University of Southern California Dissertations and Theses

Pharmacotherapy among patients with type 2 diabetes

(USC Thesis Other)

Pharmacotherapy among patients with type 2 diabetes

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content PHARMACOTHERAPY AMONG PATIENTS WITH TYPE 2 DIABETES
BY
Si Xuan
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)
December 2020

ii

ACKNOWLEDGMENTS

First and foremost, I would like to express my sincere gratitude to my advisor and
dissertation chair, Dr. Jeffrey McCombs, for mentoring me with his exceptional academic
excellence and admirable personality. Dr. McCombs has always been supportive in my personal
career development. His guidance and encouragement is invaluable and what I have learned from
him over the past few years will be a lifetime treasure for me.

My sincere thanks also go to my dissertation committee members, Drs. Rory Kim and
Bill Padula, for being patient and advising me throughout this dissertation work. Their helpful
insights, comments, and suggestions truly shaped my research work.

I’m also very grateful to Drs. Seth Seabury and Rebecca Myerson, for their helpful
feedback on my dissertation proposal. My thanks are also extended to all my colleagues and
friends at USC.

Finally, I would like to sincerely thank my parents for all their endless love and
encouragement. They have always been supporting me with their best wishes. I would like to
thank my beloved husband, Junjie Ma. He has always been there encouraging me and standing
by me through the whole journey.

iii

TABLE OF CONTENTS

ACKNOWLEDGMENTS ............................................................................................................ ii
LIST OF TABLES ........................................................................................................................ v
LIST OF FIGURES .................................................................................................................... vii
ABSTRACT ................................................................................................................................ viii
CHAPTER 1. INTRODUCTION ................................................................................................ 1
Chapter 1 References .............................................................................................................................. 7
CHAPTER 2. Treatment Duration Among Newly Treated T2D Patients and Its Impact on
Patient Outcomes ........................................................................................................................ 13
2.1 Methods ............................................................................................................................................ 13
2.1.1 Study population .........................................................................................................................................13
2.1.2 Definition of an initial treatment regimen ...................................................................................................13
2.1.3 Key explanatory variables (part 1) ..............................................................................................................14
2.1.4 Outcomes and covariates (part 1) ................................................................................................................15
2.1.5 Statistical analyses (part 1) .........................................................................................................................15
2.1.6 Sensitivity analyses (part 1) ........................................................................................................................16
2.1.7 Key explanatory variables (part 2) ..............................................................................................................16
2.1.8 Outcomes and covariates (part 2) ................................................................................................................16
2.1.9 Statistical analyses (part 2) .........................................................................................................................17
2.2 Results .............................................................................................................................................. 18
2.2.1 Baseline patient characteristics for patients beginning their initial drug therapy for T2D .........................18
2.2.2 Predictors of duration of continuous therapy ..............................................................................................19
2.2.3 The impact of duration of drug therapy on the risk of clinical events and cost over the first year following
the initiation of T2D drug therapy .......................................................................................................................19
2.3 Discussion ......................................................................................................................................... 21
2.4 Conclusions ...................................................................................................................................... 24
Chapter 2 References ............................................................................................................................ 25
Chapter 2 Appendix .............................................................................................................................. 39
Chapter 2 Appendix references ........................................................................................................... 41
CHAPTER 3. Treatment Duration Among Previously Treated T2D Patients and Its Impact
on Patient Outcomes ................................................................................................................... 42
3.1 METHODS ...................................................................................................................................... 42
3.1.1 Study population .........................................................................................................................................42
3.1.2 Definition of the second episode of treatment ............................................................................................42
3.1.3 Key explanatory variables (part 1) ..............................................................................................................43
3.1.4 Outcomes and other covariates (part 1) ......................................................................................................44
3.1.5 Statistical analyses (part 1) .........................................................................................................................44
3.1.6 Sensitivity analyses (part 1) ........................................................................................................................44

iv

3.1.7 Key explanatory variables (part 2) ..............................................................................................................45
3.1.8 Outcomes and covariates (part 2) ................................................................................................................45
3.1.9 Statistical analyses (part 2) .........................................................................................................................46
3.2 Results .............................................................................................................................................. 47
3.2.1 Baseline patient characteristics for patients beginning their second episode of drug therapy for T2D ......47
3.2.2 Predictors of duration of continuous therapy starting from the second episode of treatment .....................48
3.2.3 The impact of duration of drug therapy on the risk of clinical events and cost over the first year following
the initiation of second episode of T2D drug therapy ..........................................................................................49
3.3 Discussion ......................................................................................................................................... 50
3.4 Conclusions ...................................................................................................................................... 53
Chapter 3 References ............................................................................................................................ 54
Chapter 3 Appendix .............................................................................................................................. 65
Chapter 3 Appendix references ........................................................................................................... 66
CHAPTER 4. Delaying Treatment initiation for Type 2 Diabetes Patients Following Their
Initial A1c Test >= 7% ............................................................................................................... 67
4.1 Methods ............................................................................................................................................ 67
4.1.1 Study population .........................................................................................................................................67
4.1.2 Explanatory Variables (part 1) ....................................................................................................................67
4.1.3 Outcomes (part 1) ........................................................................................................................................68
4.1.4 Statistical analyses (part 1) .........................................................................................................................68
4.1.5 Key explanatory variables (part 2) ..............................................................................................................68
4.1.6 Outcomes and covariates (part 2) ................................................................................................................69
4.1.7 Statistical analyses (part 2) .........................................................................................................................69
4.2 Results .............................................................................................................................................. 69
4.2.1 Study Sample Selection and Baseline Characteristics ................................................................................69
4.2.2 Predictors of treatment initiation .................................................................................................................70
4.2.3 The impact of treatment delay on the risk of clinical events ......................................................................72
4.3 Discussion ......................................................................................................................................... 72
4.4 Conclusions ...................................................................................................................................... 75
Chapter 4 References ............................................................................................................................ 77
CHAPTER 5. Conclusions ......................................................................................................... 94

v

LIST OF TABLES

Table 2.1. Patient Characteristics by Drug Therapy Outcome ..................................................... 30
Table 2.2. Patient characteristics impacting duration that patients stayed on any drug among
patients Initiating an Initial Episode of Drug Therapy for T2D ................................................... 32
Table 2.3a. Sensitivity Analysis .................................................................................................... 34
Table 2.3b. Sensitivity Analysis ................................................................................................... 35
Table 2.4. Impact of Duration on Clinical Event Risk ................................................................. 36
Table 2.5. Impact of Duration on Clinical Event Risk (follow up capped at the first year) ......... 37
Table 2.6. Impact of Duration during the First Year on the Change in Health Care Costs by Type
of Service ...................................................................................................................................... 37
Table 2.7. Impact of Duration, PDC>=0.8 during the First Year on Health Care Costs by Type of
Service (GLMs) ............................................................................................................................ 37
Table 2.8. Impact of Duration, PDC>=0.8 during the First Year on Health Care Costs by Type of
Service (OLSs) .............................................................................................................................. 38
Table 2.A1. Results of tests for Schoenfeld residuals of treatment duration ................................ 39
Table 2.A2. Modified Park Test (duration in months) .................................................................. 39
Table 3.1. Patient characteristics by drug therapy outcomes ........................................................ 58
Table 3.2. Patient characteristics impacting duration that patients stayed on any drug among
patients starting the second episode of drug therapy for T2D ...................................................... 60
Table 3.3a. Sensitivity Analysis .................................................................................................... 62
Table 3.3b. Sensitivity Analysis ................................................................................................... 62
Table 3.4. Impact of Duration on Clinical Event Risk ................................................................. 62
Table 3.5. Impact of Duration during the First Year on Health Care Costs by Type of Service
(GLMs) ......................................................................................................................................... 63
Table 3.6. Impact of Duration during the First Year on Health Care Costs by Type of Service
(OLSs) ........................................................................................................................................... 63
Table 3.7. Impact of Duration during the First Year on the Change in Health Care Costs by Type
of Service ...................................................................................................................................... 63
Table 3.A1. Results of tests for Schoenfeld residuals of treatment duration ................................ 65
Table 3.A2a. Modified Park Test (duration in months) ................................................................ 65

vi

Table 3.A2b. Modified Park Test (PDC ³ 80%) .......................................................................... 66
Table 4.1. Patient Characteristics .................................................................................................. 80
Table 4.2. Patient characteristics associated with treatment initiation: ........................................ 83
Comparison of Base Model with interaction term Models ........................................................... 83
Table 4.3. Patient characteristics associated with days of delay in treatment initiation: .............. 85
Comparison of Base Model with interaction term Models ........................................................... 85
Table 4.4a. Impact of treatment delay on risk of clinical outcomes ............................................. 88
Table 4.4b. Impact of treatment delay on risk of clinical outcomes ............................................. 89
Table 4.4c. Impact of treatment delay on risk of clinical outcomes ............................................. 91

vii

LIST OF FIGURES

Figure 2.1. Selection of Patients ................................................................................................... 29
Figure 3.1. Selection of Patients with Second Episodes of Therapy ............................................ 57
Figure 4.1. Selection of Patients with First Test of A1c >= 7% ................................................... 79

viii

ABSTRACT

Type 2 diabetes (T2D) is a costly, chronic, and progressive disease characterized by
insufficient insulin production, insulin resistance or both. Treatment typically requires
pharmacologic antidiabetic therapies. Persistence with T2D medications is critical to achieving
optimal glycemic control and preventing adverse clinical events. Although a wide array of
medications is available for treating type 2 diabetes (T2D), approximately half of patients with
T2D discontinue therapy. Lack of persistence is an important barrier to glycemic control and is a
common problem in patients with T2D. Previous studies have linked non-persistence and
medication nonadherence with more T2D-related complications, increased use of healthcare
resources, and higher medical costs. However, prior studies have been limited to assess
persistence and adherence in specific drug class or specific treatment phase. In addition, a clear
temporal relationship between discontinuation status and future clinical events have not been
established in the literature. The American Diabetes Association (ADA) guidelines recommend
metformin and lifestyle modifications upon initial diagnosis, aiming to reduce the patient’s A1C
to < 7%. Achieving A1c targets has been shown to reduce microvascular complications, and may
be associated with long-term cardiovascular benefits. However, despite the importance of timely
glycemic control, a high proportion of patients experience delays in treatment initiation and
clinical inertia. Previous studies mostly focused on treatment intensification following the initial
treatment. Relatively fewer studies evaluated delays in initiating initial treatment after the
patient’s initial elevated A1c or the effects of delaying treatment on subsequent clinical events.
This dissertation focused on persistence with treatment episodes and delays in initiating
treatment.

ix

The first two papers investigated factors associated with treatment persistence, and
estimated the impact of persistence on clinical and cost outcomes among patients with T2D. The
first paper evaluated treatment persistence following patient initial treatment episodes; paper 2
focused on persistence following patient second treatment episodes. Paper 1 finds newly treated
patients who are persistent on their initial antidiabetic treatment regimens enjoyed significantly
reduced risks of adverse health outcomes and lower outpatient/inpatient costs. Paper 2 finds
higher duration of antidiabetic medications following the second treatment attempt can
significantly improve clinical outcomes at any post-index time and 1-year post-index medical
costs. These results emphasize the importance of persistence for patients with T2D and the need
to assist them with remaining persistent from the early phase of therapy.
Paper 3 documented patient characteristics associated with delays in initiating drug
therapy following the patient’s first observed A1C of 7% or higher, and estimated the impact of
delays in initiating therapy on the risk of T2D-related complications. We find once A1C is 7% or
higher, physicians are balancing age and health status of patients when deciding whether or not
initiate treatment. Delayed treatment among newly treated patients with T2D can significantly
increase the risk of clinical outcomes at any post-index time. These results emphasize the
importance of timely treatment initiation from the onset of T2D and the benefits of early tight
glycemic controls on preventing long-term diabetes-related complications.

1

CHAPTER 1. INTRODUCTION
Type 2 diabetes (T2D) is a complex, chronic, and progressive disease characterized by a
relative insufficiency of insulin caused by insufficient insulin production, insulin resistance or
both. Treatment of diabetes typically requires therapeutic lifestyle changes as well as
pharmacologic antidiabetic therapies. T2D is a global pandemic. In 2015, about 400 million
people worldwide were estimated to have T2D
1
. In the U.S., T2D affects more than 27 million
people of all ages and is the seventh leading cause of death
2
. Approximately 25% of all US
adults over age 65 have T2D
2
. With population aging in the US, 55 million Americans are
projected to have T2D by 2030
3
. Uncontrolled T2D is associated with the development of
macrovascular and microvascular complications that lead to significant morbidity and mortality
2
. Patients with T2D were estimated to have relative 1.4 to 5.8 higher risk of developing stroke
4
.
Rates of cardiovascular death were about 1.7 times higher in T2D patients than those without
T2D
2
. T2D is also associated with a substantial economic burden. In 2017, the total estimated
costs of T2D in the US was about $290 billion, with about $203 billion in direct medical costs
and $87 billion in lost productivity
5
.
Metformin is the preferred first-line therapy for T2D according to the American Diabetes
Association (ADA) guidelines. However, in instances when metformin is contraindicated or not
well tolerated, patients may take other antidiabetic medications such as sulfonylureas (SUs) or
thiazolidinedione (TZDs).
As T2D progresses, if initial antidiabetic medications (ADMs) fail to achieve the desired
glycemic goals after 3 months, the ADA guideline recommends a treatment intensification by
adding a second ADM. The recommended second-line treatment includes sulfonylureas, TZDs,
2

SGLT2 inhibitors, dipeptidyl peptidase 4 (DPP-4 inhibitors), glucagon-like peptidase 1 agonists
(GLP-1 RAs), and insulin
6
. Appropriate selection of second-line antidiabetic agent should
consider patient comorbidities, safety, and cost.
Adherence to prescribed medication regimens or more specifically, persistence with the
medication over time, is critical to achieve optimal glycemic targets and prevent complications
from T2D. Persistence refers to continuation of therapy which is measured as the time from the
initial filling until treatment termination or a substantial gap in refills. Adherence represents how
often the patient takes the drug as prescribed, which is usually measured either by a proportion of
days covered (PDC), or by a medication possession ratio (MPR). The PDC calculates the
percentage of days in a given period that were covered by pharmacy claims, whereas the MPR
measures the total days of supply by adding up all claims within the given period. Despite a
myriad of benefits associated with medication persistence and adherence for patients with T2D,
medication non-persistence and non-adherence are particularly widespread issues
7
. Previous
studies have found that diabetes has among the lowest treatment adherence rates. In the
literature, the rates of medication adherence to ADMs have generally measured using MPR or
PDC and have ranged from 65% to 85%, though the estimates vary considerably based on the
metric used
8-13
. The adherence rates have not improved significantly over time and might be
even lower after the first year of treatment
8,14-16
. Of the 34 studies included in a systematic
review, persistence (proportion of patients without a substantial treatment gap) ranged from 41%
to 81% with a mean of 56%
13
.
Montvida et al explored treatment patterns after initial regimens in the US
17
. They found
that nearly half of patients initiated the second ADM, and SU had the least discontinuation rate
within 1 year compared with other second-line non-insulin ADMs. However treatment patterns
3

evaluated in this study were limited to treatment intensifications with initial therapy of
metformin. Prescribing patterns and persistence once patients changed initial therapy or restart
discontinued treatment have not been studied thoroughly.
Non-persistence with and non-adherence to ADMs have been associated with poorer
glycemic control, higher rates of hospitalization, and more complications
18-20
. A pooled analysis
of data from 3 US-based retrospective studies in patients with T2D found that patients being
persistence with treatment compared with non-persistent patients had significantly lower average
A1c levels at 1 year
21
. Carls et al., recently found that medication non-adherence might be
responsible for up to 75% of the gap between clinical efficacy in RCTs and real-world results in
lowering HbA1c levels
22
. A recent meta-analysis found that T2D patients with adherence rates
(either PDC or MPR) ³ 80% had a significant 10% lower rate of hospitalization and a significant
28% lower rate of all-cause mortality when compared with patients with poor adherence (< 80%)
23
.
For the second treatment attempt after patients’ initial therapy, patient outcomes of the
second treatment attempt have not been explored in the literature.
The ADA guidelines recommend metformin and lifestyle modifications upon initial
diagnosis, aiming to reduce the patient’s A1c to < 7%
6
. Additional ADMs may be added if the
A1c continues to remain above the recommended target
6
. Glycemic control is fundamental to
T2D management
6
. Achieving A1c targets has been shown to reduce microvascular
complications, and may be associated with long-term cardiovascular benefits
24-27
. However,
despite the importance of timely glycemic control, a high proportion of patients experience
delays in treatment initiation and clinical intertia
28-32
. A recent systematic review found that
4

median time to treatment intensification after an A1c value above target was longer than 1 year
33
.
Delays in appropriate therapy can increase patient risk. Previous studies mostly focused
on treatment intensification following the initial treatment. Relatively fewer studies evaluated the
extent of the delay in initiating initial treatment after the patient’s initial elevated A1c and its
effects on subsequent clinical events. Delays in initial treatment or timely treatment
intensification can expose patients to long periods of an elevated A1c that can negatively impact
their prognosis. In a managed care population, Ruiz-Negrón et al., found that delay in
intensifying needed treatment was associated with poor A1c outcomes over the 12-month follow
up
34
. Pantalone et al., conducted a single site study and reported that time until A1c goal
attainment was shorter among patients who received early intensification
31
. A retrospective
cohort study in UK found that a 1 year delay in receiving needed antidiabetic treatment was
associated with significantly increased risk of cardiovascular events
35
.
Prior research into the impact of medication persistence in diabetes can be improved in
three ways. First, the methods used most frequently to measure adherence with drug therapy do
not establish the temporal relationship between the patient’s treatment status and future events. A
summary of the current literature can be found in three systematic review and three recent
observational studies of the association between future clinical events and adherence in T2D
(Asche et al, 2011 included 37 studies; Khunti et al, 2017 included 8 studies; McGovern et al,
2017 included 48 studies
12,19,23
; Berkowitz et al, 2014; Lo-Ciganic WH et al, 2016; Gatwood J et
al, 2018
36-38
). Most of previous studies only investigated adherence measured by PDC or MPR.
Both PDC and MPR risk rating a patient as being adherent (PDC or MPR ³ 80%) when in fact
the patient may have stopped therapy, experienced a related adverse event and resumed
5

treatment. Persistence, which measures the duration of use before termination or substantial gap
in medication availability, is likely a better predictor of future adverse events and health care
costs.
Second, most research is limited to the patient’s initial treatment attempt and fails to
analyze the impact of non-adherence with medications used to treat the patient who has broken
with their initial therapy or who have undergone treatment intensification over time. The
estimation of possible delays in initiating treatment after patient first elevated A1c, often referred
to as ‘treatment inertia’, is largely missing in the research literature.
Third, adherence studies which use a fixed one-year or two-year MPR/PDC can only
assess the effects of their adherence measure on clinical outcomes over these fixed periods. This
method does not establish a clear temporal relationship between the treatment status of a patient
at the time of an adverse event/complication. When assessed clinical outcomes in the subsequent
years, selected patients would be healthy since only patients who are event survive the first
one/two years are included in the analyses. Treatment duration measured starting with the
initiation of drug therapy can be measured as a time-dependent measure. Only two studies
identified to use a time-varying adherence on the risk of clinical outcomes, however, both used
MPR
20,39
. The association between adherence and health care costs is less sensitive to the time-
dependent measurement of adherence. Most studies evaluate the impact of adherence measured
over a fixed period of time (MPR/PDC) and health care costs measured over the same study
period. This method will be used in the research proposed here.
The dissertation research is divided into three studies. First, we evaluated persistence
with initial episodes of drug therapy for T2D and associated patient outcomes. Second, we
examined persistence with second episodes of therapy and associated outcomes. Third, we
6

investigated delays in initiating the initial therapy following patient first A1c>=7%. The specific
research objectives are listed as follows.
Paper 1: Treatment duration among newly treated T2D patients and its impact on patient
outcomes
The first part of this research is to evaluate initial episodes of drug therapy for T2D. This
analysis is divided into two parts. First, we examine the predictors of treatment persistence as a
function of the initial antidiabetic drug regimen used by the patient. In the second part of the
paper we estimate the impact of treatment duration on the risk of T2D-related complications and
health care cost among patients with T2D. The results of this research should be valuable to
provide insights into duration and time to insulin initiation as a function of the drugs selected and
be useful for health professionals aiming at improving patient persistence with drug therapy.
Paper 2: Treatment duration among previously treated T2D patients and its impact on
patient outcomes
The second part of this research is to document the treatment outcomes achieved once
T2D patients transition to a second episode of drug therapy. The analysis is divided into two
parts. First, we examine the predictors of persistence with drug therapy as a function of patient
characteristics, the outcomes achieved during the patient’s initial treatment attempt and the
characteristics of the patient’s second episode of the antidiabetic drug regimen. In the second
part of the paper we estimate the impact of treatment duration on the risk of serious T2D-related
health events, hospitalizations and health care cost. The results of this research provide valuable
insights into duration as a function of the drugs selected and is useful for health professionals
aiming at improving patient persistence with prescribed therapies.
7

Paper 3: Delaying treatment initiation for type 2 diabetes patients following their initial
A1c test >= 7%
The third part of this research is divided into two parts. First, we examine patient
characteristics associated with delays in drug therapy initiation following the patient’s first
observed A1c >= 7%. Second, we estimate the impact of delays in therapy initiation on the risk
of T2D-related complications. The results of this research will provide insights into factors
leading to delays in initial treatment and document the potential health risks created by delaying
treatment. These results will be useful for health professionals, health care organizations, health
maintenance organizations (HMOs), managed care organizations and insurers to develop better
clinical care guidelines aimed at preventing treatment delays.
All three analyses in this research used data derived from the Optum’s De-identified
Clinformatics® Data Mart database covering January 1, 2007, to June 30, 2018. The Optum
claims database is a national and longitudinal database covering approximately 50 million
commercial and Medicare Advantage enrollees. The data include medical claims, pharmacy
claims, outpatient lab results for a subset of patients (~25%), and member eligibility data.
Chapter 1 References

1. International Diabetes Federation. IDF Diabetes Atlas. 7th ed. Brussels, Belgium,
International Diabetes Federation. 2015.
2. National Diabetes Statistics Report, 2017: estimates of Diabetes and Its Burden in the
United States. 2017; https://www.cdc.gov/diabetes/pdfs/data/statistics/national-diabetes-
statistics-report.pdf. Accessed 7/30, 2019.
8

3. Rowley WR, Bezold C, Arikan Y, Byrne E, Krohe S. Diabetes 2030: insights from
yesterday, today, and future trends. Population health management. 2017;20(1):6-12.
4. Tun NN, Anrunagirinathan G, Munshi SK, JM P. Diabetes mellitus and stroke: a clinical
update. World J Diabetes. 2017;8(6):235-248.
5. Association AD. Economic costs of diabetes in the US in 2017. Diabetes Care.
2018;41(5):917-928.
6. American Diabetes Association. 9. Pharmacologic approaches to glycemic treatment:
standards of medical care in diabetes—2019. Diabetes Care. 2019;42(Supplement
1):S90-S102.
7. Giugliano D, Maiorino MI, Bellastella G, Esposito K. Glycemic control in type 2
diabetes: from medication nonadherence to residual vascular risk. Endocrine.
2018;61(1):23-27.
8. Cramer JA. A systematic review of adherence with medications for diabetes. Diabetes
Care. 2004;27(5):1218-1224.
9. Rubin RR. Adherence to pharmacologic therapy in patients with type 2 diabetes mellitus.
The American journal of medicine. 2005;118(5):27-34.
10. Bailey C, Kodack M. Patient adherence to medication requirements for therapy of type 2
diabetes. Int J Clin Pract. 2011;65(3):314-322.
11. DiMatteo MR. Variations in patients' adherence to medical recommendations: a
quantitative review of 50 years of research. Med Care. 2004:200-209.
12. McGovern A, Tippu Z, Hinton W, Munro N, Whyte M, de Lusignan S. Comparison of
medication adherence and persistence in type 2 diabetes: A systematic review and meta‐
analysis. Diabetes, Obesity and Metabolism. 2018;20(4):1040-1043.
9

13. Iglay K, Cartier SE, Rosen VM, et al. Meta-analysis of studies examining medication
adherence, persistence, and discontinuation of oral antihyperglycemic agents in type 2
diabetes. Curr Med Res Opin. 2015;31(7):1283-1296.
14. Donnan PT, MacDonald TM, Morris AD. Adherence to prescribed oral hypoglycaemic
medication in a population of patients with Type 2 diabetes: a retrospective cohort study.
Diabet Med. 2002;19(4):279-284.
15. Odegard PS, Capoccia K. Medication taking and diabetes. The Diabetes Educator.
2007;33(6):1014-1029.
16. Capoccia K, Odegard PS, Letassy N. Medication adherence with diabetes medication: a
systematic review of the literature. The Diabetes Educator. 2016;42(1):34-71.
17. Montvida O, Shaw J, Atherton JJ, Stringer F, Paul SK. Long-term trends in antidiabetes
drug usage in the US: real-world evidence in patients newly diagnosed with type 2
diabetes. Diabetes Care. 2018;41(1):69-78.
18. Ho PM, Rumsfeld JS, Masoudi FA, et al. Effect of medication nonadherence on
hospitalization and mortality among patients with diabetes mellitus. Arch Intern Med.
2006;166(17):1836-1841.
19. Asche C, LaFleur J, Conner C. A review of diabetes treatment adherence and the
association with clinical and economic outcomes. Clin Ther. 2011;33(1):74-109.
20. Andrew PY, Yanni FY, Nichol MB. Estimating the effect of medication adherence on
health outcomes among patients with type 2 diabetes—an application of marginal
structural models. Value Health. 2010;13(8):1038-1045.
21. Wei W, Pan C, Xie L, Baser O. Real-world insulin treatment persistence among patients
with type 2 diabetes. Endocr Pract. 2014;20(1):52-61.
10

22. Carls GS, Tuttle E, Tan R-D, et al. Understanding the gap between efficacy in
randomized controlled trials and effectiveness in real-world use of GLP-1 RA and DPP-4
therapies in patients with type 2 diabetes. Diabetes Care. 2017;40(11):1469-1478.
23. Khunti K, Seidu S, Kunutsor S, Davies M. Association between adherence to
pharmacotherapy and outcomes in type 2 diabetes: a meta-analysis. Diabetes Care.
2017;40(11):1588-1596.
24. Holman RR, Paul SK, Bethel MA, Matthews DR, Neil HAW. 10-year follow-up of
intensive glucose control in type 2 diabetes. N Engl J Med. 2008;359(15):1577-1589.
25. Duckworth W, Abraira C, Moritz T, et al. Glucose control and vascular complications in
veterans with type 2 diabetes. N Engl J Med. 2009;360(2):129-139.
26. Group AC. Intensive blood glucose control and vascular outcomes in patients with type 2
diabetes. N Engl J Med. 2008;358(24):2560-2572.
27. Ismail-Beigi F, Craven T, Banerji MA, et al. Effect of intensive treatment of
hyperglycaemia on microvascular outcomes in type 2 diabetes: an analysis of the
ACCORD randomised trial. The Lancet. 2010;376(9739):419-430.
28. Rajpathak SN, Rajgopalan S, Engel SS. Impact of time to treatment intensification on
glycemic goal attainment among patients with type 2 diabetes failing metformin
monotherapy. J Diabetes Complications. 2014;28(6):831-835.
29. Davis J, Chavez B, Juarez DT. Adjustments to diabetes medications in response to
increases in hemoglobin A1c: an epidemiologic study. Ann Pharmacother.
2014;48(1):41-47.
11

30. Lin J, Zhou S, Wei W, Pan C, Lingohr-Smith M, Levin P. Does clinical inertia vary by
personalized A1c goal? A study of predictors and prevalence of clinical inertia in a US
managed-care setting. Endocr Pract. 2015;22(2):151-161.
31. Pantalone KM, Wells BJ, Chagin KM, et al. Intensification of diabetes therapy and time
until A1C goal attainment among patients with newly diagnosed type 2 diabetes who fail
metformin monotherapy within a large integrated health system. Diabetes Care.
2016;39(9):1527-1534.
32. Zhang Q, Rajagopalan S, Marrett E, Davies M, Radican L, Engel S. Time to treatment
initiation with oral antihyperglycaemic therapy in US patients with newly diagnosed type
2 diabetes. Diabetes, Obesity and Metabolism. 2012;14(2):149-154.
33. Khunti K, Gomes MB, Pocock S, et al. Therapeutic inertia in the treatment of
hyperglycaemia in patients with type 2 diabetes: a systematic review. Diabetes, Obesity
and Metabolism. 2018;20(2):427-437.
34. Ruiz-Negrón N, Wander C, McAdam-Marx C, Pesa J, Bailey RA, Bellows BK. Factors
Associated with Diabetes-Related Clinical Inertia in a Managed Care Population and Its
Effect on Hemoglobin A1c Goal Attainment: A Claims-Based Analysis. Journal of
managed care & specialty pharmacy. 2019;25(3):304-313.
35. Paul SK, Klein K, Thorsted BL, Wolden ML, Khunti K. Delay in treatment
intensification increases the risks of cardiovascular events in patients with type 2
diabetes. Cardiovasc Diabetol. 2015;14(1):100.
36. Berkowitz SA, Krumme AA, Avorn J, et al. Initial choice of oral glucose-lowering
medication for diabetes mellitus: a patient-centered comparative effectiveness study.
JAMA internal medicine. 2014;174(12):1955-1962.
12

37. Lo-Ciganic W-H, Donohue JM, Jones BL, et al. Trajectories of diabetes medication
adherence and hospitalization risk: a retrospective cohort study in a large state Medicaid
program. J Gen Intern Med. 2016;31(9):1052-1060.
38. Gatwood J, Chisholm‐Burns M, Davis R, et al. Differences in health outcomes associated
with initial adherence to oral antidiabetes medications among veterans with
uncomplicated Type 2 diabetes: a 5‐year survival analysis. Diabet Med.
2018;35(11):1571-1579.
39. Simpson SH, Lin M, Eurich DT. Medication adherence affects risk of new diabetes
complications: a cohort study. Ann Pharmacother. 2016;50(9):741-746.

13

CHAPTER 2. Treatment Duration Among Newly Treated T2D Patients and
Its Impact on Patient Outcomes

2.1 Methods
2.1.1 Study population
We identified adults aged >= 18 years who initiated treatment with metformin, a
sulfonylurea ((glimepiride, glipizide, glyburide, tolazamide, tolbutamide), or a thiazolidinedione
(TZD) (pioglitazone, rosiglitazone), or combinations of these drugs. Patients were required to
have a minimum of 1 year of continuous enrollment before (pre-period) and after (post-period)
the first fill date of their first anti-diabetic medication (ADM) and have at least one diagnosis for
T2D (ICD-9-CM 250.x0, 250.x2; ICD-10-CM E11). In order to identify incidence users, patients
were required to be free of any prescriptions for T2D drugs in the pre-period. Patients who were
diagnosed with type 1 diabetes or gestational diabetes, had a prescription for insulin during the
pre-period, were pregnant, or had complications of pregnancy, childbirth, and the puerperium
were excluded.
2.1.2 Definition of an initial treatment regimen
Individual ADMs were grouped into one of the following drug classes: metformin,
sulfonylurea, or TZDs. A patient’s initial treatment regimen was defined based on the class or
classes of medications used in the 6-week period starting with their earliest ADM medication fill
date. This 6-week window during which to define a patient’s initial ADM treatment regimen
captures the frequent adjustments that are made to an initial treatment attempt that are common
in real world practice. The patient’s initial treatment regimen is then defined by all ADMs used
by the patient at the end of the 6 week treatment adjustment period. The first ADM class that was
14

taken by patients >= 6 weeks was the index drug class. Patients using only one drug class during
the 6-week adjustment period are definition as monotherapy patients. The index date was defined
as the first fill date of the first ADM used. Patients with initial treatment regimens which
included insulin, SGLT2 inhibitors, DPP4 inhibitors, or GLP-1 RAs were excluded from the
analysis of initial treatment regimens as these patients are unlikely to be treatment naïve patients.
Patients baseline characteristics were reported by initial drug therapy outcomes (Table
2.1). Categorical variables were summarized by frequency and percentage. Continuous variables
were reported by mean and standard deviation (SD). Statistical comparisons among different
drug cohorts were evaluated using chi-square tests for categorical variables and ANOVA tests
for continuous variables.
This study’s analysis of the newly treated T2D patients is divided into two parts. First, we
documented the factors which are correlated with treatment duration and insulin initiation. The
second set of analyses evaluate the impact of duration of continuous antidiabetic treatment on the
risk of significant clinical events related to diabetes and health care costs measured over the first
year following the index date.
Part 1. Predictors of duration of therapy following the initial regimen
2.1.3 Key explanatory variables (part 1)
The primary independent variables of interest are the drug classes used by the patient as
initial therapy and are classified into: 1) metformin alone; 2) sulfonylurea alone; 3) TZD alone;
4) metformin plus sulfonylurea; 5) metformin plus TZD; 6) sulfonylurea plus TZD; and 7)
metformin, sulfonylurea, and TZD. Metformin alone was the reference group.
15

2.1.4 Outcomes and covariates (part 1)
The primary outcomes were duration of treatment and time to discontinuation. Duration
of treatment was primarily measured as days from the index date to treatment termination, or
starting a substantial treatment gap, whichever occurred first. We defined the gap as more than
60 days without any ADM supply available to the patient. ADMs used as a second treatment
attempt include any drugs from the following classes: metformin, sulfonylurea, TZDs, SGLT-2
inhibitors, DPP-4 inhibitors, GLP-1 RAs, and insulin.
The covariates used in the statistical models included age, gender, health plan type, index
year, whether a baseline A1C test was conducted, diagnostic history (a comprehensive diagnostic
profile was created using all diagnoses reported on the patient’s paid claims for services used in
the pre-period), history of prescription drug use (mix of all non-diabetic medications filled by the
patient in the 1-year prior to index date broken down by AHFS pharmacologic-therapeutic
classes), previous hospital admission, and healthcare costs at baseline, measured over the one-
year before the patient’s index date. A subset of patients for whom a baseline A1c test was
conducted will have these lab results recorded in the data. These patients will be used for a
sensitivity analyses which will evaluate the actual value of the patient’s baseline A1c impact
outcomes or changes any estimated effects in the primary analysis.
2.1.5 Statistical analyses (part 1)
OLS models were used to model the association between index ADMs and duration of
treatment. A Cox model was also use to document the factors associated with time to
discontinuation. All models included individual variables for all baseline patient characteristics
listed in Table 2.1.
16

2.1.6 Sensitivity analyses (part 1)
Sensitivity analyses were performed using a subset of patients who had recorded
laboratory value for their baseline A1C to estimate the association between patient characteristics
and treatment duration.
Part 2. Impact of duration of antidiabetic treatment on risk of clinical events and health care
costs
2.1.7 Key explanatory variables (part 2)
The primary independent variable of interest was duration of antidiabetic treatment,
which was calculated as days from the index date to treatment discontinuation, death, or last
available fill date, whichever occurred first. We defined discontinuation as a gap of more than 60
days without any ADM supply available to the patient. ADMs include any drugs from the
following classes: metformin, sulfonylurea, TZDs, SGLT-2 inhibitors, DPP-4 inhibitors, GLP-1
RAs, and insulin. We also measured PDC over the first year following the initial treatment.
2.1.8 Outcomes and covariates (part 2)
Clinical outcomes were measured as time to clinical events associated with T2D,
measured in days from the index date. The primary outcome was time to any of the following
events: stroke, or myocardial infarction (MI). We also examined each of these events
individually as secondary outcomes. Other secondary outcomes included time to hospitalization,
either CVD related or all cause hospitalizations.
Economic outcomes were measured using cost measured over the first year following the
index date broken down by type of service.
17

The covariates used in the statistical models included age, gender, health plan type, index
year, if baseline A1C recorded, diagnostic history (a comprehensive diagnostic profile was
created using all diagnoses reported on the patient’s paid claims for services used in the pre-
period), history of prescription drug use (mix of all non-diabetic medications filled by the patient
in the 1-year prior to index date broken down by AHFS pharmacologic-therapeutic classes),
previous hospital admission, and healthcare costs at baseline, measured over the one-year before
the patient’s index date.
2.1.9 Statistical analyses (part 2)
To account for the time-varying treatment duration, we used a time-varying multivariable
Cox model to examine the association between duration of antidiabetic treatment and risk of
developing cardiovascular conditions and other T2D related events. We altered the days of
duration by diving its value by 30 to transform the estimated coefficient from the Cox model into
an estimate of the effect of adhering for an additional month. We chose a 30-day time window
because a 30-day supply is the most common quantity dispensed of ADMs. All models included
individual variables for all baseline patient characteristics listed in Table 2.1.
For economic outcomes, we use multivariable OLS and generalized linear models
(GLMs) to assess the relationship between treatment duration during the first year and all-cause
healthcare expenditures in the 1-year post-period. The Modified Park test was used to determine
family distribution and corresponding link function. Cost outcomes included total healthcare
costs and are also broken down into outpatient, inpatient, and prescription drug costs. Treatment
duration was measured within this 1-year post-period. All cost outcomes were individually
assessed. All models included individual variables for all baseline patient characteristics listed in
Table 2.1.
18

2.2 Results
2.2.1 Baseline patient characteristics for patients beginning their initial drug therapy for T2D
A total of 315,515 patients met the inclusion and exclusion criteria for this study (Figure
2.1). The most significant inclusion/exclusion criteria were the requirement for 2 years of data
around the index date and the restriction of episodes to those using metformin, a sulfonylurea or
a TZD.
The most important result documented in Figure 2.1 is the distribution of patients across
the possible drug therapy outcomes. Nearly half of all patients stayed on the initial treatment
until their data expire (average duration of 681 days), and another 8% change therapies with no
break in treatment (average duration 722 days). The remaining 44% of patients discontinue their
initial therapy (mean duration 397 days). The risk of adverse events for patients who discontinue
therapy is of concern. While most patients who discontinue therapy restarted treatment, over
25% of these patients did so after experiencing a stroke, AMI or hospitalization. A similar
pattern of event risk exists for patients who discontinue treatment with no evidence of a second
treatment attempt (16.5% of patients discontinuing treatment). Over 28% of these patients
(n=6,570) experienced an adverse clinical event after discontinuing treatment.
Baseline characteristics of patients initiating their first treatment broken down by the drug
therapy outcome achieved by the patient (Table 2.1). The baseline characteristics varied
significantly across the drug therapy outcomes. Metformin monotherapy was the most commonly
used treatment in the initial regimen. However, among patients who switched to or added a new
drug without a break in therapy, 73% initiated with combination therapy using metformin and
sulfonylurea. Patients who discontinued the initial treatment were significantly younger, had
19

lower baseline healthcare costs, were less likely to have been hospitalized in the year prior to
starting their initial treatment episode than other patients.
2.2.2 Predictors of duration of continuous therapy
The multivariable analyses of duration of therapy are displayed in Table 2.2. Two
different analytic approaches are reported: an OLS estimate of days of continuous therapy and a
Cox model of the likelihood of discontinuation. Patients starting with sulfonylurea and TZD
monotherapy were associated with significantly fewer days of continuous therapy than those
with metformin monotherapy but were also estimated to have a lower probability of
discontinuing therapy. Patients with all types of combination therapy were associated with
significantly more days of therapy and lower risk of discontinuation. Patients who were older,
enrolled with an HMO, did not have prior hospitalization, had lower prior medical costs, and had
higher prior drug costs were associated with significantly longer duration of therapy and lower
risk of discontinuation.
The results in Table 2.2 comparing the effects of alternative initial drug regimens may be
sensitivity to the baseline A1c level of the patients in each drug therapy class. The sensitivity
analyses of days of therapy and likelihood of discontinuation presented in Tables 2.3a and 2.3b
are very consistent with the primary analysis of therapy duration though the effect size of
specific initial drug regimens were attenuated. We found that higher baseline A1C was
associated with significantly longer duration of therapy and a lower risk of discontinuation.
2.2.3 The impact of duration of drug therapy on the risk of clinical events and cost over the first
year following the initiation of T2D drug therapy

20

Results from the multivariable analyses of event risk, adjusting for other risk factors, are
summarized in Table 2.4. A One month increase in treatment duration decreased the hazard rate
of stroke or AMI by 3.3% (an absolute risk reduction (ARR) of 0.363% per additional month of
therapy), stroke by 3.4% (ARR of 0.289% per month of duration), AMI by 2.5% (ARR of
0.088% per month of duration), all-cause hospitalization by 3% (ARR of 0.816% per month),
and CVD-related hospitalization by 2.5% (ARR of 0.205% per month).
The impact of duration on patient outcomes was re-estimated considering only those
events that occurred within one year of the index date for initial therapy. A Cox model events
within one year as a function of month of therapy up to one year were compared with the effects
of one-year PDC on the likelihood of an event at any time in the first year. For one-year PDC,
we had 44.2% of patients having the PDC >=80%. We found that one month increase in duration
in the first year was associated with significantly lower likelihood of having each the clinical
outcomes. However, the logistic model of events found that patients with a PDC> 80% were
more likely to experience an event in the first year. This may indicate that patients return to
therapy after and event which occurred here in over 6% of all patients. We also estimate a Cox
model using events during the first year and found results similar to the logit model with duration
as the variable measuring persistence (Table 2.5). The effects in the first year became larger.
The relationship between treatment duration and health care costs was estimated using
difference-in-difference models and OLS estimation methods (Table 2.6). One month increase in
the treatment duration was associated with a significant reduction in the year-to-year change in
medical costs of $150. The year-to-year increase in prescription drug costs were higher per
month of persistence by $231.

21

The relationship between treatment duration and health care costs was estimated using
GLMs and OLS models, respectively (Table 2.7 and Table 2.8). One month increase in the
treatment duration was associated with a significantly reduced outpatient costs of $49.39 and
reduced inpatient costs of $118.72 in the first year following the index date. The prescription
drug costs for these patients increased by $293.96 as each one month increase in treatment
duration. Similar results were found using OLS models.
2.3 Discussion
Lifestyle change in combination with metformin monotherapy is generally recommended
as a preferred first-line treatment for patients with T2D. Due to the progressive nature of T2D,
most patients will require a combination of medications over time. This progression to more
complex treatment regimens and disease management protocols are challenging for patients.
Non-adherence and non-persistence to prescribed ADMs are common and remain a barrier to
optimal health outcomes. For example, a meta-analysis of 40 studies found that only 67.9% of
patients having an MPR>=80%, and 56.2% of patients being persistent with oral ADMs
1
. A
systematic review of 27 studies that evaluated adherence rates to ADMs revealed that only 22%
of studies reported >=80% adherence among patients
2
. Our study found that 44.3% of patients
discontinued treatment with an average duration of therapy just over one year following the
initial treatment.
McGovern et al. used a systematic review to compare medication adherence across
different ADMs and reported that sulfonylureas and TZDs had better adherence (MPR or PDC)
compared with metformin. However, several other studies found treatment persistence at one
year was lower with sulfonylureas and TZDs than with metformin
3-5
. The present study found
mixed results. Patients starting with sulfonylurea or TZD monotherapy were associated with
22

shorter duration of continuous therapy than those initiating with metformin monotherapy, but
were also found to exhibit a lower likelihood of discontinuing treatment. This study also
confirmed that prescribing patterns were consistent with current clinical guidelines that
recommend metformin as the preferred first-line treatment.
Prior research on the relationship between polypharmacy and adherence has been
inconclusive, with some studies reporting that adherence is inversely related to the number of
medications used
6-8
and other studies finding that there is no such relationship
9,10
. We found
that patients starting with a combination therapy have longer duration of therapy than patients
with monotherapy and are less likely to discontinue treatment. In a sensitivity analysis, we also
found that an increased A1c at baseline was correlated with significantly longer duration of
therapy. However, including baseline A1c as an independent variable did not significantly
change the estimated effects of alternative regimens to metformin, including combination
therapy. Initiation of combination therapy and higher baseline A1c would be indicators of more
advanced or poorly controlled T2D. It is possible that these patients may be more likely to
experience symptoms of poor glycemic control and/or may receive more intensive educational
interventions than patients with relatively less complicated disease.
Persistence to treatment is a key element associated with the effectiveness of
pharmacological therapies in patients with T2D. Similar to earlier diabetes studies
11-18
, our study
shows that patients with longer duration of therapy following the initial treatment are at reduced
risk of stroke or AMI, and hospitalization. Importantly, these findings advance our understanding
of the impact of non-persistence by suggesting that patients with early tight diabetes control may
have both early and long-term benefits. In our analysis, a one month increase in treatment
duration reduced risk of AMI, stroke and hospitalization within the study period. Our results
23

highlight the importance of establishing the role of treatment persistence in preventing
complications earlier, from the onset of ADMs. A long-term observational follow-up of the UK
Prospective Diabetes Study (UKPDS) suggested that early glycemic control may have durable
effects in reducing risk of T2D-related complications
19
. A recent analysis of newly diagnosed
T2D patients also found that glycemic control during the first year after diagnosis was
significantly associated with a decrease in future risk for diabetic complications and mortality
20
.
Taken together, these findings suggest that treating T2D patients more intensively at an earlier
stage may result in long-term improvements in patient health outcomes. Prior research also
suggest that medication non-persistence is a behavior that is challenging to correct once
established
21,22
. Therefore, healthcare professionals should work with patients from the
beginning of therapy to ensure that barriers to persistence are identified and reduced in order to
limit the impact of non-persistence on poor future outcomes.
Our study shows that patients with longer duration of therapy had lower medical costs
during the first year following their initial treatment. Consistent with our findings, research
generally has demonstrated that medication adherence was inversely associated with costs of
outpatient care, ER visits, and hospitalization
23-25
. As expected, we did find that prescription
drug costs increased with longer duration of therapy. However, this increase in prescription drug
costs was larger than offsetting savings in medical costs during the first year of treatment.
Improved medication persistence may significantly impact long term health care costs as each
month of persistence significantly reduced the risk of major cardiovascular events and
hospitalization. As T2D progresses, the risk of complications increases and cost savings
associated with reductions in the risk of complications would be expected to be more significant.
Jha, et al. suggested that reduced risk of hospitalization or emergency department visits because
24

of improved medication adherence was projected to save $4.7 billion annually
26
. In a large VA
cohort study which examined the longitudinal effects of medication adherence on potential cost
savings
27
, higher adherence was associated with higher pharmacy costs over the five years of
treatment which were offset by $661 million to $1.16 billion annual cost savings.
Limitations
Our study was subject to several limitations. Studies using real-world data to compare
alternative treatments may suffer from treatment selection bias. Patients who initiated different
treatment regimens may also be different in other important but unmeasured ways. Patients who
are persistent with their medications may be different than non-persistent patients. The primary
defense against potential treatment selection bias is to more fully document the observable
characteristics of the patients in the study. This was achieved by including the multiple baseline
characteristics in our statistical models, including a complete diagnostic and prescription drug
profiles based at baseline. Second, reasons for treatment discontinuation or any additional health
benefits (e.g., reduced symptom severity) resulting from increased persistence could not be
examined using claims databases. Finally, death caused by CVD is an important outcome for
patients with T2D that should be considered. We could not estimate the association between
treatment duration and CVD-related mortality since that mortality information was not available
in the database.
2.4 Conclusions
Longer continuous duration of antidiabetic medications among newly treated patients
with T2D can significantly improve clinical outcomes at any post-index time and 1-year post-
index medical costs. These results emphasize the importance of persistence for patients with T2D
and the need to assist them with remaining persistent from the outset of therapy.
25

Chapter 2 References
1. Iglay K, Cartier SE, Rosen VM, et al. Meta-analysis of studies examining medication
adherence, persistence, and discontinuation of oral antihyperglycemic agents in type 2
diabetes. Curr Med Res Opin. 2015;31(7):1283-1296.
2. Krass I, Schieback P, Dhippayom T. Adherence to diabetes medication: a systematic
review. Diabet Med. 2015;32(6):725-737.
3. Grimes RT, Bennett K, Tilson L, Usher C, Smith SM, Henman MC. Initial therapy,
persistence and regimen change in a cohort of newly treated type 2 diabetes patients. Br J
Clin Pharmacol. 2015;79(6):1000-1009.
4. Jermendy G, Wittmann I, Nagy L, et al. Persistence of initial oral antidiabetic treatment
in patients with type 2 diabetes mellitus. Medical science monitor: international medical
journal of experimental and clinical research. 2012;18(2):CR72.
5. Plat A, Penning-van Beest F, Kessabi S, Groot M, Herings R. Change of initial oral
antidiabetic therapy in type 2 diabetic patients. Pharm World Sci. 2009;31(6):622.
6. Cramer JA. A systematic review of adherence with medications for diabetes. Diabetes
Care. 2004;27(5):1218-1224.
7. Emslie-Smith A, Dowall J, Morris A. The problem of polypharmacy in type 2 diabetes.
The British Journal of Diabetes & Vascular Disease. 2003;3(1):54-56.
8. Dailey G, Kim MS, Lian JF. Patient compliance and persistence with antihyperglycemic
drug regimens: evaluation of a medicaid patient population with type 2 diabetes mellitus.
Clin Ther. 2001;23(8):1311-1320.
9. Jin J, Sklar GE, Oh VMS, Li SC. Factors affecting therapeutic compliance: A review
from the patient’s perspective. Ther Clin Risk Manag. 2008;4(1):269.
26

10. Grant RW, Devita NG, Singer DE, Meigs JB. Polypharmacy and medication adherence in
patients with type 2 diabetes. Diabetes Care. 2003;26(5):1408-1412.
11. Ho PM, Rumsfeld JS, Masoudi FA, et al. Effect of medication nonadherence on
hospitalization and mortality among patients with diabetes mellitus. Arch Intern Med.
2006;166(17):1836-1841.
12. Asche C, LaFleur J, Conner C. A review of diabetes treatment adherence and the
association with clinical and economic outcomes. Clin Ther. 2011;33(1):74-109.
13. Andrew PY, Yanni FY, Nichol MB. Estimating the effect of medication adherence on
health outcomes among patients with type 2 diabetes—an application of marginal
structural models. Value Health. 2010;13(8):1038-1045.
14. Zhu VJ, Tu W, Rosenman MB, Overhage JM. Nonadherence to Oral Antihyperglycemic
Agents: Subsequent Hospitalization and Mortality among Patients with Type 2 Diabetes
in Clinical Practice. Paper presented at: MedInfo2015.
15. Huber CA, Rapold R, Brüngger B, Reich O, Rosemann T. One-year adherence to oral
antihyperglycemic medication and risk prediction of patient outcomes for adults with
diabetes mellitus: An observational study. Medicine. 2016;95(26).
16. Feldman BS, Cohen-Stavi CJ, Leibowitz M, et al. Defining the role of medication
adherence in poor glycemic control among a general adult population with diabetes.
PLoS One. 2014;9(9).
17. Trinacty CM, Adams AS, Soumerai SB, et al. Racial differences in long-term adherence
to oral antidiabetic drug therapy: a longitudinal cohort study. BMC Health Serv Res.
2009;9(1):24.
27

18. Gatwood J, Chisholm‐Burns M, Davis R, et al. Differences in health outcomes associated
with initial adherence to oral antidiabetes medications among veterans with
uncomplicated Type 2 diabetes: a 5‐year survival analysis. Diabet Med.
2018;35(11):1571-1579.
19. Holman RR, Paul SK, Bethel MA, Matthews DR, Neil HAW. 10-year follow-up of
intensive glucose control in type 2 diabetes. N Engl J Med. 2008;359(15):1577-1589.
20. Laiteerapong N, Ham SA, Gao Y, et al. The legacy effect in type 2 diabetes: impact of
early glycemic control on future complications (the Diabetes & Aging Study). Diabetes
Care. 2019;42(3):416-426.
21. Gatwood JD, Chisholm-Burns M, Davis R, et al. Disparities in initial oral antidiabetic
medication adherence among veterans with incident diabetes. Journal of managed care &
specialty pharmacy. 2018;24(4):379-389.
22. Saundankar V, Peng X, Fu H, et al. Predictors of change in adherence status from 1 year
to the next among patients with type 2 diabetes mellitus on oral antidiabetes drugs.
Journal of managed care & specialty pharmacy. 2016;22(5):467-482.
23. Nasseh K, Frazee SG, Visaria J, Vlahiotis A, Tian Y. Cost of medication nonadherence
associated with diabetes, hypertension, and dyslipidemia. Am J Pharm Benefits.
2012;4(2):e41-e47.
24. Roebuck MC, Liberman JN, Gemmill-Toyama M, Brennan TA. Medication adherence
leads to lower health care use and costs despite increased drug spending. Health Aff
(Millwood). 2011;30(1):91-99.
28

25. Breitscheidel L, Stamenitis S, Dippel F-W, Schöffski O. Economic impact of compliance
to treatment with antidiabetes medication in type 2 diabetes mellitus: a review paper. J
Med Econ. 2010;13(1):8-15.
26. Jha AK, Aubert RE, Yao J, Teagarden JR, Epstein RS. Greater adherence to diabetes
drugs is linked to less hospital use and could save nearly $5 billion annually. Health Aff
(Millwood). 2012;31(8):1836-1846.
27. Egede LE, Gebregziabher M, Dismuke CE, et al. Medication nonadherence in diabetes:
longitudinal effects on costs and potential cost savings from improvement. Diabetes
Care. 2012;35(12):2533-2539.
29

Figure 2.1. Selection of Patients

T2D patients with antidiabetic medications
(n = 4,144,805)
Initial antidiabetic drug class(es) had 6-week continuous
supply (index date identified)
(n = 3,854,667)
>=1 year pre-index and >=1 year post-index enrollment
(n = 584,754)
Aged >=18 years, non-pregnancy, cost<500K
(n = 525,858)
Initial antidiabetic treatment with metformin,
sulfonylurea, and TZDs
(n = 315,515)
Changed initial
Rx, with no
break: 25,794
(8.2%)
Ave. Duration:
722 Days
Discontinued
initial Rx:
139,724 (44.3%)
Ave. Duration:
397 Days
Stayed on initial,
no changes, till
data ran out:
149,997 (47.5%)
Ave. Duration:
681 Days
Discontinued initial Rx
and started second Rx:
116,616 (83.5%)
No Event: 71,568
Before event: 15,342
After event: 29,706

Discontinued initial Rx
with no 2
nd
Rx:
23,108 (16.5%)
Event post: 6,570
No post event: 16,538
30

Table 2.1. Patient Characteristics by Drug Therapy Outcome

Stay on Initial
Drug Regimen
Changed Initial
Regimen with No
Break
Discontinued
Initial Drug
Regimen

N. of patients: 149,997 47.5% 25,794 8.2% 139,724 44.3%

n/mean %/sd n/mean %/sd n/mean %/sd p value
Initial regimen <0.01
Metformin only 115,335 76.9% 609 2.4% 113,711 81.4%
Sulfonylurea only 9,927 6.6% 553 2.14% 8,215 5.9%
TZD only 1,853 1.3% 113 0.4% 1,114 0.8%
Metformin +
Sulfonylurea 18,112 12.1% 18,821 73.0% 13,141 9.4%
Metformin + TZD 2,917 1.9% 2,107 8.2% 2,280 1.6%
Sulfonylurea +
TZD 537 0.4% 478 1.9% 380 0.3%
All three drugs 1,316 0.9% 3,113 12.1% 883 0.6%
Healthcare Cost
($/year)
Medical costs
Year Prior to Index 5959.5 19349.0 5305.9 16840.5 5213.1 16356.5 <0.01
Year Post Index 6351.9 19298.1 5648.1 17084.9 5760.0 18195.3 <0.01
Drug costs
Year Prior to Index 9089.2 21546.3 6725.2 16174.6 7970.4 19007.7 <0.01
Year Post Index 11907.3 24758.4 11137.8 19287.0 10290.4 21599.3 <0.01
Total costs
Year Prior to Index 15048.7 30336.3 12031.1 24425.6 13183.5 26325.9 <0.01
Year Post Index 18259.1 33107.9 16785.9 27499.0 16050.4 29847.1 <0.01
Age 60.51 14.6 61.05 12.8 58.15 14.8 <0.01
Age group

18-34 years 9,108 6.1% 623 2.4% 10,621 7.6% <0.01
35-44 years 12,944 8.6% 2,248 8.7% 15,237 10.9%

45-54 years 25,927 17.3% 5,115 19.8% 27,854 19.9%

55-64 years 35,828 23.9% 6,351 24.6% 33,427 23.9%

65-79 years 53,116 35.4% 9,681 37.5% 43,643 31.2%

>=80 years 13,074 8.7% 1,776 6.9% 8,942 6.4%

Female 73,178 48.8% 15095 58.5% 64,946 46.5% <0.01
Health Plan

EPO 10,492 7.0% 1,825 7.1% 12,188 8.7% <0.01
HMO 39,964 26.6% 7,526 29.2% 34,875 25.0%

INDEMNITY
2,385 1.6% 382 1.5% 1,814 1.3%

POS 58,751 39.2% 9,264 35.9% 59,655 42.7%

PPO 9,117 6.1% 1445 5.6% 7,220 5.2%

OTHER 29,288 19.5% 5,352 20.8% 23,972 17.2%

A1c test at
baseline 41,186 27.5% 5,478 21.2% 35,610 25.5% <0.01
Prior
hospitalization 15,615 10.4% 2,699 10.5% 12,748 9.1% <0.01
Baseline
Diagnoses

Infectious diseases 19,698 13.1% 3542 13.7% 19,836 14.2% <0.01
Neoplasms 36,564 24.4% 5,270 20.4% 31,288 22.4% <0.01
31

Endocrine
disorders 123,742 82.5% 19,703 76.4% 114,005 81.6% <0.01
Blood diseases 22,617 15.1% 3,134 12.2% 19,241 13.8% <0.01
Mental disorders 4,0892 27.3% 5,590 21.7% 35,962 25.7% <0.01
Nervous system 70,289 46.9% 11,608 45.0% 64,117 45.9% <0.01
Eye diseases 14,524 9.7% 868 3.4% 6,129 4.4% <0.01
Ear diseases 5,316 3.5% 272 1.1% 2,461 1.8% <0.01
Circulatory
disorders 109,622 73.1% 19,037 73.8% 97,572 69.8% <0.01
Respiratory system 63,277 42.2% 9,658 37.4% 58,891 42.2% <0.01
Digestive system 50,244 33.5% 7,517 29.1% 44,805 32.1% <0.01
Genitourinary
system 62,573 41.7% 9,094 35.3% 57,500 41.2% <0.01
Skin diseases 46,650 31.1% 6,672 25.9% 40,723 29.2% <0.01
Musculoskeletal 82,664 55.1% 12,785 49.6% 74,862 53.6% <0.01
Congenital
anomalies 5,006 3.4% 687 2.7% 4,408 3.2% <0.01
Injury/Poisoning 34,323 22.9% 5,379 20.9% 31,172 22.3% <0.01
Overweight/obesity 34,222 22.8% 4,536 17.6% 30,975 22.2% <0.01
Tobacco use 11,613 7.7% 2,140 8.3% 11,082 7.9% <0.01
Baseline
Medications

Antihistamines
7,228 4.8% 1,296 5.0% 7,870 5.6% <0.01
Anti-infectives 77,113 51.4% 11,451 44.4% 73,034 52.3% <0.01
Antineoplastic 3,503 2.3% 402 1.6% 2,887 2.1% <0.01
Autonomic Drugs 38,385 25.6% 5,307 20.6% 34,539 24.7% <0.01
Anti-coagulation 15,192 10.1% 2,287 8.9% 11,243 8.1% <0.01
Cardiovascular 115,165 76.8% 19,619 76.1% 103,850 74.3% <0.01
Central Nervous
System 84,233 56.2% 12,546 48.6% 78,505 56.2% <0.01
Electrolytic 37,391 24.9% 5,499 21.3% 31,121 22.3% <0.01
Respiratory Tract 20,257 13.5% 2,676 10.4% 19,265 13.8% <0.01
Eye, Ear, Nose,
and Throat 35,258 23.5% 4,765 18.5% 31,713 22.7% <0.01
Gastrointestinal 41,803 27.9% 5,559 21.6% 35,513 25.4% <0.01
Skin and Mucous
Membrane 32,539 21.7% 4,430 17.2% 29,814 21.3% <0.01
Smooth Muscle
Relaxants 4,215 2.8% 582 2.3% 3,534 2.5% <0.01
Vitamins 7,488 5.0% 744 2.9% 7,910 5.7% <0.01
TZD=thiazolidinedione; EPO=exclusive provider organization; HMO=health maintenance
organization; POS=point-of-service; PPO=preferred provider organization; SD=standard
deviation.
32

Table 2.2. Patient characteristics impacting duration that patients stayed on any drug
among patients Initiating an Initial Episode of Drug Therapy for T2D
N = 315,515

Days of uninterrupted
therapy
Time to discontinuation
OLS model Cox model
Parameters coefficient P value HR p value
Intercept 742.00 <.0001

Diab drug initiated in
the 1st episode

Metformin only
Reference Group
Sulfonylurea only
-47.78 <.0001
0.959 0.0006
TZD only
-89.80 <.0001
0.774 <.0001
Metformin + Sulfonylurea
113.10 <.0001
0.962 <.0001
Metformin + TZD
63.28 <.0001
0.967 0.0373
Sulfonylurea + TZD
43.72 0.0023
0.968 0.382
All three drugs
205.63 <.0001
0.846 <.0001
A1c=yes 0.84 0.7058
1.047 <.0001
Index year -31.43 <.0001
0.969 <.0001
Female -2.64 0.1915
0.961 <.0001
Health plan (vs. PPO)

EPO -44.46 <.0001
1.101 <.0001
HMO 37.09 <.0001
0.94 <.0001
INDEMNITY 187.21 <.0001
0.808 <.0001
OTHER 97.16 <.0001
0.993 0.5627
POS -15.54 0.0006
1.036 0.0045
Age group (vs. >=80)

18-34 years -240.21 <.0001
1.907 <.0001
35-44 years -149.71 <.0001
1.562 <.0001
45-54 years -73.11 <.0001
1.319 <.0001
55-64 years -32.51 <.0001
1.163 <.0001
65-79 years 33.44 <.0001
1.056 <.0001
Prior hospitalization=1 -16.39 <.0001
0.963 0.0017
Prior medical costs (vs.
<=1000)

1000 - 5000 -8.16 0.0021
0.99 0.1569
5000 - 10000 -27.46 <.0001
1.023 0.0357
>10000 -35.13 <.0001
1.035 0.002
Prior drug costs (vs.
<=1000)

33

1000 - 5000 33.53 <.0001
0.929 <.0001
5000 - 10000 63.22 <.0001
0.859 <.0001
>10000 98.00 <.0001
0.759 <.0001
Baseline diagnoses

Infectious diseases -3.31 0.2563
1.023 0.0029
Neoplasms 13.50 <.0001
0.969 <.0001
Endocrine disorders 13.31 <.0001
1.015 0.0398
Blood diseases -24.46 <.0001
1.03 0.0002
Mental disorders -27.24 <.0001
1.008 0.2709
Nervous system 21.40 <.0001
0.985 0.0064
Eye diseases -103.58 <.0001
0.743 <.0001
Ear diseases -40.87 <.0001
0.886 <.0001
Circulatory disorders 6.21 0.0174
0.989 0.1103
Respiratory system -6.33 0.0054
1.007 0.2693
Digestive system -9.81 <.0001
1.03 <.0001
Genitourinary system -13.16 <.0001
1.032 <.0001
Skin diseases 6.28 0.0065
0.964 <.0001
Musculoskeletal -12.70 <.0001
1.034 <.0001
Congenital anomalies -6.58 0.0075
1.012 0.0845
Injury/Poisoning 15.67 0.0040
0.957 0.0033
Overweight/obesity 0.36 0.8835
1.016 0.0151
Tobacco use 15.97 <.0001
1.024 0.0242
Baseline drug profile

Antihistamines -22.56 <.0001
1.017 0.1487
Anti-infectives -16.65 <.0001
1.055 <.0001
Antineoplastic -24.35 0.0003
1.006 0.7394
Autonomic Drugs -17.47 <.0001
1.036 <.0001
Anti-coagulation -10.89 <.0001
0.979 0.0409
Cardiovascular 36.15 <.0001
0.961 <.0001
Central Nervous System
Agents
-24.43 <.0001
1.066 <.0001
Electrolytic agents 14.74 <.0001
0.936 <.0001
Respiratory Tract -14.47 <.0001
1.036 <.0001
Eye, Ear, Nose, and
Throat (EENT)
8.56 0.0004
1.011 0.0873
Gastrointestinal 1.14 0.6466
0.99 0.1363
Skin and Mucous
Membrane Agents
-8.14 0.0012
1.022 0.0017
Smooth Muscle Relaxants -19.41 0.0013
1.033 0.0536
Vitamins -38.97 <.0001
1.119 <.0001
TZD=thiazolidinedione; EPO=exclusive provider organization; HMO=health maintenance organization;
POS=point-of-service; PPO=preferred provider organization.
34

Table 2.3a. Sensitivity Analysis
Patient characteristics impacting duration that patients stayed on any drug
Primary OLS Model Sensitivity Analysis (OLS)
N = 315,515 N = 82,274
Parameters coefficient p value coefficient p value
Intercept 742.00 <.0001
718.72 <.0001
Diab drug initiated in
the 1st episode

Metformin only
Reference Group
Sulfonylurea only
-47.78 <.0001
-36.47 <.0001
TZD only
-89.80 <.0001
-55.07 0.0025
Metformin +
Sulfonylurea
113.10 <.0001
70.04 <.0001
Metformin + TZD
63.28 <.0001
47.05 <.0001
Sulfonylurea + TZD
43.72 0.0023
88.08 0.0042
All three drugs
205.63 <.0001
149.91 <.0001
A1c=yes VS A1c value 0.84 0.7058
10.58 <.0001
index year -31.43 <.0001
-34.05 <.0001
female -2.64 0.1915
0.51 0.8844
Health plan (vs. PPO)

EPO -44.46 <.0001
-55.43 <.0001
HMO 37.09 <.0001
20.00 0.0085
INDEMNITY 187.21 <.0001
147.58 <.0001
OTHER 97.16 <.0001
43.39 <.0001
POS -15.54 0.0006
-35.58 <.0001
Age group (vs. >=80)

18-34 years -240.21 <.0001
-223.90 <.0001
35-44 years -149.71 <.0001
-142.32 <.0001
45-54 years -73.11 <.0001
-78.49 <.0001
55-64 years -32.51 <.0001
-43.29 <.0001
65-79 years 33.44 <.0001
21.03 0.0041
TZD=thiazolidinedione; EPO=exclusive provider organization; HMO=health maintenance organization;
POS=point-of-service; PPO=preferred provider organization.

35

Table 2.3b. Sensitivity Analysis
Primary model analysis
(Cox)
Sensitivity Analysis (Cox)
N = 315,515 N = 82,274
Parameters HR p value HR p value
Diab drug initiated in the
1st episode

Metformin only
Reference group
Sulfonylurea only
0.959 0.0006 1.001 0.9785
TZD only
0.774 <.0001 0.673 <.0001
Metformin + Sulfonylurea
0.962 <.0001 1.067 <.0001
Metformin + TZD
0.967 0.0373 0.985 0.6606
Sulfonylurea + TZD
0.968 0.382 0.91 0.3031
All three drugs
0.846 <.0001 0.966 0.4142
A1c=yes VS A1c value
1.047 <.0001 0.958 <.0001
index year
0.969 <.0001 0.967 <.0001
female
0.961 <.0001 0.983 0.1264
Health plan (vs. PPO)

EPO
1.101 <.0001 1.079 0.0078
HMO
0.94 <.0001 0.951 0.0354
INDEMNITY
0.808 <.0001 0.813 0.0305
OTHER
0.993 0.5627 0.924 0.0019
POS
1.036 0.0045 1.005 0.8311
Age group (vs. >=80)

18-34 years
1.907 <.0001 1.764 <.0001
35-44 years
1.562 <.0001 1.456 <.0001
45-54 years
1.319 <.0001 1.271 <.0001
55-64 years
1.163 <.0001 1.139 <.0001
65-79 years
1.056 <.0001 1.012 0.6199
TZD=thiazolidinedione; EPO=exclusive provider organization; HMO=health maintenance organization;
POS=point-of-service; PPO=preferred provider organization.

36

Table 2.4. Impact of Duration on Clinical Event Risk
Cox Model
Logit Model of Event in 1
st
Year
Duration (months)
Duration
(up to 12 months)
1-Year PDC >= 80%
(44.2%)
Outcomes:
Hazard
ratio p value Odds ratio p value Odds ratio p value
Stroke or AMI 0.967 <.0001 0.718 <.0001 1.036 0.0009
Stroke 0.966 <.0001 0.721 <.0001 1.042 0.0213
AMI 0.975 <.0001 0.712 <.0001 1.005 0.0003
All-cause
hospitalization 0.970 <.0001 0.743 <.0001 0.86 <.0001
CVD-related
hospitalization 0.975 <.0001 0.742 <.0001 0.863 <.0001
AMI=Acute Myocardial Infarction; CVD=Cardiovascular disease.

37

Table 2.5. Impact of Duration on Clinical Event Risk (follow up capped at the first year)
Cox Model
Duration (capped at first year)
Outcomes: Hazard ratio p value
Stroke or AMI 0.711 <.0001
Stroke 0.717 <.0001
AMI 0.696 <.0001
All-cause
hospitalization 0.744 <.0001
CVD-related
hospitalization 0.742 <.0001
AMI=Acute Myocardial Infarction; CVD=Cardiovascular disease.

Table 2.6. Impact of Duration during the First Year on the Change in Health Care Costs by
Type of Service

Duration
(in months)

estimates p-value
Change in medical costs -$150 <.0001
Change in drug costs $231 <.0001
Change in total costs $80 <.0001
AMI=Acute Myocardial Infarction; CVD=Cardiovascular disease.

Table 2.7. Impact of Duration, PDC>=0.8 during the First Year on Health Care Costs by
Type of Service (GLMs)

Duration
(in months) PDC>=0.8

estimates p-value estimates p-value
Total health care costs $103.79 <.0001 $910.34 <.0001
Outpatient costs -$49.39 <.0001 -$442.62 <.0001
Inpatient costs -$118.72 <.0001 -$1043.86 <.0001
Prescription drug costs $293.96 <.0001 $2367.40 <.0001
Total costs = outpatient costs + inpatient costs + drug costs
38

Table 2.8. Impact of Duration, PDC>=0.8 during the First Year on Health Care Costs by
Type of Service (OLSs)

Duration
(in months) PDC>=0.8

estimates p-value estimates p-value
Total health care costs $95.98 <.0001 $855.76 <.0001
Outpatient costs -$50.77 <.0001 -$463.07 <.0001
Inpatient costs -$125.81 <.0001 -$1101.22 <.0001
Prescription drug costs $272.56 <.0001 $2420.05 <.0001
Total costs = outpatient costs + inpatient costs + drug costs

39

Chapter 2 Appendix

Time-varying treatment duration
The method used to assess the time-varying treatment duration is to examine the
Schoenfeld residuals
1,2
. The Schoenfeld residual computed with each covariate is defined at the
observed event times as the difference between covariate for observation and the weighted
average of the covariate values for all subjects still at risk when observation experiences the
event.

Table 2.A1. Results of tests for Schoenfeld residuals of treatment duration
Rh0 Chi2 df Prob>chi2
Duration (time to nonfatal stroke or AMI) 0.27847 20212.42 1 <.0001
Duration (time to nonfatal stroke) 0.27455 19475.42 1 <.0001
Duration (time to nonfatal AMI) 0.26177 17470 1 <.0001
Duration (time to all-cause hospitalization) 0.32080 27948.79 1 <.0001
Duration (time to CVD-related hospitalization) 0.27259 19030.06 1 <.0001
AMI=Acute Myocardial Infarction; CVD=Cardiovascular disease; df=degree of freedom; Prob=probability.

As summarized in table S1, the correlations between the Schoenfeld residuals of
treatment duration and survival times were all significant, indicating that treatment duration was
a time-varying exposure variable. Therefore, time-varying Cox models should be used to account
for this.
Family distribution and link function in GLMs
The family distribution and corresponding link function used in GLMs were determined
by Modified Park Test. Test coefficient close to 0: Gaussian NLLS distribution; close to 1:
Poisson distribution; close to 2: Gamma distribution; close to 3: Inverse Gaussian or Wald
3
.
Table 2.A2. Modified Park Test (duration in months)
Expenditures l Distribution used Link function used
Total expenditures 1.194382 Poisson Log
Outpatient expenditures 1.352503 Poisson Log
40

Inpatient expenditures 1.153771 Poisson Log
Drug expenditures 1.12534 Poisson Log

41

Chapter 2 Appendix references
1. Grambsch PM, Therneau TM. Proportional hazards tests and diagnostics based on
weighted residuals. Biometrika. 1994;81(3):515-526.
2. Zhang Z, Reinikainen J, Adeleke KA, Pieterse ME, Groothuis-Oudshoorn CG. Time-
varying covariates and coefficients in Cox regression models. Annals of translational
medicine. 2018;6(7).
3. Manning WG, Mullahy J. Estimating log models: to transform or not to transform? J
Health Econ. 2001;20(4):461-494.

42

CHAPTER 3. Treatment Duration Among Previously Treated T2D Patients
and Its Impact on Patient Outcomes

3.1 METHODS
3.1.1 Study population
From our original sample of 315,515 adult patients beginning their initial treatment
regimen, we identified patients who initiated the second treatment attempt. All patients had a
minimum of 1 year of continuous enrollment before (pre-period) starting their initial regimen
and were required to have one year of data after (post-period) the first fill date of their second
anti-diabetic medication (ADM). As before, patients were required to have at least one diagnosis
for T2D (ICD-9-CM 250.x0, 250.x2; ICD-10-CM E11). Patients who were diagnosed with type
1 diabetes or gestational diabetes, had a prescription for insulin during the pre-period, were
pregnant, or had complications of pregnancy, childbirth, and the puerperium were excluded.
3.1.2 Definition of the second episode of treatment
The first fill date of the second episode was the index date for this analysis. Patients who
initiated any additional drug classes within the one week of the index date of the second episodes
are definition as combination therapy patients. Individual ADMs used in the patient’s second
treatment attempt were grouped into one of the following drug classes: metformin, sulfonylurea
(glimepiride, glipizide, glyburide, tolazamide, tolbutamide), thiazolidinedione (TZD)
(pioglitazone, rosiglitazone), SGLT2-inhibitors (canagliflozin, dapagliflozin, empagliflozin,
ertugliflozin), DPP4-inhibitors (sitagliptin, saxagliptin, linagliptin, alogliptin), GLP-RA
43

(exenatide, lixisenatide, liraglutide, dulaglutide, semaglutide, albiglutide), or combinations of
these drugs.
This study’s analysis of T2D patients with second treatment attempt is divided into two
parts. First, we documented the factors which are correlated with treatment duration. The second
set of analyses evaluate the impact of duration of continuous antidiabetic treatment on the risk of
significant clinical events related to diabetes and health care costs measured over the first year
following the index date of the second episode.
Part 1. Predictors of duration of therapy following the second episode of treatment
3.1.3 Key explanatory variables (part 1)
Duration of treatment during the patient’s second treatment attempt will likely depend on
their treatment history from their initial episode. To capture this, we defined three types of
second episode of treatment:
• Augmentation: the second episode was added to the initial regimen which was continued
for at least 60 days;
• Switch: switched to the second episode from the initial treatment
o Switch without a break: the second episode was initiated within 60 days following
the discontinuation of the initial treatment
o Switch with a break: the second episode was initiated > 60 days following the
discontinuation of the initial treatment
• Restart: Initiation of a second treatment attempt using the previous drug regimen.

44

The other independent variables of interest are the number of days off therapy following
the end of the initial treatment and the number of days that the patient stayed on their initial
treatment regimen without break in therapy.
3.1.4 Outcomes and other covariates (part 1)
In the analysis of factors associated with treatment duration, the primary outcome is
duration of treatment measured as days from the index date of the second episode to the date all
treatment terminated, or to the start of a treatment gap of more than 60 days, whichever occurred
first. This method allows for patients switching medications without a break in therapy to have
their count of uninterrupted therapy to include persistence on their new therapy.
The covariates used in the statistical models included age, gender, health plan type, a
time trend based on the index year, whether a baseline A1c test was conducted, diagnostic
history, prescription drug profile at baseline reflecting the AHFS pharmacologic-therapeutic
classes used by the patient, and healthcare costs in the measured over the one-year before the
patient’s index date.
3.1.5 Statistical analyses (part 1)
OLS models were used to model the association between index ADMs and duration of
treatment. A Cox model was also use to document the factors associated with time to
discontinuation. All models included individual variables for all baseline patient characteristics
listed in Table 3.1.
3.1.6 Sensitivity analyses (part 1)
A subset of patients for whom a baseline A1c test was conducted will have these lab
results recorded in the data. These patients will be used for a sensitivity analyses which will
45

evaluate the actual value of the patient’s baseline A1c impact outcomes or changes any estimated
effects in the primary analysis.
Part 2. Impact of duration of antidiabetic treatment on risk of clinical events and health care
costs
3.1.7 Key explanatory variables (part 2)
The primary independent variable of interest was duration of antidiabetic treatment,
which was calculated as days from the index date to treatment discontinuation, death, or last
available fill date, whichever occurred first. We defined discontinuation as a gap of more than 60
days without any ADM supply available to the patient. ADMs include any drugs from the
following classes: metformin, sulfonylurea, TZDs, SGLT-2 inhibitors, DPP-4 inhibitors, GLP-1
RAs, and insulin. We also measured PDC over the first year following the second treatment
attempt.
3.1.8 Outcomes and covariates (part 2)
Clinical outcomes were measured as time to clinical events associated with T2D,
measured in days from the index date. The primary outcome was time to any of the following
events: stroke, or myocardial infarction (MI). We also examined each of these events
individually as secondary outcomes. Other secondary outcomes included time to hospitalization,
either CVD related or all cause hospitalizations.
Economic outcomes were measured using cost measured over the first year following the
index date broken down by type of service.
The covariates used in the statistical models included age, gender, health plan type, index
year, type of the second episode, continuous duration of the initial treatment, days off therapy
46

following the end of the initial treatment, prior history of stroke or AMI, if baseline A1C
recorded, diagnostic history (a comprehensive diagnostic profile was created using all diagnoses
reported on the patient’s paid claims for services used in the pre-period), history of prescription
drug use (mix of all non-diabetic medications filled by the patient in the 1-year prior to index
date broken down by AHFS pharmacologic-therapeutic classes), and healthcare costs at baseline,
measured over the one-year before the patient’s index date.
3.1.9 Statistical analyses (part 2)
To account for the time-varying treatment duration, we used a time-varying multivariable
Cox model to examine the association between duration of antidiabetic treatment and risk of
developing cardiovascular conditions and other T2D related events. We altered the days of
duration by diving its value by 30 to transform the estimated coefficient from the Cox model into
an estimate of the effect of adhering for an additional month. We chose a 30-day time window
because a 30-day supply is the most common quantity dispensed of ADMs. All models included
individual variables for all baseline patient characteristics listed in Table 3.1.
For economic outcomes, we use multivariable OLS and generalized linear models
(GLMs) to assess the relationship between treatment duration during the first year and all-cause
healthcare expenditures in the 1-year post-period. The Modified Park test was used to determine
family distribution and corresponding link function. Cost outcomes included total healthcare
costs and are also broken down into outpatient, inpatient, and prescription drug costs.
Additionally, we use multivariable OLS and a difference-in-difference design were used to
assess the relationship between treatment duration during the first year and the year-to-year
change in healthcare expenditures between the post-index and pre-index years. Treatment
duration was measured within this 1-year post-period. All cost outcomes were individually
47

assessed. All models included individual variables for all baseline patient characteristics listed in
Table 3.1.
3.2 Results
3.2.1 Baseline patient characteristics for patients beginning their second episode of drug
therapy for T2D
Selection of patients for the study is outlined in Figure 3.1. A total of 315,515 patients
initiated the initial treatment with metformin, sulfonylureas, and TZDs. Patients with initial
treatment regimens which included insulin, SGLT2 inhibitors, DPP4 inhibitors, or GLP-1 RAs
were excluded from as these patients are unlikely to be treatment naïve patients. Among them,
nearly half of the patients stayed on the initial treatment until the end of study period, and 45%
of them initiated the second episode. A total of 97,358 patients initiating a second treatment
episode met all of the inclusion and exclusion criteria for this study.
For the type of the second episode, most patients (76%) whose second treatment attempt
was a switch from their initial treatment. The vast majority of whom did so after a break in
therapy exceeding 60 days. An additional 17% of them augmented their initial episode treatment
with additional ADMs, and 7% of second episode was restarting their previous ADMs used in
the initial treatment.
Patients baseline characteristics were reported by type of the second episode outcomes
(Table 3.1). Categorical variables were summarized by frequency and percentage. Continuous
variables were reported by mean and standard deviation (SD). Statistical comparisons among
different drug cohorts were evaluated using chi-square tests for categorical variables and
ANOVA tests for continuous variables. The baseline characteristics varied significantly across
the type of the second episode. ADMs (metformin, sulfonylurea, TZDs) that were used in the
48

initial treatment were still commonly used in the second attempt. The newer ADM classes
(SGLT2i, DPP4i, GLP) were less likely to be initiated in the second episode. Patients whose
second episode as an augmentation were older, had higher baseline healthcare costs, were more
likely to have stroke or AMI in the year prior to starting their second treatment episode, and had
medication and diagnostic profiles consistent with poorer health status.
3.2.2 Predictors of duration of continuous therapy starting from the second episode of
treatment
The multivariable analyses of duration of therapy are displayed in Table 3.2. Patients
starting the second episode as an add-on with the initial treatment (augmentation) were
associated with significantly longer duration of continuous therapy (138 days) and lower risk of
discontinuation (HR = 0.756) than those starting their 2
nd
episode as a switch. Patients who
started 2
nd
episode with the treatment used previously (restarter episodes) also had significantly
lower risk of discontinuation. Patients who achieved longer duration of therapy during their
initial treatment attempt were significantly more likely to have longer duration of therapy during
their 2
nd
episode and lower risk of discontinuation. One month increase in the delay between
when the patient terminated their initial treatment and starting their 2
nd
episode was associated
with significantly shorter duration of therapy. Patients who were older, enrolled with an HMO,
and had higher prior drug costs were associated with significantly longer duration of therapy in
their 2
nd
episode and lower risk of discontinuation.
The results in Table 3.2 comparing the effects of alternative types of 2
nd
episode may be
sensitivity to the baseline A1c level. The sensitivity analyses of days of therapy and likelihood of
discontinuation presented in Tables 3.3a and 3.3b are very consistent with the primary analysis
of therapy duration. We found that baseline A1c values had no significant association with
49

duration of therapy and risk of discontinuation and the estimated effects of other treatment
parameters did not change significantly.
3.2.3 The impact of duration of drug therapy on the risk of clinical events and cost over the
first year following the initiation of second episode of T2D drug therapy
Results from the multivariable analyses of event risk, adjusting for other risk factors, are
summarized in Table 3.4. We found that one month increase in duration of therapy was
associated with significantly lower risk for measured clinical outcomes. One month increase in
treatment duration decreased the risk of stroke or AMI by 11.3% (an absolute risk reduction
(ARR) of 1.32% per additional month of therapy), stroke by 4.4% (ARR of 0.394% per month
of duration), AMI by 3.2% (ARR of 0.119% per month of duration), all-cause hospitalization by
3.5% (ARR of 0.934% per month), and CVD-related hospitalization by 3.1% (ARR of 0.269%
per month).
These estimated of duration were re-estimated considering only those events that
occurred within one year of the index date for initial therapy. A logit model was performed to
evaluate the likelihood of an event within one year as a function of month of therapy up to one
year. We found that one month increase in duration in the first year was associated with
significantly lower likelihood of having all the clinical outcomes. We also estimate a Cox model
using events during the first year and these results are very consistent with results from the logit
model with duration as the variable measuring persistence. The effects in the first year became
larger.
The relationship between treatment duration and health care costs was estimated using
GLMs and OLS models, respectively (Table 3.5 and Table 3.6). One month increase in the
treatment duration was associated with a significantly reduced outpatient costs of $21.49 and
50

reduced inpatient costs of $68.41 in the first year following the index date of the second
treatment attempt. The prescription drug costs for these patients increased by $310.51 as each
one month increase in treatment duration. Similar results were found using OLS models.
The relationship between treatment duration and health care costs was estimated using
difference-in-difference models and OLS estimation methods (Table 3.7). One month increase in
the treatment duration was associated with a significant reduction in the year-to-year change in
medical costs of $108.74. The year-to-year increase in prescription drug costs were higher per
month of persistence by $242.76.
3.3 Discussion
T2D typically requires adjustments to pharmacologic antidiabetic therapies over time as
insulin production declines and/or insulin resistance increases
1
. After the first T2D treatment
attempt, 45% of the patients changed their initial therapy (augmented or switched), or restarted
discontinued treatment. Among them, for most patients (82%), this change was made after a gap
in therapy exceeding 60 days. These patients were found to have significantly fewer duration of
continuous therapy and also more likely discontinue their treatment. Prior research indicate that
medication non-persistence is a behavior that is challenging to correct once established
2,3
. We
also found that patients with shorter duration of initial therapy were more likely to be non-
persistent with the second treatment attempt. And our results suggested that the longer period
patients without any ADMs on hand, the shorter duration of subsequent therapy patients had.
Those findings emphasize the importance of establishing treatment persistence from the
beginning of the therapy.
Non-adherence and non-persistence to prescribed ADMs are common and remain a
barrier to optimal health outcomes. For example, a meta-analysis of 40 studies found that only
51

67.9% of patients having an MPR>=80%, and 56.2% of patients being persistent with oral
ADMs
4
. A systematic review of 27 studies that evaluated adherence rates to ADMs revealed that
only 22% of studies reported >=80% adherence among patients
5
. Our study found that 57% of
patients discontinued treatment with an average duration of therapy nearly one year following
the second treatment attempt.
Persistence to treatment is a key element associated with the effectiveness of
pharmacological therapies in patients with T2D. Similar to earlier diabetes studies
6-13
, our study
shows that patients with longer duration of therapy following the second treatment attempt are at
reduced risk of stroke or AMI, and hospitalization. Importantly, these findings advance our
understanding of the impact of non-persistence by suggesting that patients with early tight
diabetes control may have both early and long-term benefits. In our analysis, a one month
increase in treatment duration reduced risk of AMI, stroke and hospitalization within the study
period. Our results highlight the importance of establishing the role of treatment persistence in
preventing complications earlier, from the onset of ADMs. A long-term observational follow-up
of the UK Prospective Diabetes Study (UKPDS) suggested that early glycemic control may have
durable effects in reducing risk of T2D-related complications
14
. A recent analysis of newly
diagnosed T2D patients also found that glycemic control during the first year after diagnosis was
significantly associated with a decrease in future risk for diabetic complications and mortality
15
.
Taken together, these findings suggest that treating T2D patients more intensively at an earlier
stage may result in long-term improvements in patient health outcomes. Healthcare professionals
should work with patients from the beginning of therapy to ensure that barriers to persistence are
identified and reduced in order to limit the impact of non-persistence on poor future outcomes.
52

Our study shows that patients with longer duration of therapy had lower medical costs
during the first year following their second treatment attempt. Consistent with our findings,
research generally has demonstrated that medication adherence was inversely associated with
costs of outpatient care, ER visits, and hospitalization
16-18
. As expected, we did find that
prescription drug costs increased with longer duration of therapy. However, this increase in
prescription drug costs was larger than offsetting savings in medical costs during the first year of
treatment. Improved medication persistence may significantly impact long term health care costs
as each month of persistence significantly reduced the risk of major cardiovascular events and
hospitalization. As T2D progresses, the risk of complications increases and cost savings
associated with reductions in the risk of complications would be expected to be more significant.
Jha, et al. suggested that reduced risk of hospitalization or emergency department visits because
of improved medication adherence was projected to save $4.7 billion annually
19
. In a large VA
cohort study which examined the longitudinal effects of medication adherence on potential cost
savings
20
, higher adherence was associated with higher pharmacy costs over the five years of
treatment which were offset by $661 million to $1.16 billion annual cost savings.
Limitations
Our study was subject to several limitations. Studies using real-world data to compare
alternative treatments may suffer from treatment selection bias. Patients who initiated different
treatment regimens may also be different in other important but unmeasured ways. Patients who
are persistent with their medications may be different than non-persistent patients. The primary
defense against potential treatment selection bias is to more fully document the observable
characteristics of the patients in the study. This was achieved by including the multiple baseline
characteristics in our statistical models, including a complete diagnostic and prescription drug
53

profiles based at baseline. Second, reasons for treatment discontinuation or any additional health
benefits (e.g., reduced symptom severity) resulting from increased persistence could not be
examined using claims databases. Finally, death caused by CVD is an important outcome for
patients with T2D that should be considered. We could not estimate the association between
treatment duration and CVD-related mortality since that mortality information was not available
in the database.
3.4 Conclusions
Longer continuous duration of antidiabetic medications following the second treatment
attempt can significantly improve clinical outcomes at any post-index time and 1-year post-index
medical costs. These results emphasize the importance of persistence for patients with T2D and
the need to assist them with remaining persistent from the early phase of therapy.

54

Chapter 3 References
1. American Diabetes Association. 9. Pharmacologic approaches to glycemic treatment:
standards of medical care in diabetes—2019. Diabetes Care. 2019;42(Supplement
1):S90-S102.
2. Gatwood JD, Chisholm-Burns M, Davis R, et al. Disparities in initial oral antidiabetic
medication adherence among veterans with incident diabetes. Journal of managed care &
specialty pharmacy. 2018;24(4):379-389.
3. Saundankar V, Peng X, Fu H, et al. Predictors of change in adherence status from 1 year
to the next among patients with type 2 diabetes mellitus on oral antidiabetes drugs.
Journal of managed care & specialty pharmacy. 2016;22(5):467-482.
4. Iglay K, Cartier SE, Rosen VM, et al. Meta-analysis of studies examining medication
adherence, persistence, and discontinuation of oral antihyperglycemic agents in type 2
diabetes. Curr Med Res Opin. 2015;31(7):1283-1296.
5. Krass I, Schieback P, Dhippayom T. Adherence to diabetes medication: a systematic
review. Diabet Med. 2015;32(6):725-737.
6. Ho PM, Rumsfeld JS, Masoudi FA, et al. Effect of medication nonadherence on
hospitalization and mortality among patients with diabetes mellitus. Arch Intern Med.
2006;166(17):1836-1841.
7. Asche C, LaFleur J, Conner C. A review of diabetes treatment adherence and the
association with clinical and economic outcomes. Clin Ther. 2011;33(1):74-109.
8. Andrew PY, Yanni FY, Nichol MB. Estimating the effect of medication adherence on
health outcomes among patients with type 2 diabetes—an application of marginal
structural models. Value Health. 2010;13(8):1038-1045.
55

9. Zhu VJ, Tu W, Rosenman MB, Overhage JM. Nonadherence to Oral Antihyperglycemic
Agents: Subsequent Hospitalization and Mortality among Patients with Type 2 Diabetes
in Clinical Practice. Paper presented at: MedInfo2015.
10. Huber CA, Rapold R, Brüngger B, Reich O, Rosemann T. One-year adherence to oral
antihyperglycemic medication and risk prediction of patient outcomes for adults with
diabetes mellitus: An observational study. Medicine. 2016;95(26).
11. Feldman BS, Cohen-Stavi CJ, Leibowitz M, et al. Defining the role of medication
adherence in poor glycemic control among a general adult population with diabetes.
PLoS One. 2014;9(9).
12. Trinacty CM, Adams AS, Soumerai SB, et al. Racial differences in long-term adherence
to oral antidiabetic drug therapy: a longitudinal cohort study. BMC Health Serv Res.
2009;9(1):24.
13. Gatwood J, Chisholm‐Burns M, Davis R, et al. Differences in health outcomes associated
with initial adherence to oral antidiabetes medications among veterans with
uncomplicated Type 2 diabetes: a 5‐year survival analysis. Diabet Med.
2018;35(11):1571-1579.
14. Holman RR, Paul SK, Bethel MA, Matthews DR, Neil HAW. 10-year follow-up of
intensive glucose control in type 2 diabetes. N Engl J Med. 2008;359(15):1577-1589.
15. Laiteerapong N, Ham SA, Gao Y, et al. The legacy effect in type 2 diabetes: impact of
early glycemic control on future complications (the Diabetes & Aging Study). Diabetes
Care. 2019;42(3):416-426.
56

16. Nasseh K, Frazee SG, Visaria J, Vlahiotis A, Tian Y. Cost of medication nonadherence
associated with diabetes, hypertension, and dyslipidemia. Am J Pharm Benefits.
2012;4(2):e41-e47.
17. Roebuck MC, Liberman JN, Gemmill-Toyama M, Brennan TA. Medication adherence
leads to lower health care use and costs despite increased drug spending. Health Aff
(Millwood). 2011;30(1):91-99.
18. Breitscheidel L, Stamenitis S, Dippel F-W, Schöffski O. Economic impact of compliance
to treatment with antidiabetes medication in type 2 diabetes mellitus: a review paper. J
Med Econ. 2010;13(1):8-15.
19. Jha AK, Aubert RE, Yao J, Teagarden JR, Epstein RS. Greater adherence to diabetes
drugs is linked to less hospital use and could save nearly $5 billion annually. Health Aff
(Millwood). 2012;31(8):1836-1846.
20. Egede LE, Gebregziabher M, Dismuke CE, et al. Medication nonadherence in diabetes:
longitudinal effects on costs and potential cost savings from improvement. Diabetes
Care. 2012;35(12):2533-2539.

57

Figure 3.1. Selection of Patients with Second Episodes of Therapy

Initial antidiabetic treatment with metformin,
sulfonylurea, and TZDs
(n = 315,515)
Augmentation:
16870 (17.33%)
Restart: 6434
(6.61%)
No 2
nd
treatment attempt:
173,105
• Stayed on initial, no
changes, till data ran
out: 149,997
• Discontinued initial
Rx, no 2
nd
episode:
23,108
Patients with 2nd treatment attempt:
142,410
• Discontinued, started 2
nd
, no
events: 78,686
• Discontinued, started 2
nd
, before
event: 18,092
• Discontinued, started 2
nd
, after
event: 20,238
• Started 2
nd
with no break: 25,394
STUDY SAMPLE: Second episode of treatment with at least 1-year pre
and post data, insulin excluded, no pregnancy and costs < 500K: 97,358

Switch:
74054
(76.06%)
Switch without
a break: 616
(0.83%)
Switch with a
break: 73438
(99.17%)
58

Table 3.1. Patient characteristics by drug therapy outcomes
Type of 2
nd

episode:
Restart previous
treatment Augmentation
Switch to a new
treatment

N. of patients: 6434 6.61% 16870 17.33% 74054 76.06%
n/mean %/sd n/mean %/sd n/mean %/sd p value
Rx initiated in the
2
nd
episode
met mono 1285 19.97% 12088 71.65% 66185 89.37% <.0001
sulf mono 736 11.44% 3989 23.65% 6858 9.26%
TZD mono 389 6.05% 543 3.22% 824 1.11%
SGLT2i mono 0 0.00% 9 0.05% 3 0.00%
DPP4i mono 0 0.00% 80 0.47% 17 0.02%
GLP mono 0 0.00% 98 0.58% 15 0.02%
Combo that
included met
3905 60.69% 57 0.34% 149 0.20%
Combo without met 119 1.85% 6 0.04% 3 0.00%
Healthcare Cost
($/year)
Medical costs
Year Prior to Index 5279.00 18447.51 5394.45 15982.53 5768.82 19320.69 <.0001
Year Post Index 5485.72 17124.61 5886.67 18305.79 5869.45 19701.37 <.0001
Drug costs
Year Prior to Index 9788.87 15999.01 9274.66 18460.37 9170.94 19433.42 <.0001
Year Post Index 11878.39 21083.49 10547.52 20599.57 10008.73 21796.87 <.0001
Total costs
Year Prior to Index 15067.88 25754.71 14669.11 25703.42 14939.76 28929.52 <.0001
Year Post Index 17364.11 28571.52 16434.18 29258.29 15878.17 30937.80 <.0001
Time OFF Rx until
2
nd
initiated, mean,
sd 4.88 8.75 0.00 0.29 8.60 11.34 <.0001
Age, mean (sd)
years 61.95 12.86 62.95 12.84 61.71 14.44 <.0001
Age group, n (%)

18-34 years 126 1.96% 300 1.78% 3474 4.69% <.0001
35-44 years 508 7.90% 1188 7.04% 6069 8.20%

45-54 years 1231 19.13% 2985 17.69% 12500 16.88%

55-64 years 1641 25.51% 4199 24.89% 17822 24.07%

65-79 years 2396 37.24% 6598 39.11% 26563 35.87%

>=80 years 532 8.27% 1600 9.48% 7626 10.30%

Female, n (%) 3925 61.00% 9768 57.90% 33928 45.82% <.0001
Health Plan, n (%)

EPO 561 8.72% 1159 6.87% 5858 7.91% <.0001
HMO 1699 26.41% 4878 28.92% 20022 27.04%

INDEMNITY 71 1.10% 280 1.66% 1225 1.65%

POS 2379 36.98% 5959 35.32% 28462 38.43%

PPO 299 4.65% 921 5.46% 3895 5.26%

OTHER 1425 22.15% 3673 21.77% 14592 19.70%

Prior medical costs

<=1000 4283 66.57% 10388 61.58% 45750 61.78% <.0001
1000 - 5000 1010 15.70% 3076 18.23% 13467 18.19%

5000 - 10000 399 6.20% 1195 7.08% 5187 7.00%

>10000 742 11.53% 2211 13.11% 9650 13.03%

59

Prior drug costs

<=1000 1098 17.07% 2664 15.79% 13687 18.48% <.0001
1000 - 5000 2173 33.77% 6556 38.86% 27949 37.74%

5000 - 10000 1173 18.23% 3105 18.41% 13205 17.83%

>10000 1990 30.93% 4545 26.94% 19213 25.94%

Had prior stroke
or AMI
446 6.93% 1229 7.29% 4869 6.57%
0.0031
Baseline Diagnoses

Infectious diseases 897 13.94% 2385 14.14% 10016 13.53% 0.0892
Neoplasms 1136 17.66% 3773 22.37% 17654 23.84% <.0001
Endocrine disorders 4938 76.75% 13919 82.51% 61536 83.10% <.0001
Blood diseases 844 13.12% 2447 14.51% 11172 15.09% <.0001
Mental disorders 1218 18.93% 3819 22.64% 19087 25.77% <.0001
Nervous system 2876 44.70% 8418 49.90% 36260 48.96% <.0001
Eye diseases 246 3.82% 1076 6.38% 5983 8.08% <.0001
Ear diseases 63 0.98% 328 1.94% 2013 2.72% <.0001
Circulatory
disorders
4752 73.86% 13306 78.87% 54646 73.79%
<.0001
Respiratory system 2154 33.48% 6491 38.48% 30304 40.92% <.0001
Digestive system 1679 26.10% 5187 30.75% 24054 32.48% <.0001
Genitourinary
system
2115 32.87% 6467 38.33% 30421 41.08%
<.0001
Skin diseases 1540 23.94% 4697 27.84% 21912 29.59% <.0001
Musculoskeletal 3044 47.31% 8778 52.03% 40638 54.88% <.0001
Congenital
anomalies
153 2.38% 501 2.97% 2432 3.28%
<.0001
Injury/Poisoning 1275 19.82% 3629 21.51% 17003 22.96% <.0001
Overweight/obesity 999 15.53% 3299 19.56% 15777 21.30% <.0001
Tobacco use 452 7.03% 1311 7.77% 5260 7.10% 0.0084
Baseline
Medications

Antihistamines 352 5.47% 788 4.67% 3423 4.62% 0.0084
Anti-infectives 2917 45.34% 8493 50.34% 39176 52.90% <.0001
Antineoplastic 96 1.49% 331 1.96% 1836 2.48% <.0001
Autonomic Drugs 1351 21.00% 4172 24.73% 19621 26.50% <.0001
Anti-coagulation 614 9.54% 1952 11.57% 7680 10.37% <.0001
Cardiovascular 5570 86.57% 14859 88.08% 61666 83.27% <.0001
Central Nervous
System Agents
3315 51.52% 9548 56.60% 43903 59.29%
<.0001
Electrolytic agents 1415 21.99% 4436 26.30% 18161 24.52% <.0001
Respiratory Tract 659 10.24% 1904 11.29% 10191 13.76% <.0001
Eye, Ear, Nose, and
Throat (EENT)
1193 18.54% 3821 22.65% 18292 24.70%
<.0001
Gastrointestinal 1476 22.94% 4629 27.44% 21781 29.41% <.0001
Skin and Mucous
Membrane Agents
1151 17.89% 3494 20.71% 16607 22.43%
<.0001
Smooth Muscle
Relaxants
149 2.32% 482 2.86% 2353 3.18%
0.0002
Vitamins 224 3.48% 645 3.82% 4244 5.73% <.0001
Met=metformin; sulf=sulfonylurea; TZD=thiazolidinedione; EPO=exclusive provider
organization; HMO=health maintenance organization; POS=point-of-service; PPO=preferred
provider organization; SD=standard deviation.
60

Table 3.2. Patient characteristics impacting duration that patients stayed on any drug
among patients starting the second episode of drug therapy for T2D
N = 97,358

Days of uninterrupted
therapy
Time to discontinuation
OLS model Cox model
Parameters coefficient P value HR P value
Intercept
573.81 <.0001
Type of 2
nd
episode
(vs. switch)

Restart
103.32 <.0001 0.799 <.0001
Augmentation
138.50 <.0001 0.756 <.0001
Had stroke or AMI
before 4.86 0.4381 0.973 0.1452
Time OFF Rx until 2
nd

initiated (in months) -1.04 <.0001 1.005 <.0001
Duration of initial
regimen (in months) 3.88 <.0001 0.987 <.0001
A1c=yes
4.05 0.2682 0.996 0.6823
index year
-22.49 <.0001 0.965 <.0001
female
-7.51 0.0162 0.942 <.0001
Health plan (vs. PPO)
0.00

EPO
-37.01 <.0001 0.996 0.8692
HMO
8.29 0.2404 1.009 0.6497
INDEMNITY
80.60 <.0001 0.874 0.0007
OTHER
52.68 <.0001 0.998 0.9149
POS
-12.33 0.0916 0.965 0.0905
Age group (vs. >=80)

18-34 years
-155.13 <.0001 1.585 <.0001
35-44 years
-109.45 <.0001 1.423 <.0001
45-54 years
-55.88 <.0001 1.266 <.0001
55-64 years
-21.94 0.0007 1.124 <.0001
65-79 years
32.87 <.0001 1.035 0.0304
Prior medical costs
(vs. <=1000)

1000 - 5000
-16.72 <.0001 1.005 0.6865
5000 - 10000
-13.31 0.0332 0.995 0.7725
>10000
-12.37 0.0241 0.945 0.0004
Prior drug costs (vs.
<=1000)

1000 - 5000
13.64 0.0028 0.945 <.0001
5000 - 10000
22.00 <.0001 0.921 <.0001
61

>10000
47.96 <.0001 0.844 <.0001
Baseline diagnoses

Infectious diseases
-1.66 0.7143 1.026 0.0423
Neoplasms
18.60 <.0001 0.96 0.0002
Endocrine disorders
-8.44 0.0518 1.04 0.0018
Blood diseases
-19.85 <.0001 1.013 0.3169
Mental disorders
-25.16 <.0001 1.012 0.3102
Nervous system
30.39 <.0001 0.982 0.0598
Eye diseases
-87.87 <.0001 0.898 <.0001
Ear diseases
-39.82 <.0001 0.944 0.0876
Circulatory disorders
-9.54 0.0210 1.05 <.0001
Respiratory system
-7.56 0.0349 1.008 0.4308
Digestive system
-6.91 0.0646 1.025 0.0204
Genitourinary system
-10.88 0.0008 1.027 0.0044
Skin diseases
8.28 0.0236 0.958 <.0001
Musculoskeletal
-12.68 0.0002 1.032 0.0012
Congenital anomalies
0.91 0.8135 0.995 0.6365
Injury/Poisoning
-3.18 0.7091 0.995 0.829
Overweight/obesity
0.89 0.8177 0.994 0.5749
Tobacco use
-0.89 0.8870 1.033 0.0701
Baseline drug profile

Antihistamines
-12.13 0.0914 1.007 0.7377
Anti-infectives
-11.46 0.0008 1.042 <.0001
Antineoplastic
-16.66 0.0945 0.993 0.8152
Autonomic Drugs
-7.35 0.0476 1.014 0.1968
Anti-coagulation
-1.07 0.8386 0.957 0.0051
Cardiovascular
23.45 <.0001 0.956 0.0011
Central Nervous
System Agents -23.30 <.0001 1.051 <.0001
Electrolytic agents
5.78 0.1101 0.951 <.0001
Respiratory Tract
-4.71 0.3327 1.002 0.902
Eye, Ear, Nose, and
Throat (EENT) 1.24 0.7368 1.036 0.0008
Gastrointestinal
8.81 0.0192 0.976 0.0265
Skin and Mucous
Membrane Agents -5.81 0.1330 1.028 0.0125
Smooth Muscle
Relaxants -5.48 0.5289 0.983 0.4987
Vitamins
-23.41 0.0005 1.07 0.0004
EPO=exclusive provider organization; HMO=health maintenance organization; POS=point-of-service;
PPO=preferred provider organization.

62

Table 3.3a. Sensitivity Analysis
Primary model analysis
(OLS)
Sensitivity analysis (OLS)
N = 97,358 N = 21,281
Parameters coefficient P value coefficient p value
Intercept
573.81 <.0001 652.65 <.0001
Type of 2
nd
episode
(vs. switch)

Restart
103.32 <.0001 87.12 <.0001
Augmentation
138.50 <.0001 139.43 <.0001
Had stroke or AMI
before 4.86 0.4381 -4.06 0.7385
Time OFF Rx until
2
nd
initiated (in
months) -1.04 <.0001 -0.91 0.0006
Duration of initial
regimen (in months) 3.88 <.0001 3.37 <.0001
A1c=yes vs A1c
value 4.05 0.2682 -3.96 0.1163

Table 3.3b. Sensitivity Analysis
Primary model analysis
(Cox)
Sensitivity analysis (Cox)
N = 97,358 N = 21,281
Parameters HR P value HR p value
Type of 2
nd
episode (vs.
switch)

Restart
0.799 <.0001 0.8 <.0001
Augmentation
0.756 <.0001 0.736 <.0001
Had stroke or AMI
before 0.973 0.1452 0.982 0.6519
Time OFF Rx until 2
nd

initiated (in months) 1.005 <.0001 1.005 <.0001
Duration of initial
regimen (in months) 0.987 <.0001 0.987 <.0001
A1c=yes vs A1c value
0.996 0.6823 1.003 0.7018

Table 3.4. Impact of Duration on Clinical Event Risk
Cox Model
Logit Model of Event in 1
st

Year
Cox Model

Duration (months)
Duration
(up to 12 months)
Duration (capped at
first year)
63

Outcomes:
Hazard
ratio p value Odds ratio p value
Hazard
ratio p value
Stroke or AMI 0.887 <.0001 0.731 <.0001 0.734
<.0001
Stroke 0.956 <.0001 0.726 <.0001 0.728 <.0001
AMI 0.968 <.0001 0.744 <.0001 0.741 <.0001
All-cause
hospitalization 0.965 <.0001 0.757 <.0001 0.761 <.0001
CVD-related
hospitalization 0.969 <.0001 0.758 <.0001 0.759 <.0001
AMI=acute myocardial infarction; CVD=cardiovascular disease.

Table 3.5. Impact of Duration during the First Year on Health Care Costs by Type of
Service (GLMs)
Explanatory variable:
Duration
(in months) PDC>=0.8
Outcomes: estimates p-value estimates p-value
Total health care costs $210.74 <.0001 $1647.92 <.0001
Outpatient costs -$21.49 0.035 -$282.43 0.001
Inpatient costs -$68.41 <.0001 -$684.98 <.0001
Prescription drug costs $310.51 <.0001 $2587.16 <.0001
Total costs = outpatient costs + inpatient costs + drug costs
Table 3.6. Impact of Duration during the First Year on Health Care Costs by Type of
Service (OLSs)
Explanatory variable:
Duration
(in months) PDC>=0.8
Outcomes: estimates p-value estimates p-value
Total health care costs $197.99 <.0001 $1604.02 <.0001
Outpatient costs -$23.10 0.025 -$317.31 <.0001
Inpatient costs -$69.16 <.0001 -$730.96 <.0001
Prescription drug costs $290.25 <.0001 $2652.29 <.0001
Total costs = outpatient costs + inpatient costs + drug costs

Table 3.7. Impact of Duration during the First Year on the Change in Health Care Costs by
Type of Service

Duration
(in months)

estimates p-value
64

Change in medical costs -$108.74 <.0001
Change in drug costs $242.76 <.0001
Change in total costs $134.02 <.0001
65

Chapter 3 Appendix
Time-varying treatment duration
The method used to assess the time-varying treatment duration is to examine the
Schoenfeld residuals
1,2
. The Schoenfeld residual computed with each covariate is defined at the
observed event times as the difference between covariate for observation and the weighted
average of the covariate values for all subjects still at risk when observation experiences the
event.

Table 3.A1. Results of tests for Schoenfeld residuals of treatment duration
Rh0 Chi2 df Prob>chi2
Duration (time to stroke or AMI) 0.25338 5118.94 1 <.0001
Duration (time to stroke) 0.23757 4426.48 1 <.0001
Duration (time to AMI) 0.22065 3755.44 1 <.0001
Duration (time to all-cause hospitalization) 0.27525 6158.71 1 <.0001
Duration (time to CVD-related hospitalization) 0.23093 4131.84 1 <.0001
AMI=Acute Myocardial Infarction; CVD=Cardiovascular disease; df=degree of freedom; Prob=probability.

As summarized in table S1, the correlations between the Schoenfeld residuals of
treatment duration and survival times were all significant, indicating that treatment duration was
a time-varying exposure variable. Therefore, time-varying Cox models should be used to account
for this.
Family distribution and link function in GLMs
The family distribution and corresponding link function used in GLMs were determined
by Modified Park Test. Test coefficient close to 0: Gaussian NLLS distribution; close to 1:
Poisson distribution; close to 2: Gamma distribution; close to 3: Inverse Gaussian or Wald
3
.
Table 3.A2a. Modified Park Test (duration in months)
Expenditures l Distribution used Link function used
Total expenditures 1.164963 Poisson Log
Outpatient expenditures 1.489868 Poisson Log
Inpatient expenditures 1.137312 Poisson Log
66

Drug expenditures 1.100348 Poisson Log

Table 3.A2b. Modified Park Test (PDC ³ 80%)
Expenditures l Distribution used Link function used
Total expenditures 1.174427 Poisson Log
Outpatient expenditures 1.487941 Poisson Log
Inpatient expenditures 1.133374 Poisson Log
Drug expenditures 1.130221 Poisson Log

Chapter 3 Appendix references
1. Grambsch PM, Therneau TM. Proportional hazards tests and diagnostics based on
weighted residuals. Biometrika. 1994;81(3):515-526.
2. Zhang Z, Reinikainen J, Adeleke KA, Pieterse ME, Groothuis-Oudshoorn CG. Time-
varying covariates and coefficients in Cox regression models. Annals of translational
medicine. 2018;6(7).
3. Manning WG, Mullahy J. Estimating log models: to transform or not to transform? J
Health Econ. 2001;20(4):461-494.

67

CHAPTER 4. Delaying Treatment initiation for Type 2 Diabetes Patients
Following Their Initial A1c Test >= 7%

4.1 Methods
4.1.1 Study population
We identified adults aged >= 18 years with at least one diagnosis for T2D (ICD-9-CM
250.x0, 250.x2; ICD-10-CM E11). The laboratory test results for the subset of patients were
then used to identify the first recorded A1c >=7% (index date). To better ensure that this test was
the patient’s first elevated A1c, we required patients to have a minimum of 1 year of continuous
enrollment before the index A1c > 7% and that the patient had no prescriptions for T2D
medications filled in the pre-period. Patients were also required to have a minimum of one year
of post-index enrollment. Patients who were diagnosed with type 1 diabetes or gestational
diabetes, were pregnant, or had complications of pregnancy, childbirth, and the puerperium were
excluded.
Part 1. Predictors of Delays in the Initiation of Treatment Following an Elevated A1c
4.1.2 Explanatory Variables (part 1)
The patient baseline demographic and clinical characteristics were assessed based on
paid claims date recorded during the 1-year pre-index period. Demographic characteristics
included age, sex, type of health insurance, and index year. Clinical measures included index
A1c, hospitalization, healthcare costs, diagnostic history, and history of prescription drug use.
The patient’s diagnostic profile was created using all diagnostic codes reported on the patient’s
paid claims for services used in the pre-period. The patient’s drug use history was based on all
68

non-diabetic medications filled by the patient in the 1-year prior to index date broken down by
AHFS pharmacologic-therapeutic classes.
4.1.3 Outcomes (part 1)
The primary outcome for part 1 was time to treatment initiation, which was measured as
days from the index date to the first fill date for metformin, sulfonylurea, TZDs, SGLT-2
inhibitors, DPP-4 inhibitors, GLP-1 RAs, or insulin.
4.1.4 Statistical analyses (part 1)
Multivariable OLS models and Cox proportional hazards models were used to evaluate
the associations between patient characteristics and initial treatment initiation.
Part 2. Impact of delays in treatment initiation on risk of clinical events
4.1.5 Key explanatory variables (part 2)
Delays in treatment initiation were measured in three ways for use as the key explanatory
variable for the analysis in Part 2. First, we created a dichotomous variable indication whether or
not the patient started drug treatment prior to any adverse clinical event or hospitalization. The
comparison group for this analysis were patients not filling an ADM at all (or only following
their first clinical event). In the second specification of the Cox models of event risk, the time to
initiation was entered as a time-varying exposure variable. Finally, we divided the study
population into 4 groups depending on the delays in initiating treatment: less than 30 days, 30-
180 days (6 months), after 6 months and never treated which served as the comparison group in
the Cox analyses.
69

4.1.6 Outcomes and covariates (part 2)
Outcomes for the analysis of the impact of delays in therapy were measured as time to
clinical events associated with T2D, measured in days from the date of the index A1c test >=7%.
The primary clinical outcome events were stroke, or myocardial infarction (MI) and time to all-
cause hospitalization.
The covariates used in the statistical models included age, gender, health plan type, index
year, index A1c, hospitalization, healthcare costs, diagnostic history (a comprehensive diagnostic
profile was created using all diagnoses reported on the patient’s paid claims for services used in
the pre-period), and history of prescription drug use (mix of all non-diabetic medications filled
by the patient in the 1-year prior to index date broken down by AHFS pharmacologic-therapeutic
classes).
4.1.7 Statistical analyses (part 2)
Multivariable Cox models were used to examine the association between delays in
treatment initiation and the risk of T2D-related clinical outcomes.
4.2 Results
4.2.1 Study Sample Selection and Baseline Characteristics
Selection of patients for the study is outlined in Figure 1. We initially identified 1.3
million T2D patients with an elevated A1c. We excluded the 67% of patients who were treated
with a T2D drug therapy prior to their elevated A1c. The remaining 33% of patients with an
initial A1c test > 7% were then screened for having one year of data prior to and following the
index A1c test. Finally, patients under the age of 18, diagnosed with type 1 diabetes or
gestational diabetes, were pregnant, or had complications of pregnancy, childbirth, and the
puerperium were excluded. A total of 132,925 patients met all inclusion and exclusion criteria
70

and were included in this study. Of these, a total of 91,379 patients (69%) were treated following
their first elevated A1c, including 7,217 patients (26%) who initiated treatment only after
experiencing a stroke, AMI, or hospitalization in the post-index period. An additional 21% of all
patients were never treated of whom 13,060 (31%) experienced an event after their elevated A1c.
Baseline characteristics of patients are presented in Table 1 broken down by treatment
initiation status. The baseline characteristics varied significantly across these groups. Patients
who were treated within a month of their index A1c > 7% had both the lowest average age and
the highest average index A1c level. Patients who were treated more than 6 months following
their index elevated A1c had the lowest index A1c. Patients who were never treated were older,
more likely to have a prior stroke or AMI, more likely to be hospitalized, had the highest rates of
reported diagnoses in the pre-index period but had fewer prescription drugs filled during the pre-
period. These results suggest that patient age, index A1c level and comorbidity status are inter-
related in their impact on the treatment decision. To investigate these relationships, multivariable
models documenting the predictors of treatment initiation will include sensitivity analyses in
which the interaction term between age, baseline A1c and health status are explored. To simplify
the interpretation of these results, patient health status was measured using the Charlson
Comorbidity Index (CCI) which is based on diagnostic mix. Index A1c will be specified as a
continuous variable and both variables will be interacted with the patient age distribution.
4.2.2 Predictors of treatment initiation
The multivariable analyses of treatment initiation are displayed in Table 2. In the base
model, CCI, age, and index A1c were entered as separate categorical variables. The first
alternative model added interaction terms of age and CCI and age and the patient’s index A1c.
71

The second alternative model expanded these two interaction terms by measuring age as a
categorical variable.
In the base model, we found that as index A1c increased, patients were more likely to
initiate treatment. Baseline CCI score was also a significant predictor of treatment initiation, but
patients with higher score were less likely to receive treatment. Age had a significant impact on
the treatment decision. The hazard of receiving treatment was lowest in patients who aged 18-23
years, increased through age 36-41, then decreased monotonically thereafter.
After adding the interaction terms (alternative model 1), we found that the likelihood of
treatment decreased with age and increased with baseline A1c, but now increased with the CCI .
The interaction term between age and index A1c indicates that the age effect on the likelihood of
treatment becomes more pronounced as the baseline A1c increases. Or stated differently, the
A1c effect becomes more significant with age. The opposite effect was found for the age-CCI
interaction. The impact of age becomes less pronounced with increasing CCI or, conversely, the
effect of the CCI increasing the likelihood of treatment becomes less pronounced with age.
To make things more complicated, the second alternative specification of the interaction
terms indicates that the interaction effects are not monotonic in their direction of effect. In the
alternative model 2, the hazard of receiving treatment peaked at age of 30-35, and all the hazard
ratios of age groups were higher than the base model. The effects of CCI on treatment initiation
in each age group showed similar pattern as in the model 2. For patients aged below 36, as index
A1c increased, older patients were significantly less likely to receive treatment. When patients
aged over 36, as index A1c increased, patients in older age groups were more likely to receive
treatment. However, increase in A1c was associated with lower likelihood of treatment when
patients reached 80 or older.
72

Across the three models, patients who were enrolled in POS, had higher medical costs,
and had prescription drugs in the pre-period were significantly more likely to receive treatment
after having an elevated A1c.
We found similar results as using the multivariable OLS models (Table 3). Higher index
A1c was associated with fewer days of treatment initiation but the effect was less pronounced
with age. CCI with age also displayed similar pattern as in the Cox models.
4.2.3 The impact of treatment delay on the risk of clinical events
Results from the multivariable analyses of event risk, adjusting for other risk factors, are
summarized in Table 3a-c. Timely treatment initiation (Table 4a) significantly decreased the
hazard rate of stroke or AMI by 55.1% (an absolute risk reduction (ARR) of 6.8% ), stroke by
56.4% (ARR of 5.2%), AMI by 53.4% (ARR of 1.7%), and all-cause hospitalization by 49.4%
(ARR of 13.7%). One month increase in delaying treatment was associated with significantly
higher risk of stroke, AMI, and all-cause hospitalization (Table 4b). Compared to those who
were never treated, treated patients were associated with significantly lower risk of all the
clinical outcomes (Table 4c), even those patients who delayed therapy beyond 6 months. What
is particularly curious is that impact of the patients treated after 6 months. As a group, these
patients exhibit the lowest risk of events.
4.3 Discussion
The primary goal of T2D treatment is reduction of blood glucose to prevent long-term
microvascular and macrovascular complications. In the clinical guidelines, A1c of >=7% is
generally served as a call to action to initiate or change therapy. Despite the importance of
glycemic control, many patients do not receive treatment in a timely manner. The failure of
receiving treatment when clinically indicated has been termed “clinical inertia”. Prior studies
73

have indicated that clinical inertia is evident at early stages of the disease. Marrett, et al. found
that clinical inertia affected one-third of patients with T2D
1
. In a study of Spanish primary-care
practitioners, clinical inertia was greater in patients treated with only lifestyle changes or
monotherapy than in those with more complex therapy
2
. In the present study, we found that 67%
of patients with an initial A1c > 7% were already using T2D medications. However, more than
half of previously untreated patients had no treatment initiated within 6 months following A1c of
>=7%. This finding is concerning given that the population had relatively high average baseline
A1c levels (mean: 8.6%) and that all had insurance.
We observed an inverse relationship between index A1c level and time to treatment
initiation. A shorter time to initiation was significantly associated with higher A1c levels that
indicated a more severe hyperglycemic status. Our results suggest that A1c levels that are
slightly above goal are less likely to be addressed in a timely manner versus those that clearly
remain too high. This finding is consistent with the literature that higher index A1c was
associated with a greater likelihood of initiation or intensification
3-7
. We also found that as age
increased, the hazard of treatment initiation decreased. And older patients with increasing CCI
(poorer comorbidity status) were also associated with lower likelihood of treatment initiation.
When interacting age with index A1c, we found that although higher A1c was associated with
earlier treatment initiation, the likelihood of initiation decreased as age increased. Our results
indicate that once A1c is 7% or higher, physicians are balancing age and health status of patients
when deciding whether or not initiate treatment. Older patients may have more comorbidities
that may serve as barriers to treatment initiation or they may have higher individualized A1c
goals. In the clinical guidelines, less stringent A1c goals (such as <8%) are recommended for
patients with high risk of hypoglycemia, limited life expectancy, and severe comorbid conditions
74

8
. Other significant factors included prior medical costs and prior prescription drug use. We
found that higher medical costs and use of prescription drugs in the pre-period were associated
with earlier treatment initiation. These two factors probably indicate more severe disease that
requires more physician visits, which may increase the likelihood of treatment initiation.
Delays in treatment initiation can mean that patients may live with suboptimal glycemic
control for years, and this can lead to adverse long-term outcomes. Pantalone, et al. showed that
among newly diagnosed T2D patients, time until A1c goal attainment was longer in patients who
received late intensification of therapy after metformin failure
9
. A large cohort study in UK
found that delayed intensification along with poor glycemic control was associated with
significantly higher risks of AMI, heart failure, and stroke
4
. However, prior research mainly
focused on delays in treatment intensification following initial treatment.
The present study showed that delays in initiating the first-line treatment after elevated
A1c could also significantly increase risks of AMI, stroke, and all-cause hospitalization. Failure
to intervene early when it is indicated may have a lasting effect on risks of developing diabetes-
related complications. Our results highlight the importance of timely treatment initiation in
preventing complications earlier, from the onset of ADMs. A long-term observational follow-up
of the UK Prospective Diabetes Study (UKPDS) suggested that early glycemic control may have
durable effects in reducing risk of T2D-related complications
10
. A recent analysis of newly
diagnosed T2D patients also found that glycemic control during the first year after diagnosis was
significantly associated with a decrease in future risk for diabetic complications and mortality
11
.
Taken together, these findings suggest that treating T2D patients more intensively at an earlier
stage may result in long-term improvements in patient health outcomes.
75

Healthcare professionals should work with patients from the beginning of therapy to
ensure that challenges of clinical inertia are addressed in order to limit its impact on poor future
outcomes. Current performance measures for diabetes care address patient A1c below certain
targets (e.g., <7%)
12
. Our findings indicate that there may be a need to incorporate measures of
time to treatment initiation into performance measures. As availability of health data systems and
patient-reported data grow, quality measures related to timely initiation of treatment may
catalyze new efforts to address clinical inertia and improve patient health outcomes.
Limitations
Our study was subject to several limitations. Patients’ A1c goals may have been different
or individualized. However, such individual goals were not available in the database. Generally,
there is an consensus in current clinical guidelines that an A1c of >=7% is a “action threshold”
8
;
thus the results at this cut point could be generalizable to the newly treated population with T2D.
Second, reasons for delayed initiation or any additional health benefits (e.g., reduced symptom
severity) resulting from earlier initiation could not be examined using claims databases. Finally,
death caused by CVD is an important outcome for patients with T2D that should be considered.
We could not estimate the association between treatment duration and CVD-related mortality
since that mortality information was not available in the database.
4.4 Conclusions
While we found that 67% of patients with an initial A1c >= 7% were using T2D drug
therapy, we also observed that fewer than half of untreated patients with T2D received initial
treatment after an A1c of >=8%. Delayed treatment among newly treated patients with T2D can
significantly increase the risk of clinical outcomes at any post-index time. These results
76

emphasize the importance of timely treatment initiation from the onset of T2D and the benefits
of early tight glycemic controls on preventing long-term diabetes-related complications.

77

Chapter 4 References
1. Marrett E, Zhang Q, Kanitscheider C, Davies MJ, Radican L, Feinglos MN. Physician
reasons for nonpharmacologic treatment of hyperglycemia in older patients newly
diagnosed with type 2 diabetes mellitus. Diabetes Ther. 2012;3(1):5.
2. Mata-Cases M, Benito-Badorrey B, Roura-Olmeda P, et al. Clinical inertia in the
treatment of hyperglycemia in type 2 diabetes patients in primary care. Curr Med Res
Opin. 2013;29(11):1495-1502.
3. Lin J, Zhou S, Wei W, Pan C, Lingohr-Smith M, Levin P. Does clinical inertia vary by
personalized A1c goal? A study of predictors and prevalence of clinical inertia in a US
managed-care setting. Endocr Pract. 2016;22(2):151-161.
4. Paul SK, Klein K, Thorsted BL, Wolden ML, Khunti K. Delay in treatment
intensification increases the risks of cardiovascular events in patients with type 2
diabetes. Cardiovasc Diabetol. 2015;14(1):100.
5. Raebel MA, Ellis JL, Schroeder EB, et al. Intensification of antihyperglycemic therapy
among patients with incident diabetes: a Surveillance Prevention and Management of
Diabetes Mellitus (SUPREME‐DM) study. Pharmacoepidemiol Drug Saf.
2014;23(7):699-710.
6. Fu A, Qiu Y, Davies M, Radican L, Engel S. Treatment intensification in patients with
type 2 diabetes who failed metformin monotherapy. Diabetes, Obesity and Metabolism.
2011;13(8):765-769.
7. Pantalone KM, Hobbs TM, Wells BJ, et al. Clinical characteristics, complications,
comorbidities and treatment patterns among patients with type 2 diabetes mellitus in a
78

large integrated health system. BMJ Open Diabetes Research and Care.
2015;3(1):e000093.
8. American Diabetes Association. 9. Pharmacologic approaches to glycemic treatment:
standards of medical care in diabetes—2019. Diabetes Care. 2019;42(Supplement
1):S90-S102.
9. Pantalone KM, Wells BJ, Chagin KM, et al. Intensification of diabetes therapy and time
until A1C goal attainment among patients with newly diagnosed type 2 diabetes who fail
metformin monotherapy within a large integrated health system. Diabetes Care.
2016;39(9):1527-1534.
10. Holman RR, Paul SK, Bethel MA, Matthews DR, Neil HAW. 10-year follow-up of
intensive glucose control in type 2 diabetes. N Engl J Med. 2008;359(15):1577-1589.
11. Laiteerapong N, Ham SA, Gao Y, et al. The legacy effect in type 2 diabetes: impact of
early glycemic control on future complications (the Diabetes & Aging Study). Diabetes
Care. 2019;42(3):416-426.
12. National Committee for Quality Assurance. Comprehensive diabetes care (CDC). 2019;
https://www.ncqa.org/hedis/measures/comprehensive-diabetes-care/. Accessed July 30,
2020.

79

Figure 4.1. Selection of Patients with First Test of A1c >= 7%

T2D patients with A1c>=7%:
n=1,298,001
Treated post first A1c >= 7%:
N = 91,379
• Event after Rx: 21,036
(23.0%)
• Event before Rx: 7,217
(7.9%)
• No events: 63,126
(69.1%)

874,436 Treated prior to first
A1c>=7%
• 822,786 treated after the
elevated A1C (94.1%)
• 51,650 Not treated after
the elevated A1C (5.9%)
423,565 patients were newly
treated or never treated after
first A1c >= 7%
STUDY SAMPLE after inclusion/exclusion criteria applied: N = 132,925
Never treated post first A1c
>= 7%
N = 41,546
• Event: 13,060 (31.4%)
• No events: 28,486
(68.6%)

67.4%
32.6%
31.3%
68.7%
>=1 year pre-index and >=1 year post-index enrollment
(n =134,655 )
Aged >=18 years, non-pregnancy
(n =132,925 )
80

Table 4.1. Patient Characteristics

Treated within 30
days
Treated 30 days –
6 months
Treated > 6
months Never treated
N. of patients: 45485 34.2% 16036 12.1% 29858 22.5% 41546 31.3%

Median follow up
(days) 1025 1012 1319 857

n/mean %/sd n/mean %/sd n/mean %/sd n/mean %/sd P value
Index A1c level
9.29 2.13 8.57 1.74 8.07 1.40 8.13 1.63 <.0001
Index A1c group <.0001
7% - 8% 18663 41.0% 9049 56.4% 21103 70.7% 29190 70.3%
8% - 9% 7287 16.0% 2716 16.9% 4083 13.7% 5809 14.0%
9% - 10% 4956 10.9% 1450 9.0% 1842 6.2% 2554 6.2%
>10% 14579 32.1% 2821 17.6% 2830 9.5% 3993 9.6%
Age 56.03 13.34 59.03 13.36 62.22 13.14 67.03 14.06 <.0001
Age group <.0001
18-23 years 233 0.5% 56 0.4% 81 0.3% 387 0.9%
24-29 years 665 1.5% 166 1.0% 197 0.7% 309 0.7%
30-35 years 1843 4.1% 452 2.8% 529 1.8% 630 1.5%
36-41 years 3739 8.2% 942 5.9% 1228 4.1% 1082 2.6%
42-47 years 5929 13.0% 1592 9.9% 2263 7.6% 1934 4.7%
48-53 years 7568 16.6% 2382 14.9% 3603 12.1% 2807 6.8%
54-59 years 7697 16.9% 2675 16.7% 4296 14.4% 3699 8.9%
60-65 years 5740 12.6% 2168 13.5% 3859 12.9% 3903 9.4%
66-71 years 5974 13.1% 2549 15.9% 5707 19.1% 8882 21.4%
72-79 years 4185 9.2% 2055 12.8% 5488 18.4% 10078 24.3%
>=80 years 1912 4.2% 999 6.2% 2607 8.7% 7835 18.9%
Female 18485 40.6% 7217 45.0% 13380 44.8% 18597 44.8% <.0001
Health Plan <.0001
EPO 4787 10.5% 1555 9.7% 2442 8.2% 2420 5.8%
HMO 10220 22.5% 4449 27.7% 10836 36.3% 13143 31.6%
INDEMNITY
128 0.3% 39 0.2% 96 0.3% 130 0.3%
POS 22539 49.6% 6678 41.6% 10101 33.8% 9047 21.8%
PPO 2010 4.4% 866 5.4% 1373 4.6% 1618 3.9%
OTHER
5801 12.8% 2449 15.3% 5010 16.8% 15188 36.6%
Prior medical
costs <.0001
<=1000 9726 21.4% 3229 20.1% 6157 20.6% 10567 25.4%
1000 – 5000 7808 17.2% 2529 15.8% 4538 15.2% 5620 13.5%
5000 – 10000 5123 11.3% 1635 10.2% 3056 10.2% 3927 9.5%
>10000 22828 50.2% 8643 53.9% 16107 54.0% 21432 51.6%
Prior drug costs <.0001
<=1000 20509 45.1% 7159 44.6% 14971 50.1% 25673 61.8%
1000 – 5000 12004 26.4% 3954 24.7% 6781 22.7% 7997 19.3%
5000 – 10000 5133 11.3% 1882 11.7% 3232 10.8% 3334 8.0%
>10000 7839 17.2% 3041 19.0% 4874 16.3% 4542 10.9%
Had prior stroke
or AMI 1015 2.2% 565 3.5% 897 3.0% 1874 4.5% <.0001
Had prior all-
cause
hospitalization 2776 6.1% 1279 8.0% 2000 6.7% 4441 10.7% <.0001
Charlson
comorbidity
index, mean 1.51 10.60 1.73 13.10 1.77 13.20 2.05 15.50 <.0001
81

Charlson
comorbidity index
group <.0001
1-3 42,718 93.9% 14,453 90.1% 26609 89.1% 34606 83.3%
4-5 2184 4.8% 1165 7.3% 2483 8.3% 5136 12.4%
6-7 472 1.0% 332 2.1% 630 2.1% 1477 3.6%
>=8 111 0.2% 86 0.5% 136 0.5% 327 0.8%
Baseline
Diagnoses
Infectious diseases 5446 12.0% 2117 13.2% 3825 12.8% 6000 14.4% <.0001
Neoplasms 8081 17.8% 3197 19.9% 6114 20.5% 10710 25.8% <.0001
Endocrine
disorders 32340 71.1% 12315 76.8% 22609 75.7% 31687 76.3% <.0001
Blood diseases 5057 11.1% 2340 14.6% 4078 13.7% 7940 19.1% <.0001
Mental disorders 9914 21.8% 3697 23.1% 5828 19.5% 9029 21.7% <.0001
Nervous system 17118 37.6% 6681 41.7% 12977 43.5% 19893 47.9% <.0001
Eye diseases 2459 5.4% 1114 7.0% 1591 5.3% 5586 13.5% <.0001
Ear diseases 1037 2.3% 405 2.5% 500 1.7% 1697 4.1% <.0001
Circulatory
disorders 29728 65.4% 11571 72.2% 21392 71.7% 30546 73.5% <.0001
Respiratory system 17461 38.4% 6344 39.6% 10815 36.2% 16040 38.6% <.0001
Digestive system 12085 26.6% 4736 29.5% 8472 28.4% 12726 30.6% <.0001
Genitourinary
system 14855 32.7% 5937 37.0% 11122 37.3% 17863 43.0% <.0001
Skin diseases 11085 24.4% 4088 25.5% 7697 25.8% 13133 31.6% <.0001
Musculoskeletal 20716 45.5% 7973 49.7% 14428 48.3% 22351 53.8% <.0001
Congenital
anomalies 1039 2.3% 432 2.7% 832 2.8% 1374 3.3% <.0001
Injury/Poisoning 8497 18.7% 3271 20.4% 5869 19.7% 9547 23.0% <.0001
Overweight/obesity 9459 20.8% 3351 20.9% 5474 18.3% 7460 18.0% <.0001
Tobacco use 3300 7.3% 1154 7.2% 1873 6.3% 2588 6.2% <.0001
Baseline
Medications
Antihistamines
1901 4.2% 621 3.9% 953 3.2% 700 1.7% <.0001
Anti-infectives 22079 48.5% 7165 44.7% 11062 37.1% 11460 27.6% <.0001
Antineoplastic 741 1.6% 290 1.8% 420 1.4% 575 1.4% <.0001
Autonomic Drugs 9890 21.7% 3356 20.9% 4916 16.5% 5326 12.8% <.0001
Anti-coagulation 2992 6.6% 1298 8.1% 2061 6.9% 2556 6.2% <.0001
Cardiovascular 26932 59.2% 9287 57.9% 15227 51.0% 14721 35.4% <.0001
Central Nervous
System Agents 21804 47.9% 7346 45.8% 11175 37.4% 11432 27.5% <.0001
Electrolytic agents 7943 17.5% 3116 19.4% 4666 15.6% 5156 12.4% <.0001
Respiratory Tract 5947 13.1% 1908 11.9% 2713 9.1% 2797 6.7% <.0001
Eye, Ear, Nose,
and Throat (EENT) 8787 19.3% 3010 18.8% 4802 16.1% 5403 13.0% <.0001
Gastrointestinal 9072 20.0% 3313 20.7% 5143 17.2% 5633 13.6% <.0001
Skin and Mucous
Membrane Agents 8427 18.5% 2856 17.8% 4402 14.7% 4679 11.3% <.0001
Smooth Muscle
Relaxants 754 1.7% 293 1.8% 436 1.5% 572 1.4% <.0001
Vitamins 1235 2.7% 474 3.0% 713 2.4% 734 1.8% <.0001
T2D Medication
Used <.0001
Metformin 35308 77.6% 11176 69.7% 19277 64.6% NA NA
Sulfonylurea 5461 12.0% 2166 13.5% 4751 15.9% NA NA
TZD 675 1.5% 321 2.0% 642 2.2% NA NA
82

SGLT2i 202 0.4% 165 1.0% 350 1.2% NA NA
DPP4i 1037 2.3% 712 4.4% 1258 4.2% NA NA
GLP 261 0.6% 173 1.1% 352 1.2% NA NA
Insulin 2541 5.6% 1323 8.3% 3228 10.8% NA NA
TZD=thiazolidinedione; EPO=exclusive provider organization; HMO=health maintenance organization;
POS=point-of-service; PPO=preferred provider organization; SD=standard deviation.

83

Table 4.2. Patient characteristics associated with treatment initiation:
Comparison of Base Model with interaction term Models
N=132,925

Outcome: time to
treatment
Cox model
(Base)
Cox model
(alternative model 1)
Cox model (alternative
model 2)
Parameters HR p value HR p value HR p value
Comparison of results across variables use in interactions
Age group (vs. 18-23)

24-29 years
2.155 <.0001 5.441 <.0001
30-35 years
2.329 <.0001 6.355 <.0001
36-41 years
2.47 <.0001 2.574 <.0001
42-47 years
2.288 <.0001 1.980 0.0026
48-53 years
2.146 <.0001 1.366 0.1666
54-59 years
1.939 <.0001 1.447 0.101
60-65 years
1.797 <.0001 1.174 0.4774
66-71 years
1.573 <.0001 1.721 0.0163
72-79 years
1.35 <.0001 1.582 0.0438
>=80 years
1.008 0.8864 1.336 0.2136
Index A1c group (vs.
7%-8%)

8% - 9%
1.383 <.0001

9% - 10%
1.679 <.0001

>10%
2.237 <.0001

Charlson index
group (vs. 1)

2
0.917 <.0001

3
0.852 <.0001

4-6
0.768 <.0001

>=7
0.715 <.0001

AGE (in years)
0.976 <.0001

Index A1c
1.057 <.0001
1.103
<.0001
Charlson index
1.061 <.0001 1.231 0.0635
Age * Index A1c
1.001 <.0001
Age * Charlson
Index

0.998 <.0001
Age group * Index
A1c

24-29 years*A1c
0.937 0.0005
30-35 years* A1c
0.936 <.0001
36-41 years* A1c
1.035 0.0364
84

42-47 years* A1c
1.048 0.004
48-53 years* A1c
1.085 <.0001
54-59 years* A1c
1.067 <.0001
60-65 years* A1c
1.086 <.0001
66-71 years* A1c
1.023 0.1729
72-79 years* A1c
1.011 0.5022
>=80 years* A1c
0.990 0.5724
Age group*Charlson

24-29 years*charlson
0.806 0.0872
30-35 years*charlson
0.747 0.012
36-41 years*charlson
0.735 0.0068
42-47 years*charlson
0.776 0.0247
48-53 years*charlson
0.775 0.0234
54-59 years*charlson
0.766 0.0175
60-65 years*charlson
0.754 0.012
66-71 years*charlson
0.751 0.0108
72-79 years*charlson
0.756 0.0126
>=80 years*charlson
0.766 0.0174
Comparison of results across variables NOT use in interactions
Index year
1.005 0.0005 1.001 0.2844 1.003 0.0131
Female
0.963 <.0001 0.941 <.0001 0.954 <.0001
Health plan (vs.
POS)
EPO
0.904 <.0001 0.902 <.0001 0.9 <.0001
HMO
0.917 <.0001 0.869 <.0001 0.907 <.0001
INDEMNITY
0.793 0.0002 0.784 <.0001 0.795 0.0002
OTHER
0.698 <.0001 0.655 <.0001 0.691 <.0001
PPO
0.968 0.0528 0.937 0.0001 0.959 0.0154
Prior medical costs
(vs. <=1000)

1000 - 5000
1.087 <.0001 1.099 <.0001 1.086 <.0001
5000 - 10000
1.062 <.0001 1.065 <.0001 1.054 <.0001
>10000
1.054 <.0001 1.055 <.0001 1.051 <.0001
Prior drug costs (vs.
<=1000)
1000 - 5000
1.052 <.0001 1.022 0.0304 1.032 0.0016
5000 - 10000
1.007 0.5778 0.971 0.0264 0.983 0.2048
>10000
1.013 0.3146 0.982 0.1738 0.995 0.6852
Baseline drug profile

Antihistamines
1.015 0.4192 1.022 0.2182 1.02 0.2789
Anti-infectives
1.13 <.0001 1.131 <.0001 1.129 <.0001
85

Antineoplastic
1.011 0.6801 1.019 0.4923 1.018 0.5034
Autonomic Drugs
1.079 <.0001 1.081 <.0001 1.079 <.0001
Anti-coagulation
1.071 <.0001 1.06 <.0001 1.072 <.0001
Cardiovascular
1.582 <.0001 1.604 <.0001 1.568 <.0001
Central Nervous
System Agents 1.148 <.0001 1.156 <.0001 1.146 <.0001
Electrolytic agents
1.09 <.0001 1.083 <.0001 1.086 <.0001
Respiratory Tract
1 0.9699 1.006 0.5966 0.998 0.8926
Eye, Ear, Nose, and
Throat (EENT) 1.034 0.0003 1.02 0.0294 1.027 0.0038
Gastrointestinal
1.072 <.0001 1.065 <.0001 1.067 <.0001
Skin and Mucous
Membrane Agents 1.038 <.0001 1.037 0.0001 1.038 <.0001
Smooth Muscle
Relaxants 1.077 0.0051 1.056 0.0423 1.071 0.01
Vitamins
0.911 <.0001 0.896 <.0001 0.909 <.0001
EPO=exclusive provider organization; HMO=health maintenance organization; POS=point-of-service;
PPO=preferred provider organization.

Table 4.3. Patient characteristics associated with days of delay in treatment initiation:
Comparison of Base Model with interaction term Models
N=132,925

Outcome: time to
treatment
OLS model
(Base)
OLS model
(alternative model 1)
OLS model (alternative
model 2)
Parameters estimate p value estimate p value estimate p value
Comparison of results across variables use in interactions
Age group (vs. 18-23)

24-29 years
-301.80 <.0001 -807.03 <.0001
30-35 years
-311.12 <.0001 -833.97 <.0001
36-41 years
-323.81 <.0001 -446.18 <.0001
42-47 years
-296.75 <.0001 -381.30 0.0001
48-53 years
-269.72 <.0001 -241.72 0.0148
54-59 years
-224.97 <.0001 -228.70 0.0208
60-65 years
-206.78 <.0001 -158.23 0.1117
66-71 years
-109.00 <.0001 -203.60 0.0398
72-79 years
-41.69 0.0512 -109.01 0.2736
>=80 years
-7.55 0.7274 -101.94 0.3172
86

Index A1c group (vs.
7%-8%)

8% - 9%
-131.52 <.0001

9% - 10%
-202.77 <.0001

>10%
-291.03 <.0001

Charlson index
group (vs. 1)

2
25.37 <.0001

3
43.77 <.0001

4-6
75.30 <.0001

>=7
64.67 <.0001

AGE (in years)
-21.21 <.0001

Index A1c
12.03 <.0001 -46.83 <.0001
Charlson index
34.17 <.0001 -114.77 0.0147
Age * Index A1c
-0.63 <.0001
Age * Charlson
Index

-0.23 0.0168
Age group * Index
A1c

24-29 years*A1c
35.21 0.0003
30-35 years* A1c
35.95 <.0001
36-41 years* A1c
-7.22 0.3866
42-47 years* A1c
-7.87 0.3319
48-53 years* A1c
-20.53 0.0104
54-59 years* A1c
-18.31 0.0219
60-65 years* A1c
-24.66 0.0023
66-71 years* A1c
-5.63 0.4841
72-79 years* A1c
-5.52 0.4992
>=80 years* A1c
-0.38 0.9649
Age group*Charlson

24-29 years*charlson
124.71 0.0227
30-35 years*charlson
142.29 0.0039
36-41 years*charlson
157.31 0.0011
42-47 years*charlson
133.35 0.005
48-53 years*charlson
136.11 0.004
54-59 years*charlson
144.89 0.0022
60-65 years*charlson
144.15 0.0023
66-71 years*charlson
136.00 0.0039
72-79 years*charlson
124.61 0.0082
>=80 years*charlson
121.87 0.0097
Comparison of results across variables NOT use in interactions
Index year
-50.22 <.0001 -49.38 <.0001 -49.93 <.0001
87

Female
9.27 0.0039 15.08 <.0001 12.27 0.0001
Health plan (vs.
POS)
EPO
33.98 <.0001 36.07 <.0001 35.58 <.0001
HMO
76.29 <.0001 106.18 <.0001 75.28 <.0001
INDEMNITY
88.58 0.0021 99.46 0.0006 89.46 0.0019
OTHER
137.73 <.0001 176.27 <.0001 136.13 <.0001
PPO
2.85 0.7302 27.99 0.0006 1.54 0.8529
Prior medical costs
(vs. <=1000)

1000 - 5000
-25.40 <.0001 -28.72 <.0001 -24.65 <.0001
5000 - 10000
-7.06 0.2351 -9.04 0.1301 -5.43 0.3626
>10000
18.27 <.0001 16.66 0.0001 18.83 <.0001
Prior drug costs (vs.
<=1000)
1000 - 5000
-44.35 <.0001 -38.85 <.0001 -40.24 <.0001
5000 - 10000
-40.28 <.0001 -33.94 <.0001 -35.52 <.0001
>10000
-53.53 <.0001 -50.08 <.0001 -51.37 <.0001
Baseline drug profile

Antihistamines
-26.61 0.0037 -29.64 0.0013 -26.94 0.0033
Anti-infectives
-55.45 <.0001 -57.33 <.0001 -56.15 <.0001
Antineoplastic
-1.41 0.9127 -1.39 0.9145 -3.18 0.8053
Autonomic Drugs
-35.05 <.0001 -36.18 <.0001 -35.09 <.0001
Anti-coagulation
-23.67 0.0004 -18.13 0.0074 -23.00 0.0007
Cardiovascular
-178.35 <.0001 -183.21 <.0001 -176.40 <.0001
Central Nervous
System Agents -65.37 <.0001 -69.30 <.0001 -65.37 <.0001
Electrolytic agents
-40.01 <.0001 -38.52 <.0001 -38.83 <.0001
Respiratory Tract
9.30 0.1016 5.64 0.3218 8.76 0.1225
Eye, Ear, Nose, and
Throat (EENT) -13.16 0.0034 -9.09 0.0442 -11.17 0.0131
Gastrointestinal
-23.92 <.0001 -23.11 <.0001 -23.79 <.0001
Skin and Mucous
Membrane Agents -19.13 <.0001 -18.09 <.0001 -18.85 <.0001
Smooth Muscle
Relaxants -33.77 0.0083 -29.23 0.0231 -30.93 0.0158
Vitamins
23.99 0.0207 24.13 0.0206 25.33 0.0148
EPO=exclusive provider organization; HMO=health maintenance organization; POS=point-of-service;
PPO=preferred provider organization.

88

Table 4.4a. Impact of treatment delay on risk of clinical outcomes
N=132,925

Cox model Cox model Cox model Cox model
Outcome: Stroke or AMI
(event=16,315)
Stroke
(event=12,192)
AMI
(event=4,123)
All-cause
hospitalization
(event=36,791)
Parameters HR p value HR p value HR p
value
HR p value
Treated on time
(vs. delayed or
never treated)
0.449 <.0001 0.436 <.0001 0.466 <.0001 0.506 <.0001
Index A1c group
(vs. 7%-8%)

8% - 9%
1.127 <.0001 1.092 0.0009 1.231 <.0001 1.146 <.0001
9% - 10%
1.257 <.0001 1.18 <.0001 1.454 <.0001 1.229 <.0001
>10%
1.464 <.0001 1.376 <.0001 1.664 <.0001 1.409 <.0001
index year
0.933 <.0001 0.895 <.0001 1.082 <.0001 1.015 <.0001
female
0.832 <.0001 0.926 <.0001 0.625 <.0001 0.956 <.0001
Health plan (vs.
POS)
EPO
0.99 0.8075 1.002 0.9701 0.992 0.9163 0.986 0.548
HMO
1.27 <.0001 1.265 <.0001 1.284 <.0001 1.328 <.0001
INDEMNITY
0.85 0.2155 0.854 0.2971 0.901 0.6923 0.974 0.7551
OTHER
1.155 <.0001 1.197 <.0001 0.998 0.9755 1.136 <.0001
PPO
1.376 <.0001 1.46 <.0001 1.067 0.4397 1.298 <.0001
Age group (vs.
18-23)
24-29 years
4.776 0.0367 8.048 0.0421 1.414 0.7772 1.411 0.0012
30-35 years
8.076 0.0034 10.777 0.0172 5.615 0.0915 1.131 0.1983
36-41 years
11.497 0.0005 13.096 0.0095 10.78 0.0181 0.868 0.124
42-47 years
15.748 <.0001 18.804 0.003 13.477 0.0095 0.837 0.0484
48-53 years
19.862 <.0001 25.999 0.001 13.896 0.0086 0.891 0.1957
54-59 years
25.802 <.0001 34.083 0.0004 17.448 0.0043 0.984 0.8565
60-65 years
30.842 <.0001 43.997 0.0001 16.082 0.0056 1.092 0.3199
66-71 years
35.36 <.0001 49.538 <.0001 19.242 0.0032 1.152 0.1115
72-79 years
40.77 <.0001 58.591 <.0001 19.967 0.0028 1.236 0.0174
>=80 years
46.48 <.0001 65.462 <.0001 24.231 0.0015 1.57 <.0001
Prior medical
costs (vs.
<=1000)
1000 - 5000
1.286 <.0001 1.374 <.0001 0.987 0.8976 0.784 <.0001
5000 - 10000
1.501 <.0001 1.545 <.0001 1.311 0.0078 1.312 <.0001
>10000
3.59 <.0001 3.184 <.0001 4.649 <.0001 6.692 <.0001
89

Prior drug costs
(vs. <=1000)
1000 - 5000
0.873 <.0001 0.858 <.0001 0.931 0.1591 0.892 <.0001
5000 - 10000
0.871 <.0001 0.895 0.0036 0.843 0.0099 0.916 <.0001
>10000
0.903 0.0014 0.931 0.0536 0.837 0.0058 0.97 0.1509
Charlson index
group (vs. 1)
2
1.445 <.0001 1.508 <.0001 1.214 <.0001 1.16 <.0001
3
1.672 <.0001 1.759 <.0001 1.322 <.0001 1.355 <.0001
4-6
1.834 <.0001 1.877 <.0001 1.519 <.0001 1.528 <.0001
>=7
2.207 <.0001 2.323 <.0001 1.612 <.0001 1.891 <.0001
Baseline drug
profile
Antihistamines
0.953 0.2845 1.047 0.3546 0.686 0.0006 1.039 0.1727
Anti-infectives
0.971 0.1493 0.971 0.2078 0.989 0.7843 1.007 0.6094
Antineoplastic
0.855 0.0114 0.854 0.0256 0.895 0.3918 1.039 0.3049
Autonomic Drugs
1.03 0.1981 1.013 0.6086 1.075 0.1102 1.112 <.0001
Anti-coagulation
1.475 <.0001 1.409 <.0001 1.512 <.0001 1.151 <.0001
Cardiovascular
1.174 <.0001 1.208 <.0001 1.083 0.0912 0.979 0.1748
Central Nervous
System Agents 1.044 0.0338 1.067 0.0062 0.996 0.9272 1.136 <.0001
Electrolytic
agents 1.031 0.1632 0.989 0.6462 1.176 0.0002 1.193 <.0001
Respiratory Tract
0.927 0.0153 0.894 0.0021 1.029 0.6288 0.957 0.022
Eye, Ear, Nose,
and Throat
(EENT) 0.974 0.2372 0.989 0.6594 0.943 0.2028 0.932 <.0001
Gastrointestinal
1.014 0.5206 1.022 0.4011 0.974 0.5545 1.027 0.0682
Skin and Mucous
Membrane Agents 1 0.9979 1.002 0.9544 0.985 0.7497 1.01 0.5171
Smooth Muscle
Relaxants 1.055 0.3146 1.058 0.3485 0.97 0.7972 1.13 0.0005
Vitamins
0.967 0.5374 0.993 0.9079 0.93 0.5161 1.044 0.1986
AMI=acute myocardial infarction; EPO=exclusive provider organization; HMO=health maintenance organization;
POS=point-of-service; PPO=preferred provider organization.

Table 4.4b. Impact of treatment delay on risk of clinical outcomes
N=132,925
90

Cox model Cox model Cox model Cox model
Outcome: Stroke or AMI
(event=16,315)
Stroke
(event=12,192)
AMI
(event=4,123)
All-cause
hospitalization
(event=36,791)
Parameters HR p value HR p value HR p value HR p value
Time to
treatment
(in months)
1.01 <.0001 1.012 <.0001 1.006 <.0001 1.011 <.0001
Index A1c group
(vs. 7%-8%)
8% - 9%
1.043 0.0663 1.003 0.924 1.155 0.0013 1.07 <.0001
9% - 10%
1.115 0.0004 1.037 0.3155 1.318 <.0001 1.104 <.0001
>10%
1.227 <.0001 1.14 <.0001 1.436 <.0001 1.205 <.0001
Index year
0.942 <.0001 0.904 <.0001 1.096 <.0001 1.02 <.0001
Female
0.835 <.0001 0.931 0.0001 0.629 <.0001 0.956 <.0001
Health plan (vs.
POS)
EPO
1.037 0.3571 1.05 0.2909 1.031 0.6839 1.014 0.5601
HMO
1.297 <.0001 1.289 <.0001 1.302 <.0001 1.362 <.0001
INDEMNITY
0.899 0.4184 0.898 0.4783 0.95 0.8446 1.027 0.7468
OTHER
1.26 <.0001 1.3 <.0001 1.104 0.099 1.212 <.0001
PPO
1.409 <.0001 1.493 <.0001 1.091 0.2996 1.305 <.0001
Age group (vs.
18-23)
24-29 years
3.997 0.0644 6.738 0.0639 1.176 0.8949 1.205 0.0804
30-35 years
6.66 0.0079 8.902 0.0291 4.555 0.1381 0.941 0.5224
36-41 years
9.455 0.0015 10.796 0.0168 8.676 0.0317 0.725 0.0005
42-47 years
13.175 0.0003 15.853 0.0054 10.912 0.0172 0.709 0.0001
48-53 years
17.062 <.0001 22.531 0.0017 11.498 0.0148 0.77 0.0032
54-59 years
22.844 <.0001 30.395 0.0006 14.824 0.0071 0.874 0.127
60-65 years
27.865 <.0001 40.184 0.0002 13.917 0.0086 0.992 0.9305
66-71 years
33.203 <.0001 47.036 0.0001 17.308 0.0044 1.086 0.3538
72-79 years
39.707 <.0001 57.953 <.0001 18.531 0.0036 1.204 0.0367
>=80 years
47.905 <.0001 68.393 <.0001 23.885 0.0015 1.605 <.0001
Prior medical
costs (vs.
<=1000)
1000 - 5000
1.223 <.0001 1.3 <.0001 0.945 0.5652 0.757 <.0001
5000 - 10000
1.438 <.0001 1.475 <.0001 1.263 0.0217 1.275 <.0001
>10000
3.448 <.0001 3.036 <.0001 4.485 <.0001 6.604 <.0001
Prior drug costs
(vs. <=1000)
1000 - 5000
0.846 <.0001 0.83 <.0001 0.915 0.0775 0.866 <.0001
5000 - 10000
0.853 <.0001 0.874 0.0004 0.833 0.0056 0.896 <.0001
91

>10000
0.883 <.0001 0.909 0.0095 0.828 0.0033 0.945 0.0061
Charlson index
group (vs. 1)
2
1.49 <.0001 1.551 <.0001 1.252 <.0001 1.191 <.0001
3
1.757 <.0001 1.851 <.0001 1.389 <.0001 1.405 <.0001
4-6
1.979 <.0001 2.023 <.0001 1.637 <.0001 1.628 <.0001
>=7
2.38 <.0001 2.512 <.0001 1.76 <.0001 2.023 <.0001
Baseline drug
profile
Antihistamines
0.961 0.3747 1.059 0.2495 0.687 0.0006 1.038 0.1765
Anti-infectives
0.939 0.0015 0.937 0.0051 0.957 0.2741 0.977 0.0754
Antineoplastic
0.866 0.0202 0.862 0.036 0.898 0.4093 1.05 0.1865
Autonomic Drugs
1.002 0.923 0.988 0.6505 1.051 0.2721 1.091 <.0001
Anti-coagulation
1.48 <.0001 1.413 <.0001 1.493 <.0001 1.148 <.0001
Cardiovascular
1.001 0.9505 1.02 0.4561 0.935 0.1417 0.868 <.0001
Central Nervous
System Agents 0.996 0.8495 1.014 0.5406 0.957 0.2687 1.09 <.0001
Electrolytic
agents 1.023 0.3033 0.981 0.4489 1.161 0.0006 1.18 <.0001
Respiratory Tract
0.931 0.0222 0.899 0.0032 1.032 0.6018 0.962 0.042
Eye, Ear, Nose,
and Throat
(EENT) 0.967 0.1314 0.982 0.4835 0.933 0.135 0.921 <.0001
Gastrointestinal
0.994 0.7773 1 0.9863 0.953 0.2855 1.008 0.5701
Skin and Mucous
Membrane Agents 0.985 0.5209 0.987 0.6166 0.971 0.5354 1 0.9909
Smooth Muscle
Relaxants 1.054 0.3279 1.068 0.2701 0.963 0.7536 1.124 0.0008
Vitamins
0.986 0.7898 1.011 0.8613 0.952 0.6635 1.057 0.0944
AMI=acute myocardial infarction; EPO=exclusive provider organization; HMO=health maintenance organization;
POS=point-of-service; PPO=preferred provider organization.

Table 4.4c. Impact of treatment delay on risk of clinical outcomes
N=132,925

Cox model Cox model Cox model Cox model
Outcome: Stroke or AMI
(event=16,315)
Stroke
(event=12,192)
AMI
(event=4,123)
All-cause
hospitalization
(event=36,791)
Parameters HR p value HR p value HR p value HR p value
Time to treatment
(vs. no treatment)

Month 1 0.53 <.0001 0.522 <.0001 0.531 <.0001 0.586 <.0001
92

Months 2 – 6
0.531 <.0001 0.517 <.0001 0.548 <.0001 0.593 <.0001
Month 7 and
beyond 0.351 <.0001 0.338 <.0001 0.378 <.0001 0.392 <.0001
Index A1c group
(vs. 7%-8%)

8% - 9%
1.095 <.0001 1.058 0.0326 1.201 <.0001 1.112 <.0001
9% - 10%
1.197 <.0001 1.121 0.0018 1.398 <.0001 1.169 <.0001
>10%
1.358 <.0001 1.27 <.0001 1.566 <.0001 1.308 <.0001
index year
0.931 <.0001 0.893 <.0001 1.079 <.0001 1.013 <.0001
female
0.834 <.0001 0.929 <.0001 0.627 <.0001 0.959 <.0001
Health plan (vs.
POS)
EPO
0.999 0.9879 1.011 0.8095 0.999 0.9937 0.992 0.7167
HMO
1.291 <.0001 1.287 <.0001 1.301 <.0001 1.344 <.0001
INDEMNITY
0.863 0.2617 0.867 0.348 0.912 0.7265 0.989 0.8965
OTHER
1.157 <.0001 1.199 <.0001 0.998 0.9672 1.132 <.0001
PPO
1.389 <.0001 1.474 <.0001 1.076 0.3833 1.301 <.0001
Age group (vs. 18-
23)
24-29 years
4.665 0.0398 7.848 0.0451 1.389 0.7883 1.385 0.0022
30-35 years
7.822 0.0039 10.428 0.0191 5.462 0.0969 1.095 0.3431
36-41 years
11.144 0.0006 12.677 0.0106 10.496 0.0194 0.841 0.0608
42-47 years
15.395 0.0001 18.39 0.0033 13.179 0.0101 0.817 0.0249
48-53 years
19.636 <.0001 25.727 0.001 13.724 0.0089 0.877 0.1409
54-59 years
25.663 <.0001 33.958 0.0004 17.334 0.0044 0.976 0.7853
60-65 years
31 <.0001 44.35 0.0001 16.113 0.0055 1.094 0.3094
66-71 years
35.721 <.0001 50.167 <.0001 19.372 0.0031 1.159 0.096
72-79 years
41.486 <.0001 59.794 <.0001 20.211 0.0027 1.253 0.0113
>=80 years
47.122 <.0001 66.569 <.0001 24.476 0.0014 1.583 <.0001
Prior medical costs
(vs. <=1000)
1000 - 5000
1.289 <.0001 1.378 <.0001 0.988 0.9014 0.785 <.0001
5000 - 10000
1.504 <.0001 1.55 <.0001 1.311 0.0077 1.313 <.0001
>10000
3.601 <.0001 3.197 <.0001 4.656 <.0001 6.703 <.0001
Prior drug costs
(vs. <=1000)
1000 - 5000
0.867 <.0001 0.851 <.0001 0.927 0.1321 0.886 <.0001
5000 - 10000
0.866 <.0001 0.89 0.0021 0.84 0.0085 0.911 <.0001
>10000 0.894 0.0005 0.92 0.0247 0.832 0.0042 0.962 0.0615
Charlson index
group (vs. 1)
2
1.445 <.0001 1.508 <.0001 1.217 <.0001 1.162 <.0001
3
1.676 <.0001 1.762 <.0001 1.328 <.0001 1.358 <.0001
93

4-6
1.842 <.0001 1.888 <.0001 1.527 <.0001 1.534 <.0001
>=7
2.223 <.0001 2.342 <.0001 1.626 <.0001 1.902 <.0001
Baseline drug
profile
Antihistamines
0.95 0.2504 1.044 0.3867 0.682 0.0005 1.036 0.2108
Anti-infectives
0.966 0.0866 0.966 0.134 0.985 0.7003 1 0.9868
Antineoplastic
0.854 0.0109 0.854 0.0258 0.894 0.3877 1.037 0.3263
Autonomic Drugs
1.022 0.3345 1.007 0.8011 1.068 0.1431 1.107 <.0001
Anti-coagulation
1.48 <.0001 1.414 <.0001 1.515 <.0001 1.156 <.0001
Cardiovascular
1.142 <.0001 1.172 <.0001 1.059 0.2247 0.959 0.0067
Central Nervous
System Agents 1.032 0.1181 1.054 0.0253 0.987 0.7474 1.125 <.0001
Electrolytic agents
1.027 0.2272 0.985 0.542 1.171 0.0003 1.185 <.0001
Respiratory Tract
0.925 0.0124 0.893 0.0018 1.028 0.6495 0.954 0.0146
Eye, Ear, Nose, and
Throat (EENT) 0.969 0.1627 0.984 0.5243 0.939 0.1762 0.928 <.0001
Gastrointestinal
1.014 0.5373 1.021 0.4082 0.972 0.5317 1.026 0.0749
Skin and Mucous
Membrane Agents 0.998 0.9169 1 0.987 0.982 0.6952 1.006 0.6781
Smooth Muscle
Relaxants 1.049 0.3718 1.052 0.3959 0.965 0.7672 1.123 0.0009
Vitamins
0.974 0.634 1.001 0.9917 0.934 0.5405 1.045 0.1838
AMI=acute myocardial infarction; EPO=exclusive provider organization; HMO=health maintenance organization;
POS=point-of-service; PPO=preferred provider organization.
94

CHAPTER 5. Conclusions

Lifestyle change in combination with metformin monotherapy is generally recommended
as a preferred first-line treatment for patients with T2D. Due to the progressive nature of T2D,
most patients will require a combination of medications over time. This progression to more
complex treatment regimens and disease management protocols are challenging for patients.
Non-adherence and non-persistence to prescribed ADMs are common and remain a barrier to
optimal health outcomes. On the other hand, the primary goal of T2D treatment is reduction of
blood glucose to prevent long-term microvascular and macrovascular complications. In the
clinical guidelines, A1c of >=7% is generally served as a call to action to initiate or change
therapy. Despite the importance of glycemic control, many patients do not receive treatment in a
timely manner. The failure of receiving treatment when clinically indicated has been termed
“clinical inertia”.
In Chapter 2, we evaluated the factors associated with treatment persistence, including
the combination of medications used as initial therapy, and estimated the impact of persistence
on clinical and cost outcomes among newly treated T2D patients. Our study found that 44.3% of
patients discontinued treatment with an average duration of therapy just over one year following
the initial treatment. This study also confirmed that prescribing patterns were consistent with
current clinical guidelines that recommend metformin as the preferred first-line treatment. Our
study shows that patients with longer duration of therapy following the initial treatment are at
reduced risk of stroke or AMI, and hospitalization. Importantly, these findings advance our
understanding of the impact of non-persistence by suggesting that patients with early tight
diabetes control may have both early and long-term benefits. In our analysis, a one month
95

increase in treatment duration reduced risk of AMI, stroke and hospitalization within the study
period. Our results highlight the importance of establishing the role of treatment persistence in
preventing complications earlier, from the onset of ADMs.
In Chapter 3, we further explored predictors of treatment persistence and estimated the
impact of persistence on clinical and cost outcomes among T2D patients changing drug therapy.
T2D typically requires adjustments to pharmacologic antidiabetic therapies over time as insulin
production declines and/or insulin resistance increases. After the first T2D treatment attempt,
45% of the patients changed their initial therapy (augmented or switched), or restarted
discontinued treatment. Our study finds that 57% of patients discontinued treatment with an
average duration of therapy nearly one year following the second treatment attempt. To be
consistent with the findings in Chapter 2, our study shows that patients with longer duration of
therapy following the second treatment attempt are at reduced risk of stroke or AMI, and
hospitalization. Taken together, these findings suggest that treating T2D patients more
intensively at an earlier stage may result in long-term improvements in patient health outcomes.
Healthcare professionals should work with patients from the beginning of therapy to ensure that
barriers to persistence are identified and reduced in order to limit the impact of non-persistence
on poor future outcomes.
Prior studies have indicated that clinical inertia is evident at early stages of the disease. In
Chapter 4, we finds that more than half of patients had no treatment initiated within 6 months
following A1C of >=7%. This finding is concerning given that the population had relatively high
average baseline A1C levels (mean: 8.6%) and that all had insurance. Results in Chapter 4 shows
that delays in initiating the first-line treatment after elevated A1c could also significantly
increase risks of AMI, stroke, and all-cause hospitalization. Failure to intervene early when it is
96

indicated may have a lasting effect on risks of developing diabetes-related complications. Our
results highlight the importance of timely treatment initiation in preventing complications earlier,
from the onset of ADMs.
Throughout this dissertation, we focused on the impacts of treatment persistence and
delays in starting treatment on patient outcomes in T2D. Our research work suggests that treating
T2D patients more intensively at an earlier stage may result in long-term improvements in
patient health outcomes. Healthcare professionals should work with patients from the beginning
of therapy to ensure that challenges of clinical inertia are addressed in order to limit its impact on
poor future outcomes. In addition, current performance measures for diabetes care address
patient A1c below certain targets (e.g., <7%). Our findings indicate that there may be a need to
incorporate measures of time to treatment initiation into performance measures. As availability
of health data systems and patient-reported data grow, quality measures related to timely
initiation of treatment may catalyze new efforts to address clinical inertia and improve patient
health outcomes.

Abstract (if available)

Abstract Type 2 diabetes (T2D) is a costly, chronic, and progressive disease characterized by insufficient insulin production, insulin resistance or both. Treatment typically requires pharmacologic antidiabetic therapies. Persistence with T2D medications is critical to achieving optimal glycemic control and preventing adverse clinical events. Although a wide array of medications is available for treating type 2 diabetes (T2D), approximately half of patients with T2D discontinue therapy. Lack of persistence is an important barrier to glycemic control and is a common problem in patients with T2D. Previous studies have linked non-persistence and medication nonadherence with more T2D-related complications, increased use of healthcare resources, and higher medical costs. However, prior studies have been limited to assess persistence and adherence in specific drug class or specific treatment phase. In addition, a clear temporal relationship between discontinuation status and future clinical events have not been established in the literature. The American Diabetes Association (ADA) guidelines recommend metformin and lifestyle modifications upon initial diagnosis, aiming to reduce the patient’s A1C to < 7%. Achieving A1c targets has been shown to reduce microvascular complications, and may be associated with long-term cardiovascular benefits. However, despite the importance of timely glycemic control, a high proportion of patients experience delays in treatment initiation and clinical inertia. Previous studies mostly focused on treatment intensification following the initial treatment. Relatively fewer studies evaluated delays in initiating initial treatment after the patient’s initial elevated A1c or the effects of delaying treatment on subsequent clinical events. This dissertation focused on persistence with treatment episodes and delays in initiating treatment. ❧ The first two papers investigated factors associated with treatment persistence, and estimated the impact of persistence on clinical and cost outcomes among patients with T2D. The first paper evaluated treatment persistence following patient initial treatment episodes

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Documenting the impact of under-treatment for chronic disorders

PDF

Outcomes of antibiotic use among children with acute respiratory tract infections

PDF

Health economics and outcomes research for informed decision making in rapidly evolving therapeutic areas

PDF

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

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

PDF

Understanding primary nonadherence to medications and its associated healthcare outcomes: a retrospective analysis of electronic medical records in an integrated healthcare setting

PDF

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

PDF

The price and clinical impact of specialty drug management in the US healthcare system

PDF

Renin-angiotensin system modulation for the prevention and treatment of metabolic dysfunction

PDF

Hepatitis C in the post-interferon era: selected essays in health economics

PDF

Assessment of the impact of second-generation antipscyhotics in Medi-Cal patients with bipolar disorder using panel data fixed effect models

PDF

Molecular mechanisms of young-onset type 2 diabetes: integration of diet and multi-omics biomarkers

PDF

Compliance study of second-generation antipsychotics on patients with schizophrenia

PDF

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

PDF

Treatment of hepatitis C in era of direct acting antivirals (DAAs): selected essays in health economics

PDF

Evaluating cancer treatments with health economics methods

PDF

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

PDF

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

PDF

Three essays in health economics

Asset Metadata

Creator Xuan, Si (author)

Core Title Pharmacotherapy among patients with type 2 diabetes

School School of Pharmacy

Degree Doctor of Philosophy

Degree Program Health Economics

Publication Date 10/19/2020

Defense Date 08/28/2020

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

Tag clinical inertia,Diabetes,OAI-PMH Harvest,persistence

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor McCombs, Jeffrey (committee chair)

Creator Email sxuan@usc.edu,xuan0261@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-384551

Unique identifier UC11666526

Identifier etd-XuanSi-9049.pdf (filename),usctheses-c89-384551 (legacy record id)

Legacy Identifier etd-XuanSi-9049.pdf

Dmrecord 384551

Document Type Dissertation

Rights Xuan, Si

Type texts

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

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

Repository Name University of Southern California Digital Library

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