic may play a role in access to these
medications, thereby affecting treatment patterns. Our results are in line with previous
studies which showed that AD therapy initiation with a generic improved adherence.
60

OLS regression model* Cox proportional hazards
model
Independent Variables OLS Coeff. p value Hazard Ratio Pr > ChiSq
race
white ref. group - ref. group -
unknown
19.9280 0.086 0.9042 0.093
hispanic
-22.0319 < 0.0001 1.1507 <0.0001
black
-14.0533 0.003 1.1176 <0.0001
asian
-7.9189 0.411 1.0733 0.255
education
unknown
-44.1667 < 0.0001 1.3417 <0.0001
less high school
-32.7447 < 0.0001 1.2608 <0.0001
high school diploma
-3.7910 0.410 1.0530 0.054
less college
-2.0494 0.603 1.0216 0.349
college plus ref. group - ref. group -
*R
2
= 0.14; Abbreviations: OLS, ordinary least squares; Coeff., coefficient; Pr, probability;
ChiSq, chi-square; Rx, drug; mo., month; ref., reference; mo., month.
47

Our findings also suggest that patients with chronic pain disorders had a slightly
shorter duration of initial therapy and a significantly higher risk of discontinuation.
We found that physician-related factors also influence persistence; patients who were
prescribed their initial prescription by their primary care or mental health provider were
significantly more likely to discontinue their medication as compared to those who were
prescribed in an inpatient or outpatient facility.

48

Chapter 7: Conclusions and Policy Implications

Consistent with prior literature, we found that multiple demographic, clinical, and
provider-related factors influenced duration of therapy. For example, patients who were
younger, less educated, and of Hispanic or Black origin had a significantly increased
risk of discontinuing therapy. Certain comorbid conditions increased the risk of
discontinuation. Patients with chronic pain conditions, as well as those with drug and
alcohol dependence (albeit statistically non-significant) were at greater risk of
discontinuation.
This analysis also found that persistence varies across therapeutic class-
patients on TCAs and AAPs were more likely to discontinue treatment. Given the new
guidance issued by the FDA in 2018 for the use of ketamine analogues for patients who
remain treatment-resistant, repeating our analysis using data that includes this new
class of rapid-action antidepressants will provide a better understanding to providers of
the current landscape, thus contributing to more effective treatment decisions and
improved patient outcomes.
We also compared the type of follow-on episodes to a patient’s initial therapy.
Our study adds to the growing evidence that adjunctive/augment therapy may serve as
a more effective treatment approach for a certain subset of patients. Additional research
is warranted as to what agents are more effective and tolerable for long-term use in
treating depression.

Health system and provider factors may also play a role in treatment
discontinuation. Generic AD therapy initiation decreased the risk of discontinuation.
49

To expand on the current body of literature that links lower out-of-pocket costs to
generic medication to increased adherence, it would be informative to further examine
provider preferences, and whether patients who were prescribed branded were at
higher risk for discontinuing therapy. Moreover, if the beneficial effect of prescribing
generics to improve persistence outweighs the possible quality and effectiveness
concerns that countries with less strict drug policy regulations may face.

50

References

1. World Health Organization. Depression: Let’s Talk. 2021; http://www.who.int/
news-room/fact-sheets/detail/depression. Accessed February 24, 2022.

2. National Institute of Mental Health (NIMH). Major Depression. 2022;
https://www.nimh.nih.gov/health/statistics/major-depression.shtml. Accessed
February 24, 2022.

3. Centers for Disease Control and Prevention (CDC), National Center for Injury
Prevention and Control. 10 leading causes of death by age group, United States
– 2018. 2018; https://www.cdc.gov/injury/wisqars/pdf/leading_ causes_of_death
_by_age_group_2018-508.pdf. Accessed February 24, 2022.

4. Strakowski SM, Adler CM, Delbello MP. Is depression simply a nonspecific
response to brain injury? Curr Psychiatry Rep. 2013;15:386.

5. Bartoli F, Carrà G, Crocamo C, et al. Association between depression and
neuropathy in people with type 2 diabetes: a meta-analysis. Int J Geriatr
Psychiatry. 2016;31:829-36.

6. Katon W, Pedersen HS, Ribe AR, et al. Effect of depression and diabetes
mellitus on the risk for dementia: a national population-based cohort study.
JAMA Psychiatry. 2015;72:612-9.

7. Saleh, N. Adjunctive antidepressants for bipolar depression linked to reduced
rehospitalization. MDLinx Psychiatry. 2018; https://www.mdlinx.com/
psychiatry/article/1647. Accessed February 22, 2022.

8. American Psychiatric Association. Practice guideline for the treatment of patients
with major depressive disorder. 3rd (edition). 2010; https://psychiatryonline.org
/pb/assets/raw/sitewide/practice_guidelines/guidelines/mdd.pdf. Accessed
February 20, 2022.

9. Richelson E. Multi-modality: a new approach for the treatment of major
depressive disorder. Int J of Neuropsychopharmacol. 2013;16(6):1433-1442.

10. Gaynes BN, Warden D, Trivedi MH, et al. What did STAR*D teach us? Results
from a large-scale, practical, clinical trial for patients with depression.
Psychiatr Serv. 2009;60:1439–1445.

11. Nierenberg AA, Husain MM, Trivedi MH, et al. Residual symptoms after
remission of major depressive disorder with citalopram and risk of relapse: a
STAR*D report. Psychol Med. 2010;40:41–50.
51

12. Greenberg PE, Fournier AA, Sisitsky T, et al. The Economic Burden of Adults
with Major Depressive Disorder in the United States (2005 and 2010).
J Clin Psychiatry. 2015; 76(2):155-162.

13. Rush AJ, Trivedi MH, Wisniewski SR, Nierenberg AA, et al. Acute and longer-
term outcomes in depressed outpatients requiring one or several treatment
steps: a STAR*D report. Am J Psychiatry. 2006;163:1905–1917.

14. Trivedi MH, Lin EH, Katon WJ. Consensus recommendations for improving
adherence, self-management, and outcomes in patients with depression.
CNS Spectr. 2007;12:S1–27.

15. Ashton K, Jamerson BD, Weinstein WL, Wagoner C. Antidepressant-related
adverse effects impacting treatment compliance: results of a patient survey.
Curr Ther Res. 2005;66:96–106.

16. National Institute of Mental Health (NIMH). Depression. 2018;
https://www.nimh.nih.gov/health/topics/depression/index.shtml. Accessed
February 4 2022.

17. Dunlop BW, Nemeroff CB. The role of dopamine in the pathophysiology of
depression. Arch Gen Psychiatry. 2007;64(3):327-37.

18. Fekadu N, Shibeshi W, Engidawork E. Major Depressive Disorder:
Pathophysiology and Clinical Management. J Depress Anxiety. 2017;6:255.

19. American Psychiatric Association (APA). Depression. In: Diagnostic and
Statistical Manual of Mental Disorders, 5th edition (DSM-5). Washington, DC:
American Psychiatric Publishing. 2013; https://dsm.psychiatryonline.org/
doi/abs/10.1176/ appi.books. 9780890425596.dsm04. Accessed
February 4, 2022.

20. Hasin DS, Sarvet AL, Meyers JL, et al. Epidemiology of Adult DSM-5 Major
Depressive Disorder and Its Specifiers in the United States. JAMA Psychiatry.
2018;75(4):336-346.

21. Thapar A, Collishaw S, Thapar AK. Depression in adolescence. Lancet. 2012;
379(9820): 1058-1067.

22. Mocking RJT, Figueroa CA, Rive MM, et al. Vulnerability for new episodes in
recurrent major depressive disorder: protocol for the longitudinal DELTA-
neuroimaging cohort study. BMJ Open. 2016;6:e009510.

52

23. Steinert C, Hofmann M, Kruse J, et al. The prospective long-term course of adult
depression in general practice and the community. A systematic literature review.
J Affect Disord. 2014;152–154:65–75.

24. Eaton WW, Shao H, Nestadt G, et al. Population-based study of first onset and
chronicity in major depressive disorder. Arch Gen Psychiatry. 2008;65:513–20.

25. Hardeveld F, Spijker J, De Graaf R, et al. Recurrence of major depressive
disorder and its predictors in the general population: results from the Netherlands
Mental Health Survey and Incidence Study (NEMESIS). Psychol Med.
2013;43:39–48.

26. Hardeveld F, Spijker J, De Graaf R, et al. Prevalence and predictors of
recurrence of major depressive disorder in the adult population. Acta Psychiatr
Scand. 2010;122:184–91.

27. Cramer JA, Roy A, Burrell A, et al. Medication compliance and persistence:
terminology and definitions. Value Health. 2008;11(1):44–7.

28. Kim-Romo DN, Rascati KL, Richards KM, et al. Medication Adherence and
Persistence in Patients with Severe Major Depressive Disorder with Psychotic
Features: Antidepressant and Second-Generation Antipsychotic Therapy Versus
Antidepressant Monotherapy. J Manag Care Spec Pharm. 2016;22(5):588-96.

29. Hansen RS, Dusetzina SB, Dominik RC, et al. Prescription refill records as a
screening tool to identify antidepressant non-adherence. Pharmacoepidemiol
Drug Saf. 2010;19:33-37.

30. Zimmerman M, Clark HL, Multach MD, et al. Symptom severity and the
generalizability of antidepressant efficacy trials: changes during the past 20
years. J Clin Psychopharmacol. 2016;36;153-6.

31. Qaseem A, Barry MJ, Kansagara D, Clinical Guidelines Committee of the
American College of Physicians. Nonpharmacologic versus pharmacologic
treatment of adult patients with major depressive disorder: a clinical practice
guideline from the American College of Physicians. Ann Intern Med.
2016;164:350-9.

32. Dopheide JA. Depression: More reasons and options for recovery.
The Rx Consultant. 2018; Vol XXVII, No 3.

33. Dupuy JM, Ostacher MJ, Huffman J, et al. A critical review of pharmacotherapy
for major depressive disorder. Int J Neuropsychoph. 2011;14:1417–1431.

53

34. Köhler-Forsberg O, Larsen ER, Buttenschøn HN, et al. Effect of antidepressant
switching between nortriptyline and escitalopram after a failed first
antidepressant treatment among patients with major depressive disorder.
Br J Psychiatry. 2019; 215(2):494-501.

35. de Sousa RT, Zanetti MV, Brunoni AR, et al. Challenging Treatment-Resistant
Major Depressive Disorder: A Roadmap for Improved Therapeutics.
Curr Neuropharmacol. 2015;13:616-635.

36. Nelson JC. Managing treatment-resistant major depression. J Clin Psychiatry.
2003;64 Suppl 1:5–12.

37. Citrome L. The ABC’s of dopamine receptor partial agonists: aripiprazole,
brexpiprazole, and cariprazine: the 15 min challenge to sort these agents out.
Int J Clin Pract. 2015;69:1211-20.

38. Crossley NA, Bauer M. Acceleration and augmentation of antidepressants with
lithium for depressive disorders: two meta-analyses of randomized, placebo-
controlled trials. J Clin Psychiatry. 2007;68:935–940.

39. Hannan N, Hamzah Z, Akinpeloye HO, Meagher D. Venlafaxinemirtazapine
combination in the treatment of persistent depressive illness.
J Psychopharmacol. 2007;21:161–164.

40. Malhi GS, Ng F, Berk M. Dual-dual action? Combining venlafaxine and
mirtazapine in the treatment of depression. Aust N Z J Psychiatry.
2008;42:346–349.

41. Al-Harbi KS. Treatment-resistant depression: therapeutic trends, challenges,
and future directions. Patient Prefer Adher. 2012:6:369-388.

42. Lam RW, Kennedy SH, Grigoriadis S, et al. Canadian Network for Mood and
Anxiety Treatments (CANMAT) clinical guidelines for the management of major
depressive disorder in adults. III. Pharmacotherapy. J Affect Disord.
2009;117(Suppl 1):S26-S43.

43. Culpepper L, Muskin PR, Stahl SM. Major Depressive Disorder: Understanding
the Significance of Residual Symptoms and Balancing Efficacy with Tolerability.
Am J Med. 2015; 128:S1-S15.

44. Lee MS, Lee HY, Kang SG, et al. Variables influencing antidepressant
medication adherence for treating outpatients with depressive disorders.
J Affect Disord. 2010;123:216-221.

54

45. American Association of Suicidology. Some Facts About Suicide and
Depression. 2009; https://www.cga.ct.gov/asaferconnecticut/tmy/0129/
Some%20Facts%20About%20Suicide%20and%20Depression%20-
%20Article.pdf. Accessed February 4, 2022.

46. Moller HJ. Suicide, suicidality and suicide prevention in affective disorders.
Acta Psychiatr Scand Suppl. 2003:73–80.

47. Carter MW, Reymann MR. ED use by older adults attempting suicide.
Am. J. Emerg. Med. 2014;32(6):535–540.

48. Baldessarini RJ, Tondo L, Davis P, et al. Decreased risk of suicides and attempts
during long-term lithium treatment: a meta-analytic review. Bipolar Disord.
2007;9:314.

49. Bressi SK, Marcus SC, Solomon PL. The impact of psychiatric comorbidity on
general hospital length of stay. Psychiatr Q. 2006;77(3):203–209.

50. Citrome L, Jain R, Tung A, et al. Prevalence, treatment patterns, and stay
characteristics associated with hospitalizations for major depressive disorder.
J Affect Disord. 2019;249:378-384.

51. Gramaglia C, Feggi A, Bergamasco P, et al. Clinical characteristics associated
with suicide attempts in clinical settings: a comparison of suicidal and non-
suicidal depressed inpatients. Front Psychiatry. 2016;7:109.

52. Ho SC, Chong HY, Chaiyakunapruk N, et al. Clinical and economic impact of
non-adherence to antidepressants in major depressive disorder: A systematic
review. J of Affect Disord. 2016:193:1-10.

53. Osterberg L, Blaschke T. Adherence to Medication. New Engl J Med.
2005;353:487–497.

54. Keyloun KR, Hepp Z, Hansen RN, et al. Adherence and Persistence Across
Antidepressant Therapeutic Classes: A Retrospective Claims Analysis Among
Insured US Patients with Major Depressive Disorder. CNS Drugs.
2017;31:421-432.

55. Sheehan DV, Keene MS, Eaddy M, Krulewicz S, Kraus JE, Carpenter DJ.
Differences in medication adherence and healthcare resource utilization patterns:
older versus newer antidepressant agents in patients with depression and/or
anxiety disorders. CNS Drug. 2008;22(11):963–973.

55

56. Serna MC, Cruz I, Real J, Gasco E, Galvan L. Duration and adherence of
antidepressant treatment (2003–2007) based on prescription database.
Eur Psychiatry. 2010;25(4):206–213.

57. Lui X, Chen Y. Adherence and persistence with branded antidepressants and
generic SSRIs among managed care patients with major depressive disorder.
Clinicoecon Outcomes Res. 2011;3:63-72.

58. Murphy E, Hou L, Maher BS, et al. Race Genetic Ancestry and Response to
Antidepressant Treatment for Major Depression. Neuropsychopharmacol.
2013;38: 2598-2606.

59. Henssler J, Alexander D, Schwarzer G, et al. Combining Antidepressants vs
Antidepressant Monotherapy for Treatment of Patients with Acute Depression-
A Systematic Review and Meta-analysis. JAMA Psychiatry. 2022; e214313.

60. Barbui C, Conti V. Adherence to generic v. brand antidepressant treatment and
the key role of health system factors. Epidemiol Psychiatr Sci. 2015;24(1):23-6.

Abstract (if available)

Abstract Depression is a common illness, affecting 280 million people worldwide. It is associated with mood fluctuations or loss of interest in pleasurable activities of everyday life. It is recurrent in nature, making it a long-term condition, which poses clinical challenges that may lead to increased health resource utilization, and furthermore, a burden to our economy. Furthermore, the majority of patients discontinue treatment prior to the recommended treatment length to achieve symptom remission. The overarching goal of this dissertation is to document treatment patterns of antidepressant (AD) therapy and to examine the effect on clinical and economic outcomes. The dissertation comprises of three objectives to explore these issues: 1) examining the factors associated with persistence with initial treatment and the impact of persistence on hospitalization risks and costs, with special attention to the therapeutic class of initial drug therapy, 2) determining the factors associated with persistence with second treatment episodes and the impact on hospitalization risks and costs, with a particular interest in the type of treatment episode and factors from a patients’ initial treatment, and 3) to document the potential impact of generic/branded therapy and examine medication- and disease-related factors associated with persistence in initial therapy. These objectives will be addressed using de-identified claims from Optum Clinformatics® DataMart Database.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Pharmacotherapy among patients with type 2 diabetes

PDF

Three essays on comparative economic and clinical outcomes of antipsychotics using retrospective claims data

PDF

Effects of a formulary expansion on the use of atypical antipsychotics and health care services by patients with schizophrenia in the California Medicaid Program

PDF

Physician behavior and low-value care

PDF

Crohn’s disease: health outcomes and resource utilization in the biologic era

PDF

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

PDF

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

PDF

Essays in health economics

PDF

Evaluating treatment options for metastatic, castration-resistant prostate cancer: a comprehensive value assessment

PDF

The impact of treatment decisions and adherence on outcomes in small hereditary disease populations

PDF

Outcomes of antibiotic use among children with acute respiratory tract infections

PDF

Discriminating changes in health using patient-reported outcomes

PDF

Assessing value defects in limb preservation care

PDF

An evaluation of risk predictors for disease progression in Veteran Affairs patients with chronic hepatitis C

PDF

Three essays on emerging issues in hemophilia care

PDF

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

PDF

Advances and applications for economic evaluation methods in health technology assessment (HTA)

PDF

Exploring mental health care utilization and academic persistence among college students: evaluating racial and stigma-related disparities

PDF

Documenting the impact of under-treatment for chronic disorders

PDF

Essays in pharmaceutical and health economics

Asset Metadata

Creator Seyedin, Roxanna (author)

Core Title Impact of medication persistence on clinical and health care cost outcomes in patients with major depressive disorder using retrospective claims data

School School of Pharmacy

Degree Doctor of Philosophy

Degree Program Health Economics

Degree Conferral Date 2022-05

Publication Date 04/13/2022

Defense Date 03/07/2022

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

Tag Depression,OAI-PMH Harvest,persistence,retrospective claims analysis

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Padula, William (committee chair), Dopheide, Julie (committee member), Nichol, Michael (committee member)

Creator Email roxyseyedin@gmail.com,rseyedin@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC110937307

Unique identifier UC110937307

Document Type Dissertation

Format application/pdf (imt)

Rights Seyedin, Roxanna

Type texts

Source 20220413-usctheses-batch-923 (batch), 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 author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.

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

Repository Email cisadmin@lib.usc.edu