ESSAYS ON THE ECONOMICS OF ORGAN TRANSPLANTATION

By

Bethany I Lemont

A DISSERTATION

Michigan State University

in partial fulﬁllment of the requirements

Submitted to

for the degree of

Economics – Doctor of Philosophy

2019

ABSTRACT

ESSAYS ON THE ECONOMICS OF ORGAN TRANSPLANTATION

By

Bethany I Lemont

Chapter 1: The Eﬀect of Share 35 for Kidneys on Pediatric Transplant Candidates
Due to the shortage of donated kidneys in the United States, allocation policy is used to balance
equity and eﬃciency in distribution of these scarce resources. I analyze the eﬀect of a change in
the allocation system for deceased donor kidneys called “Share 35” that gave priority to pediatric
kidney transplant candidates over adult candidates for young deceased donor kidneys. Using data
from the Scientiﬁc Registry of Transplant Recipients on all kidney transplant candidates in the
US, I show that providing priority for pediatric candidates increased the average kidney quality
by 30% and increased the likelihood of being transplanted within a year by 20 percentage points
for pediatric candidates who are predicted to receive a deceased donor kidney in absence of Share
35. However, this policy also created an incentive for pediatric candidates with a living donor
available to them to forgo using that donor to use a deceased donor instead. These candidates, who
I estimate switch donor types, experience worse average kidney quality but no change in wait time.
Additionally using the limited long run follow-up data available, the policy does not appear to have
the unintended consequence of candidates with available living donors strategically “saving” their
living donor for their second transplants when they would no longer have pediatric priority.

Chapter 2: The Eﬀects of the Aﬀordable Care Act on the Demand for Organ Transplants
Many potential transplant candidates are unable to be registered on the wait list to receive an organ
for transplant due to their lack of insurance. With the introduction of the Aﬀordable Care Act,
some of these potential candidates obtained insurance coverage through state Medicaid expansions
or through the introduction of the private insurance marketplace. In this paper, I estimate the eﬀect
of this increase in availability of insurance coverage on these potential transplant candidates’ wait
list decisions and transplant outcomes using a diﬀerence in diﬀerences model. I ﬁnd that the state

expansions of Medicaid increased monthly wait list registrations by candidates insured through
Medicaid by about 50 percent on average for all organs, but I ﬁnd no eﬀect of the marketplace
on registrations. Additionally, I ﬁnd that for candidates insured through Medicaid, the Medicaid
expansion led to an increase in monthly deceased donor transplants for all organs, and a doubling
of monthly living donor liver transplants.

Chapter 3: Opioids and Organs: How Overdoses Aﬀect the Supply of Donors, Waiting Lists,
and Transplant Outcomes (with Stacy Dickert-Conlin, Todd Elder, and Keith Teltser)
As the number of fatal drug overdoses has rapidly grown in recent years, patients awaiting organ
transplants may be the unintended beneﬁciaries. In 2017, 70,237 people died due to a drug overdose,
5,795 transplant candidates died while waiting for an organ, and an additional 6,363 candidates
were removed from waiting lists because they were too sick to accept a transplant. In this paper,
we use mortality data from the National Vital Statistics System, merged with restricted-use data on
transplant candidates and recipients from the Scientiﬁc Registry of Transplant Recipients, to study
the extent to which the recent growth in fatal drug overdoses impacts the supply of deceased organ
donations and transplants. We ﬁnd that each opioid overdose death generates 0.019 additional organ
donors, resulting in 0.053 additional organ transplants. Nearly all of this association is concentrated
among donors aged 18-49, who account for the majority of opioid overdose victims. Somewhat
surprisingly, opioid-driven supply shocks induce limited demand-side responses.

ACKNOWLEDGEMENTS

I am sincerely grateful to Stacy Dickert-Conlin for providing mentorship, support, encouragement,
and guidance to me during this endeavor. I additionally thank Mike Conlin and John Goddeeris
for their comments, feedback, and support. I also acknowledge my co-authors on my third chapter,
Todd Elder, Keith Teltser, and Stacy Dickert-Conlin. I am grateful to Lori Jean Nichols, whose
enthusiasm and encouragement has been invaluable. Finally, I thank my parents for encouraging
me to follow my passions, and for the countless ways you’ve supported me as I’ve done so.

The data reported here have been supplied by the Hennepin Healthcare Research Institute
(HHRI) as the contractor for the Scientiﬁc Registry of Transplant Recipients (SRTR). The inter-
pretation and reporting of these data are the responsibility of the author(s) and in no way should be
seen as an oﬃcial policy of or interpretation by the SRTR or the U.S. Government.

iv

TABLE OF CONTENTS

LIST OF TABLES .
LIST OF FIGURES .

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

ix

CHAPTER 1 THE EFFECT OF SHARE 35 FOR KIDNEYS ON PEDIATRIC TRANS-

.
.

.
.

.
.

. .
.
.

PLANT CANDIDATES . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.1
Introduction .
.
1.2 Environment .
.
1.3 Conceptual Framework .
1.4 Data and Descriptive Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.5 Eﬀect of Share 35 on Pediatric Kidney Transplant Recipients . . . . . . . . . . .

1
1
5
8
. 11
. 15
1.5.1 Eﬀect of Share 35 on all Pediatric Recipients
. . . . . . . . . . . . . . . . 15
1.5.2 Crowd out in Living Donors for First Transplant . . . . . . . . . . . . . . . 18
1.5.3 Eﬀect of Share 35 on New Pediatric Transplants . . . . . . . . . . . . . . . 20
1.5.4 Eﬀect of Share 35 on Stayers . . . . . . . . . . . . . . . . . . . . . . . . . 21
. . . . . . . . . . . . . . . . . . . . . . . 25
1.5.5 Eﬀect of Share 35 for Switchers
. . . . . . . . . . 27
1.5.6
. 29
. 30

. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Potential Saving of Living Donors for Second Transplant

1.6 Robustness . .
1.7 Conclusion .

. .
. .

.
.

.
.

.
.

.
.

.
.

CHAPTER 2 THE EFFECTS OF THE AFFORDABLE CARE ACT ON THE DE-

.
.

.
.

.
.

.
.

. .
.
.

2.1
Introduction .
2.2 Environment .

2.2.1 Transplantation in the United States
2.2.2 Health Insurance before the Aﬀordable Care Act
2.2.3 Health Insurance after the Aﬀordable Care Act

MAND FOR ORGAN TRANSPLANTS . . . . . . . . . . . . . . . . . . . 32
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
. . . . . . . . . . . . . . . . . . . . . 34
. . . . . . . . . . . . . . 35
. . . . . . . . . . . . . . . 37
2.3 Data .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.4 Wait List Registrations
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.5 Deceased Donor Transplants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.6 Living Donor Transplants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.7 Conclusion .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

. .

. .

. .

. .

.
.

.

.

.

.

.

.

.

CHAPTER 3 OPIOIDS AND ORGANS: HOW OVERDOSES AFFECT THE SUPPLY
OF DONORS, WAITING LISTS, AND TRANSPLANT OUTCOMES
(with Stacy Dickert-Conlin, Todd Elder, and Keith Teltser) . . . . . . . . . . 49
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Introduction .
Institutional Details/Literature Review . . . . . . . . . . . . . . . . . . . . . . .
. 50
3.2.1 The Opioid Epidemic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.2.2 Link to Organ Donation and Transplantation . . . . . . . . . . . . . . . . . 51
3.2.3 Transplants
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

3.1
3.2

. .

.

.

.

.

.

.

.

.

v

. .

.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.5.1 Eﬀects of Supply Shocks on Waitlist Additions
3.5.2 Living Donor Crowd-Out
3.5.3 Discussion . .
.
.

3.3 Data and Descriptive Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.4 Do Overdoses Lead to Increases in the Supply of Organ Donors and Transplants? . 57
3.4.1 The Eﬀects of Opioid Overdoses on Transplants . . . . . . . . . . . . . . . 59
3.4.2 Discussion . .
. 60
3.5 Behavioral Responses to the Change in the Supply of Organs . . . . . . . . . . . . 61
. . . . . . . . . . . . . . . 61
. 63
.
. 64
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
APPENDIX A TABLES AND FIGURES FOR CHAPTER 1 . . . . . . . . . . . . . . 67
APPENDIX B TABLES AND FIGURES FOR CHAPTER 2 . . . . . . . . . . . . . . 89
APPENDIX C TABLES AND FIGURES FOR CHAPTER 3 . . . . . . . . . . . . . . 102
APPENDIX D APPENDICES FOR CHAPTER 1 . . . . . . . . . . . . . . . . . . . . 116
APPENDIX E APPENDICES FOR CHAPTER 2 . . . . . . . . . . . . . . . . . . . . 120
APPENDICES FOR CHAPTER 3 . . . . . . . . . . . . . . . . . . . . 124
APPENDIX F
.
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
REFERENCES .

. .
.
.
.
.

. .
.
.

3.6 Conclusion . .
.

APPENDICES .

.

.

.

.

.

.

.

.

.

.

.

vi

LIST OF TABLES

Table A.1: Characteristics of Pediatric, Young Adult, and Adult Kidney Candidates Before

Share 35 .

.

.

. .

. .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

Table A.2: Estimates of Overall Share 35 Eﬀect on Wait Time and Quality . . . . . . . . . . 68

Table A.3: Estimates of Coeﬃcients for Living Donor Use at First Transplant

. . . . . . . . 69

Table A.4: Characteristics of Pediatric and Young Adult Deceased Donor Kidney Recip-

ients Before and After Share 35 . . . . . . . . . . . . . . . . . . . . . . . . . . 70

Table A.5: Characteristics of Pediatric and Young Adult Living Donor Kidney Recipients

Before and After Share 35 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

Table A.6: Estimates of Share 35 Eﬀect on Deceased Donor Stayers . . . . . . . . . . . . . 72

Table A.7: Estimates of Switching Eﬀect on Wait Time and Quality . . . . . . . . . . . . . 73

Table A.8: Characteristics of Deceased Donor Recipients at Their Second Transplant

. . . . 74

Table A.9: Estimates of Coeﬃcients for Living Donor Use at Second Transplant . . . . . . . 75
Table A.10:Robustness: Coeﬃcient on Share ∗ Pediatric

. . . . . . . . . . . . . . . . . . 76

Table B.1: State Decisions to Expand Medicaid or Include Transplants as an EHB . . . . . . 89

Table B.2: Transplant Candidate Characteristics . . . . . . . . . . . . . . . . . . . . . . .

. 90

Table B.3: Monthly Wait List Additions per 1 Million Population . . . . . . . . . . . . . . 93

Table B.4: Monthly Deceased Donor Transplants per 1 Million Population . . . . . . . . . . 94

Table B.5: Monthly Living Donor Transplants per 1 Million Population . . . . . . . . . . . 95

Table C.1: Average Number of Organs Transplanted per Donor, by Mechanism of Death .

. 102

Table C.2: Average Number of Organs Transplanted per Donor, by Mechanism of Death .

. 103

Table C.3: Organ Donors by Year, Gender, and Mechanism . . . . . . . . . . . . . . . . .

. 104

Table C.4: Estimates of the Eﬀect of Opioid Overdose Deaths on Organ Donations Due

to Drug Intoxication, by Gender and Age

. . . . . . . . . . . . . . . . . . . . . 105

vii

Table C.5: Estimates of the Eﬀect of Opioid Overdose Deaths on Organ Transplants, by Organ106

Table C.6: Estimates of the Eﬀect of Opioid Overdose Deaths on Waiting List Additions

by In- Versus Out-of-Area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

Table C.7: Estimates of the Eﬀect of Opioid Overdose Deaths on Waiting List Additions

by In- Versus Out-of-Area and by Multilisting Status . . . . . . . . . . . . . .

. 108

Table C.8: Estimates of the Eﬀect of Opioid Overdose Deaths on Living-Donor Transplants 109

Table D.1: Wait Time Diﬀerences .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. 116

Table D.2: Estimates of Share 35 Eﬀect on Deceased Donor Stayers . . . . . . . . . . . . . 117

Table E.1: State Medicaid Childless Individual Income Eligibility Levels as a Percent of

the Federal Poverty Level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

Table E.2: State Medicaid Family of Three Income Eligibility Levels as a Percent of the

Federal Poverty Level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122

Table F.1: Crosswalk of Organ Procurement Organizations and Abbreviations . . . . . . . . 124

viii

LIST OF FIGURES

Figure A.1: Deceased Donor Kidney Transplants by Donor Age . . . . . . . . . . . . . . . 77

Figure A.2: Kidney Transplants by Donor Type . . . . . . . . . . . . . . . . . . . . . . .

. 77

Figure A.3: Transplant Wait Time Less than a Year Among All Donor Types

. . . . . . . . 78

Figure A.4: Average KDPI Among All Donor Types

. . . . . . . . . . . . . . . . . . . .

. 78

Figure A.5: Event Study Estimates: Wait Time Among All Donor Types . . . . . . . . . . . 79

Figure A.6: Event Study Estimates: Average KDPI Among All Donor Types . . . . . . . . . 80

Figure A.7: Percent of First Transplants That Use a Living Donor

. . . . . . . . . . . . . . 81

Figure A.8: Event Study Estimates: First Transplant Uses Living Donor . . . . . . . . . . . 82

Figure A.9: Pediatric Kidney Transplants

. . . . . . . . . . . . . . . . . . . . . . . . . .

. 82

Figure A.10:Percent of Pediatric Candidates who Receive a Transplant . . . . . . . . . . .

. 83

Figure A.11:Treatment Choice by Year . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

Figure A.12:Wait Time Less than a Year Among Deceased Donor Recipients . . . . . . . . . 84

Figure A.13:Average KDPI Among Deceased Donor Recipients

. . . . . . . . . . . . . . . 84

Figure A.14:Event Study Estimates: Wait Time Among Deceased Donor Stayers . . . . . . . 85

Figure A.15:Event Study Estimates: Average KDPI Among Deceased Donor Stayers

. . .

. 86

Figure A.16:Percent of Second Transplants That Use a Living Donor That Used a Deceased

Donor for the First Transplant . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

Figure A.17:Number of Wait List Additions by Age . . . . . . . . . . . . . . . . . . . . . . 88

Figure B.1: Registrations by Medicaid Expansion Status . . . . . . . . . . . . . . . . . .

. 96

Figure B.2: Wait List Addition Trends by Organ . . . . . . . . . . . . . . . . . . . . . . .

. 97

Figure B.3: Deceased Donor Transplants by Medicaid Expansion Status . . . . . . . . . . . 98

ix

Figure B.4: Deceased Donor Transplant Trends by Organ . . . . . . . . . . . . . . . . . . . 99

Figure B.5: Living Donor Transplant Trends by Organ . . . . . . . . . . . . . . . . . . .

. 100

Figure B.6: Living Donor Transplants by Medicaid Expansion Status

. . . . . . . . . . . . 101

Figure C.1: Donations by Mechanism of Death . . . . . . . . . . . . . . . . . . . . . . . . 110

Figure C.2: Number of Donors by Source . . . . . . . . . . . . . . . . . . . . . . . . . .

. 111

Figure C.3: Organ Donors by Year, Age, and Mechanism . . . . . . . . . . . . . . . . . . . 112

Figure C.4: Opioid Overdoses and Organ Donors by 2016 Quintile of Opioid Overdose Levels113

Figure C.5: Young Donor Recovery Rates . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

Figure C.6: Young Donor Discard Rates . . . . . . . . . . . . . . . . . . . . . . . . . . .

. 115

Figure D.1: Hypothetical Wait List Change for a Kidney from a Donor Younger than 35 . . . 118

Figure D.2: Number of Deceased Donors Who Are Younger Than 35 . . . . . . . . . . . . 119

Figure D.3: Density of Propensity Scores Before and After Matching . . . . . . . . . . . . 119

x

CHAPTER 1

THE EFFECT OF SHARE 35 FOR KIDNEYS ON PEDIATRIC TRANSPLANT

CANDIDATES

“The current system was not conceived as a uniﬁed allocation policy. When kidney transplantation
began it was only oﬀered to younger patients with minimal underlying comorbidities. Because
all potential candidates looked more or less “the same,” time spent waiting for a transplant was a
reasonable proxy for need. Over time, as immunosuppression and patient management improved
and the criteria for kidney transplantation became more inclusive, transplantation became the gold
standard for treating end stage renal failure. With this development, the number of patients waiting
for a kidney transplant swelled. In response to the increasing demand for organs, donor service
areas (DSA) and [the United Network for Organ Sharing (UNOS)] regions began to modify the
allocation system, through the incorporation policy variances, largely in an eﬀort to address what
was perceived as local problems with organ availability. The result was an allocation system that
placed an overemphasis on waiting time, progressively downplayed biologic and immunologic
considerations, failed to adapt to the changing needs of the patients being listed, and morphed
into an overlapping patchwork of variances that rendered the implementation of any meaningful
changes nearly impossible from an information technology perspective.” - Richard Formica, MD,
Yale University School of Medicine (2014)1

1.1 Introduction

Without a price mechanism to allocate deceased donor kidneys, the United States government
mandates that the United Network for Organ Sharing [UNOS] creates an allocation policy that
balances eﬃciency and equity in distributing this scarce resource. The diﬀerence between the
supply of and the demand for these kidneys is huge. In 2017, more than 35,000 candidates entered
the waiting list to receive a deceased donor kidney, but surgeons performed fewer than 15,000

1Source:

https://www.myast.org/about-ast/presidents-blog/time-change-new

-kidney-allocation-system

1

deceased donor transplants. This shortage highlights the critical eﬀect that the allocation system
has on candidates’ lives: changes to the allocation system can determine whether a candidate
ultimately receives a kidney or not.

In 2014, UNOS overhauled the kidney allocation system to implement an allocation scheme
that matches the individual expected survival of a deceased donor kidney to the individual expected
survival of a patient, which are both measured as continuous variables based on medical factors.
This system replaced one that primarily allocated kidneys ﬁrst to patients who had waited the longest,
regardless of the characteristics of the patient or the donor kidney. But, there was one element of
the pre-2014 system that included this “survival matching” in the allocation system. Speciﬁcally,
Share 35, which UNOS introduced in 2005, gave pediatric candidates priority for young deceased
donor kidneys. Share 35 recognized the longer post-transplant expected survival of children and
gave them priority for kidneys that were thought of as lasting longer after transplantation. Pediatric
transplant patients were a crude proxy for patients with high expected survival after transplantation
and donors under the age of 35 were a crude proxy for high expected survival of a deceased donor
kidney. Analyzing this ﬁrst rudimentary attempt at longevity matching, provides insights into the
intended and unintended consequences of the long run changes in the broader allocation reforms
of 2014.

The goals of Share 35 were to have children waiting fewer days before a deceased donor
transplant and also to allocate them deceased donor kidneys that were expected to last longer after
transplantation (Moudgil et al., 2013). Giving pediatric patients priority for kidneys from young
deceased donors reduces the cost of receiving a deceased donor transplant by reducing the wait
time associated with it. Share 35 also increases the expected beneﬁt of receiving a deceased donor
transplant for pediatric patients by increasing the quality of the transplant. Unsurprisingly, the
number of pediatric transplants that occurred using deceased donors that were less than 35 years
old increased dramatically from 260 in 2004 to 552 in 2006. This represents an increase in the
percent of pediatric deceased donor transplants that used donors who are less than 35 from 70
percent to 98 percent. Additionally, the median wait time for a pediatric deceased donor transplant

2

dropped from about 400 days in 2004 to only 300 days in 2006.

At ﬁrst glance, these trends suggest that Share 35 achieved its intended eﬀects of reducing
wait time and improving kidney quality for pediatric candidates.2 However, candidates can make
behavioral changes in response to policy changes that can cause unintended consequences. Many
examples of these behavioral changes have been previously documented: Sweeney (2011) and
Choi (2019) ﬁnd that candidates who receive Panel Reactive Antibody [PRA] priority in kidney
allocation, which is based on their likelihood ﬁnding a match, are less likely to use a living donor
as a result. Howard (2011) also ﬁnds that candidates are less likely to use a living donor when their
expected wait time decreases. Fernandez et al. (2013) ﬁnd that a decrease in the supply of deceased
donors leads to more living donors being used. Finally, Dickert-Conlin et al. (forthcoming) ﬁnd
that an increase in the supply of deceased donors decreases the number of living donor transplants
and that candidates are more likely to register in areas with increases in supply. It is likely that this
behavioral change is also present in pediatric candidates’ response to Share 35: between 2004 and
2006, the percent of annual pediatric transplants that used living donors dropped from 51 percent
to 35 percent.

In this paper, I investigate whether a reallocation of high quality deceased donor kidneys from
adults to children aﬀects pediatric candidate behavior and their outcomes, as measured by wait
time and kidney quality. Furthermore, I explore what the implications of the changes in pediatric
candidates’ behavior are for the eﬀectiveness of the allocation policy.

Share 35 reduces expected wait time and improves expected kidney quality of a deceased donor
transplant for pediatric patient. As a result, candidates who do not have a living donor, and need to
use a deceased donor, should experience shorter wait times and improved kidney quality. However,
because asking someone to give up a kidney may be emotionally costly, this expected improvement
2In fact, existing medical literature documents this reduction of wait time and improvement in
kidney quality (Agarwal et al., 2009; Abraham et al., 2009). However, these estimates are not
causal and do not account for the change in composition of pediatric deceased donor recipients
after Share 35 that results from some pediatric candidates switching from using living to deceased
donors. In this paper, I ﬁnd that this change in composition counteracts the expected improvements
in candidates’ outcomes.

3

to deceased donor transplants from Share 35 may also cause some pediatric candidates to now
use a deceased donor when without Share 35 they would have used a living donor. For most
pediatric patients, this living donor would have been a parent. These parents, who are involved
in the kidney transplant decision process, now avoid the costs of giving up their kidney. Living
donor kidney transplants require less wait time and typically involve higher quality kidneys than
deceased donor transplants (Hardy et al., 2009; Dale-Shall et al., 2009; Mercy Health, 2018),
so these candidates who switch are potentially worse oﬀ with respect to wait time and kidney
quality. Because candidates who receive a deceased donor transplant are likely better oﬀ but
candidates who switch away from using a living donor are likely worse oﬀ with respect to wait time
and kidney quality, the overall eﬀects of the policy change on pediatric candidates’ outcomes are
ambiguous. Additionally, because children typically have many more years to live after receiving
a transplant, Share 35 may incentivize children to “save” their willing living donor for a future
necessary additional transplant when they no longer have pediatric priority.

I use data from the Scientiﬁc Registry of Transplant Recipients [SRTR] from 1998 to 2019 that
includes the universe of all transplant candidates, recipients, deceased donors, and living donors
in the United States. By comparing the outcomes of pediatric candidates to young adults’ before
and after Share 35, I ﬁnd that Share 35 increases the likelihood that a pediatric candidate receives a
transplant within a year by 10 percentage points and slightly improves average kidney quality. By
additionally exploiting diﬀerences in observable characteristics between deceased and living donor
transplant recipients, I also estimate that pediatric deceased donor recipients who would have used
a deceased donor in absence of Share 35 are 20 percentage points more likely to receive a transplant
within a year and have an average improvement in kidney quality of 8.4 points.
In contrast, I
estimate that candidates who switch from using a living to deceased donor are transplanted with
kidneys that are 11.8 points lower quality on average, but do not have any change in their wait time.
Finally, using variation in eligibility for Share 35 priority at ﬁrst transplant, I ﬁnd that candidates
who had Share 35 priority for their ﬁrst transplant are not signiﬁcantly more likely to use a living
donor for their second transplant although they are 17 percentage points less likely to use a living

4

donor for their ﬁrst transplant.

The paper proceeds as follows:

In section 1.2, I provide an overview of pediatric kidney
transplants and a brief explanation of deceased donor kidney allocation. In section 1.3, I present a
conceptual framework which motivates my empirical methods. In section 1.4, I describe the data
that I use and provide descriptive statistics. In section 1.5, I estimate the eﬀects of Share 35 on
pediatric candidates’ transplant outcomes and behavior. In section 1.6 I show that my results are
robust to excluding groups of candidates that could potentially manipulate their treatment status.
Finally, section 1.7 concludes.

1.2 Environment

When someone’s kidneys stop functioning, they are diagnosed with End-Stage Renal Disease
[ESRD] (MedlinePlus, 2015). Although most adults develop ESRD due to diabetes or hypertension,
most pediatric patients develop ESRD due to glomerulonephritis or congenital diseases (SRTR,
2014; USRDS, 2018). Patients with ESRD have two treatment options: dialysis or a kidney
transplant (MedlinePlus, 2015). If a patient choses dialysis, a dialysis machine will clean their
blood, replicating the work that their functioning kidneys would do. However, patients using dialysis
typically need three treatments a week, where each treatment lasts for about four hours.(NKF, 2019)
A kidney transplant is associated with better survival and quality of life than dialysis, so many
patients try to receive a kidney transplant instead of using dialysis long-term (SRTR, 2014). For
pediatric patients with ESRD, kidney transplants also oﬀer another beneﬁt over dialysis: improved
growth and development (Smith & Dharnidharka, 2014; Sharma et al., 2013). Kidney specialists
recommend pediatric patients pursue transplants instead of using dialysis long-term because of
these developmental concerns (Smith & Dharnidharka, 2014; Sharma et al., 2013).

Before transplantation, physicians evaluate match quality between potential donor kidneys and
the intended transplant recipient. A necessary condition for receiving a donor kidney is that it
must have a compatible blood type with the intended recipient. Potential kidney matches are also
evaluated for how many of the six human leukocyte antigen [HLA] mismatches they have. A

5

higher number of HLA mismatches increases the chance of graft failure, which occurs when the
patient’s transplanted kidney stops functioning or is rejected by the patient’s body. Recipients take
immunosuprressive drugs after transplantation to reduce the risk of graft failure. These drugs do
not completely eliminate graft failure; 51% of deceased donor kidney transplant recipients and 34%
of living donor kidney recipients experience graft failure at some point within 10 years after their
transplant (SRTR, 2014). Because most pediatric patients will experience graft failure at some
point, this creates the need for multiple kidney transplants over their lifetime.

Kidneys used in transplants come from one of two sources: living donors or deceased donors.
Because the body only needs one working kidney, one of a living person’s two kidneys can be
donated to a transplant candidate if they are compatible. Most of the time, living donors are family
members of the transplant candidate (SRTR, 2014). Many patients who use a living donor do
not register on the deceased donor waiting list. Because these patients do not need to wait for a
deceased donor kidney oﬀer and they only need to wait for their potential living donor to be evaluated
for suitability for transplant, living donor recipients typically spend less time waiting to receive
a transplant. Many transplant centers prefer living donor kidneys over deceased donor kidneys
because living donor kidneys are associated with better patient outcomes after transplantation
(SRTR, 2014).

If a transplant candidate wants to receive a deceased donor kidney, they must register on the
waiting list.3 Because of the shortage of deceased donor kidneys, most patients will spend a
signiﬁcant amount of time waiting for a kidney before they can undergo transplantation. For
example, the median wait time for adults who registered in 2000 was 724 days and the median wait
time for pediatric candidates was 342 days.4

For the purpose of deceased donor kidney allocation, the United States is divided into 58
Donation Service Areas [DSAs] within 11 regions (UNOS, 2017a). When an Organ Procurement
3Until September 1, 2014, if a candidate was receiving a living donor kidney, they were not
required to register on the waiting list. Source: https://optn.transplant.hrsa.gov/media/
1281/policynotice_20140815.pdf

4Source: Author’s calculations from SRTR data

6

Organization [OPO], the governing body of the DSA, collects a deceased donor kidney, they enter
the donor’s information into the United Network for Organ Sharing’s [UNOS] system. This system
also contains information on all kidney transplant candidates currently on the waiting list. Using
this information, UNOS creates a unique listing of all blood type compatible candidates for that
kidney (UNOS, 2017a). The candidates on the list are ranked according to the current allocation
policy, and the highest ranked candidate on the list is oﬀered the kidney ﬁrst (UNOS, 2017a). At
the time of Share 35, allocation policy oﬀered kidneys ﬁrst to candidates who were registered in the
same DSA as the donated kidney. If the kidney was not accepted by someone in the same DSA as
the kidney, the list was expanded to candidates in the same region as the donated kidney. If no one
in the same region accepted the kidney, the list was expanded to all other compatible candidates.
Within each of these geographic categories, candidates were further split into categories based on
the number of HLA mismatches they have with the donated kidney, and then within each of these
subgroups, candidates were ranked by the number of points they have. Candidates earned points for
many diﬀerent reasons. For most candidates, a majority of their points would be earned from the
amount of time they spent waiting to receive a transplant.5 If the ﬁrst person on a list turned down
an oﬀered kidney, the OPO oﬀered the kidney to the next person on the list. The OPO continued
making oﬀers down the list until a candidate accepted or the kidney was discarded due to spending
too much time outside of the body.

Due to concerns that pediatric candidates were not receiving transplants fast enough and that they
were frequently turning down oﬀers for deceased donor kidneys from older donors, OPTN revised
the allocation system on September 28, 20056 to give additional priority to pediatric candidates
5Candidates can register on the waiting list as soon as they obtain approval from a transplant
center. However, adult candidates are listed as “inactive” and do not accrue waiting time until their
kidney function reaches a suﬃciently low level. Pediatric candidates do not have this restriction
and accrue waiting time as soon as they are placed on the list.

6This change to the policy document can ﬁrst be seen in the November 19, 2004 document with
the note that “The amendment ... shall be implemented pending distribution of appropriate notice
and programming on the UNOS Computer, if and as applicable.” The policy change also appears
with the same note in the June 24, 2005 policy document. I follow Smith et al. (2012) who list
September 28, 2005 as the start date of the policy.

7

for kidneys from deceased donors under 35 years of age (Smith et al., 2012).7 Candidates received
pediatric status if they were under 18 years old at the time that they registered on the waiting list and
retained their pediatric status until they received a transplant or were otherwise removed from the
waiting list (OPTN Policy 3.5.11.5.1; UNOS, 2010).8 When a kidney from a deceased donor under
35 years of age became available, in each subgroup of geographic category and HLA mismatches,
the allocation rule placed pediatric patients above adult patients, regardless of the number of points
assigned to the adult candidates (OPTN Policy 3.5.11.5.1).9 This allocation change is illustrated
using a hypothetical wait list in Appendix Figure D.1.

1.3 Conceptual Framework

Suppose a pediatric candidate needs a kidney transplant. If the parent of the pediatric candidate
has a compatible blood type with their child, the parent can decide to donate one of their kidneys
to their child.10 To make this decision, the parent compares the costs and beneﬁts of donating their
own kidney to the costs and beneﬁts of their child receiving a deceased donor kidney. The parent
cares about three aspects of the donated kidney that their child receives: the expected wait time
associated with the kidney (less time on dialysis is better), the expected quality (expected graft
survival) associated with the kidney, and the costs associated with the donation of the kidney. If
a parent donates one of their kidneys, they have emotional costs from going through surgery and
7The age of 35 was picked to try to give pediatric patients priority for kidneys with the best
expected graft survival. Since there was no oﬃcial measure of expected graft survival at the
time, the age of 35 was selected as a proxy for expected graft function (Anne Paschke, personal
communication, March 22, 2017).

8Organ allocation policy from the past is not publicly available. I received the deceased kidney
allocation policy from November 18, 2005 through correspondence with UNOS. This section is
based oﬀ of information contained in that policy document.

9If there were multiple pediatric patients in a group for a kidney from a donor who was less
than 35 years old, candidates who were less than 10 years old at the time of the match received one
additional point (OPTN Policy 3.5.11.5.1).

10If a parent is incompatible with their child, the child will either need to wait to receive a
deceased donor kidney or the family can choose to seek out a diﬀerent compatible living donor. For
simplicity I do not model the search for a diﬀerent living donor. Approximately 80% of pediatric
living donors are the parent of the patient so this simple model is representative of a majority of
pediatric candidates.

8

from losing a kidney.11 If a child receives a deceased donor kidney, the family does not have any
costs associated with that donation.12 Because of this, suppose the cost for a parent of donating
their own kidney is higher than the cost of using a deceased donor.

By giving pediatric candidates priority for kidneys under 35 years old, Share 35 reduces the
expected wait time for a deceased donor kidney and improves the expected quality of a deceased
donor kidney, all else equal.13 There are at least two margins along which all else may not be
equal: pediatric candidates may change the types of kidneys they are willing to accept after Share
35 or the deceased donor waiting list may change. If pediatric candidates expect to receive more
kidney oﬀers, they may become more selective with the types of kidneys they will accept.
If
Share 35 induces more pediatric candidates to register on the wait list then these marginal pediatric
wait list additions could be placed above originally wait listed pediatric candidates. Both of these
changes in candidate behavior could counteract the expected reduction in wait time from Share 35.
However, because pediatric candidates have an incentive to avoid extended dialysis use (Smith &
Dharnidharka, 2014; Sharma et al., 2013), and because the eﬀects of these two behavioral changes
should be small compared to the mechanical eﬀect of Share 35, pediatric candidates should still
expect to have shorter wait times with deceased donor transplants after Share 35. Because of this, I
hypothesize that after Share 35, pediatric deceased donor recipients will on average receive higher
quality kidneys and shorter wait times. In section 1.5.4 I empirically test this hypothesis.

To justify why I might expect to see additional pediatric candidates join the wait list for a
deceased donor kidney, suppose a parent is considering donated their kidney to their child. They
chose to donate their kidney if their expected utility from donating is greater than their expected
utility from their child joining the wait list and receiving a deceased donor kidney. Then a parent
will donate if the beneﬁts of a better quality kidney and shorter wait time for their child outweigh
11The child’s insurance will pay for the surgery that removes the kidney from the living donor

but will not cover any lost wages while the donor is recovering from the surgery.

12Transplant recipients do not pay for the donated organs that they receive.
13Appendix Figure D.2 shows that there were enough deceased donors younger than 35 for every
pediatric candidate to receive a kidney from one so pediatric candidates should not have concerns
about there not being enough kidneys from donors less than 35.

9

the additional cost of donating their own kidney. After Share 35, the expected quality and wait time
for a deceased donor kidney improves, which increases the expected utility from a deceased donor
kidney without changing the expected utility of a parent donating their own kidney. Because of
this, I hypothesize that Share 35 reduces the number of living donor transplants. I empirically test
this hypothesis in Section 1.5.2.

Living donor kidneys are associated with better quality and shorter wait times than deceased
donor kidneys before Share 35 (SRTR, 2014). Because Share 35 improves the expected wait time
and kidney quality of deceased donor kidneys, pediatric candidates who have their parent no longer
donate to them as a result of Share 35 have an ambiguous eﬀect on their wait time and kidney
quality. I estimate this eﬀect empirically in Section 1.5.5.

Because I have two competing eﬀects of Share 35 on pediatric candidates: one that improves
deceased donor transplants, and one that induces some candidates to switch from potentially higher
quality and shorter wait time living donors to deceased donors, the overall eﬀect of Share 35 on
pediatric candidates’ outcomes is ambiguous. I empirically estimate this eﬀect in Section 1.5.1.

The decision of a parent to donate to their child may also incorporate the expected future
subsequent transplants of the child. Consider instead if candidates need two transplants: one as
a pediatric candidate and one as an adult. Parents decide both if and when they want to donate
by maximizing their total expected utility of the two transplants. This leaves a parent with three
options: never donate, donate when their child has pediatric status, or donate when their child is
an adult. If Share 35 improves expected outcomes for pediatric deceased donor transplants, then
parents have less of an incentive to donate when their child is pediatric. If a parent does not donate
when their child is pediatric, the parent will be available to donate when their child is an adult.
Because of this, I hypothesize that after Share 35, pediatric candidates should be more likely to use
a living donor when they are an adult. I empirically test this in Section 1.5.6.

10

1.4 Data and Descriptive Statistics

This study uses data from the Scientiﬁc Registry of Transplant Recipients (SRTR). The SRTR
data system includes data on all donors, waitlisted candidates, and transplant recipients in the
US, submitted by the members of the Organ Procurement and Transplantation Network (OPTN).
The Health Resources and Services Administration (HRSA), US Department of Health and Human
Services, provides oversight to the activities of the OPTN and SRTR contractors. The SRTR dataset
contains individual level information on all kidney transplant candidates and transplant recipients
in the United States. From October 1987 to March 2019, I observe basic demographic information
and clinical information about every candidate who registers on the waiting list. Additionally, I
observe candidate information at any transplants they receive and the characteristics of the donor. I
also can link all recipients to any follow up visits, future waiting list registrations, and transplants.
All candidates registered on the wait list are followed until their death regardless of transplant
status, upon which I observe the date and cause of death.

I restrict my sample to kidney-only transplant recipients who are 0 to 21 years old.14 Addition-
ally, I restrict my sample to ﬁrst time transplant recipients because the choice of living or deceased
donor is more similar among candidates without prior transplants than for candidates with prior
transplants.15 I deﬁne a candidate as being “pediatric” if they are less than 18 and “young adult” if
they are 18 to 21 years old. I also restrict my sample to candidates who receive a transplant between
the years 1998 and 2013 due to inconsistency in the recording of key variables before 1998 and
the substantial change in the allocation rules in 2014.16 I deﬁne each candidate as belonging to the
“Pre-Share 35 cohort” if they received a transplant before September 28, 2005 or as belonging to
14Because dual-organs are allocated diﬀerently from only kidneys, I exclude candidates who are
listed for dual-organs from my sample. These dual-organ registrations are rare among pediatric
candidates; between October 1987 and March 2018, there have only been 118 kidney-pancreas
pediatric registrations.

1515% of pediatric candidates before Share 35 had prior kidney transplants. 25% of young adult

candidates had prior kidney transplants.

16I further restrict my sample to recipients who are not missing information on their diabetes

type. This removes 38 recipients from my sample.

11

the “Share 35 cohort” if they received a transplant after September 28, 2005.17

In this paper, I evaluate Share 35’s eﬀect on two outcomes for transplant recipients: wait
time and kidney quality.18 A candidate’s wait time captures the amount of time a candidate spends
waiting before their transplant. I observe transplant dates for all candidates who receive a transplant,
regardless of their donor type, but because not all living donor recipients register on the waiting list
before their transplant, I use the candidate’s HLA typing test date for the start of wait time instead
of registration on the waiting list.19 Additionally, I use this date to deﬁne a candidate’s age.20 The
HLA typing test is a blood test that is one of the ﬁrst things that is done when someone needs a
transplant, regardless of whether they plan to obtain a deceased or living donor transplant(NIDDK,
2018). The results of this test are used to determine the number HLA mismatches a candidate has
with a potential donor kidney.

To measure kidney quality I combine two measures of quality: the Kidney Donor Proﬁle Index
[KDPI] and the Living Kidney Donor Proﬁle Index [LKDPI].21 For deceased donor kidneys I use
the oﬃcial measure of kidney quality, KDPI. This measure is used by the new kidney allocation
17Some candidates register in the Pre-Share 35 era and receive a transplant in the Share 35 era.
I deﬁne these candidates as part of the Share 35 cohort since they receive Share 35 priority after it
is introduced.

18Previous papers (Choi, 2019; Dickert-Conlin et al., forthcoming; Teltser, forthcoming) use
graft survival instead of kidney quality. I use kidney quality because it is only calculated from
donor factors so it is independent of the transplant recipient. Graft survival can reﬂect both kidney
quality and characteristics or behaviors of the recipient.

19I present summary statistics of my measure of wait time and the measure of wait time that
uses registration date in Appendix Table D.1. The ﬁrst column looks at how the two measures
diﬀer for deceased donor recipients. The second column shows the two measures for living donor
recipients who registered on the waiting list before their transplant. For both groups of recipients,
the average test date measure of wait time is longer than the average registration version of wait
time, however, the average diﬀerence between the two measures is not statistically diﬀerent from
zero. Additionally, the average diﬀerence between the two measures across donor types is not
statistically signiﬁcant.

20I do not observe a candidate’s age at this date or a candidate’s birthday. Instead I observe a
candidate’s age in months at their transplant date and subtract from this age the number of months
between the transplant date and the HLA test date.

21Some candidates in the early years of my data are missing information in variables needed to
compute KDPI and LKDPI. I impute values for donor weight, height, creatinine level, and systolic
blood pressure using sample means conditioned on donor age, gender, and race.

12

system introduced in 2014 to help match kidneys with longer expected graft survival to patients
with longer expected post-transplant survival (OPTN, 2017).22 The KDPI uses ten donor factors
to compare the quality of a deceased donor kidney to all other recovered deceased donor kidneys
(OPTN/UNOS, 2016). These ten factors include donor age, height, weight, ethnicity, and other
medical characteristics of the donor (OPTN/UNOS, 2016).23 The KDPI incorporates all of these
factors into a single number that ranges from 0 to 100 where lower KDPI scores are associated
with a lower risk of post transplant graft failure (OPTN/UNOS, 2016). One limitation of using
this measure of kidney quality is that KDPI “is not a precise enough tool to diﬀerentiate with
high conﬁdence the quality of kidney donors with only slight diﬀerences in KDPI” (OPTN/UNOS,
2016).

For living donor kidney quality I use the measure proposed by Massie et al.

in their 2016
article. By extending the KDPI, they allow comparisons between both living and deceased donor
kidneys.24 The LKDPI is constructed on the same scale as the KDPI, but, unlike the KDPI, LKDPI
is not restricted to lie between 0 and 100. Massie et al. allow the LKDPI to be unbounded to
account for the fact that living donor kidneys have better expected graft survival than deceased
donor kidneys.25 Hereafter I refer to both KDPI and LKDPI as KDPI. A kidney with a lower
KDPI score represents a better quality kidney with respect to expected graft survival, regardless of
whether the donor was deceased or living.

To diﬀerence out any other changes that might be happening over time I use young adults as
a comparison group for the pediatrics because they are close in age. There is no control group in
this setting that is completely unaﬀected by Share 35. When one candidate is moved up the wait
22Although the measure was oﬃcially introduced in 2014, I can use the formula to also compute

this measure for kidneys used before the measure was introduced.

23Speciﬁcally, the KDPI calculation includes the age, height, weight, ethnicity, history of hyper-
tension, history of diabetes, cause of death, serum creatine, Hepatitis C Virus status, and donation
after circulatory death status of the deceased donor (OPTN/UNOS, 2016).

24The LKDPI calculation includes the following donor factors: age, eGFR, race, history of
cigarette use, gender, blood type, relation to potential recipient, HLA mismatches with the potential
recipient, and the donor/recipient weight ratio (Massie et al., 2016).

25Among pediatric and young adult living donor recipients in my sample, LKDPI ranges from

-67.9 to 84.2.

13

list, mechanically, at least one candidate must be moved down the wait list. Because young adults
are only aﬀected by Share 35 through pediatric patients potentially being moved above them on the
wait list, the eﬀect of Share 35 on them is likely to be very small.26 Additionally, although OPTN
deﬁnes a patient as “pediatric” if they are under 18, the United States Renal Data System [USRDS]
deﬁnes a patient as “pediatric” if they are under 22 (SRTR, 2014; USRDS, 2018). Several pediatric
transplant centers also advertise that they accept “young adult” patients.27 Table A.1 presents
summary statistics on pediatric and young adult candidates in my sample. For comparison, I
also include summary statistics on adult candidates (older than 21 years old) who are not in my
estimation sample. Although all three groups are similar in terms of sex, race, blood type, and
insurance distributions, adult candidates are on average more than 35 years older than pediatric
candidates. Additionally, adult candidates are 27 percentage points more likely to have diabetes
than pediatric or young adult candidates. Because of these large diﬀerences between pediatric and
adult candidates, I restrict my comparison group to only young adults, instead of including all adult
candidates.

26Appendix Figure D.1 illustrates this with a hypothetical wait list change as a result of Share 35.
The pediatric candidates are moved up signiﬁcantly on the wait list but the young adult only moves
down one position. Note that this hypothetical wait list only contains 15 candidates. In reality, wait
lists usually contain at least 100 candidates, and many wait lists have several thousand candidates
on them (Source: https://optn.transplant.hrsa.gov/data/view-data-reports/build
-advanced/). Since real wait lists are much longer than this example, the eﬀect of being moved
up the wait list for pediatric candidates would be even larger than this ﬁgure illustrates, and the
eﬀect of young adult candidates being moved down would be even smaller in proportion to the size
of the wait list.

27For example, Seattle Children’s Kidney Transplant Program accepts ages up to 21, Driscoll’s
Children’s Hospital Kidney Transplant Program and Kidney Center say their recipients “range in
age from 1 to 21 years old”, and Texas Children’s Hospital “has transplanted kidney in children
from newborns to young adults”. Additionally, children up to age 21 are eligible for Medicare
coverage of their ESRD treatment (Source: https://www.medicare.gov/people-like-me/
esrd/children-and-esrd.html)

14

1.5 Eﬀect of Share 35 on Pediatric Kidney Transplant Recipients

1.5.1 Eﬀect of Share 35 on all Pediatric Recipients

My conceptual framework implies that the overall eﬀect of Share 35 on wait time and quality
for pediatric candidates is ambiguous because of two competing eﬀects:
the improvement of
deceased donor transplants for pediatric recipients because of Share 35 and the behavioral change
of candidates switching from using a living to a deceased donor. The panel on the left in Figure
A.1 shows the improvement from Share 35: approximately 200 more pediatric candidates were
receiving transplants that used younger deceased donor kidneys every year while there was not any
change in the amount of young adults receiving young deceased donor kidneys. But, as is evident
in the panel on the left in Figure A.2, the behavioral change as a result of Share 35 is signiﬁcant:
it appears that about 100 pediatric candidates switch from using living donors to using deceased
donors instead every year while there is no change in the living donor use of young adults. These
two eﬀects potentially counteract each other when considering the eﬀect of Share 35 on all pediatric
recipients. For example, Figure A.3 shows the percent of pediatric and young adult recipients in
my sample who are transplanted within a year over time. In the ﬁgure, pediatric wait time does
not appear to change after Share 35 but young adult wait time increases. Figure A.4 illustrates the
trends in average KDPI across years among recipients in the same sample. Pediatric and young
adults trend similarly with respect to kidney quality both before and after Share 35 is introduced.
In order to obtain estimates of these eﬀects I now move to a regression model. To diﬀerence out
unobserved changes that are common across transplant recipients, like changes in deceased donor
kidney supply or quality, I use a diﬀerence in diﬀerences model where my comparison group is

15

young adult kidney transplant recipients.28 Speciﬁcally, I estimate the following model:

yi = β1Sharei + β2Pediatrici + β3Sharei ∗ Pediatrici + xiγ + ui,

(1.1)

where yi = 1 if recipient i was transplanted within 3, 6, 12, or 18 months or where yi is the KDPI
of the kidney that was used in recipient i’s transplant. Sharei is a dummy variable that indicates
whether individual i received a transplant before or after Share 35. Pediatrici =1 if individual i is
less than 18 years old when they start waiting for a transplant. β3 is my coeﬃcient of interest; it
represents the causal eﬀect of Share 35 on wait time or kidney quality. xi includes age at start of
wait time, gender, insurance type, race, blood type, transplant year, diabetes status, highest PRA
score, citizenship status, functional status, diagnosis that lead to ESRD, DSA level ﬁxed eﬀects, an
indicator for if candidate i was ever inactive on the wait list, and an indicator for whether candidate
i opted into receiving Expanded Criteria Kidney[ECK] oﬀers.29

Table A.2 shows the estimated coeﬃcients of equation (1.1) on the entire sample of pediatric
and young adult deceased and living kidney recipients. Columns 1 through 4 show that Share 35
signiﬁcantly improved wait time before transplant for pediatric recipients relative to young adults.
The estimated coeﬃcient on Share ∗ Pediatric in Column 1 implies that Share 35 increased the
28By using a sample of only transplant recipients, there is the potential for bias in my estimates if
Share 35 changes who among pediatric and young adult candidates receives a transplant. I am not
concerned about this change for pediatric candidates because an overwhelming majority of them
were receiving transplants before Share 35. In Section 1.5.3 I elaborate on this argument. The main
concern that one might have about young adult recipients is that by giving pediatric candidates
priority, young adult candidates are moved down the wait list, and therefore might be less likely
to receive a transplant. The right hand panel of Figure A.1 illustrates that this is not the case.
Young adults receive the same number of deceased donor transplants after Share 35 as they did
before it was introduced. Additionally one might believe that young adult candidates might have
an increased incentive to ﬁnd a living donor after Share 35. The right hand panel of Figure A.2
shows that even if this is the case, they are not able to ﬁnd these living donors; young adults are
receiving approximately the same number of living donor transplants before and after Share 35.

29ECK status kidneys were introduced on October 30, 2002 to address the growing shortage of
deceased donor kidneys. ECK status kidneys are older than kidneys that were previously accepted
as deceased donor kidneys. Only 6% of pediatric deceased donor recipients between 2002 and
2013 opted into receiving these oﬀers. However, among young adult deceased donor recipients in
this same time period, 25% opted into these oﬀers. This diﬀerence in opt in rate is not surprising
since it is logical that older candidates would be more willing to accept older kidneys.

16

likelihood of a pediatric patient receiving a transplant within 3 months, relative to a young adult
recipient, by 4 percentage points. Compared to the percent of pre Share 35 pediatric recipients
receiving a transplant within 3 months of 16%, this is a substantial increase in likelihood. Columns
2, 3, and 4 show even larger increases in likelihood receiving a transplant within 6, 12, or 18 months
after Share 35 for pediatric recipients.

Column 5 of Table A.2 shows the estimated coeﬃcients of equation (1.1) on the same sample of
pediatric and young adult deceased and living kidney recipients, where yi = KDPIi. The estimated
coeﬃcient on Share ∗ Pediatric implies that Share 35 slightly improved average kidney quality
of pediatric recipients by 1.7 points, relative to young adult recipients. This is consistent with my
hypothesis that the improvement in quality for pediatric deceased donor recipients could counteract
any decline in quality that candidates who switch from using a living donor kidney to a deceased
donor kidney experience.

The key assumption of my diﬀerence and diﬀerence design is that pediatric and young adult
deceased donor recipients would have had parallel trends in wait time and kidney quality in the
absence of Share 35. To assess the plausibility of this assumption, I estimate a variation of equation
(1.1) that allows for the eﬀect of Share 35 to vary by transplant year. For ease of notation, I introduce
the subscript t for the year during which recipient i receives their transplant and the subscript a for
the age of recipient i:

yita = αt + δa +

βτ1(τ = t)∗ Pediatrici +

βτ1(τ = t)∗ Pediatrici +xiγ + uita (1.2)

2002

τ=1998

2014

τ=2004

If the assumption of parallel trends holds then the estimated pre Share 35 transplant year βτ’s should
be approximately zero, representing that pediatric and young adult candidates have parallel trends
in wait time or kidney quality before the policy change. Figures A.5 and A.6 plot the estimated
betas from equation (1.2), where the betas measure the change in that year compared to the base
year of 2003. I use 2003 as the base year because the policy was announced publicly before it was
implemented.30 In Figures A.5 and A.6, the coeﬃcients are approximately zero before the policy
302003 is the year before the policy was announced, and therefore is before any changes in

17

change, representing that there are no signiﬁcant diﬀerences in trends in wait time and kidney
quality between pediatric and young adult recipients before Share 35. The lack of evidence of
diﬀerential trends before Share 35 supports my assumption of parallel trends before Share 35. The
coeﬃcients start to diverge after Share 35 was introduced, representing the diﬀerential eﬀect that
the priority had for pediatric recipients over young adult ones.

These results suggest that although Share 35 improved wait time for pediatric candidates, on
average, candidates did not experience large improvements in kidney quality. One likely explanation
for the lack of quality improvement is that some candidates switched from using a living donor to
using a deceased donor, changing the composition of who receives a deceased donor transplant after
Share 35. The composition of pediatric deceased donor recipients after Share 35 may include new
transplants recipients who would not have received a transplant in absence of Share 35, deceased
donor transplant recipients who would have received deceased donor transplants in the absence
of Share 35, and deceased donor recipients who would have received living donor transplants in
absence of Share 35. I now turn to considering the eﬀect of Share 35 on each of these diﬀerent
groups of recipients separately.

1.5.2 Crowd out in Living Donors for First Transplant

Because Share 35 improves expected kidney quality and wait time for pediatric deceased donor
transplants, some pediatric candidates are likely to switch from using a living donor to using a
deceased donor instead. Figure A.7 shows that this switch was very dramatic. The percent of
pediatric ﬁrst transplants that used a living donor drops by about 20 percentage points after Share
35 while the percent for young adults remains constant. 75% of pediatric living donors before
Share 35 are the parent of the pediatric recipient and 52% of young adult living donors are parents
transplant decisions might be changed as a result of Share 35. For example, a pediatric candidate
might decide not to use their living donor after hearing that they would be receiving Share 35 priority
next year. This candidate would register on the wait list in anticipation of the implementation of
the policy but could receive and accept an oﬀer of a young deceased donor kidney before Share 35
takes eﬀect.

18

of the recipient.31 These parents would be likely to consider the future transplants that their child
might need so although the decline in living donors in Figure A.7 may be due to the improvement
in expected wait time and quality of deceased donor transplants from Share 35, it also could be due
to pediatric recipients strategically “saving” their living donors for future transplants.

To quantify this decline for pediatric candidates, I estimate equation (1.1) where yi

is
LivingDonori. LivingDonori = 1 if recipient i uses a living donor for their ﬁrst transplant
and = 0 if recipient i uses a deceased donor. The vector xi includes the same controls as it did in
Section 1.5.1. I estimate equation (1.1) using OLS on my whole sample of deceased and living
donor recipients before and after Share 35. Here β3 represents the change in likelihood of a pediatric
recipient using a living donor because of the introduction of Share 35.

The estimated coeﬃcients from equation (1.1) where yi is LivingDonori on the entire sample
of pediatric and young adult deceased and living kidney recipients are presented in Table A.3. The
estimated coeﬃcient on Share ∗ Pediatric implies that Share 35 caused a statistically signiﬁcant
decrease in the probability that a pediatric candidate uses a living donor by 17 percentage points,
relative to young adult candidates, on average. Multiplying this coeﬃcient by the approximately
800 pediatric transplants that occur each year, means that about 136 fewer pediatric transplants use
a living donor each year after Share 35. I hypothesize that this decline in the probability of using
a living donor is due to these candidates saving that living donor for a future transplant. I look for
evidence of this eﬀect in Section 1.5.6.

In order for β3 to have a causal interpretation, pediatric and young adult candidates must have
common trends in likelihood of using a living donor. To assess the plausibility of this assumption,
I estimate equation (1.2) where yita = is LivingDonori. Figure A.8 plots these estimated betas.
The estimated betas are approximately zero before Share 35, supporting the assumption that there
is no diﬀerence in trends in living donor use between pediatric and young adult recipients before
Share 35.

31Other potential living donor sources include siblings, children, other relatives, spouses, or

friends.

19

1.5.3 Eﬀect of Share 35 on New Pediatric Transplants

One assumption that I need to make for my estimation of the eﬀect of Share 35 on pediatric
candidates who always would have received a deceased donor transplant is that Share 35 did not
create any new pediatric deceased donor transplants. Because of this assumption, before moving
to estimating the eﬀects of Share 35 on wait time and kidney quality on pediatric deceased donor
recipients, I ﬁrst argue that Share 35 did not create any new pediatric transplants. Because Share
35 improved deceased donor transplants for pediatric candidates, one might expect that this led to
more pediatric candidates receiving transplants who would not have received one without Share
35. The rest of this section shows that this is unlikely. First, Figure A.9 adds an additional line onto
Figure A.2 that shows the total number of pediatric transplants over time. This ﬁgure shows that
although there are more yearly pediatric transplants after Share 35, the increase in transplants is
actually a trend that started before Share 35. Starting in 2000, the number of pediatric transplants
increased until 2006, after which the total number of transplants starts to decline. This suggests
that Share 35 did not increase the total number of transplants.32

Additionally, if Share 35 did create new transplants, they could come from two sources:
increasing the proportion of pediatric wait list registrations that lead to transplants or from inducing
more pediatric patients with ESRD to try to get a transplant earlier or even, rather than remaining on
dialysis in the near future. Figure A.10 shows the trend in the percentage of wait list registrations by
all pediatric candidates that registered on the waiting list between 1995 and 2010 that lead to either
a deceased or living donor transplant. The ﬁgure shows that not only was there not a signiﬁcant
change in the trend after Share 35 in 2005, but also that pediatric candidates already had a high rate
of transplantation before Share 35. It would be diﬃcult to create more pediatric transplants from
registrations when about 96% of these registrations already end in transplantation.33 This evidence
32Although there does appear to be a jump in transplants in 2006, this jump is exclusive to 2006.
Additionally, in Section 1.6 I show that my results are robust to excluding 2005 and 2006 so this
jump in transplants in 2006 is not driving my results.

33The remaining 4% of registrations that do not end in transplantation can be due to the candidate
still waiting, improvement to where the candidate no longer needs a transplant, deterioration to
where the candidate is too sick for a transplant, candidate dies, candidate listed in error, or refused

20

makes it unlikely that a higher proportion of registrations end in transplantation.

Next, Figure A.11 shows that there is little change in the dialysis use of pediatric ESRD patients
after Share 35.34 Furthermore, it shows that any increase in transplantation among pediatric ESRD
patients follows the general increasing trend in ESRD development in the pediatric population,
not from more pediatric patients with ESRD choosing to get a transplant instead of using dialysis
because of Share 35. Additionally, opting for a transplant instead of remaining on dialysis approx-
imately doubles the expected remaining lifetime of pediatric patients with ESRD (USRDS, 2018).
For example in 2014, for a 14- to 17-year-old child with ESRD, receiving a transplant instead of
dialysis increases their expected remaining lifetime from 20.9 years to 48.8 years (USRDS, 2018).35
Given that there already was a substantial beneﬁt to receiving a transplant over using dialysis for
pediatric patients, it is unlikely that the beneﬁts of Share 35 induced many ESRD children to try to
receive a transplant instead of using dialysis.

Because I’ve shown that it is unlikely that Share 35 creates new transplants, I proceed assuming
that the group of pediatric deceased donor recipients after Share 35 is only composed of candidates
who would have received a deceased donor transplant even in absence of Share 35 (“Stayers”) and
candidates who switched from using a living donor to using a deceased donor because of Share 35
(“Switchers”).

1.5.4 Eﬀect of Share 35 on Stayers

In this section, I estimate the eﬀect of Share 35 on “stayers”, the candidates who would have
received a deceased donor transplant without Share 35. These are the candidates who the Share
transplant.

34The data reported here have been supplied by the United States Renal Data System (USRDS).
The interpretation and reporting of these data are the responsibility of the author and in no way
should be seen as an oﬃcial policy or interpretation of the U.S. government.

35This is not a causal estimate of the survival beneﬁt of receiving a transplant over remaining
on dialysis. It is possible that patients who select into receiving a transplant also have diﬀerent
characteristics or behaviors that lead to better survival regardless of whether they receive a transplant
or use dialysis. However, it is unlikely that these characteristics or behaviors would account for the
entirety of the 28 year survival beneﬁt of transplantation over using dialysis.

21

35 policy change was intended to beneﬁt. Figures A.12 and A.13 show the trend in likelihood
of transplantation within a year and the average KDPI over time for all pediatric and young adult
deceased donor recipients. From Figure A.12, it appears that there was a decrease in wait time for
pediatric recipients after Share 35. Figure A.13 shows that there is a improvement in the average
kidney quality for pediatric recipients. The rest of this section quantiﬁes these eﬀects for pediatric
recipients who are more similar to pre-Share 35 deceased donor recipients.

First, note that I cannot directly observe “stayers” and “switchers” because these designations are
based on counterfactual outcomes. However, looking at Tables A.4 and A.5, there are observable
diﬀerences between deceased and living donor recipients. For example, comparing the pre Share 35
pediatric recipients in column 1 in these two tables, pediatric living donor recipients are on average
more likely to be white, have type A blood, and use private insurance than pediatric deceased donor
recipients. To take advantage of these observable diﬀerences, I use a two-step estimation technique.
I ﬁrst calculate weights for the post Share 35 pediatric deceased donor recipients so that they are
more similar with respect to observable characteristics to pre-Share 35 pediatric deceased donor
recipients. Next, I use those weights in the diﬀerence in diﬀerence model in equation (1.1).36

To generate the weights I use propensity-score matching[PSM].37 The basic idea behind match-
ing methods is that comparison units with similar observable characteristics to a treated unit can be
used to estimate a counterfactual version of the treated unit. PSM reduces the number of dimensions
that a unit needs to be matched on by estimating an estimated probability of being treated called
the propensity score. Comparison units with a close propensity score to the propensity score of
a treated unit are matched to that treated unit. I use a variation of PSM called kernel matching,
which instead of forming pairs of treated and comparison units, creates weights for all comparison
units for each treated unit, where comparison units with closer propensity scores to the treated
unit receive larger weights.38
I restrict my sample for the matching step to pediatric deceased
36Stuart et al. (2014) use a similar methodology but allow both treatment and comparison groups
to change as a result of a policy change. An and Winship (2017) also combine propensity score
matching with a diﬀerence in diﬀerence model but they use it in a panel data setting.

37Imbens and Wooldridge (2009) provides an overview of propensity score matching methods.
38For more information on kernel matching, see Jann, B. (2017). kmatch: Stata module for

22

donor transplant recipients before and after Share 35. I assign “treatment” status to recipients who
received a transplant before Share 35. This means that I am using post Share 35 pediatric deceased
donor recipients to estimate what outcomes for a pre Share 35 pediatric deceased donor recipient
would be in the post Share 35 period. I estimate propensity-scores using a logit model and match
observations based on the following recipient characteristics: age, gender, insurance type, race,
blood type, diabetes status, highest PRA score, citizenship status, functional status, diagnosis that
lead to ESRD, DSA, an indicator for if candidate i was ever inactive on the wait list, and an indicator
for whether candidate i opted into receiving Expanded Criteria Kidney[ECK] oﬀers. 39 For each
treated unit, all comparison units are given a weight that is a function of how close their propensity
score is to that of the treated unit.

The second step of my estimation procedure is to estimate equation (1.1) using the sample
of pediatric and young adult deceased donor recipients before and after Share 35, weighting the
observations with the weights I create in step 1. Pre Share 35 pediatric recipients receive a weight
of 1. Post Share 35 pediatric recipients receive a weight equal to the sum of all the weights they
receive for each pre Share 35 pediatric recipient. This re-weights the post-Share 35 group so that it is
similar to the pre Share 35 pediatric recipient group with respect to observable characteristics. This
weighting is done to counteract the composition change that results in the post Share 35 pediatric
group containing both Stayers and Switchers because of Share 35. Since the young adult recipients
do not experience a compositional change as a result of Share 35, all young adult recipients receive
a weight of 1. Because the weights used in this step are estimated, I bootstrap standard errors over
both steps.
multivariate-distance and propensity-score matching. Available from https://ideas.repec
.org/c/boc/bocode/s458346.html.

39Figure D.3 shows how well the matching in the ﬁrst step rebalances the recipients’ propen-
In particular, the red line shows the distribution of propensity scores for “treated”
sity scores.
individuals. These are pre Share 35 pediatric deceased donor recipients. The blue line shows the
propensity score distribution for post Share 35 pediatric deceased donor recipients. Note that a
higher propensity score represents a higher predicted likelihood of being a pre Share 35 deceased
donor recipient. The right hand panel of Figure D.3 shows that the propensity score matching
successfully re-weights the post Share 35 recipients so that their propensity score distribution is
more similar to that of the pre Share 35 recipients.

23

Table A.6 presents the estimated coeﬃcients from the two step estimation procedure: equation
(1.1) is estimated, weighting the observations with the weights I created from PSM. xi includes
contains the same controls that I used in the matching step plus an additional control for transplant
year. The coeﬃcient on Share ∗ Pediatric implies that Share 35 increased the likelihood that a
pediatric deceased donor recipient was transplanted within a three months by 6 percentage points,
within six months by 15 percentage points, within a year by 20 percentage points, and within 18
months by 22 percentage points, compared to young adult deceased donor recipients. These are
substantial eﬀects, especially considering only 44% of pediatric deceased donor recipients were
transplanted within a year before Share 35. The coeﬃcient on Share ∗ Pediatric in Column 5
implies that Share 35 improved the quality of kidneys used in pediatric deceased donor transplants
by about 8.4 points on average. This represents a 30% improvement in quality, given the pre Share
35 pediatric sample mean quality of 28.

Table D.2 compares these estimates to the estimates of equation (1.1) without re-weighting the
observations. The estimated unweighted coeﬃcients on Share ∗ Pediatric are smaller, in general,
than the weighted one. This suggests that the stayers have shorter wait times for deceased donor
transplants and accept higher quality deceased donor kidneys than switchers, on average.

The event study analysis shown in Figures A.14 and A.15 supports the assumption of parallel
trends in wait time and quality between pediatric and young adult recipients before Share 35.
However, the introduction of the matching procedure adds another key assumption that must hold
for my estimated coeﬃcients to have a causal interpretation. With the matching, I am assuming that
post Share 35 pediatric recipients have the same outcomes as what pre Share 35 pediatric recipients
would have in the post period, conditional on observable characteristics. Because the set of post
Share 35 pediatric recipients has both Stayers and Switchers in it, this could potentially violate this
assumption. However, both Stayers and Switchers are receiving the same kidney oﬀers through the
same allocation system, conditional on observables like on being in the same locality and having
the same blood type. This would make it diﬃcult for Stayers and Switchers have have diﬀerent
deceased donor kidney quality or wait time unless they have diﬀerences in kidney oﬀer acceptance

24

rules. (For example, if Switchers were more willing to turn down oﬀers in order to wait for a higher
quality oﬀer.) This is unlikely to be the case because Share 35 oﬀers all pediatric candidates high
quality kidneys that they should be willing to accept. It is unlikely that candidates would be willing
to wait longer, without functioning kidneys, in order to receive an oﬀer that would be even higher
quality.

These results imply that Share 35 did achieve its goals among pediatric candidates who are
similar to pre-Share 35 candidates. These candidates experienced reduced wait times and received
higher quality kidneys. These results are especially encouraging given the large number of pediatric
candidates who switched from living donors to deceased donors; there was the potential that the
addition of these pediatric deceased donor candidates could crowd out the beneﬁts for the candidates
who the policy was intended to beneﬁt.

1.5.5 Eﬀect of Share 35 for Switchers

In this section, I estimate the eﬀect of Share 35 on “Switchers”: the recipients who were induced to
use a deceased donor instead of a living donor by Share 35. To do this, I follow the methodology
introduced in Jones (2015). First, consider the following model of the eﬀect of Share 35 on wait
time or kidney quality for a pediatric recipient:

Yi = βSharei + xiγ + ei

(1.3)

where Sharei = 1 if recipient i received a deceased or living donor transplant after Share 35 and
= 0 if they received a transplant before Share 35. β in this equation represents the average eﬀect
of Share 35 on all pediatric transplant recipients. Note that there are three diﬀerent subgroups of
pediatric recipients: the Stayers, the Switchers, and the living donor recipients who would use a
living donor with or without the introduction of Share 35. Since β is the average eﬀect, it can be
rewritten as the weighted average eﬀect on these three diﬀerent subgroups of pediatric recipients:

β = θStayer βStayer + θSwitcher βSwitcher + θ AlwaysLiving βAlwaysLiving

(1.4)

25

where θStayer represents the proportion of pediatric recipients that are Stayers, θSwitcher is the
proportion that are Switchers, and θ AlwaysLiving is the proportion that would always use a living
donor. βStayer, βSwitcher, and βAlwaysLiving represent the eﬀect of Share 35 on each of these
subgroups.

Now, assume βAlwaysLiving = 0 because Share 35 only aﬀects wait time or quality for recipients

who use deceased donors. Then equation (1.4) can be rearranged to solve for βSwitcher:

βSwitcher =

β

θSwitcher

− βStayer

θStayer
θSwitcher

(1.5)

Therefore, I can estimate βSwitcher by estimating the average eﬀect of Share 35 on pediatric
recipients, the eﬀect of Share 35 on Stayers, and the proportion of recipients who are Stayers and
Switchers. I previously estimated β in Section 1.5.1 and the eﬀect of Share 35 on Stayer in Section
1.5.4. Finally, to estimate θStayer and θSwitcher I estimate the following model, using the my entire
sample of pediatric and young adult recipients:

DeceasedDonori = αY A + αPPediatrici + θSwitcher Sharei ∗ Pediatrici + γSharei + ui

(1.6)

In this model, θStayer = αY A + αP.

Table A.7 contains my estimates of βSwitcher using this process. Standard errors are boot-
strapped over the whole process. The estimates on wait time are not statistically signiﬁcant at
conventional levels but they imply that pediatric candidates who switched from using a living
donor to using a deceased one as a result of Share 35 are able to receive transplants faster as a result
of this switch. However, looking at column 5, these candidates received transplants with kidneys
that were, on average, 11.8 points worse quality than what they would have received otherwise.
This is a substantial reduction in quality for these candidates. Given how highly transplant teams
value higher quality living donor kidneys, it seems unlikely that Share 35 would induce so many
candidates to switch unless they are strategically saving their living donor for their next transplant
or unless the costs of ﬁnding or using a living donor are very large. In the next section, I investigate
whether any candidates who switch are saving their living donor for their next transplant.

26

1.5.6 Potential Saving of Living Donors for Second Transplant

Now that I’ve discussed how switchers are aﬀected by their switching donor types at their ﬁrst
transplant, I estimate whether they are more likely to use a living donor for their second transplant.
To do this, I use data on the second transplant of deceased donor recipients in my original sample.
Table A.8 shows that although my original sample of deceased donor transplant recipients contained
about 8,600 pediatric and young adult recipients, I only observe the second transplant of 1,696
of them. Unsurprisingly, given that this sample contains recipients whose ﬁrst transplant had the
worse graft survival, these recipients’ ﬁrst transplant used slightly worse quality kidneys than the
average among all deceased donor recipients in my sample. My hypothesis is that the decrease
in the likelihood that a pediatric patient uses a living donor for their ﬁrst transplant from Section
1.5.2 is because these patients save their living donor for their next transplant. Figure A.16 shows
the trend in the likelihood that pediatric and young adult recipients who used a deceased donor for
their ﬁrst transplant use a living donor for their second transplant. This ﬁgure shows that pediatric
and young adult deceased donor recipients had similar likelihood of using living donors for their
second transplants before Share 35. However, after Share 35, it does look like pediatric recipients
are slightly more likely to use living donors while young adults are slightly less likely to use living
donors for their second transplants. In order to estimate this potential eﬀect, I ﬁrst estimate the
following diﬀerence in diﬀerences model using the second transplants of deceased donor recipients
in my sample from Section 1.5.2:

LivingDonori = β1ShareFirsti + β2PediatricFirsti + β3ShareFirsti∗PediatricFirsti +xiγ +ui
(1.7)
where ShareFirsti = 1 if recipient i’s ﬁrst transplant occurred after Share 35 and PediatricFirsti =
1 if recipient i was considered pediatric at their ﬁrst transplant. The vector xi includes age at start
of wait time, gender, insurance type, race, blood type, transplant year, diabetes status, citizenship
status, functional status, diagnosis that lead to ESRD, DSA level ﬁxed eﬀects, an indicator for
receiving Share 35 priority for the second transplant, and an indicator for whether the recipient

27

receives EPTS priority in the new allocation system.40 β3 represents the diﬀerence in likelihood of
using a living donor for their second transplant between pediatric and young adult candidates who
used a deceased donor for their ﬁrst transplant after Share 35. If some candidates save their living
donor for their second transplant, I expect ˆβ3 to be positive.

Table A.9, column 1 shows the estimated coeﬃcients from equation (1.7) using OLS on the
second transplant observations of candidates who were pediatric or a young adult in my sample who
used a deceased donor for their ﬁrst transplant. The coeﬃcient on ShareFirst ∗ PediatricFirst is
positive but not statistically signiﬁcant, implying that Share 35 might have increased the likelihood
that deceased donor recipient use a living donor for their second transplant.

Because I only observe approximately 20% of second transplants for the deceased donor
recipients in my original sample, my estimates will be biased if selection into receiving a second
transplant earlier is correlated with receiving a transplant after Share 35. This is likely to be
true because recipients who have earlier graft failure might exhibit worse compliance with post-
transplant care, and this reduced compliance might be correlated with whether or not they had a
parent as a willing living donor available to them at their ﬁrst transplant. Recipients with parents
willing to donate a kidneys for them might also have parents helping them adhere to their post
transplant care. There also is a correlation between family and personal health (Case & Paxson,
2002). If a recipient’s parents do not meet the criteria to be living donors because of their health,
the recipient might be unhealthier than other transplant recipients on average. Because of the
crowd out in living donors created by Share 35, deceased donor recipients after Share 35 are more
likely to have had a willing donor available to them than recipients before Share 35. Hence, it is
plausible that earlier graft failure (i.e. inclusion in my sample) is negatively correlated with having
a willing living donor available and having a willing living donor available is positively correlated
with being transplanted after Share 35. Therefore, my OLS estimate of the coeﬃcient on Share 35
40The new Kidney Allocation System introduced in 2014 gives priority to candidates who
have Estimated Post Transplant Survival [EPTS] scores less than 20. Most of the candi-
dates in my sample will still be young enough in 2019 to qualify for this priority at their
second transplant. Source: https://optn.transplant.hrsa.gov/resources/allocation
-calculators/epts-calculator/

28

is negatively biased.

To correct the bias due to the selection problem, I use a two step Heckman procedure. First,
using all observations of candidates who used a deceased donor for their ﬁrst transplant, I estimate
the probability of selection using the following Probit model:

Pr(Received Second Transplanti|zi) = Φ(ziγ)

(1.8)

where zi contains the vector xi plus the additional explanatory variables of KDPI of ﬁrst
transplant, year of the ﬁrst transplant, and an indicator for dialysis use within the ﬁrst week after
a candidate’s ﬁrst transplant. These additional variables satisfy the exclusion restriction: they are
correlated with earlier graft failure but are plausibly independent of the decision to use a living donor
for a candidate’s second transplant. Next, I use ˆγ to estimate inverse Mills ratios for candidates who
received a second transplant. Finally, I estimate equation (1.7) including these estimated inverse
Mills ratios as an additional covariate.

Table A.9, column 2 shows the estimated coeﬃcients from equation (1.7) with the estimated
Mills ratio as an additional covariate. The coeﬃcient on ShareFirst ∗ PediatricFirst is positive,
although it is smaller and still statistically insigniﬁcant. However, 80% of the recipients in my
sample who received a ﬁrst transplant have not received a second transplant yet. It is possible that
with more years of data in the future, there will be better evidence of these recipients saving their
living donor for their second transplant.

1.6 Robustness

One might be concerned that candidates can aﬀect their treatment status and that this manipu-
lation is driving my results. Two potential ways a candidate could manipulate their treatment status
is by registering on the wait list before they turn 18, or by waiting to receive a transplant until after
Share 35. Figure A.17 shows that eighteen year olds are still registering at a similar rate after Share
35 while the number of seventeen year old registrations increases. This suggests that the increase
in seventeen year old registrations is not driven by young adults who might register earlier to get
Share 35 priority, but rather by pediatric candidates who are choosing deceased donors over living

29

donors. Additionally, kidney transplant candidates gain wait list points for how long they have been
on the waiting list. These points move them higher up the kidney oﬀer list; therefore, it has always
been optimal to register on the wait list as soon as possible. To provide additional evidence that this
manipulation is not driving my results, I re-estimate my speciﬁcations from Sections 1.5.1, 1.5.2,
and 1.5.4 excluding all candidates aged 17 or 18. Column 2 of Table A.10 presents the estimated
coeﬃcient on Share ∗ Pediatric from this re-estimation. These coeﬃcients are very similar to the
original coeﬃcients in column 1, providing evidence that it is unlikely that candidates manipulating
their registration age is driving my results.

It is also unlikely that pediatric candidates wait until after Share 35 to accept a deceased donor
kidney. Waiting until after the policy takes eﬀect would involve candidates waiting longer, which
their transplant team would not recommend. However, as a speciﬁcation check, columns 3 in Table
A.10 addresses any concerns that candidates may be manipulating their treatment status by waiting
to receive a transplant by omitting candidates who were transplanted within one year on either side
of Share 35’s introduction. Because median wait time is approximately one year, this eliminates
any candidates who might change their behavior to take advantage of Share 35. The estimated
coeﬃcients on Share ∗ Pediatric in column 3 are again very similar to the original estimated
coeﬃcient in column 1 suggesting that candidates manipulating their treatment status through their
transplant year does not drive the results.

1.7 Conclusion

In this paper, I am the ﬁrst to analyze the eﬀect of the Share 35 policy separately on deceased
donor pediatric recipients who always would have used a deceased donor, and on those who were
induced to switch from living to deceased donors as a result of Share 35. I ﬁnd that the Share 35
policy change improved the kidney quality and reduced wait time for pediatric candidates who are
predicted to have received a deceased donor kidney in absence of the policy change but does reduce
kidney quality for pediatric patients who switch from using a living donor to using a deceased
donor kidney because of Share 35. This means that although Share 35 achieves its intended eﬀect

30

for its intended recipients, it induces some pediatric patients to substitute away from living donors,
resulting in worse kidney quality for these candidates. Although there is strong evidence of pediatric
patients being less likely to use a living donor after Share 35, I ﬁnd very little evidence of these
patients saving that donor for their second transplant.

Creating policy that balances eﬃciency and equity in allocating a growing shortage of donated
kidneys is challenging, especially when there are unintended eﬀects that result from the policy.
The population aﬀected by Share 35 is small; pediatric candidates were only about 3% of the new
candidates added to the wait list in 2004. Because pediatric candidates are such a small proportion
of the wait list, the distortion in living donor use might not be too concerning as a side eﬀect
of Share 35, given the improvements I ﬁnd to wait time and quality for deceased donor pediatric
recipients. But, now there are more candidates on the wait list and in 2014 the allocation system
was revised to give priority to candidates in the top 20% of Expected Post Transplant Survival
[EPTS]. This is a much larger group of candidates than the pediatric candidates aﬀected by Share
35, but since the calculation for EPTS is heavily based oﬀ of age, these candidates should be similar
to the pediatric candidates aﬀected by Share 35. If the proportion of candidates who switch when
given EPTS priority is similar to what it is for candidates given Share 35 priority, then we should
anticipate a larger crowd out in living donors from the new EPTS priority. 20% of candidates is a
very large proportion of the wait list to have potentially crowd out of using living donors. My crowd
out estimate implies that the EPTS priority potentially crowds out 3,224 living donor transplants.41
When these candidates who have living donors available to them receive deceased donor
kidneys, this denies the candidates on the wait list without living donors the opportunity to receive
those deceased donor kidneys. Although some priority rules improve survival matching between
candidates and deceased donor kidneys, this improvement comes with a cost of reducing the total
number of transplants that can take place and a change in the equity in allocation.

413,224=(current number of kidney wait list candidates)*(percent of candidates given EPTS

priority)*(estimated crowd out from Share 35 priority)=94,835*0.2*0.17

Source for current number of kidney wait list candidates: https://optn.transplant.hrsa

.gov/data/view-data-reports/national-data/

31

CHAPTER 2

THE EFFECTS OF THE AFFORDABLE CARE ACT ON THE DEMAND FOR ORGAN

TRANSPLANTS

2.1 Introduction

Before the passage of the Aﬀordable Care Act [ACA] in 2010, many low income, childless
households had diﬃculty ﬁnding insurance coverage if they were unemployed or if their employers
did not sponsor a plan for them. One of the goals of the ACA was to introduce more options for
these households to obtain insurance coverage. This was done primarily through state expansions
of Medicaid and through the introduction of a centralized marketplace for private insurance plans.
Previous literature documents that the ACA was successful at reducing the number of uninsured
(Simon et al., 2017; Duggan et al., 2017; Miller & Wherry, 2019; Courtemanche et al., 2017;
Sommers et al., 2015; Wherry & Miller, 2016) and that the Medicaid expansions were successful
at increasing some forms of preventative and dental care, increasing diagnosis of chronic health
conditions, and increasing access to timely care for some common surgical conditions, like ap-
pendectomy surgeries, among low-income childless adults (Simon et al., 2017; Wherry & Miller,
2016; Loehrer et al., 2018).1 However, some individuals have chronic illnesses and may need more
than preventative care.

In this paper, I look at the eﬀects of increased access to insurance from the ACA for individuals
who have organ failure and need an organ transplant. Low income individuals can develop organ
failure (Crews et al., 2014; Clark et al., 2009; Raju et al., 2019; Scaglione et al., 2015). For most
of these individuals, an organ transplant is their only treatment option. However, these transplants
are expensive and most transplant centers require patients to have insurance coverage (Thibodeau
et al., 2013; Laurentine & Bramstedt, 2010; Glueckert et al., 2013). These low income individuals
1Further evidence that uses the Oregon Health Insurance Experiment shows that an increase in
access to Oregon’s Medicaid program increases cancer screenings, dental care, and preventive care
(Wright et al., 2016; Baicker et al., 2018; Finkelstein et al., 2012).

32

are likely to have been able to gain insurance through the state Medicaid eligibility expansions
or the private insurance marketplace subsidies introduced with the ACA. Previous literature ﬁnds
that the states that expanded Medicaid on January 1, 2014 had increased wait list registrations for
transplants (Tumin et al., 2017; Breathett et al., 2017; Ising et al., 2018; Trivedi et al., 2018; Harhay
et al., 2018; Oliveira et al., 2016). Although some of these papers use a diﬀerence in diﬀerences
design to causally look at the eﬀects of Medicaid expansion, some simply compare pre ACA counts
of Medicaid insured listings to post ACA counts of listings. Additionally, this literature has not
yet looked at whether these registrations lead to transplants and no papers look at the eﬀects of the
private insurance marketplace on wait list registrations or transplants.

In this paper, I use a diﬀerence in diﬀerences model based on timing of Medicaid expansions to
access whether Medicaid expansions increase wait list registrations and transplants. Additionally,
I use variation in requirements for private insurance plans to cover transplants to identify the
eﬀects of the newly available subsidized private marketplace plans on registrations and transplants.
Using the universe of all transplant candidates and recipients in the United States, I ﬁnd that
the state expansions of Medicaid increased monthly wait list registrations by candidates insured
through Medicaid by 20 to 60 percent for each organ, but I ﬁnd no eﬀect of the marketplace on
privately insured individuals’ registrations.
I also ﬁnd an increase in the number of Medicaid
insured deceased donor transplants for Medicaid expansion states, although this increase is not
statistically signiﬁcant for all organs. Finally, I ﬁnd the number of Medicaid insured living donor
liver transplants after Medicaid expansion more than doubles.

The paper proceeds as follows: In section 2.2, I provide an explanation of transplantation and
allocation of organs in the United States. Additionally, I describe the changes to health insurance
that result from the ACA that would impact transplant candidates. In section 2.3, I describe the data
that I use and provide descriptive statistics. In section 2.4, I estimate the eﬀects of the Medicaid
expansions and private insurance marketplace on wait list registrations. In section 2.5, I estimate
the eﬀects the ACA on deceased donor transplants. In section 2.6 I estimate the eﬀects of the ACA
on living donor transplants. Finally, section 2.7 concludes.

33

2.2 Environment

2.2.1 Transplantation in the United States

The most commonly transplanted organs in the United States are kidneys, livers, hearts, and lungs.
In order to receive a transplant, individuals with organ failure need to be accepted at a transplant
center. Each transplant center sets its own patient acceptance criteria, which typically includes
physical, psychological, and social evaluations. Most transplant centers also require the patient to
have insurance before accepting them because the transplant procedure involves many costs. For
example, the American Kidney Fund states that kidney transplant evaluations may be denied due
to lack of adequate insurance coverage (Samoray & Satarino, 2016). Thidbodeau et al. (2013)
surveyed heart transplant centers and found that 84% of the centers that they surveyed required
insurance in order for a potential candidate to register on the waiting list. Even among centers that
would allow uninsured candidates to register, 81% of these centers required the candidate to pay
a deposit before they could be added to the waiting list (Thibodeau et al., 2013).2 In 2017, billed
charges for a transplant averaged from $415,000 for a kidney transplant to $1,382,000 for a heart
transplant (Bentley & Phillips, 2017). These costs include both upfront costs for the surgery and
expensive medications used to maintain the transplant for the rest of the recipient’s life. Once a
patient has been accepted at a transplant center, they can be registered on the waiting list. While
on the waiting list, a candidate receives organ oﬀers based on the current organ allocation policy,
which varies by organ. In general, organs are allocated ﬁrst to candidates in the same locality as
the recovered organ. There is a shortage of deceased donor organs, so candidates need to wait to
receive an organ oﬀer before their deceased donor transplant can take place.3

Because organs are allocated locally ﬁrst, this creates an incentive for candidates to strategically
choose which transplant center they list at. Data are publicly available on how quickly candidates

2The median amount that these centers required was $200,000 (Thibodeau et al., 2013).
380% of kidney candidates, 45% of heart candidates, 35% of lung candidates, and 55% of liver

candidates wait longer than a year to receive a transplant.

34

receive transplants on average from speciﬁc centers.4 This data can help candidates list at centers
in locations with faster times to transplant. Some candidates even register at multiple transplant
centers in diﬀerent localities. This behavior is commonly called “multilisting”. In order to multilist,
a candidate needs to be accepted as a patient at both centers. If the center is located a substantial
distance away from the candidate’s home, the transplant center requires that the candidate has
transportation that will allow them to travel to the center quickly in case they receive an organ oﬀer.
Kidney and liver transplant candidates can try to ﬁnd a living donor instead of, or in addition
to, waiting to receive a deceased donor organ oﬀer. By ﬁnding a willing and blood type compatible
living donor, these candidates can avoid the long wait to receive a deceased donor transplant.
Additionally, living donor transplants are associated with better post-transplant outcomes than
deceased donor transplants (SRTR, 2014). In 2014, 32% of kidney transplants and 4% of liver
transplants used a living donor.5

After receiving a transplant, candidates require immunosuppressants to prevent their body’s
immune system from attacking the transplanted organ. These medications help prevent graft
failure, where the transplanted organ ceases functioning. Billed charges for these medications can
range from $20,000 to $50,000 a year (Bentley & Phillips, 2017). Because transplant recipients
take these medications until the transplanted organ ceases functioning, a transplant recipient will
beneﬁt from insurance coverage after the transplant procedure.

2.2.2 Health Insurance before the Aﬀordable Care Act

Before the 2014 implementation of most provisions of the ACA, 90% of all privately insured
adults were covered by employer sponsored plans (Buchmueller & Monheit, 2009). Individuals
who did not have access to an employer’s group plan could buy a plan in the individual health
insurance marketplace. The non-employer sponsored plans were typically expensive and pre-
existing conditions could be used to deny an individual coverage.

4See https://www.srtr.org/reports-tools/program-specific-reports/
5Source:

https://optn.transplant.hrsa.gov/data/view-data-reports/

build-advanced/

35

If a person did not have private insurance, they might qualify for one of two public options:
Medicaid or Medicare. Medicaid primarily targeted low-income households with children.
In
order to qualify, a household must meet their state’s eligibility levels, which depend not only on
income, but also on household size. Most states did not oﬀer Medicaid to households without
children.6 For households with children, 26 states required the household’s income to be below 64
percent of the Federal Poverty Level [FPL] in 2010.7 For example, in order to qualify for Medicaid
in 2010, a family of three in Michigan would need to have an income less than $11,718.8 Cost
sharing in Medicaid programs varies by state, however there are some federal restrictions. For
example, in 2013, The Centers for Medicare and Medicaid Services [CMS] prohibited deductibles
from exceeding $2.65 and prohibited cost sharing on preferred drugs from exceeding $4.9 CMS
caps all out of pocket costs at 5 percent of a family’s income.

All US citizens are eligible to enroll in Medicare when they are 65 years old. If an individual or
their spouse has paid payroll taxes for at least 10 years, they can enroll in Medicare part A without
having to pay premiums. Medicare part A will cover 100% of the costs of an inpatient hospital
stay for a transplant after the enrollee covers an deductible ($1,364 in 2019).10 Medicare Part B
will cover 80% of outpatient costs, like transplant evaluations or transplant follow-up care, after
the enrollee pays the Part B deductible ($185 in 2019). Medicare Part B will also cover 80% of the
costs of immunosuppressants after the transplant.11

Candidates who need a kidney potentially qualify for End State Renal Disease [ESRD] Medicare
without a minimum age requirement. CMS rules require these candidates have paid payroll taxes
6Refer to Appendix Tables E.1 and E.2 for income eligibility levels by state and household size.
734 states required the household’s income to be below 100 percent of the Federal Poverty

-federal-register-references

from https://aspe.hhs.gov/prior-hhs-poverty-guidelines-and

Level.

8=(0.64)*(18310)
FPL obtained

9Source:

10Source:

-costs/index.html

https://www.medicaid.gov/medicaid/cost-sharing/out-of-pocket

https://www.cms.gov/newsroom/fact-sheets/2019-medicare-parts-b

-premiums-and-deductibles

11Source: https://www.medicare.gov/node/35206

36

for the required number of years to Social Security in order to qualify (U.S. Centers for Medicare &
Medicaid Services, 2019). The required number of years varies by how old the potential enrollee is.
Any dependents of someone who has met the contribution requirement are also eligible for ESRD
Medicare. ESRD Medicare covers 80% of the costs of dialysis treatments, without any time limits,
and if a patient opts for a kidney transplant, ESRD Medicare will cover the costs of the transplant
as outlined previously. However, ESRD Medicare only covers the costs of immunosuppressants for
36 months after the transplant.

2.2.3 Health Insurance after the Aﬀordable Care Act

The ACA introduced three major changes to health insurance that potentially aﬀected transplant
candidates. First, the ACA mandated that all insurance plans must cover ten Essential Health
Beneﬁts [EHB]. States have discretion over which services were included in the 10 EHB categories.
States determined these services by choosing an existing health plan available in that state as the
“EHB Benchmark Plan”. Table B.1 shows whether or not a state’s benchmark plan included
coverage for transplants. Although most states’ benchmark plans included coverage for transplants
starting in 2014, eight states’ plans did not until 2017.

Second, the ACA encouraged states to expand the income level at which households were
eligible for Medicaid to 138 percent of the FPL by providing states funding for newly eligible
enrollees. This expansion would especially beneﬁt childless individuals who were previously
ineligible for Medicaid at any income level in most states. 25 states expanded their eligibility level
for Medicaid on January 1, 2014, but an additional 12 states expanded their eligibility level since
then.12 Because many potential transplant candidates have low incomes (Crews et al., 2014; Clark
et al., 2009; Scaglione et al., 2015; Raju et al., 2019), these expanded eligibility levels make it more
12Note that although for most states, this “expansion” increased the income eligibility level for
Medicaid for childless individuals, some states reduced their income eligibility level Medicaid for
households with children. This would result in it being harder for these households to be eligible
for Medicaid after the expansion. However, these households who are no longer eligible for their
state’s Medicaid program would be eligible for subsidies on the newly introduced Marketplace for
private insurance.

37

likely for potential transplant candidates, especially ones without children, to be eligible to receive
Medicaid coverage, and in turn register at a transplant center.

Finally, the ACA introduced a centralized marketplace for private insurance plans. Unlike
private insurance plans of the past, the ACA required these plans to cover the EHBs outlined in
the state’s benchmark plan and these plans cannot deny individuals with pre-existing conditions
coverage. Plans on the marketplace are categorized into Bronze, Silver, Gold, or Platinum plans
based on how the insurance company splits costs with the purchaser. Cost sharing ranges from
the purchaser paying 40 percent of costs with a Bronze plan to 10 percent with a Platinum plan.
Individuals who need a transplant are likely to want to buy a plan with reduced cost-sharing because
they would expect to have higher medical costs.13 Tax subsidies are available to help cover the
cost of the premium of these plans for households with incomes between 100 and 400 percent of
the FPL, adjusted for the household’s size. These tax subsidies are calculated so that a household
does not spend more than a certain percentage of their household income on the premium for the
second lowest cost Silver plan in their county. This percentage ranges from 3 percent of income
for households at 133 percent of the FPL to 9.5 percent for households at 400 percent of the FPL.
This subsidy is given to the household, regardless of whether they choose to buy a silver plan or a
bronze, gold, or platinum plan instead.

2.3 Data

This study uses data from the Scientiﬁc Registry of Transplant Recipients (SRTR). The SRTR
data system includes data on all donors, waitlisted candidates, and transplant recipients in the US,
submitted by the members of the Organ Procurement and Transplantation Network (OPTN). The
Health Resources and Services Administration (HRSA), US Department of Health and Human
Services, provides oversight to the activities of the OPTN and SRTR contractors.

For each transplant candidate from 2008 to 2018, I observe their age, primary source of payment
13There is variation in how many diﬀerent insurance companies oﬀer plans on the marketplace
and also in which categories of plans are oﬀered in each county. Future work will explore how this
availability of plans eﬀects transplant candidates ability to register on the wait list.

38

for the transplant, and state of permanent residence.14 For each transplant candidate I also observe
their wait list registration date and the date of their transplant, if they receive one. Table B.2 shows
that 72 to 82 percent of candidates are between the ages of 18 and 64. These candidates are of an age
that would beneﬁt the most from the provisions of the ACA because they are too young to qualify
for Medicare. The racial distribution of candidates varies by organ needed. Although 82 percent of
lung candidates are white, only 46 percent of kidney candidates are white. Additionally, with the
exception of kidney candidates, who are eligible for Medicare before they turn 65, approximately 50
percent of candidates had private insurance as their primary source of payment and depending on
the organ needed, 7 to 17 percent of candidates had Medicaid insurance before the ACA. Because
kidney candidates can use dialysis to help replace the function of their kidneys before they receive
a transplant, kidney candidate mortality is low compared to other organ candidates. Only 4 percent
of kidney candidates die within a year of registering on the wait list. However, for other organs,
approximately 15 percent of candidates die within a year of listing. For candidates who need a
heart, liver, or lung, receiving a transplant quickly can be a deciding factor in whether they live or
die. Because of these diﬀerences between the candidates of diﬀerent organ types, I expect that my
estimates of the ACA on transplant candidates will also vary by organ type needed.

I use population estimates from the National Cancer Institute to create population adjusted counts
of wait list registrations and transplants for each state.15 Population counts are only available at
a yearly level, so I linearly interpolate the increase in population between years into estimated
monthly counts. I also use information obtained from the Henry J Kaiser Family Foundation on
the Medicaid expansion dates of each state.16

14Although I can also observe the secondary source of payment for some candidates, for over
60% of candidates it is unobserved. For this reason, I am unable to look at changes in wait
listing or transplants among candidates who might have gained private or Medicaid insurance
as their secondary insurance as a result of the ACA. However, the Medicaid expansions and
private insurance marketplace subsidies were targeted at beneﬁting people who would otherwise
be uninsured. These candidates would gain private or Medicaid insurance as their primary source
of payment which I can observe reliably.

15Source: https://seer.cancer.gov/popdata/download.html
16Source: https://www.kff.org/health-reform/state-indicator/state-activity

39

2.4 Wait List Registrations

With the introduction of the ACA, many potential transplant candidates have new options for
insurance coverage through either the Medicaid expansions or through the Marketplace for private
insurance.17 This creates an opportunity not only for previously uninsured individuals to gain
coverage, but also for individuals who already had coverage to switch to being insured through
one of these channels instead. I expect that a majority of new registrations come from previously
uninsured candidates who are gaining insurance from the ACA, not from candidates switching
from other insurance types to Medicaid or Marketplace private insurance.18 Courtmanche et al.
(2017) provides some support for this assumption. They estimate that only 27 percent of people
who bought private insurance on the Marketplace already had some form of insurance coverage.
If these people are switching insurance coverage to a private plan on the Marketplace, it is likely
that they are doing so because the Marketplace plans oﬀer better coverage or lower expected costs
to the individual. If this is the case, these individuals’ prior insurance coverage might not have
been good enough to allow them to register to receive a transplant, which would mean that these
are still new registrations. Additionally, Courtmanche et al. ﬁnd that increase in Medicaid insured
individuals after Medicaid expansion does not come from a crowd out of private insurance.19 I

-around-expanding-medicaid-under-the-affordable-care-act/

17This is conditional on transplant centers being willing to accept Medicaid patients or patients
with insurance from the Marketplace. This not necessarily the case. Medicaid payments lower than
those from Medicare for many procedures (Mabry et al., 2016). There are also documented cases of
people with insurance from the Marketplace having diﬃculties ﬁnding physicians who would accept
their insurance.
(See https://www.nytimes.com/2016/05/15/sunday-review/sorry-we
-dont-take-obamacare.html)

18Future work will explore if I see any evidence of candidates switching between insurance types
as a result of the ACA. To do this, I will compare the insurance type that a candidate uses when
they register on the wait list before the ACA to the insurance type that a candidate uses when they
receive a transplant after the ACA.

19One might also be concerned that individuals are switching insurance from Medicare due to
disability to private insurance or Medicaid.
I do not ﬁnd any evidence of this when I look for
a change in the number of registrations by candidates who are less than 65 years old who have
insurance through Medicare. Additionally, I ﬁnd no evidence of a decrease in the number of
candidates who join the wait list with their primary source of payment listed as self, donation, or
free care after the ACA. This suggests that candidates are also not switching from these payment

40

hypothesize that new candidates that join the wait list after the ACA should increase the number
of private insurance registrations in all states and also increase Medicaid insurance registrations in
expansion states. Future work will also explore how existing candidates respond to the expected
increase in Medicaid and privately insured candidates on the wait list. If candidates expect that
wait lists will get longer in Medicaid expansion states, candidates who are not Medicaid insured
might choose to register in non-expansion states.

I expect Medicaid expansion to increase the number of Medicaid insured candidates on the
wait list in states that expand Medicaid, but not in non-expansion states. Column 1 of Figure B.1
shows that this seems to be the case: on average, Medicaid expansion states have an increase in
Medicaid insured candidates after the ACA, but non-expansion states do not have a similar increase
in registrations. I also expect that both Medicaid expansion states and non-expansion states should
see an increase in the number of privately insured candidates after transplants are declared an
EHB and due to the increase in the availability and aﬀordability of private insurance plans on the
Marketplace. With the exception of heart candidates, Column 2 of Figure B.1 does not show any
noticeable increases in privately insured candidates after the ACA.20

To estimate the eﬀects of the changes implemented with the ACA on the monthly number of

registrations, I estimate the following model for each transplantable organ separately21:

W Lst = β1EHBst + β2PostMedicaidE xpansionst + γxst + µs + θt + ust,

(2.1)

where yst is either the monthly number of Medicaid or private insured wait list registrations per 1
million state population. EHBst = 1 if state s deﬁnes a transplant as an Essential Health Beneﬁt
in their Benchmark Plan in month t and PostMedicaidE xpansionst is a dummy variable that
equals one after a state expands their Medicaid program. xst = 1 if state s requires adults to wear
types to private insurance or Medicaid.

20Figure B.1 also highlights that even before the ACA, states that expanded Medicaid eligibility
had more than twice the number of transplant candidates per population than states that did not
expand Medicaid. Future work will explore how these wait list size diﬀerences interact with the
Medicaid expansions.

21Results for intestines and pancreas transplants are not included in this paper due to the small

number of recipients of these transplants. Results for these organs are available upon request.

41

motorcycle helmets in month m. µs and θt are state and month-year ﬁxed eﬀects. β1 represents
the eﬀects of states mandating coverage for transplants in private insurance plans after the ACA.
β2 represents the eﬀects of a state expanding its Medicaid program to cover individuals up to 138
percent of the FPL.

Table B.3 presents estimates of equation (2.1). The coeﬃcient in row 1, column 1 for kidneys
implies that after the ACA requires private insurance plans on the marketplace to cover transplants,
on average, there is no change in Medicaid registrations, but, row 2 shows there is an increase in
Medicaid registrations after a state expands their Medicaid eligibility. Furthermore, the estimated
increases in Medicaid registrations is substantial; These estimates for kidneys implies that on
average, after a state expands their Medicaid program, monthly Medicaid wait list registrations per
million state residents increases by 0.29. This represents a 59 percent increase in registrations,
comparing this eﬀect size to the pre-ACA monthly average number of registrations of 0.49. The
estimated increases in Medicaid registrations for other organs are similar. After Medicaid expansion
there is a 19 percent increase in Medicaid insured heart candidates, a 60% increase in Medicaid
insured lung candidates, and a 41 percent increase in the Medicaid insured liver candidates. These
estimates suggest that the Medicaid expansions allowed signiﬁcantly more low-income individuals
to gain Medicaid coverage and register on the wait list for a transplant for all organs as a result.

Row 1, Column 2 for kidneys in Table B.3 shows that there is no signiﬁcant eﬀect of mandating
transplants as an EHB on the monthly privately insured wait list registrations per million state
residents. This also holds true for for heart, lung, and liver candidates. The coeﬃcient in row 2 for
kidneys and lungs implies that after a state expands their Medicaid program, on average, there is no
signiﬁcant increase monthly wait list registrations per million state residents by privately insured
candidates. Eﬀects for hearts and livers are negative and statistically signiﬁcant. These negative
eﬀects could result from one of the following scenarios: The ﬁrst scenario is privately insured
candidates are not able to, or choose not to, register on the wait list after the Medicaid expansion.
This seems unlikely because transplant centers do not have limits on how many candidates they
can accept. Furthermore, candidates are unlikely to willingly choose not to pursue a lifesaving

42

transplant as a result of Medicaid expansions. The second scenario is that there is a diﬀerential
eﬀect of the private insurance marketplace between expansion and non-expansion states. Then, the
negative coeﬃcient on Medicaid expansion could represent that candidates in non-expansion states
are more likely to gain private insurance coverage than candidates in Medicaid expansion states.
Figure B.1 lends some support to this theory: it does appear that there is an increase in private
insurance registrations in non-expansion states after 2014 that is not present in expansion states.
Future work will explore alternative speciﬁcations that would allow a diﬀerential eﬀect of EHB
designation on candidates in expansion and non-expansion states. Future work will also explore
whether Medicaid expansions induce some non-Medicaid insured candidates to register at centers
that are not in expansion states. It is possible that non-Medicaid insured candidates recognize that
Medicaid expansion will lead to more Medicaid candidates joining the wait lists and as a result
non-Medicaid candidates might seek out centers that they expect will have shorter wait lists.

In order to interpret my Medicaid expansion results as causal, expansion and non-expansion
states need to have parallel trends in registrations. To test the plausibility of this assumption, I
estimate the following event study model, which allows the Medicaid expansion treatment eﬀect to
diﬀer in every year:

yst = β1EHBst +

y∈Y

βy1(t ∈ y) ∗ E xpansions + γxst + µs + θ y + ust,

(2.2)

This equation is identical to equation (2.1) except PostMedicaidE xpansionst has been replaced
with a series of dummy variables for each year (1(t ∈ y)) interacted with E xpansions which equals
1 if state s expanded Medicaid. Y includes all years from 2008 to 2018 except 2013, the year
before the introduction of the ACA. Therefore βy represents the diﬀerence between expansion and
non-expansion states in year y, normalized the diﬀerence in 2013. If expansion and non-expansion
states have parallel trends in registrations, estimates of βy should be approximately zero before
2014, representing that expansion and non-expansion states do not diﬀer diﬀerentially over time
before 2014. Figure B.2 plots the estimated βy’s for Medicaid registrations and private insurance
registrations by organ. The estimated βy’s before 2014 are approximately zero supporting my
assumption of parallel trends. Furthermore, in column 1, the estimated βy’s after 2014 are visually

43

more positive than before 2014, which is consistent with the estimated coeﬃcients in Table B.3.
Note that although this ﬁgure supports my assumption of parallel trends, a state’s decision to expand
their Medicaid eligibility or deﬁne transplants as an EHB could be correlated with diﬀerential organ
demand or supply trends between expansion and non-expansion states. Future work will explore
alternative speciﬁcations that are robust to these types of diﬀerential trends.

2.5 Deceased Donor Transplants

Although I ﬁnd an increase in Medicaid candidates as a result of Medicaid expansion, joining
the wait list does not guarantee that a candidate will receive a transplant. There is a shortage of
deceased donor organs, so candidates need to wait to be oﬀered an organ before they can receive a
transplant. It is also possible that while candidates are waiting for an organ oﬀer, they could become
to sick to receive a transplant.22 As a result, although I expect to see an increase in the number of
transplants occurring for Medicaid candidates after Medicaid expansion, I expect the increase in
transplants to be smaller in magnitude than the increase in the number of candidates who joined the
wait list. Additionally, because the supply of organs is constrained, if more Medicaid candidates
receive transplants, this means that fewer candidates with other insurance types would be able to
receive transplants.23

Figure B.3 illustrates the diﬀerences between yearly private or Medicaid deceased donor trans-
plants over time by a state’s Medicaid expansion status. These ﬁgures preview my results: there
appears to be a positive eﬀect of Medicaid expansion on Medicaid insured deceased donor trans-
plants, but Medicaid expansion does not have clearly visible eﬀect on privately insured deceased
donor transplants.

Table B.4 estimates equation (2.1) where yst is now the monthly number of Medicaid insured
or private insured deceased donor transplants per 1 million state population. As expected, there
22Future work will look at how the health of candidates on the wait list changes as a result of the

new Medicaid candidates.

23It is possible that the additional candidates added after the ACA are more willing to accept
“worse” quality organs that would not have been accepted by other candidates, resulting in an
increase in the supply of organs. Future work will explore this possibility.

44

is no increase in private insurance transplants, likely because there is no increase in private
insurance registrations after the introduction of the Marketplace. For all organs, the estimated
eﬀect of Medicaid expansions on deceased donor transplants is positive.24 Although this increase
is only statistically signiﬁcant for lung and liver transplants, the estimated increase in transplants
is clinically signiﬁcant: the increases in transplants represents a 28 percent increase in the number
of Medicaid deceased donor kidney transplants, a 7 percent increase for heart transplants, a 47
percent increase for lungs, and a 41 percent increase for liver transplants. Additionally, these
increases imply that 12 percent of the new kidney Medicaid candidates, 25 percent of new heart
candidates, 58 percent of new lung candidates, and 54 percent of new liver candidates are able to
receive transplants as a result of the Medicaid expansions. Table B.4 also provides some evidence
that the new Medicaid candidates crowd out transplants for candidates with private insurance. For
all organs, I estimate a negative eﬀect of Medicaid expansion on the number of privately insured
transplants of a similar magnitude to the increase in Medicaid transplants, supporting my theory
that these private insurance transplants are being crowded out.

As discussed in the previous section, the assumption of parallel trends needs hold in order for
these estimates to be causal. To access the plausibility of this assumption, I estimate equation
equation (2.2), where yst is now the monthly number of Medicaid insured or privately insured
deceased donor transplants per 1 million state population. Figure A.4, shows the estimated trends
in number of transplants between expansion and non-expansion states are approximately parallel
before the ACA, supporting this assumption. Additionally the ﬁgure shows that over time the eﬀect
of the Medicaid expansions on transplants increases. This is likely due to fact that candidates need
to wait on the wait list to receive an organ oﬀer before they can receive a transplant.

24As shown in Table B.2, candidates can frequently need to wait longer than a year to receive a
transplant after joining the wait list. As a result of this, I expect the eﬀect of Medicaid expansion
on transplants to increase over time. Figure B.3 supports this hypothesis, as does Figure B.4. Note
that although I expect the magnitude of the eﬀect increase over time, the functional form I estimate
averages the size of the eﬀect in all post expansion years.

45

2.6 Living Donor Transplants

Patients who need a kidney or liver transplant have the option of ﬁnding someone currently
living who is willing to donate one of these organs to them. These living donors need to be blood
type compatible with the potential recipient and the living donor usually needs to have insurance
coverage for after they donate.25 Some patients who need a transplant before the ACA and who did
not have insurance are likely to have had someone who was willing to donate to them, but due to the
candidate’s lack of insurance, the transplant did not take place. If these candidates gain insurance
after the ACA, these living donor transplants are more likely to occur. Moreover, these transplants
represent new transplants that increase the total number of transplants because they do not use one
of the limited number of deceased donor organs.

Table B.5 presents estimates of equation (2.1) where yst is the monthly number of Medicaid
insured or privately insured living donor transplants of kidneys or livers.
I ﬁnd that Medicaid
expansion more than doubles the number of Medicaid insured liver living donor transplants. Addi-
tionally, although the eﬀect on Medicaid insured kidney transplants is not statistically signiﬁcant, it
does represent a 35% increase in the number of Medicaid insured kidney living donor transplants.
Figure B.5 supports the assumption of parallel trends. Although the estimated eﬀect of Medicaid
expansion on privately insured living donor kidney transplants is negative, Figure B.6 makes it
clear that this is not a causal eﬀect because expansion and non-expansion states do not have parallel
trends in privately insured living donor kidney transplants. This pretrend is also visible in Figure
B5. Future work will explore diﬀerent speciﬁcations that address this diﬀerential pretrend.

2.7 Conclusion

One aim of the ACA was to increase insurance coverage for previously uninsured low income
individuals. The increase in access to insurance should allow individuals to receive needed
transplants if they experience organ failure. Using the universe of organ transplant candidates
25Although the organ recipient’s insurance will pay for the donor’s surgery to remove the organ for
transplant, some transplant centers want the donor to have insurance in case they have complications
after the surgery.

46

and recipients, I ﬁnd that state Medicaid eligibility expansions allowed about 50 percent more
Medicaid insured candidates to register on the wait list to receive a transplant. This increase in
access to Medicaid insurance also allows more Medicaid deceased donor transplants to take place
and a signiﬁcant increase in living donor liver transplants to occur. However, I do not ﬁnd evidence
that the introduction of the private insurance marketplace increases the number of individuals with
private insurance registering.

More work should be done to understand why there is no eﬀect of the marketplace on private
insurance registrations and should explore heterogeneity in these results by the proportion of
uninsured individuals in counties and also by availability of diﬀerent metal level marketplace plans
across counties.26 Future work might also look at additional outcomes like patient and graft
survival, or the wait time of candidates before transplant.27

The increase in the ability of low income individuals to register on transplant wait lists after the
ACA is especially encouraging considering the organ allocation policy is created to be “equitable”.
Although the chances of receiving an organ are equal across insurance sources after candidates
join the waitlist, candidates face diﬀerent barriers to being waitlisted based on the source of their
insurance. Uninsured individuals do not even have access to the pool of organs. Ansell et al. (2014)
highlight this disparity in their blog post, stating “Studies estimate the uninsured receive less than
1 percent of all organs but comprise almost 20 percent of all organ donors”. With truly equitable
26I also might explore heterogeneity by change in the income eligibility level for Medicaid.
For example, Vermont actually decreased their eligibility level when they “expanded” Medicaid
eligibility so this might lead to a diﬀerent eﬀect even though it’s an “expansion” state. An additional
example is Wisconsin, which isn’t an “expansion” state, but they started oﬀering Medicaid coverage
for childless individuals with income up to 100 percent of the FPL around the same time as the
introduction of the ACA.

27Even though I don’t see an eﬀect of the marketplace on private insurance wait list registrations,
it might have an eﬀect on candidates who receive transplants with other insurance types. For
example, if a candidate loses their insurance after they receive a transplant, they might be able to
purchase a plan on the marketplace that would help them pay for the immunosuppressants they need
to help maintain their transplant. This would be especially relevant for ESRD Medicare kidney
transplant recipients who lose Medicare coverage 3 years after they receive their transplant.

Existing candidates on the wait list might be made worse oﬀ by the inﬂow of new Medicaid

candidate registrations, leading to longer wait times for them.

47

allocation, all individuals should have equal access to the organs being allocated. By increasing
access to Medicaid insurance, the ACA helps improve equity in the organ allocation system.

48

CHAPTER 3

OPIOIDS AND ORGANS: HOW OVERDOSES AFFECT THE SUPPLY OF DONORS,
WAITING LISTS, AND TRANSPLANT OUTCOMES (with Stacy Dickert-Conlin, Todd

Elder, and Keith Teltser)

3.1 Introduction

The Organ Procurement and Transplant Network (OPTN) proclaimed that “US organ transplant
and deceased donors set new records in 2016.”1 The source of the record-setting number of donors
and transplants includes a dramatic increase in donors who died of drug intoxications. Figure C.1
shows that the share of all deceased donors who experienced drug intoxication rose from less than 1
percent in 1995 to 13.1 percent in 2018. In Figure C.2 we see that the number of donations arising
from drug intoxication is quickly approaching the number that arise from motor vehicle donations.
The news is sobering in the midst of an opioid epidemic, in which the number of drug overdose
deaths more than tripled in the 15 years since 1999 (Centers for Disease Control Prevention, 2018).
Quantifying the eﬀect of the increase in opioid overdoses on donations and transplants leads
to signiﬁcant equity and eﬃciency questions, given the allocation system of organs in the United
States. Notably, donated organs from deceased donors are allocated ﬁrst, in most cases, to transplant
candidates in the geographic region where the organ came from. Given that the opioid epidemic is
more widespread in some parts of the United States, OPTN’s often competing goals of increasing
“the number of transplants”, providing “equity in access to transplants”, and improving “waitlisted
patient, living donor, and transplant recipient outcomes” (OPTN, 2017) may be further tested. For
example, in Massachusetts opioid overdose deaths rose by 248 percent between 2010 and 2017,
while the number of donors from drug intoxication rose by 389 percent. In contrast, opioid overdose
deaths rose by only 40 percent in Iowa over the same period, and the number of donors from drug
intoxication remained unchanged. Diﬀerential growth rates have the potential to diﬀerentially
1https://optn.transplant.hrsa.gov/news/us-organ-transplants-and-deceased

-donors-set-new-records-in-2016/

49

impact waiting list behaviors and outcomes of transplant candidates.

In this paper, we use mortality data from the National Vital Statistics System and restricted-use
data on transplant candidates and recipients from the Scientiﬁc Registry of Transplant Recipients
to study the extent to which the recent growth in fatal drug overdoses aﬀects the supply of deceased
organ donations. We ﬁnd that every 100 additional opioid deaths leads to 1.9 additional donors,
allowing 5.3 additional transplants to take place. With the increase in donations and transplants
arising from the opioid epidemic, we further consider whether organ transplant candidates respond
to the increased supply. Local administrators allocate organs from deceased donors within their
geographic areas via a waiting list process. Existing research (Dickert-Conlin et al., forthcoming;
Fernandez et al., 2013; Lemont, 2019; Choi, 2019) shows that transplant candidates respond
dramatically to supply shocks by increasing their likelihood of listing in areas with the supply
increases and, in the case of kidneys, choosing to pursue an organ from a deceased donor rather
than from a living donor. We ﬁnd surprisingly little measurable response to the opioid related
shocks. In particular, there is no evidence that the higher supply of organs from drug intoxication
deaths crowds out living donors. Overall wait list additions do increase in response to increased
organ donors in a geographic area, but this is primarily driven by kidney, liver, and lung candidates,
although the response from kidney candidates is not statistically signiﬁcant.

3.2 Institutional Details/Literature Review

3.2.1 The Opioid Epidemic

The number of drug overdose deaths more than tripled from 1999 to 2017 (CDC WONDER,
2017).2 Opioids represent a major contributor, as 68 percent of the 70,237 drug overdose deaths
in the U.S. in 2017 involved an opioid.3 This is up from 55 percent of the 38,329 overdose deaths
2Source: United States Department of Health and Human Services (US DHHS), Centers for
Disease Control and Prevention (CDC), National Center for Health Statistics (NCHS), Multiple
Cause of Death 1999-2017 on CDC WONDER Online Database, released 2018. Data are compiled
from data provided by the 57 vital statistics jurisdictions through the Vital Statistics Cooperative
Program. Accessedathttp://wonder.cdc.gov/mcd-icd10.html.

3Source: Authors’ calculations from CDC WONDER data

50

in 2010, which is an aggregate increase of 126 percent over those seven years. This growth from
1999 to 2017 was overwhelmingly driven by deaths involving non-methadone synthetic opioids
(for example, fentanyl) and heroin (CDC WONDER, 2017).

While the epidemic is a matter of national concern, there is considerable cross-state variation
in opioid overdose death rates (CDC WONDER, 2017). In 2017, the states with the highest rates of
opioid overdose deaths were West Virginia, Ohio, and District of Columbia (460, 370, and 350 per
million population, respectively). The states experiencing the most rapid per-capita growth from
2010 to 2017 were District of Columbia, New Jersey, and Connecticut (518%, 421%, and 329%,
respectively). Meanwhile, the states with the lowest opioid overdose death rates were Nebraska,
Montana, and Hawaii (30, 40, and 40, per million population, respectively). The states with the least
growth from 2010 to 2017 were Montana, Hawaii, and Oklahoma who all experienced declining
opioid overdose death rates (-35%, -31%, and -27%, respectively).

3.2.2 Link to Organ Donation and Transplantation

Almost all deceased organ donors are brain dead at the time of organ recovery, which means that
all brain function has irreversibly stopped. Because the heart continues to beat after brain death
occurs, current medical technology allows for respiration via a ventilator, so that the internal organs
receive oxygenated blood and remain viable for transplantation.4 When a person experiences an
opioid overdose their breathing often slows or stops, which causes hypoxia (reduced oxygen to the
brain) and can cause brain death.5 We hypothesize that there is a mechanical relationship between
opioid overdose deaths and organ donation: more brain deaths due to opioid overdoses will increase
the number of organ donors.

Initial published links between the opioid epidemic and organ donation were made circumstan-
4In contrast, organs deteriorate rapidly following cardiac deaths and are therefore unsuitable for
transplantation except in extraordinary circumstances in which patients with non-survivable brain
injuries who are not brain dead because they retain some minimal brain stem function become
donors. See http://www.organtransplants.org/understanding/death/.
5See https://www.drugabuse.gov/publications/drugfacts/heroin

51

tially or anecdotally. For example, the April 2016 issue of Journal of Transplantation reprinted
a Morbidity and Mortality Weekly Report from the Center for Disease Control and Prevention
entitled “Increases in Drug and Opioid Overdose Deaths-United States. 2000-2014” (Rudd et al.,
2016), which is entirely about the rise in overdose deaths, with the following abstract inserted
before the reprint: “Deaths from opioid overdoses are increasing in the United States, which may
have an impact on the organ donor pool.” Around the same time in 2016, numerous newspaper
articles appeared that highlighted the link between drug overdoses and organ donors and addressed
the health of drug overdose donors. The Washington Post (Izadi, 2016) quotes David Klassen, the
chief medical oﬃcer for the United Network for Organ Sharing as saying, “[T]ruthfully, people
who are dying of drug overdoses are young and tend to be otherwise healthy. In many ways, they
are ... potentially excellent donors, from an organ quality standpoint.” In the same article Klassen
cautions that long-term drug use can have damaging eﬀects on organs. Seelye (2016) in the New
York Times and Wenner (2016) in PennLive.com report that victims of drug overdose may be higher
risk because they practice other risky behaviors that are associated with HIV and hepatitis C, but
even these diseases can be treated or cured in the small chance that the transplant recipient contracts
the disease.6 Further, the risk of contracting a disease from the organ may be preferable to the
outcome of death if the transplant candidate does not receive an organ at all.7

Goldberg et al.’s (2016) Personal Viewpoint article in the American Journal of Transplantation
systematically documents the rise in the number of donors whose deaths are due to drug intoxication
between 2003 and 2014 and shows vast geographic and organ diﬀerences in the changes in organ
donation from drug overdose. They also document that donors who died from a drug overdose have
the highest overall donation rate, yet do not have the highest mean number of organs transplanted
6Additionally, OPTN reports that the risk of contracting these diseases from an organ is smaller
than the individual’s lifetime risk of dying from a traﬃc accident. (https://optn.transplant
.hrsa.gov/media/2270/dtac\_guidance\_risks\_201706.pdf)

7The webpage organdonor.gov notes that “[T]here are very few conditions that would prevent
someone from being an organ eye or tissue donor - such as HIV infection, active cancer, or a
systemic infection. Even with an illness, you may be able to donate your organs or tissues.” Note
that the risk of waiting for a lower risk organ diﬀers by organ. Kidney patients have dialysis, which
allows them to survive without a transplant, but other patients who need other organs do not.

52

per donor. They conclude by urging the medical community to maximize the utilization of potential
donors in the face of organs from drug overdose donors often being labeled as “increased risk”.
In a “Special Article” for the Transplantation Journal, Weiner et al. (2017) make the same plea,
suggesting that “due to concerns over disease transmission (HIV, hepatitis B, and hepatitis C virus),
these donors are underused by the transplant community.”8

Of course, for a potential organ donor to become an actual organ donor, the individual or
Individuals can do so by indicating their preference
their family must consent to the donation.
on their driver’s license or by registering on a state registry. First person consent laws, which
exist in most states, require health care professionals to abide by the potential donor’s consent to
recover organs. If no such indication exists, health care professionals will seek permission from the
potential donor’s next-of-kin.9 Wenner (2016) interviews the vice president for clinical services
for a Pennsylvania Organ Procurement Organization (a local administrative body of OPTN) who
reports that 83 percent of potential donors who died of drug overdose ultimately become donors
while only 63 percent of the general population of potential donors do. He hypothesizes that this is
in part due to higher donor registrations rates among overdose victims, who are more likely to be
from young, white, higher income backgrounds than the overall population: 46 percent of overdose
victims are registered donors, compared to 29 percent of the overall population of potential donors.
In addition, he suggests that family members may be more likely to consent to donation after a
drug overdose so that something positive comes from a tragic experience.10, 11 After consent for

8See

the

reference

to Volk

in

this

article

and

the

following:

https://

www.healio.com/hepatology/transplantation/news/online/\%7Bc95305eb-691c
-409c-9cad-d184ed02ade9\%7D/organs-donated-after-drug-overdose-safe-for
-transplantation

https://consultqd.clevelandclinic.org/harvesting-life-from-a-deadly

-epidemic-protocol-ups-heart-transplants-from-overdose-deaths/

https://annals.org/aim/fullarticle/2678899/drug-overdose-epidemic

-deceased-donor-transplantation-united-states-national-registry.

9Howard (2007) estimates that donation rates among all potential donors range from 51 to 60

percent, primarily due to low consent rates.

10Siminoﬀ et al. (2001) ﬁnd that a family was more likely to consent to donation if the death

was due to trauma compared with non-trauma related deaths.

11Future work might consider whether the probability transplant candidates/physicians accept

53

donation has been given, the organ needs to be oﬀered to potential recipients.

3.2.3 Transplants

Transplant candidates seeking a deceased donor transplant must ﬁrst register with a transplant
center in order to join the national waiting list. Each transplant center is located in one of 58
donation service areas (DSAs), which broadly follow state boundaries, although large states have
multiple DSAs, while some DSAs include multiple states or portions of states. When a deceased
donor organ becomes available, it is ﬁrst oﬀered to candidates who are registered in the same DSA
from which the organ was recovered.

Because organs are allocated ﬁrst to these local candidates, transplant candidates may also want
to register on multiple waiting lists in diﬀerent DSAs, a process known as “multilisting”. In order
to do this, candidates need to be able to arrive in time to the center to receive a transplant if an
organ was to become available and they also need to be accepted at the additional center. Some
centers do not accept candidates who have already been registered at another center.12 Although
only 6 percent of all candidates multilist, those who do generally have higher resources than those
who do not, with higher education and employment and lower Medicaid insurance coverage.

3.3 Data and Descriptive Statistics

This study uses data from the Scientiﬁc Registry of Transplant Recipients (SRTR). The SRTR
data system includes data on all donors, waitlisted candidates, and transplant recipients in the US,
submitted by the members of the Organ Procurement and Transplantation Network (OPTN). The
Health Resources and Services Administration (HRSA), US Department of Health and Human
Services, provides oversight to the activities of the OPTN and SRTR contractors. The SRTR data,
which comes from hospitals and immunology laboratories, include detailed donor level data such as
organs from drug intoxication deaths evolves over time and how this aﬀects the rate at which
donated organs are transplanted.

http://optn.transplant.hrsa.gov/learn/about-transplantation/

12See

transplant-process/ for more details.

54

cause, circumstance, and mechanism of death, age, gender, geographic location of donation, which
organs were recovered, discarded, and transplanted, and relevant health markers. SRTR data also
include transplant candidate-level information such as time spent on the waitlist, transplant centers
at which each candidate is registered, health markers, demographics such as zip code of residence,
reason for leaving the waitlist, and follow-up health data for those who receive a transplant. All
donors can be matched with the transplant recipient when a transplant took place. In our analysis
of the SRTR data, we use the years 2000, when data began being consistently collected on our
variables of interest, through calendar year 2018.13

The mechanism of death is a key variable that includes the following categories: drug in-
toxication, gunshot/stab wounds, asphyxiation, blunt injury, cardiovascular, drowning, electrical,
intracranial hemorrhage/stroke, natural causes, seizure, SIDS, none of the above, or not reported.
We count donors whose mechanism of death is coded as “drug intoxication” as donors who arose
from a drug overdose. In addition, SRTR codes a cause of death and circumstance of death for
each donor. Not surprisingly, approximately 93 percent of all donors whose death is coded as a
“drug intoxication” receive “anoxia” (deﬁned as injury to the brain due to lack of oxygen) as their
cause of death; “cerebrovascular/stroke” and “other” make up the remainder. Among the donors
whose mechanism of death is “drug intoxication”, almost half have no circumstance of death listed,
almost 30 percent are listed as non-motor vehicle accidents, approximately 20 percent are suicides
and the small remainder list natural causes as their circumstance of death.

Figure C.3 shows the dramatic rise in the number of organ donors whose death results from
drug intoxication (DI). While the number of donors from all other mechanisms combined have been
relatively ﬂat and even falling in some years over the past decade, the donors from drug intoxication
deaths have been dramatically rising in every year. Figure C.3 also supports the argument that
donors whose death results from drug intoxication are often young, meaning that these organs are
good organs to use in transplants.

Table C.1 shows that an average of 3.138 organs are transplanted per donor who died due to drug
13We have SRTR data through 2018, but the 2018 Mortality data are not yet available.

55

intoxication, which is consistent with these donors being young and otherwise healthy. However,
Table C.1 also shows that organs transplanted per donor who died due to drug intoxication is
relatively low compared to many other mechanisms of death. Among deaths that occur from a
gunshot wound and blunt injuries, 4.340 and 3.687 organs are transplanted per donor, respectively.14
We also use Vital Statistics mortality micro data with county identiﬁers and population data
from the National Cancer Institute at the individual level from 1999 to 2017 from the National
Vital Statistics System (NVSS) multiple cause of death mortality ﬁles.15 These data include every
reported death of a resident in the United States and for each death, the ﬁles record information
obtained from the individual’s death certiﬁcate. This information includes the individual’s county
of residence and their cause of death. We map these counties into the Donation Service Area
that covers that county. Following the CDC, opioid-related deaths are indicated by the following
International Classiﬁcation of Disease, Tenth Revision (ICD-10) multiple cause of death codes:
T40.0 (opium), T40.1 (heroin), T40.2 (other opioids), T40.3 (methadone), T40.4 (other synthetic
narcotics), and T40.6 (other and unspeciﬁed narcotics) (Rudd et al., 2017). Again, following the
CDC, we classify a death as an opioid overdose if it is opioid-related and the underlying cause of
death code is one of the following: X40-44 (unintentional drug poisoning), X60-64 (intentional self-
poisoning using drugs), X85 (homicide by drugs), or Y10-14 (drug poisoning with undetermined
intent).

The two panels of Figure C.4 preview our regression results.

In both panels, we divide 57
DSAs into quintiles based on their 2016 opioid deaths per capita as calculated in the Vital Statistics
Data.16 In the top panel, we show the DSAs in the top quintile of opioid deaths in 2016 also tended
to be in the top quintile of opioid deaths between 2000 and 2010. However, the diﬀerences between
14Goldberg (2016) reports a ﬁnding similar to this one; that the organs per donor are higher for
asphyxiation, gunshot wounds, and blunt injuries than for drug intoxication deaths in spite of drug
intoxication deaths leading to higher donation consent rates.

15Data are from the Multiple Cause of Death Files with County Identiﬁers, 1999-2017. as
compiled from data provided by the 57 vital statistics jurisdictions through the Vital Statistics
Cooperative Program.

16There are 58 DSAs in the United States; however, we do not have mortality data for the DSA

that includes Puerto Rico for all years in our sample, so we exclude this DSA from our analysis.

56

the top quintile and the other four began to widen dramatically around 2010. For example, prior
to 2010, the ratio of opioid deaths per capita between the top quintile and bottom quintile ranged
between 1.9 and 2.9. By 2016, the number of deaths per capita in the top quintile was 302, while
the number of deaths per capita in the bottom quintile was 52, which increased the ratio of deaths
to over 6.3.

The bottom panel shows that the DSA-level correlation between opioid deaths and DI organ
donors is positive, although not perfectly so. For example, the DSAs in the top quintile of 2016
opioid deaths per capita have the highest level of DI organ donors per capita in most years since
2010, including in 2016. Again, there has been dramatic divergence over time across groups.

3.4 Do Overdoses Lead to Increases in the Supply of Organ Donors and

Transplants?

We next turn to assessing whether drug overdoses aﬀect the supply of organ donors and
transplants. Our primary empirical strategy involves estimating DSA- and year-speciﬁc organ
donation and transplantation rates as a function of the number of deaths due to drug overdoses in
that DSA and year. We begin by estimating the following model:

Donorsst = αs + δt + γ(Deaths)st + st,

(3.1)

where Donorsst is a measure of the number of deceased organ donors, s indexes the DSA, t indexes
the year, and Deathsst is a measure of the number of drug-related deaths in that DSA and year.
All speciﬁcations include a full set of DSA and year indicators (αs and δt, respectively). We
weight each observation by the DSA’s population in that year using U.S. Census Bureau estimates.
Estimates of γ based on (1) capture the association between organ donation rates and drug-related
deaths over time within DSAs.

Table C.2 presents estimates of γ from speciﬁcation (3.1) separately for donations due to DI
and donations due to all other mechanisms. The top row presents our central estimates, in which
Donorsst includes only DI donors. In all cases, we measure Donorsst and Deathsst per million
state residents, so the estimates measure the eﬀect of one additional death on the supply of organ

57

donors. In column (1), we use the number of opioid-related deaths as our measure of Deathsst. This
variable includes all deaths in which one of the multiple causes of death is coded as opioid-related
(ICD-10 codes T40.0, T40.1, T40.2, T40.3, T40.4, and T40.6).17 The estimate of 0.019 implies
that each opioid-related death increases the supply of organ donors by 0.019, with a standard error
of 0.002.

In column (2), we use opioid overdoses as our measure of Deathsst. The set of opioid overdoses
is a subset of opioid deaths, as it includes deaths that 1) have the ICD-10 codes listed above, and 2)
have drug overdose listed as the underlying cause of death (ICD-10 codes X40-44, X60-64, X85, or
Y10-Y14). The point estimate in this column is again 0.019, with a standard error of 0.002. Finally,
column (3) shows estimates from models in which all drug overdoses is the measure of Deathsst.
This is a superset of the deaths in column (2) that also includes deaths due to anesthetics, sedatives,
and stimulants. The resulting point estimate is 0.021 (0.003). Clearly, the estimates are insensitive
to the particular measure of Deathsst that we use, and this pattern holds for all speciﬁcations that
we estimate throughout the paper. As a result, below we focus on models similar to column (1), in
which opioid deaths is our measure of Deathsst.

In all cases, the estimates in the top row of Table C.2 imply that an increase of 100 drug-related
deaths increases the supply of DI organ donors by only roughly two donors.18 Viewed in this
context, this implies that very few overdose victims eventually become organ donors. Nonetheless,
each drug-related death increases the supply of DI organ donors by roughly 1.4 percent (= 0.019 /
1.365) of the baseline donation rate of 1.365 per million state residents.

The principal threat to the internal validity of estimates based on speciﬁcations such as (3.1)
stems from diﬀerential unmeasured time trends in organ donation rates that may be correlated
17Speciﬁcally, ICD-10 code T40.0 refers to poisoning by opium, while T40.1, T40.2, T40.3,
T40.4, and T40.6 refer to poisoning by heroin, other opioids, methadone, synthetic narcotics, and
“other and unspeciﬁed narcotics”, respectively.

18An increase of 100 deaths is approximately the size of the yearly increase in opioid-related
deaths in a DSA during each year of the recent opioid crisis. In 2014, there were an average of 520
opioid-related deaths in a DSA. In 2015, the number of deaths increased to 599. In 2016 and 2017,
the number of deaths were 762 and 857.

58

with drug-related deaths. These time trends could reﬂect hospitals improving their procedures for
organ recovery, education campaigns encouraging informed consent, or endogenous law changes
in response to declining donation rates. Based on the “All Others” row of the table, we do not think
it is likely that diﬀerential trends drive the positive point estimates in the top row of the table. In
all three speciﬁcations, the point estimates on measures of drug deaths are negative. The estimates
are also small relative to the sample mean donation rate of 24.141 per million state residents.

Figure C.3 and Table C.3 show, the incidence of drug-related deaths varies widely by age and
gender, so in Table C.4 we present age- and gender-speciﬁc estimates of speciﬁcation (3.1) above.
The top row shows that each drug-related death increases the supply of male organ donors who died
due to drug intoxication by 0.0097 (0.0014) and increases the supply of female organ donors who
died due to drug intoxication by 0.0091 (0.0009). These estimates lie between 13 and 15 percent of
the respective sample means, shown in brackets. As another check on the plausibility of the results
shown above, the remaining rows in Table C.4 present estimates of equation (3.1) separately by age
categories. We expect the absolute eﬀects of opioid overdose deaths on DI donation rates to be
largest among individuals most likely to be involved in overdose deaths, who tend to be 25 to 54
years old (Hedegaard et al., 2018). The results conﬁrm this expectation: among males, roughly 91
percent (= (0.4867+0.2089) / 0.7679) of DI organ donors are in the 18 to 49 age range, and roughly
90 percent (= (0.0051+0.0031) / 0.0091) of the overall marginal eﬀect of deaths on male DI donors
is accounted for by donors in the 18 to 49 age range. Among females, the analogous fractions are
85 percent (= (0.3063+0.1993) / 0.5974) and 90 percent (= (0.0051+0.0031) / 0.0091), respectively.
For both sexes, DI organ donations are relatively uncommon (and relatively unaﬀected by opioid
overdoses) among those younger than 18 and older than 49.

3.4.1 The Eﬀects of Opioid Overdoses on Transplants

We turn next to assessing the eﬀects of opioid overdoses on the number of organ transplants.
Column (2) of Table C.5 shows the estimated eﬀect of opioid overdose deaths on total transplants
from DI donors. The estimate implies that an increase of 100 overdose deaths results in an average

59

of 5.3 more organs transplanted. Note that the number of additional transplants is roughly 2.8 times
the number of donors (1.9), which is roughly consistent with the descriptive statistics presented in
Table 2.

The remaining estimates in column (2) show organ-speciﬁc estimates. The estimate for kid-
neys implies that an increase of one additional overdose death increases the number of kidneys
transplanted by 0.029, or roughly 1.2 percent of the mean number of DI kidneys transplanted per
million persons [2.044]. DI liver transplants increase by 0.017 per million persons (1.6 percent
relative to the sample mean of 1.066), heart transplants increase by 0.006 (1.3 percent), and lung
transplants increase by 0.005 (1.9 percent). The estimates are also positive for pancreatic and
intestinal transplants but are small in magnitude in comparison to the other organs.

Columns (3) and (4) show analogous results for non-DI donors and transplants as a falsiﬁcation
test; all estimates in these columns are very small relative to their sample means (and negative in
all cases except for intestines).

3.4.2 Discussion

Although it is likely that this increase in donors and number of transplants is largely mechanical
because more deaths means there are more potential donors, there may also be some behavioral
responses to the opioid crisis. For example, transplant centers may be more willing to recover DI
organs as they learn more about their usefulness. That is, as more DI organs are transplanted over
time, transplant centers may realize that the risks of using such organs (e.g., HIV/HBV transmission)
are not as substantial as they previously believed. Figure C.5 lends support to this theory; over
time the organ recovery rates for young, drug-related death donors increases over time much faster
than the organ recovery rate for other young, non-opioid-related deaths.19 If transplant centers are
19In the future, we will explore this theory in more detail. Additional future work might put
this further into context by considering the correlation between other mechanisms of death rates
and their corresponding donation rates, including gunshot/stab wounds, asphyxiation, blunt injury,
cardiovascular, drowning, electrical, intracranial hemorrhage/stroke, natural causes, seizure, and
SIDS.

60

becoming increasingly more willing to recover opioid related organs over time, then the eﬀects that
we ﬁnd are not purely mechanical.

Transplant candidates may also respond to the opioid crisis. Ultimately, in order for a transplant
to take place, candidates must approve an organ oﬀer.
If candidates do not wish to receive DI
organs, they can refuse those oﬀers. If candidates frequently refuse DI organs, it could counteract
the mechanical increase in transplants from the increase in deaths. Organ discard rates may reﬂect
such a behavioral response; if the opioid crisis increases candidates’ beliefs that DI organs carry
an increased risk of infection, they may be increasingly likely to refuse DI organs oﬀered to them.
The more an organ is refused, the more likely it is to be discarded eventually. However, Figure C.6
suggests that this is not the case: the discard rate for young DI organs remains steady over time, and
DI organs are discarded at a similar rate to organs from donors who died from other mechanisms of
death.20 Note that these are not the only possible behavioral responses. In the following sections,
we explore how the opioid epidemic has aﬀected candidates’ waitlist behavior and living organ
donations.

3.5 Behavioral Responses to the Change in the Supply of Organs

3.5.1 Eﬀects of Supply Shocks on Waitlist Additions

In this section, we examine how a shock to the supply of deceased donor organs via drug overdose
aﬀects whether and where candidates register for a deceased organ transplant. As Dickert-Conlin
et al. (forthcoming) show, more candidates are induced to join a DSA’s waitlist following a positive
supply shock because the shock reduces expected waiting time and increases expected transplant
quality. However, because of this inﬂux of additional candidates, the net eﬀect on expected waiting
time is theoretically ambiguous. Moreover, they show that this inﬂux generates a positive externality
to neighboring DSAs, since candidates who are induced to register in multiple areas exit all waitlists
once they receive a transplant.

We test the predictions of Dickert-Conlin et al.
20This lack of a response is also supported by regressions. Results are available upon request.

(forthcoming) in the context of the opioid

61

epidemic using SRTR and NVSS data from 1999 to 2017. We estimate the following model:

Additionsst = αs + δt + γ(Deaths)st + st,

(3.2)

where Additionsst is a measure of the number of waitlist additions per capita, s indexes the DSA, t
indexes the year, and Deathsst is the number of opioid overdose deaths per capita in that DSA and
year. All speciﬁcations include a full set of DSA and year indicators (αs and δt, respectively). We
weight each observation by the DSA’s population in that year using U.S. Census Bureau estimates.
Estimates of γ based on (3.2) capture the association between waitlist registrations and opioid
overdose deaths over time within DSAs.

In Table C.6, we present estimates of the eﬀect of opioid overdose deaths on total waitlist
additions and separately by organ, including kidneys, livers, hearts, lungs, and pancreases (for
brevity, we exclude results for intestines hereafter, as intestines represent a very small segment
of the transplant market). In column (1) of the top row, the estimate implies that 100 additional
opioid overdose deaths results in an average of 15.4 more waitlist additions, with a standard error
of 6.7. This is a 9.3 percent increase relative to the mean of 165.881. The organ-speciﬁc estimates
suggest that increases in opioid overdoses mainly generate inﬂows onto waitlists for livers (21.6
percent), lungs (17.5 percent), and kidneys (6.3 percent, though this is not statistically signiﬁcant
at conventional levels). The estimated magnitudes of the wait list responses are large, especially
compared the increase in transplants that we estimated earlier:
the increase in registrations are
2.9 (=15.4/5.3) times as large as the increase in transplants. Note that these estimates capture
how waitlist additions respond to diﬀerential changes in opioid overdoses across DSAs. However,
as Dickert-Conlin et al.’s (forthcoming) conceptual framework implies, there may exist spillover
eﬀects between DSAs. As a result, the estimates in Table C.6 may understate the true eﬀects
because relatively large shocks in one DSA may induce additional waitlist registrations in another
DSA that experiences a relatively small supply shock.

Columns (2) and (3) of Table C.6 provide insight on which candidates respond to supply shocks.
Using zip code data for candidates and transplant centers, we generate separate counts of waitlist
additions for those who live inside and outside of the DSA’s boundaries. Waitlist inﬂows induced

62

by opioid overdoses are disproportionately concentrated among out-of-DSA candidates, relative
to the overall sample means. The organ-speciﬁc estimates suggest that most of the out-of-DSA
inﬂows primarily come from kidney and liver candidates, but the coeﬃcient for kidneys is not
statistically signiﬁcant at conventional levels. Additionally, liver candidates drive the increase in
in-DSA registrations, but lung and kidney candidates also see an increase in in-DSA registrations
in response to the increased opioid deaths.

The waitlist inﬂows presented in Table C.6 represent new registrations in the DSA, but not
necessarily new transplant candidates. This is because candidates may simultaneously register on
waitlists across multiple DSAs. Table C.7 diﬀerentiates the eﬀects of opioid overdose deaths on
waitlist inﬂows separately for candidates who only list at one transplant center and those who ever
multilist. From columns (1) and (4), we can see that most of the increase in waitlist registrations
come from single listers (10.8 per 100 vs 4.5 per 100 for multilisters). Among single listers,
most of the increase in registrations is from an increase in within-DSA listings, but the increase
in out-of-DSA listings is proportionally larger. Among listings by multilisters, most come from
out-of-DSA listings, but estimates are either marginally statistically signiﬁcant or not signiﬁcant at
conventional levels.

3.5.2 Living Donor Crowd-Out

Previous literature documents signiﬁcant substitution away from living donor transplants in response
to deceased donor supply shocks, especially among kidney candidates (Dickert-Conlin et al.,
forthcoming; Fernandez et al., 2013; Lemont, 2019; Choi, 2019). While earlier work shows
evidence of crowd-out, we ﬁnd no such evidence. Speciﬁcally, in Table C.8 we present estimates
from models of living-donor transplants as a function of opioid overdose deaths, ﬁnding small and
statistically insigniﬁcant estimates for all organs, kidneys, and “all organs except kidneys” (kidneys
account for nearly all living-donor transplants).

63

3.5.3 Discussion

Overall, we ﬁnd some evidence of candidates responding to the increase in supply of DI organs, but
the response is much smaller than that shown in other settings. Notably, the wait list response from
kidney candidates is smaller and we do not ﬁnd any crowd out of living kidney donors. One reason
this might be that the quality of the DI organs is lower than that of the supply shocks used to measure
responses in earlier papers, such as shocks due to increased deaths of helmetless motorcyclists. In
addition, there may be stigma associated with DI organs.

Additionally, we primarily see a response in wait listing by single listers. This increase in single
listings is likely driven by candidates choosing to list at a diﬀerent DSA than they otherwise would
have, rather than an increase in new candidates who wouldn’t have otherwise listed. This is because
an increase in opioid deaths is unlikely to induce new candidates to want or need a transplant.21
This is further supported by our lack of evidence of living donor crowd out, which would have led
to new candidates joining the waitlist.

Given that we see very little behavioral response by candidates, we expected that this should
improve patient survival. If more transplants are taking place, and no/few additional wait list can-
didates, candidates should have shorter wait times and should have better survival rates. However,
we ﬁnd little evidence of improved wait times or survival.

3.6 Conclusion

This study highlights the existence of unexpected beneﬁciaries of the opioid crisis:

those
awaiting organ transplants. Organ donations due to drug intoxication (DI) have increased more
than tenfold since 2000, and the rate of increase has accelerated in recent years as the opioid crisis
has deepened. Our central estimates suggest that 5.3 additional organ transplants occur for every
100 additional opioid overdose deaths. These eﬀects are concentrated among organ donors aged
21One exception to this might be liver candidates. Some centers are relaxing sobriety rules for
listing candidates with liver failure due to alcoholism. (Source: https://www.chicagotribune
.com/business/ct-biz-liver-transplant-sober-policy-0304-story.html) It is pos-
sible that the increase in the supply of livers from DI donors has had some eﬀect on this decision.

64

18-49, who are disproportionately likely to die of opioid overdoses.

Surprisingly, we see little demand-side response. In contrast to previous papers that ﬁnd that
kidney candidates respond strongly to supply shocks, we ﬁnd a relatively small and statistically
insigniﬁcant increase in waitlist registrations by these candidates. Additionally, they do not appear
to substitute away from living donors. Rather, we ﬁnd that liver and lung candidates drive most of
the additional waitlist registrations. It is possible that, due to lower life expectancies of liver and
lung candidates, they cannot be as selective about their organ source. The limited demand-side
response to the opioid supply shock is especially puzzling when contrasted with our evidence that
transplant centers are becoming increasingly willing to recover DI organs for transplant.

Based on our estimates, the number of transplant candidates who died while awaiting an organ
in 2015 would have been signiﬁcantly higher in the absence of the opioid epidemic. While the
crisis has undoubtedly been a tragedy, it has profoundly aﬀected the lives of thousands of transplant
recipients.

65

APPENDICES

66

APPENDIX A

TABLES AND FIGURES FOR CHAPTER 1

Table A.1: Characteristics of Pediatric, Young Adult, and Adult Kidney Candidates Before Share 35

Age

Male =1

White =1
Black =1
Hispanic =1
Other Race =1

Type A blood =1
Type B blood =1
Type AB blood =1
Type O blood =1

Does not have diabetes
Has diabetes
Unknown diabetes status

Private Insurance =1
Public Insurance =1
Other Insurance =1
0 ≤ PRA < 20
20 ≤ PRA < 80
80 ≤ PRA ≤ 100
Unknown PRA

Pre-Transplant Dialysis =1

Died while still on the wait list =1

Pediatric Young Adult Adult
10.67
48.07

19.64

0.58

0.55
0.21
0.20
0.05

0.35
0.13
0.04
0.48

0.97
0.01
0.02

0.50
0.45
0.05

0.55
0.03
0.01
0.41

0.24

0.01

0.56

0.51
0.25
0.17
0.06

0.37
0.12
0.03
0.48

0.96
0.01
0.02

0.47
0.49
0.04

0.61
0.06
0.03
0.31

0.42

0.02

0.60

0.55
0.25
0.13
0.07

0.36
0.13
0.04
0.46

0.69
0.29
0.02

0.51
0.46
0.03

0.68
0.09
0.05
0.18

0.47

0.05

Received a transplant =1
Observations

0.95
5,586

0.90
2,762

0.87
97,776

67

Notes:
This table contains pediatric (age 0 to 17), young adult (age 18 to 21), and adult (age 22+) kidney-only transplant
candidates who started waiting for a kidney transplant between 1998 and 2005 who have not had any previous kidney
transplants. The age of the candidate is how old they were at their HLA typing test. Public insurance is either Medicare,
Medicaid, CHIP, Department of VA, or “other government”. PRA score represents what percentage of the potential
donor pool a candidate has antibodies against that would cause them to reject an organ. High PRA scores are typically
a result of previous transplants or pregnancies. Other outcomes besides death on the wait list or transplant from the
wait list are patient removed due to improved or deteriorated health, candidate refused transplant, unable to contact
candidate, or candidate still waiting.

Table A.2: Estimates of Overall Share 35 Eﬀect on Wait Time and Quality

Wait Time Before Transplant

< 3 Months < 6 Months < 12 Months < 18 Months

Quality

[14.2]
-1.7*
(1.0)

[0.75]
0.11***
(0.02)

Share*Pediatric

[0.16]
0.04**
(0.01)

[0.37]
0.09***
(0.02)

[0.62]
0.10***
(0.02)

Observations

17,329

17,329

17,329

17,329

16,806

Robust Standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

Notes:
This table contains the estimated coeﬃcients of equation 1.1 where yi = KDPIi or yi = 1 if recipient i was
transplanted within 3, 6, 12, or 18 months using my entire sample of pediatric and young adult kidney transplant
recipients. The model include indicators for transplant year, age, and transplant center. Standard errors of estimates,
listed in parentheses, are robust to clustering within DSA over time. Pre Share 35 pediatric sample mean is listed in
brackets.

68

Table A.3: Estimates of Coeﬃcients for Living Donor Use at First Transplant

Share*Pediatric

All Living Donors

[0.56]
-0.17***
(0.02)

Observations
Robust standard errors in parentheses

17,355

*** p<0.01, ** p<0.05, * p<0.1

Notes:
This table contains the estimated coeﬃcients of equation 1.1 where yi = LivingDonori using my entire sample of
pediatric and young adult kidney transplant recipients. The model includes indicators for years, age, and transplant
center. Standard errors of estimates, listed in parentheses, are robust to clustering within DSA over time. Pre Share 35
pediatric sample mean is listed in brackets.

69

Table A.4: Characteristics of Pediatric and Young Adult Deceased Donor Kidney Recipients Before and

After Share 35

Pediatric

Young Adult

Pre Share 35 Post Share 35 Pre Share 35 Post Share 35

Male =1

White =1
Black =1
Hispanic =1
Other Race =1

Type A blood=1
Type B blood=1
Type AB blood=1
Type O blood=1

Does not have Diabetes
Has Diabetes
Unknown Diabetes Status

Private Insurance =1
Public Insurance =1
Other Insurance =1

Pre-Transplant Dialysis =1
0 ≤ PRA < 20
20 ≤ PRA < 80
80 ≤ PRA ≤ 100
Unknown PRA

0.57

0.42
0.30
0.21
0.07

0.31
0.14
0.03
0.52

0.97
0.01
0.02

0.34
0.63
0.04

0.83

0.85
0.05
0.01
0.08

Transplanted in less than a year
Mean KDPI
Mean Donor Age
Observations

0.44
27.9
26.2
2,481

0.57

0.37
0.25
0.31
0.06

0.30
0.13
0.03
0.53

0.98
0.00
0.01

0.31
0.68
0.01

0.78

0.50
0.04
0.01
0.45

0.58
18.9
21.0
4,175

0.55

0.39
0.36
0.18
0.07

0.37
0.10
0.04
0.49

0.96
0.01
0.03

0.23
0.75
0.02

0.93

0.76
0.11
0.05
0.08

0.22
35.6
29.9
1,018

0.58

0.35
0.33
0.24
0.08

0.37
0.12
0.04
0.47

0.97
0.03
0.01

0.20
0.79
0.00

0.93

0.60
0.12
0.03
0.25

0.19
35.7
29.2
978

Notes:
This table contains pediatric and young adult deceased donor kidney transplant recipients who had no previous kidney
transplants, and who were transplanted between 1998 and 2013. Public insurance is either Medicare, Medicaid, CHIP,
Department of VA, or “other government”. PRA score represents what percentage of the potential donor pool a
candidate has antibodies against that would cause them to reject an organ. High PRA scores are typically a result of
previous transplants or pregnancies.

70

Table A.5: Characteristics of Pediatric and Young Adult Living Donor Kidney Recipients Before and After

Share 35

Pediatric

Young Adult

Pre Share 35 Post Share 35 Pre Share 35 Post Share 35

Male =1

White =1
Black =1
Hispanic =1
Other Race =1

Type A blood=1
Type B blood=1
Type AB blood=1
Type O blood=1

Does not have Diabetes
Has Diabetes
Unknown Diabetes Status

Private Insurance =1
Public Insurance =1
Other Insurance =1

Pre-Transplant Dialysis =1
0 ≤ PRA < 20
20 ≤ PRA < 80
80 ≤ PRA ≤ 100
Unknown PRA

Transplanted in less than a year
Mean KDPI
Mean Donor Age

0.59

0.67
0.13
0.16
0.03

0.38
0.12
0.04
0.46

0.97
0.00
0.03

0.60
0.36
0.04

0.63

0.28
0.01
0.00
0.71

0.77
2.2
37.0

Parent was Donor =1
Other Relative was Donor =1
Non-Relative was Donor =1
Observations

0.75
0.18
0.08
3,157

0.60

0.67
0.10
0.19
0.05

0.37
0.12
0.04
0.47

0.98
0.00
0.01

0.58
0.41
0.01

0.60

0.24
0.01
0.00
0.74

0.74
3.9
37.0

0.65
0.19
0.16
2,479

71

0.57

0.62
0.16
0.16
0.05

0.37
0.13
0.03
0.46

0.96
0.02
0.02

0.54
0.44
0.02

0.73

0.46
0.03
0.01
0.51

0.72
6.8
38.4

0.52
0.36
0.12
1,605

0.57

0.59
0.13
0.23
0.05

0.38
0.12
0.04
0.47

0.97
0.01
0.02

0.55
0.45
0.01

0.71

0.40
0.04
0.01
0.55

0.64
7.8
38.5

0.43
0.34
0.24
1,475

Notes:
This table contains pediatric and young adult living donor kidney transplant recipients who had no previous kidney
transplants, and who were transplanted between 1998 and 2013. Public insurance is either Medicare, Medicaid, CHIP,
Department of VA, or “other government”. PRA score represents what percentage of the potential donor pool a
candidate has antibodies against that would cause them to reject an organ. High PRA scores are typically a result of
previous transplants or pregnancies. “Other Relative” is a child, identical twin, sibling, spouse, or “other relative”.
“Non-Relative” is any unrelated donor.

Table A.6: Estimates of Share 35 Eﬀect on Deceased Donor Stayers

Wait Time Before Transplant

< 3 Months < 6 Months < 12 Months < 18 Months

Share*Pediatric

[0.10]
0.06***
(0.02)

[0.22]
0.15***
(0.03)

[0.44]
0.20***
(0.03)

[0.61]
0.22***
(0.03)

Quality

[27.8]
-8.4***
(1.4)

Observations

8,634

8,634

8,634

8,634

8,634

Robust Standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

Notes:
This table contains the estimated coeﬃcients of equation 1.1 with yi = KDPIi or yi = 1 if recipient i was transplanted
within 3, 6, 12, or 18 months where each observation is weighted using the weights described in Section 1.5.4 using
only deceased donor recipients in my sample of pediatric and young adult kidney transplant recipients. All models
include indicators for years, age, and transplant center. Standard errors of estimates, listed in parentheses, are robust
to clustering within DSA over time. Standard errors are bootstrapped over both steps of estimation. Pre Share 35
pediatric deceased donor sample mean is listed in brackets.

72

Table A.7: Estimates of Switching Eﬀect on Wait Time and Quality

Wait Time Before Transplant

< 3 Months < 6 Months < 12 Months < 18 Months

ˆβSwitcher

[0.21]
0.03
(0.08)

[0.49]
0.15
(0.11)

[0.74]
0.10
(0.08)

[0.87]
0.11
(0.09)

Quality

[2.7]
11.8***
(4.0)

Observations

17,336

17,336

17,336

17,336

17,336

Bootstrapped Standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

Notes:
This table contains the estimates of βSwitcher derived using the method described in Section 1.5.5. Pre Share 35
pediatric living donor recipient sample mean is listed in brackets

73

Table A.8: Characteristics of Deceased Donor Recipients at Their Second Transplant

Pediatric at First Transplant

Young Adult at First Transplant
First Transplant First Transplant First Transplant First Transplant
Before Share 35 After Share 35 Before Share 35 After Share 35

Age

Received EPTS

Priority

Used a

Living Donor

White=1
Black=1
Hispanic=1
Other Race=1

KDPI of First
Transplant
Observations

19.46

0.28

0.19

0.43
0.30
0.22
0.05

30.60

758

17.38

0.60

0.20

0.39
0.26
0.29
0.06

19.95

514

28.91

0.12

0.18

0.41
0.34
0.18
0.07

36.83

309

26.87

0.52

0.10

0.40
0.39
0.18
0.03

41.71

105

Note:
This table includes the characteristics of recipients at their second transplant if they were pediatric or a young adult
when they received their ﬁrst deceased donor kidney transplant between 1998 and 2013.

74

Table A.9: Estimates of Coeﬃcients for Living Donor Use at Second Transplant

ShareFirst

PediatricFirst

ShareFirst*PediatricFirst

Observations
Censored Observations
Uncensored Observations

OLS

Heckman

-0.15***
(0.05)
0.05
(0.04)
0.07
(0.04)

1,690

-0.22
(0.26)
0.03
(0.03)
0.04
(0.05)

6,573
1,562

Robust standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

Notes:
This table contains the estimated coeﬃcients of equation 1.7 using the second transplant observations of pediatric and
young adult ﬁrst transplant deceased donor kidney recipients in my sample. Both models include indicators for year of
ﬁrst transplant and second transplant DSA. Standard errors of estimates, listed in parentheses, are robust to clustering
within DSA over time.

75

Table A.10: Robustness: Coeﬃcient on Share ∗ Pediatric

All Pediatric Recipients (Section 1.5.1):

Quality
Wait Time

< 3 Months
< 6 Months
< 12 Months
< 18 Months

Original

Exclude

Age 17 & 18

Exclude

2005 & 2006

-1.7*

-1.9**

0.04**
0.09***
0.10***
0.11***

0.04**
0.09***
0.09***
0.11***

-1.5*

0.05***
0.11***
0.12***
0.13***

Crowd out in Living Donors (Section 1.5.2):

-0.17***

-0.16***

-0.17***

Stayers (Section 1.5.4):

Quality
Wait Time

< 3 Months
< 6 Months
< 12 Months
< 18 Months

-8.4***

-8.7***

0.06***
0.15***
0.20***
0.22***

0.05**
0.14***
0.18***
0.21***

-8.8***

0.07***
0.17***
0.22***
0.24***

*** p<0.01, ** p<0.05, * p<0.1

Note:
This table replicates existing estimations listed in the table by excluding candidates who are age 17 or 18 in column 2
or by excluding candidates who were transplanted within one year on either side of Share 35 in column 3.

76

Figure A.1: Deceased Donor Kidney Transplants by Donor Age

Notes:
Author’s calculations from SRTR data. Candidates are considered pediatric if they are less than 18 years old at the start
of their wait time. Candidates are considered young adult if they are 18 to 21 years old at the start of their wait time.

Figure A.2: Kidney Transplants by Donor Type

Notes:
Author’s calculations from SRTR data. Candidates are considered pediatric if they are less than 18 years old at the start
of their wait time. Candidates are considered young adult if they are 18 to 21 years old at the start of their wait time.

77

0200400600800Number of Transplants19952000200520102015Year of transplantPediatric0200400600800Number of Transplants19952000200520102015Year of transplantYoung AdultDonor Younger Than 35Donor Older Than 350200400600800Number of Transplants19952000200520102015Year of transplantPediatric0200400600800Number of Transplants19952000200520102015Year of transplantYoung AdultDeceased DonorLiving DonorFigure A.3: Transplant Wait Time Less than a Year Among All Donor Types

Notes:
Author’s calculations from SRTR data. Candidates are considered pediatric if they are less than 18 years old at the start
of their wait time. Candidates are considered young adult if they are 18 to 21 years old at the start of their wait time.

Figure A.4: Average KDPI Among All Donor Types

Notes:
Author’s calculations from SRTR data. Candidates are considered pediatric if they are less than 18 years old at the start
of their wait time. Candidates are considered young adult if they are 18 to 21 years old at the start of their wait time.

78

.4.45.5.55.6.65% Transplanted in Less Than a Year19952000200520102015Year of transplantPediatricYoung Adult121416182022Average KDPI19952000200520102015Year of transplantPediatricYoung AdultFigure A.5: Event Study Estimates: Wait Time Among All Donor Types

Notes:
Author’s calculations from SRTR data. This ﬁgure plots the estimated βτ’s from equation 1.2 where yi = 1 if recipient
i was transplanted within 3, 6, 12, or 18 months using the entire sample of pediatric and young adult deceased and
living donor recipients. β2003 is omitted and therefore equal to zero.

79

-.15-.1-.050.05.1Probability Transplanted within 3 Months1998199920002001200220032004200520062007200820092010201120122013Year of transplant-.10.1.2Probability Transplanted within 6 Months1998199920002001200220032004200520062007200820092010201120122013Year of transplant-.2-.10.1.2.3Probability Transplanted within 12 Months1998199920002001200220032004200520062007200820092010201120122013Year of transplant-.2-.10.1.2.3Probability Transplanted within 18 Months1998199920002001200220032004200520062007200820092010201120122013Year of transplantFigure A.6: Event Study Estimates: Average KDPI Among All Donor Types

Notes:
Author’s calculations from SRTR data. This ﬁgure plots the estimated βτ’s from equation 1.2 where yi = KDPIi using
the entire sample of pediatric and young adult deceased and living donor recipients. β2003 is omitted and therefore
equal to zero.

80

-10-50510Average KDPI1998199920002001200220032004200520062007200820092010201120122013Year of transplantFigure A.7: Percent of First Transplants That Use a Living Donor

Notes:
Author’s calculations from SRTR data. The percent of transplants in this ﬁgure is calculated as the total number
of living donor transplants divided by the total number of deceased and living donor transplants in a given year.
Candidates are considered pediatric if they are less than 18 years old at the start of their wait time. Candidates are
considered young adult if they are 18 to 21 years old at the start of their wait time.

81

3040506070% of Transplants19952000200520102015Year of TransplantPediatricYoung AdultFigure A.8: Event Study Estimates: First Transplant Uses Living Donor

Notes:
Author’s calculations from SRTR data. This ﬁgure plots the estimated βτ’s from equation 1.2 where yi = LivingDonori
using the entire sample of pediatric and young adult deceased and living donor recipients. β2003 is omitted and therefore
equal to zero.

Figure A.9: Pediatric Kidney Transplants

Notes:
Author’s calculations from SRTR data. Candidates are considered pediatric if they are less than 18 years old at the
start of their wait time.

82

-.4-.20.2.4% of Transplants1998199920002001200220032004200520062007200820092010201120122013Year of transplant2004006008001000Number of Transplants19952000200520102015Year of transplantDeceased DonorLiving DonorEither Donor TypeFigure A.10: Percent of Pediatric Candidates who Receive a Transplant

Notes:
Author’s calculations from SRTR data. The percent of candidates who receive a transplant in this ﬁgure is calculated
as conditional on starting waiting in a given year, the total number of candidates who eventually receive a transplant
divided by the total number of wait list registrations and living donor transplants from candidates who did not register.
Candidates are considered pediatric if they are less than 18 years old at the start of their wait time.

Figure A.11: Treatment Choice by Year

Notes:
Author’s calculations from the United States Renal Data System, 2017 Annual Data Report: Epidemiology of Kidney
Disease in the United States. National Institutes of Health, National Institute of Diabetes and Digestive and Kidney
Diseases, Bethesda, MD, 2017. This ﬁgure shows the treatment choice of all current ESRD patients who are 0 to 17
years old. Transplant is counted as a “treatment” for ESRD so the total number of children with ESRD contains both
children who are using dialysis and those who have a functioning kidney transplant.

83

92949698100% Transplanted1995200020052010Year wait time started100020003000400050006000Number of Patients19952000200520102015YearTotal Number of Pediatric Patients with ESRD TransplantDialysisFigure A.12: Wait Time Less than a Year Among Deceased Donor Recipients

Notes:
Author’s calculations from SRTR data. Candidates are considered pediatric if they are less than 18 years old at the start
of their wait time. Candidates are considered young adult if they are 18 to 21 years old at the start of their wait time.

Figure A.13: Average KDPI Among Deceased Donor Recipients

Notes:
Author’s calculations from SRTR data. Candidates are considered pediatric if they are less than 18 years old at the start
of their wait time. Candidates are considered young adult if they are 18 to 21 years old at the start of their wait time.

84

.1.2.3.4.5.6% Transplanted in Less Than a Year19952000200520102015Year of transplantPediatricYoung Adult152025303540Average KDPI19952000200520102015Year of transplantPediatricYoung AdultFigure A.14: Event Study Estimates: Wait Time Among Deceased Donor Stayers

Notes:
Author’s calculations from SRTR data. This ﬁgure plots the estimated βτ’s from equation 1.2 where yi = 1 if recipient
i was transplanted within 3, 6, 12, or 18 months using the sample of pediatric and young adult deceased recipients,
weighted as described in Seciton 1.5.4. β2003 is omitted and therefore equal to zero.

85

-.2-.10.1.2Probability Transplanted within 3 Months1998199920002001200220032004200520062007200820092010201120122013Year of transplant-.2-.10.1.2.3Probability Transplanted within 6 Months1998199920002001200220032004200520062007200820092010201120122013Year of transplant-.4-.20.2.4Probability Transplanted within 12 Months1998199920002001200220032004200520062007200820092010201120122013Year of transplant-.20.2.4Probability Transplanted within 18 Months1998199920002001200220032004200520062007200820092010201120122013Year of transplantFigure A.15: Event Study Estimates: Average KDPI Among Deceased Donor Stayers

Notes:
Author’s calculations from SRTR data. This ﬁgure plots the estimated βτ’s from equation 1.2 where yi = KDPIi
using the sample of pediatric and young adult deceased recipients, weighted as described in Seciton 1.5.4. β2003 is
omitted and therefore equal to zero.

86

-20-1001020Average KDPI1998199920002001200220032004200520062007200820092010201120122013Year of transplantFigure A.16: Percent of Second Transplants That Use a Living Donor That Used a Deceased Donor for the

First Transplant

Notes:
Author’s calculations from SRTR data through March 2019. The percent of transplants in this ﬁgure is calculated as
the total number of living donor second transplants divided by the total number of deceased and living donor second
transplants in a given year. Candidates are considered pediatric if they were less than 18 years old at the start of their
wait time for their ﬁrst transplant. Candidates are considered young adult if they were 18 to 21 years old at the start of
their wait time for their ﬁrst transplant.

87

10203040% of Transplants19952000200520102015Year of First TransplantPediatricYoung AdultFigure A.17: Number of Wait List Additions by Age

Notes:
Author’s calculations from SRTR data. Age is determined at the date of the candidate’s wait list registration.

88

50100150200Number of Registrations19952000200520102015Year of registration17 Years Old18 Years OldAPPENDIX B

TABLES AND FIGURES FOR CHAPTER 2

Table B.1: State Decisions to Expand Medicaid or Include Transplants as an EHB

State
Alabama
Alaska
Arizona
Arkansas
California
Colorado
Connecticut
Delaware
District of Columbia
Florida
Georgia
Hawaii
Idaho
Illinois
Indiana
Iowa
Kansas
Kentucky
Louisiana
Maine
Maryland
Massachusetts
Michigan
Minnesota
Mississippi
Missouri
Montana
Nebraska
Nevada
New Hampshire
New Jersey
New Mexico
New York
North Carolina
North Dakota
Ohio

EHB in 2017 Medicaid Expansion Date

Yes
Yes

Yes

Yes

Yes

Yes

89

9/1/2015
1/1/2014
1/1/2014
1/1/2014
1/1/2014
1/1/2014
1/1/2014
1/1/2014

1/1/2014

TBD

1/1/2014
2/1/2015
1/1/2014

1/1/2014
7/1/2016
1/10/2019
1/1/2014
1/1/2014
4/1/2014
1/1/2014

1/1/2016

TBD

1/1/2014
8/8/2014
1/1/2014
1/1/2014
1/1/2014

1/1/2014
1/1/2014

Table B.1 (Cont’d)

Yes

Yes

1/1/2014
1/1/2015
1/1/2014

TBD

1/1/2014
1/1/2019
1/1/2014
1/1/2014

Oklahoma
Oregon
Pennsylvania
Rhode Island
South Dakota
South Carolina
Tennessee
Texas
Utah
Vermont
Virginia
Washington
West Virginia
Wisconsin
Wyoming

Notes:
State that did not choose to include transplants as an EHB in 2017 already chose to include them in 2014.
States without a Medicaid Expansion date did not choose to expand Medicaid. States with “TBD” as their
Medicaid Expansion date have not yet determined when their Medicaid expansion will take eﬀect.
Sources:
Medicaid Expansion Dates:
-activity-around-expanding-medicaid-under-the-affordable-care-act/
Essential Health Beneﬁts: https://www.cms.gov/cciio/resources/data-resources/ehb.html

https://www.kff.org/health-reform/state-indicator/state

Table B.2: Transplant Candidate Characteristics

Kidney

Heart

Before ACA After ACA Before ACA After ACA

Age < 18
18 ≤ Age ≤ 64
65 ≤ Age

White
Black
Hispanic
Other Race

Medicaid
Private

0.03
0.82
0.16

0.46
0.29
0.17
0.08

0.07
0.43

0.03
0.79
0.18

0.44
0.29
0.18
0.09

0.09
0.43

90

0.15
0.72
0.13

0.65
0.21
0.09
0.04

0.17
0.52

0.14
0.70
0.16

0.61
0.23
0.10
0.05

0.18
0.49

Other Insurance

Previously transplanted

Died within ...

6 months of listing
1 year of listing
2 years of listing
3 years of listing

Received a transplant

within a year of listing

Observations

Age < 18
18 ≤ Age ≤ 64
65 ≤ Age

White
Black
Hispanic
Other Race

Medicaid
Private
Other Insurance

Previously transplanted

Died within ...

6 months of listing
1 year of listing
2 years of listing
3 years of listing

Received a transplant

within a year of listing

Observations

Table B.2 (Cont’d)

0.50

0.14

0.02
0.04
0.07
0.11

0.48

0.13

0.02
0.03
0.07
0.11

0.31

0.05

0.10
0.14
0.18
0.22

0.33

0.04

0.10
0.13
0.19
0.24

0.20

218,031

0.24

192,154

0.56
21,841

0.54
22,694

Lung

Liver

Before ACA After ACA Before ACA After ACA

0.04
0.74
0.22

0.82
0.09
0.07
0.03

0.09
0.56
0.35

0.06

0.12
0.18
0.26
0.34

0.02
0.69
0.29

0.78
0.10
0.09
0.03

0.09
0.48
0.43

0.04

0.11
0.16
0.25
0.33

0.06
0.81
0.12

0.69
0.10
0.15
0.06

0.17
0.57
0.26

0.07

0.10
0.15
0.21
0.25

0.06
0.75
0.19

0.69
0.09
0.16
0.06

0.19
0.51
0.30

0.05

0.10
0.15
0.22
0.27

0.64
14,416

0.71
14,485

0.45
70,003

0.49
62,799

91

Table B.2 (Cont’d)

Note:
Sample includes transplant candidates registered between 2008 and 2018.

92

Table B.3: Monthly Wait List Additions per 1 Million Population

Kidney

Heart

Medicaid Private Medicaid

[0.49]
0.049
(0.081)
0.29***
(0.093)

[3.35]
0.12
(0.23)
-0.23
(0.20)

[0.15]
-0.037
(0.034)
0.028*
(0.016)

Private
[0.44]
0.018
(0.032)
-0.085**
(0.033)

Transplant
is EHB
Post Medicaid
expansion

Observations
R-squared

6,732
0.307

6,732
0.330

6,732
0.090

6,732
0.104

Lung

Liver

Medicaid Private Medicaid

[0.04]
-0.0081
(0.016)
0.024**
(0.0098)

[0.30]
-0.032
(0.031)
-0.023
(0.019)

[0.41]
-0.045
(0.045)
0.17***
(0.052)

Transplant
is EHB
Post Medicaid
expansion

Observations
R-squared

6,732
0.123

6,732
0.055

6,732
0.175
Standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1

Private
[1.43]
-0.063
(0.11)
-0.16**
(0.074)

6,732
0.183

Notes:
Author’s calculations from SRTR data. This table contains estimates of equation (2.1) where yst is number
of Medicaid insured or privately insured wait list additions per 1 million state population. Estimation
sample includes 50 states and the District of Columbia from 2008 to 2018. The unit of observation is
a state-month-year. Standard errors are clustered at the state level. Pre-ACA sample means are listed in
brackets

93

Table B.4: Monthly Deceased Donor Transplants per 1 Million Population

Kidney

Heart

Medicaid

[0.12]
0.0012
(0.024)
0.034
(0.021)

Private Medicaid
[0.71]
-0.088
(0.075)
-0.037
(0.051)

[0.10]
-0.0098
(0.012)
0.0071
(0.012)

Transplant
is EHB
Post Medicaid
expansion

Observations
R-squared

6,732
0.150

6,732
0.139

6,732
0.065

Private
[0.28]
0.011
(0.027)
-0.025
(0.023)

6,732
0.083

Transplant
is EHB
Post Medicaid
expansion

Observations
R-squared

Lung

Liver

Medicaid

[0.03]
-0.0070
(0.010)
0.014**
(0.0059)

Private Medicaid
[0.21]
-0.064**
(0.031)
-0.0092
(0.017)

[0.22]
-0.033
(0.046)
0.091***
(0.028)

Private
[0.77]
-0.059
(0.044)
-0.093***
(0.032)

6,732
0.053

6,732
0.078

6,732
0.107
Standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1

6,732
0.151

Notes:
Author’s calculations from SRTR data. This table contains estimates of equation (2.1) where yst is number of
Medicaid insured or privately insured deceased donor transplants per 1 million state population. Estimation
sample includes 50 states and the District of Columbia from 2008 to 2017. The unit of observation is
a state-month-year. Standard errors are clustered at the state level. Pre-ACA sample means are listed in
brackets

94

Table B.5: Monthly Living Donor Transplants per 1 Million Population

Kidney

Liver

Medicaid

[0.04]
-0.013
(0.011)
0.014
(0.0095)

Private Medicaid
[0.005]
[0.80]
-0.032
0.00020
(0.0026)
(0.038)
0.0067***
-0.13***
(0.040)
(0.0022)

Private
[0.03]
0.0071
(0.011)
0.0064
(0.0084)

Transplant
is EHB
Post Medicaid
expansion

Observations
R-squared

6,732
0.077

6,732
0.197

6,732
0.084

6,732
0.169

Standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1

Notes:
Author’s calculations from SRTR data. This table contains estimates of equation (2.1) where yst is number
of Medicaid insured or privately insured living donor transplants per 1 million state population. Estimation
sample includes 50 states and the District of Columbia from 2008 to 2018. The unit of observation is
a state-month-year. Standard errors are clustered at the state level. Pre-ACA sample means are listed in
brackets

95

Figure B.1: Registrations by Medicaid Expansion Status

Kidney

Heart

Lung

Liver

Note:
This ﬁgure plots the number of registrations per million population, averaged across all states who either expanded
Medicaid or did not.

96

02468Number of Registrations/1 Million Population200820102012201420162018Year of RegistrationState Expanded MedicaidState Did Not Expand MedicaidMedicaid101520253035Number of Registrations/1 Million Population200820102012201420162018Year of RegistrationState Expanded MedicaidState Did Not Expand MedicaidPrivate0.511.52Number of Registrations/1 Million Population200820102012201420162018Year of RegistrationState Expanded MedicaidState Did Not Expand MedicaidMedicaid12345Number of Registrations/1 Million Population200820102012201420162018Year of RegistrationState Expanded MedicaidState Did Not Expand MedicaidPrivate0.2.4.6Number of Registrations/1 Million Population200820102012201420162018Year of RegistrationState Expanded MedicaidState Did Not Expand MedicaidMedicaid0123Number of Registrations/1 Million Population200820102012201420162018Year of RegistrationState Expanded MedicaidState Did Not Expand MedicaidPrivate123456Number of Registrations/1 Million Population200820102012201420162018Year of RegistrationState Expanded MedicaidState Did Not Expand MedicaidMedicaid468101214Number of Registrations/1 Million Population200820102012201420162018Year of RegistrationState Expanded MedicaidState Did Not Expand MedicaidPrivateFigure B.2: Wait List Addition Trends by Organ

Kidney

Heart

Lung

Liver

Notes:
Author’s calculations from SRTR data. This ﬁgure plots the estimated βy’s from equation (2.2) where yst is number
of Medicaid insured or privately insured wait list additions per 1 million state population. Estimation sample includes
50 states and the District of Columbia from 2008 to 2018. The unit of observation is a state-month-year. Standard
errors are clustered at the state level.

97

-.4-.20.2.4.6Number of Registrations/1 Million Population20082009201020112012201320142015201620172018Year of RegistrationMedicaid-1-.50.51Number of Registrations/1 Million Population20082009201020112012201320142015201620172018Year of RegistrationPrivate-.2-.10.1.2Number of Registrations/1 Million Population20082009201020112012201320142015201620172018Year of RegistrationMedicaid-.4-.3-.2-.10.1Number of Registrations/1 Million Population20082009201020112012201320142015201620172018Year of RegistrationPrivate-.050.05.1Number of Registrations/1 Million Population20082009201020112012201320142015201620172018Year of RegistrationMedicaid-.10.1.2Number of Registrations/1 Million Population20082009201020112012201320142015201620172018Year of RegistrationPrivate-.20.2.4Number of Registrations/1 Million Population20082009201020112012201320142015201620172018Year of RegistrationMedicaid-.6-.4-.20.2.4Number of Registrations/1 Million Population20082009201020112012201320142015201620172018Year of RegistrationPrivateFigure B.3: Deceased Donor Transplants by Medicaid Expansion Status

Kidney

Heart

Lung

Liver

Note:
This ﬁgure plots the number of deceased donor transplants per million population, averaged across all states who either
expanded Medicaid or did not.

98

0.511.522.5Number of Transplants/1 Million Population200820102012201420162018Year of TransplantState Expanded MedicaidState Did Not Expand MedicaidMedicaid234567Number of Transplants/1 Million Population200820102012201420162018Year of TransplantState Expanded MedicaidState Did Not Expand MedicaidPrivate0.511.5Number of Transplants/1 Million Population200820102012201420162018Year of TransplantState Expanded MedicaidState Did Not Expand MedicaidMedicaid0123Number of Transplants/1 Million Population200820102012201420162018Year of TransplantState Expanded MedicaidState Did Not Expand MedicaidPrivate0.1.2.3.4.5Number of Transplants/1 Million Population200820102012201420162018Year of TransplantState Expanded MedicaidState Did Not Expand MedicaidMedicaid0.511.522.5Number of Transplants/1 Million Population200820102012201420162018Year of TransplantState Expanded MedicaidState Did Not Expand MedicaidPrivate.511.522.53Number of Transplants/1 Million Population200820102012201420162018Year of TransplantState Expanded MedicaidState Did Not Expand MedicaidMedicaid2468Number of Transplants/1 Million Population200820102012201420162018Year of TransplantState Expanded MedicaidState Did Not Expand MedicaidPrivateFigure B.4: Deceased Donor Transplant Trends by Organ

Kidney

Heart

Lung

Liver

Notes:
Author’s calculations from SRTR data. This ﬁgure plots the estimated βy’s from equation (2.2) where yst is number
of Medicaid insured or privately insured deceased donor transplants per 1 million state population. Estimation sample
includes 50 states and the District of Columbia from 2008 to 2018. The unit of observation is a state-month-year.
Standard errors are clustered at the state level.

99

-.1-.050.05.1.15Number of Transplants/1 Million Population20082009201020112012201320142015201620172018Year of TransplantMedicaid-.4-.20.2Number of Transplants/1 Million Population20082009201020112012201320142015201620172018Year of TransplantPrivate-.1-.050.05.1.15Number of Transplants/1 Million Population20082009201020112012201320142015201620172018Year of TransplantMedicaid-.2-.10.1.2Number of Transplants/1 Million Population20082009201020112012201320142015201620172018Year of TransplantPrivate-.050.05Number of Transplants/1 Million Population20082009201020112012201320142015201620172018Year of TransplantMedicaid-.15-.1-.050.05.1Number of Transplants/1 Million Population20082009201020112012201320142015201620172018Year of TransplantPrivate-.2-.10.1.2.3Number of Transplants/1 Million Population20082009201020112012201320142015201620172018Year of TransplantMedicaid-.4-.20.2Number of Transplants/1 Million Population20082009201020112012201320142015201620172018Year of TransplantPrivateFigure B.5: Living Donor Transplant Trends by Organ

Kidney

Liver

Notes:
Author’s calculations from SRTR data. This ﬁgure plots the estimated βy’s from equation (2.2) where yst is number
of Medicaid insured or privately insured living donor transplants per 1 million state population. Estimation sample
includes 50 states and the District of Columbia from 2008 to 2018. The unit of observation is a state-month-year.
Standard errors are clustered at the state level.

100

-.050.05.1Number of Transplants/1 Million Population20082009201020112012201320142015201620172018Year of TransplantMedicaid-.20.2.4Number of Transplants/1 Million Population20082009201020112012201320142015201620172018Year of TransplantPrivate-.020.02.04.06.08Number of Transplants/1 Million Population20082009201020112012201320142015201620172018Year of TransplantMedicaid-.2-.10.1.2Number of Transplants/1 Million Population20082009201020112012201320142015201620172018Year of TransplantPrivateFigure B.6: Living Donor Transplants by Medicaid Expansion Status

Kidney

Liver

Note:
This ﬁgure plots the number of living donor transplants per million population, averaged across all states who either
expanded Medicaid or did not.

101

0.2.4.6.8Number of Transplants/1 Million Population200820102012201420162018Year of TransplantState Expanded MedicaidState Did Not Expand MedicaidMedicaid246810Number of Transplants/1 Million Population200820102012201420162018Year of TransplantState Expanded MedicaidState Did Not Expand MedicaidPrivate0.05.1.15.2.25Number of Transplants/1 Million Population200820102012201420162018Year of TransplantState Expanded MedicaidState Did Not Expand MedicaidMedicaid0.2.4.6.81Number of Transplants/1 Million Population200820102012201420162018Year of TransplantState Expanded MedicaidState Did Not Expand MedicaidPrivateAPPENDIX C

TABLES AND FIGURES FOR CHAPTER 3

Table C.1: Average Number of Organs Transplanted per Donor, by Mechanism of Death

Mechanism
Gunshot
Blunt Injury
Stab
Asphyxiation
Seizure
Drowning
Electrical
Drug Intoxication
Other
Natural Causes
Stroke
SIDS
Cardiovascular

Kidney Liver Heart Lung Pancreas
0.354
1.775
0.256
1.738
1.735
0.224
0.198
1.698
0.150
1.575
0.128
1.767
1.756
0.144
0.102
1.529
0.137
1.500
1.418
0.079
0.078
1.282
0.118
0.704
1.286
0.055

0.582 0.713
0.465 0.388
0.388 0.399
0.376 0.334
0.356 0.415
0.415 0.117
0.356 0.211
0.360 0.356
0.344 0.288
0.236
0.255
0.182 0.315
0.734 0.036
0.155 0.168

0.892
0.812
0.769
0.758
0.714
0.756
0.711
0.784
0.713
0.668
0.751
0.604
0.664

Intestine Total
4.340
0.023
3.687
0.029
0.004
3.519
3.394
0.029
3.247
0.036
3.231
0.058
0.022
3.200
3.138
0.006
3.022
0.040
0.012
2.668
2.616
0.008
2.408
0.213
0.009
2.336

Note:
Authors’ calculations using 2000-2018 SRTR data.

102

Table C.2: Average Number of Organs Transplanted per Donor, by Mechanism of Death

Drug Deaths Measured as:

Opioid Deaths Opioid Overdoses Drug Overdoses Sample Mean
(per million)

Donor Mechanism
Drug Intoxication

(1)
0.019
(0.002)

(2)
0.019
(0.002)

(3)
0.021
(0.003)

1.365

24.141

All Others

-0.017
(0.007)

-0.018
(0.007)

-0.010
(0.005)

Notes:
All estimation samples consist of 57 DSAs from 1999 to 2017. The unit of observation is a DSA-year. All models
include indicators for years and DSAs. Standard errors, listed in parentheses, are robust to clustering with DSA over
time. All variables are measured per million DSA residents.

103

Table C.3: Organ Donors by Year, Gender, and Mechanism

Mechanism:

Mechanism:
All Others

Drug Intoxication
All Male Female All Male Female
2472
66
2468
84
2449
107
138
2603
2945
188
3092
158
230
3144
3092
268
3125
285
3141
322
342
3098
3129
473
3115
441
560
3113
3184
625
3249
848
3520
1262
1384
3504
3660
1401

3451
3531
3640
3721
4019
4345
4649
4734
4583
4560
4506
4528
4592
4601
4793
4987
5197
5403
5662

5924
5999
6089
6324
6964
7437
7793
7826
7708
7701
7604
7657
7707
7714
7977
8236
8717
8907
9322

35
47
57
75
91
79
135
148
153
168
179
238
231
309
374
501
763
798
834

31
37
50
63
97
79
95
120
132
154
163
235
210
251
251
347
499
586
567

2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018

Note:
Authors’ calculations using 2000-2018 SRTR data.

104

Table C.4: Estimates of the Eﬀect of Opioid Overdose Deaths on Organ Donations Due to Drug

Intoxication, by Gender and Age

Males

Females

Pooled

DI Organ Donors DI Organ Donors DI Organ Donors

(1)

(2)

(3)

Overall

Age Categories:

<18

18-34

35-49

50-64

65+

0.0097
(0.0014)
[0.7679]

0.0002
(0.0001)
[0.0253]

0.0050
(0.0009)
[0.4867]

0.0038
(0.0006)
[0.2089]

0.0006
(0.0002)
[0.0459]

0.0001
(0.0000)
[0.0010]

0.0091
(0.0009)
[0.5974]

0.0001
(0.0001)
[0.0211]

0.0051
(0.0005)
[0.3063]

0.0031
(0.0005)
[0.1993]

0.0008
(0.0001)
[0.0686]

0.0000
(0.0001)
[0.0021]

0.0188
(0.0021)
[1.3652]

0.0002
(0.0001)
[0.0464]

0.0102
(0.0012)
[0.7930]

0.0069
(0.0010)
[0.4082]

0.0014
(0.0002)
[0.1145]

0.0001
(0.0001)
[0.0031]

Notes:
All estimation samples consist of 57 DSAs from 1999 to 2017. The unit of observation is a DSA-year. All models
include indicators for years and DSAs. Standard errors, listed in parentheses, are robust to clustering with DSA over
time. Sample means for relevant dependent variables are listed in brackets, with all variables measured per million
DSA residents.

105

Table C.5: Estimates of the Eﬀect of Opioid Overdose Deaths on Organ Transplants, by Organ

DI Organ DI Organ Non-DI Organ Non-DI Organ
Donors
Transplants

Transplants

(1)
0.019
(0.002)
[1.365]

Overall

By Organ

Kidney

Liver

Heart

Lung

Pancreas

Intestine

(2)
0.053
(0.007)
[4.001]

0.029
(0.005)
[2.044]

0.017
(0.002)
[1.066]

0.006
(0.001)
[0.471]

0.005
(0.001)
[0.269]

0.001
(0.001)
[0.142]

0.000
(0.000)
[0.009]

Donors

(3)

(4)

-0.017
(0.007)
[24.141]

-0.045
(0.020)
[70.281]

-0.021
(0.009)
[34.935]

-0.002
(0.004)
[18.527]

-0.006
(0.002)
[7.508]

-0.003
(0.003)
[4.991]

-0.001
(0.002)
[3.888]

0.000
(0.001)
[0.431]

Notes:
All estimation samples consist of 57 DSAs from 1999 to 2017. The unit of observation is a DSA-year. All models
include indicators for years and DSAs. Standard errors, listed in parentheses, are robust to clustering with DSA over
time. Sample means for relevant dependent variables are listed in brackets, with all variables measured per million
DSA residents.

106

Table C.6: Estimates of the Eﬀect of Opioid Overdose Deaths on Waiting List Additions by In- Versus

Out-of-Area

Overall

By Organ

Kidney

Liver

Heart

Lung

Pancreas

All Additions

(1)
0.154
(0.067)
[165.881]

0.063
(0.049)
[100.242]

0.076
(0.027)
[35.237]

0.000
(0.006)
[10.844]

0.012
(0.006)
[6.850]

-0.003
(0.004)
[1.999]

In-DSA Out-of-DSA

(2)
0.089
(0.047)
[128.980]

(3)
0.065
(0.032)
[36.901]

0.029
(0.037)
[81.173]

0.054
(0.017)
[26.029]

0.000
(0.005)
[8.390]

0.008
(0.003)
[4.249]

-0.002
(0.002)
[1.362]

0.034
(0.024)
[19.069]

0.021
(0.012)
[9.209]

0.000
(0.003)
[2.455]

0.004
(0.003)
[2.601]

-0.001
(0.002)
[0.637]

Notes:
All estimation samples consist of 57 DSAs from 1999 to 2017. The unit of observation is a DSA-year. All models
include indicators for years and DSAs. Standard errors, listed in parentheses, are robust to clustering with DSA over
time. Sample means for relevant dependent variables are listed in brackets, with all variables measured per million
DSA residents.

107

Table C.7: Estimates of the Eﬀect of Opioid Overdose Deaths on Waiting List Additions by In- Versus

Out-of-Area and by Multilisting Status

No Multilistings

Multilistings

Overall

By Organ

Kidney

Liver

Heart

Lung

Pancreas

All

Additions

(1)
0.108
(0.051)
[122.559]

In-DSA

(2)
0.070
(0.039)
[101.038]

Out-of-
DSA
(3)
0.039
(0.022)
[21.521]

All

Additions

(4)
0.045
(0.025)
[43.322]

In-DSA

(5)
0.019
(0.016)
[27.942]

Out-of-
DSA
(6)
0.026
(0.014)
[15.380]

0.036
(0.036)
[70.710]

0.066
(0.024)
[29.198]

-0.001
(0.005)
[9.695]

0.013
(0.006)
[5.894]

-0.003
(0.003)
[1.248]

0.018
(0.029)
[61.917]

0.049
(0.016)
[22.431]

-0.001
(0.004)
[7.606]

0.008
(0.003)
[3.779]

-0.002
(0.001)
[0.859]

0.018
(0.016)
[8.794]

0.017
(0.010)
[6.767]

0.000
(0.002)
[2.089]

0.005
(0.003)
[2.115]

-0.001
(0.002)
[0.389]

0.027
(0.019)
[29.532]

0.011
(0.012)
[19.256]

0.016
(0.010)
[10.275]

0.010
(0.007)
[6.040]

0.001
(0.001)
[1.149]

0.000
(0.001)
[0.957]

-0.001
(0.001)
[0.751]

0.005
(0.004)
[3.598]

0.001
(0.001)
[0.784]

0.000
(0.000)
[0.471]

-0.001
(0.001)
[0.503]

0.005
(0.004)
[2.442]

0.000
(0.001)
[0.365]

-0.001
(0.001)
[0.486]

0.000
(0.001)
[0.248]

Notes:
All estimation samples consist of 57 DSAs from 1999 to 2017. The unit of observation is a DSA-year. All models
include indicators for years and DSAs. Standard errors, listed in parentheses, are robust to clustering with DSA over
time. Sample means for relevant dependent variables are listed in brackets, with all variables measured per million
DSA residents.

108

Table C.8: Estimates of the Eﬀect of Opioid Overdose Deaths on Living-Donor Transplants

All Organs Kidneys All Except Kidneys

Overall

0.002
(0.009)
[19.385]

0.000
(0.009)
[18.563]

0.002
(0.003)
[0.823]

Notes:
All estimation samples consist of 57 DSAs from 1999 to 2017. The unit of observation is a DSA-year. All models
include indicators for years and DSAs. Standard errors, listed in parentheses, are robust to clustering with DSA over
time. Sample means for relevant dependent variables are listed in brackets, with all variables measured per million
DSA residents.

109

Figure C.1: Donations by Mechanism of Death

Notes:
Authors’ calculations from the SRTR data. “All others” include Gunshot/Stab wound, Asphyxiation, Cardiovascular,
Drowning, Electrical, Natural Causes, None of the above.

110

Figure C.2: Number of Donors by Source

Notes:
Authors’ calculations from the SRTR data. In the SRTR data, Motor Vehicle Accidents are coded as a circumstance of
death and Drug Intoxications are a mechanism of death, but there is almost no overlap in the two in the data.

111

Figure C.3: Organ Donors by Year, Age, and Mechanism

Notes:
Authors’ calculations from the SRTR data. “Other Mechanisms” include Gunshot/Stab wound, Blunt Injury, Stab,
Seizure, Stroke, SIDS, Asphyxiation, Cardiovascular, Drowning, Electrical, Natural Causes, None of the above. Donors
are “young” if they are 19-49 years old.

112

Figure C.4: Opioid Overdoses and Organ Donors by 2016 Quintile of Opioid Overdose Levels

Notes:
Authors’ calculations from the Vital Statistics Mortality Data and SRTR data. The 1st quintile contains the following
DSAs: AROR, CADN, CAGS, CAOP, HIOP, IAOP, MSOP, MWOB, NEOR, TXGC, TXSA, and TXSB. The 2nd
quintile contains the following DSAs: ALOB, AZOB, CASD, CORS, GALL, INOP, LAOP, MNOP, OKOP, ORUO,
and WALC. The 3rd quintile contains the following DSAs: DCTC, FLFH, NCCM, NCNC, NYAP, NYRT, PADV,
SCOP, TNMS, UTOP, VATB, and WIUW. The 4th quintile contains the following DSAs: FLMP, FLUF, FLWC, ILIP,
MIOP, MOMA, NJTO, NMOP, NVLV, NYFL, and TNDS. The 5th quintile contains the following DSAs: KYDA,
MAOB, MDPC, NYWN, OHLB, OHLC, OHLP, OHOV, PATF, CTOP, and WIDN. See appendix Table F.1 for the full
names of the OPOs that oversee the DSAs and the state that their headquarter is located in.

113

Figure C.5: Young Donor Recovery Rates

Notes:
Authors’ calculations from the SRTR data. Young donor drug intoxication rate is calculated using donors 18 to 49 years
old whose mechanism of death was drug intoxication and using 18 to 49-year-old deaths whose death was classiﬁed
as “opioid-related” as previously deﬁned in this paper. Other young donor rate is calculated using donors 18 to 49
years old whose mechanism of death was not drug intoxication and using 18 to 49-year-old deaths whose death was
not classiﬁed as “opioid-related” as previously deﬁned in this paper. Recovery rate is calculated as the total number of
organs recovered that are intended for transplant divided by 8 times the number of deaths.

114

Figure C.6: Young Donor Discard Rates

Notes:
Authors’ calculations from the SRTR data. Young donor drug intoxication rate is calculated using donors 18 to 49
years old whose mechanism of death was drug intoxication. Other young donor rate is calculated using donors 18 to
49 years old whose mechanism of death was not drug intoxication. Discard rate is calculated as the total number of
organs recovered that are intended for transplant, but not ultimately transplanted divided by the total number of organs
recovered that are intended for transplant.

115

APPENDIX D

APPENDICES FOR CHAPTER 1

Table D.1: Wait Time Diﬀerences

Deceased Donor Registered Living Donor

Average Wait Time
(test date)

Average Wait Time
(registration date)

Average ∆ Wait Time

Observations

464
(435)

347
(392)

117
(299)

6,645

365
(384)

264
(334)

101
(253)

2,428

Notes:
This table contains pediatric kidney recipients who registered on the waiting list and received a deceased or living
donor kidney between 1995 and 2013.
Wait Time (test date)= Transplant date - HLA test date
Wait Time (Registration date)= Transplant date - Registration date
∆ Wait Time = Wait Time (test date) - Wait Time (Registration date)

116

Table D.2: Estimates of Share 35 Eﬀect on Deceased Donor Stayers

Wait Time Before Transplant

< 3 Months
Y
N

< 6 Months
Y
N

< 12 Months
Y
N

< 18 Months
Y
N

Quality

Y

N

0.06***
(0.02)

0.04**
(0.02)

0.15***
(0.03)

0.12***
(0.03)

0.20***
(0.03)

0.16***
(0.04)

0.22*** 0.21*** -8.4*** -8.9***
(0.03)
(1.2)

(0.03)

(1.4)

Weighted?
Share*Pediatric

Observations

8,634

8,602

8,602

8,634

8,602

8,634

8,540

8,602

8,634
Robust Standard errors in parentheses

8,634

*** p<0.01, ** p<0.05, * p<0.1

Notes:
This table contains the estimated coeﬃcients of equation 1.1 with yi = KDPIi or yi = 1 if recipient i was transplanted within 3, 6, 12, or 18 months where each
observation is weighted using the weights described in Section 1.5.4 or where observations have not been re-weighted, using only deceased donor recipients in my
sample of pediatric and young adult kidney transplant recipients. All models include indicators for years, age, and transplant center. Standard errors of estimates,
listed in parentheses, are robust to clustering within DSA over time. Standard errors for weighted estimates are bootstrapped over both steps of estimation.

117

Figure D.1: Hypothetical Wait List Change for a Kidney from a Donor Younger than 35

Wait List Before Share 35
1: Adult
2: Adult
3: Adult
4: Adult
5: Adult
6: Adult
7: Adult
8: Pediatric *
9: Adult
10: Young Adult *
11: Adult
12: Adult
13: Pediatric *
14: Adult
15: Adult

Wait List After Share 35
1: Pediatric up 7
2: Pediatric up 11
3: Adult
4: Adult
5: Adult
6: Adult
7: Adult
8: Adult
9: Adult
10: Adult
11: Young Adult down 1
12: Adult
13: Adult
14: Adult
15: Adult

Note:
25,814 Adult, 456 Young adult, and 851 Pediatric candidates were added to the wait list in 2004

118

Figure D.2: Number of Deceased Donors Who Are Younger Than 35

Note:
Author’s calculations from SRTR data.

Figure D.3: Density of Propensity Scores Before and After Matching

Note:
Author’s calculations from SRTR data.

119

2500300035004000Donors Younger than 3519952000200520102015Year01230.510.51RawMatched (ATT)UntreatedTreatedDensityPropensity scoreAPPENDIX E

APPENDICES FOR CHAPTER 2

Table E.1: State Medicaid Childless Individual Income Eligibility Levels as a Percent of the Federal

Poverty Level

State
Alabama
Alaska
Arizona
Arkansas
California
Colorado
Connecticut
Delaware
District of Columbia
Florida
Georgia
Hawaii
Idaho
Illinois
Indiana
Iowa
Kansas
Kentucky
Louisiana
Maine
Maryland
Massachusetts
Michigan
Minnesota
Mississippi
Missouri
Montana
Nebraska
Nevada
New Hampshire
New Jersey
New Mexico
New York
North Carolina
North Dakota

2011
0.00
0.00
1.10
0.00
0.00
0.00
0.73
1.10
2.11
0.00
0.00
1.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
1.00
0.00
0.00

2012
0.00
0.00
1.10
0.00
0.00
0.00
0.72
1.10
2.11
0.00
0.00
1.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.75
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
1.00
0.00
0.00

2013
0.00
0.00
1.00
0.00
0.00
0.20
0.70
1.10
2.11
0.00
0.00
1.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.75
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
1.00
0.00
0.00

2014
0.00
0.00
1.38
1.38
1.38
1.38
1.38
1.38
2.15
0.00
0.00
1.38
0.00
1.38
0.00
1.38
0.00
1.38
0.00
0.00
1.38
1.38
1.38
2.00
0.00
0.00
0.00
0.00
1.38
0.00
1.38
1.38
1.38
0.00
1.38

2015
0.00
0.00
1.38
1.38
1.38
1.38
1.38
1.38
2.15
0.00
0.00
1.38
0.00
1.38
0.00
1.38
0.00
1.38
0.00
0.00
1.38
1.38
1.38
1.38
0.00
0.00
0.00
0.00
1.38
1.38
1.38
1.38
1.38
0.00
1.38

120

2016
0.00
1.38
1.38
1.38
1.38
1.38
1.38
1.38
2.15
0.00
0.00
1.38
0.00
1.38
1.39
1.38
0.00
1.38
0.00
0.00
1.38
1.38
1.38
1.38
0.00
0.00
1.38
0.00
1.38
1.38
1.38
1.38
1.38
0.00
1.38

2017
0.00
1.38
1.38
1.38
1.38
1.38
1.38
1.38
2.15
0.00
0.00
1.38
0.00
1.38
1.39
1.38
0.00
1.38
1.38
0.00
1.38
1.38
1.38
1.38
0.00
0.00
1.38
0.00
1.38
1.38
1.38
1.38
1.38
0.00
1.38

2018
0.00
1.38
1.38
1.38
1.38
1.38
1.38
1.38
2.15
0.00
0.00
1.38
0.00
1.38
1.39
1.38
0.00
1.38
1.38
0.00
1.38
1.38
1.38
1.38
0.00
0.00
1.38
0.00
1.38
1.38
1.38
1.38
1.38
0.00
1.38

2019
0.00
1.38
1.38
1.38
1.38
1.38
1.38
1.38
2.15
0.00
0.00
1.38
0.00
1.38
1.39
1.38
0.00
1.38
1.38
1.38
1.38
1.38
1.38
1.38
0.00
0.00
1.38
0.00
1.38
1.38
1.38
1.38
1.38
0.00
1.38

Ohio
Oklahoma
Oregon
Pennsylvania
Rhode Island
South Carolina
South Dakota
Tennessee
Texas
Utah
Vermont
Virginia
Washington
West Virginia
Wisconsin
Wyoming

Table E.1 (Cont’d)

0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
1.50
0.00
0.00
0.00
0.00
0.00

0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
1.60
0.00
0.00
0.00
0.00
0.00

1.38
0.00
1.38
0.00
1.38
0.00
0.00
0.00
0.00
0.00
1.38
0.00
1.38
1.38
1.00
0.00

1.38
0.00
1.38
1.38
1.38
0.00
0.00
0.00
0.00
0.00
1.38
0.00
1.38
1.38
1.00
0.00

0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
1.60
0.00
0.00
0.00
0.00
0.00

1.38
0.00
1.38
1.38
1.38
0.00
0.00
0.00
0.00
0.00
1.38
0.00
1.38
1.38
1.00
0.00

1.38
0.00
1.38
1.38
1.38
0.00
0.00
0.00
0.00
0.00
1.38
0.00
1.38
1.38
1.00
0.00

1.38
0.00
1.38
1.38
1.38
0.00
0.00
0.00
0.00
0.00
1.38
0.00
1.38
1.38
1.00
0.00

1.38
0.00
1.38
1.38
1.38
0.00
0.00
0.00
0.00
0.00
1.38
1.38
1.38
1.38
1.00
0.00

Source: https://www.kff.org/medicaid/state-indicator/medicaid-income-eligibility-limits-for
-other-non-disabled-adults/

121

Table E.2: State Medicaid Family of Three Income Eligibility Levels as a Percent of the Federal Poverty Level

State
Alabama
Alaska
Arizona
Arkansas
California
Colorado
Connecticut
Delaware
District of Columbia
Florida
Georgia
Hawaii
Idaho
Illinois
Indiana
Iowa
Kansas
Kentucky
Louisiana
Maine
Maryland
Massachusetts
Michigan
Minnesota
Mississippi
Missouri
Montana
Nebraska
Nevada

2008 2009 2010 2011 2012
0.24
0.26
0.81
0.81
1.06
2.00
0.17
0.18
1.06
1.06
0.66
1.06
1.91
1.91
1.19
1.06
2.07
2.06
0.58
0.56
0.49
0.53
1.00
1.00
0.42
0.39
1.91
1.91
0.24
0.26
0.89
0.82
0.32
0.34
0.59
0.64
0.25
0.20
2.06
2.00
1.16
0.37
1.33
1.33
0.61
0.63
2.15
2.75
0.44
0.32
0.39
0.36
0.55
0.60
0.57
0.59
0.94
0.87

0.24
0.81
1.06
0.17
1.06
1.06
1.91
1.20
2.07
0.59
0.50
1.00
0.39
1.91
0.36
0.83
0.32
0.62
0.25
2.00
1.16
1.33
0.64
2.15
0.44
0.37
0.56
0.58
0.88

0.25
0.85
2.00
0.17
1.06
0.66
1.91
1.21
2.07
0.55
0.52
1.00
0.28
1.85
0.26
0.86
0.34
0.62
0.26
2.06
1.16
1.33
0.66
2.75
0.46
0.26
0.58
0.58
0.91

0.24
0.81
1.06
0.17
1.06
0.66
1.91
1.21
2.07
0.53
0.50
1.00
0.27
1.85
0.25
0.83
0.32
0.62
0.25
2.06
1.16
1.33
0.64
2.15
0.44
0.25
0.56
0.58
0.88

2013
0.23
0.78
1.06
0.16
1.06
1.06
1.91
1.20
2.06
0.56
0.48
1.38
0.37
1.39
0.24
0.80
0.31
0.57
0.24
2.00
1.22
1.33
0.64
2.15
0.29
0.35
0.54
0.58
0.84

2014
0.16
1.28
1.38
1.38
1.38
1.38
2.01
1.38
2.21
0.35
0.39
1.38
0.27
1.38
0.24
1.38
0.38
1.38
0.24
1.05
1.38
1.38
1.38
2.05
0.29
0.24
0.52
0.55
1.38

2015 2016 2017 2018 2019
0.18
0.18
1.46
1.38
1.38
1.38
1.38
1.38
1.38
1.38
1.38
1.38
1.55
2.01
1.38
1.38
2.21
2.21
0.32
0.34
0.35
0.38
1.38
1.38
0.27
0.25
1.38
1.38
1.39
0.24
1.38
1.38
0.38
0.38
1.38
1.38
1.38
0.24
1.05
1.38
1.38
1.38
1.38
1.38
1.38
1.38
1.38
1.38
0.26
0.28
0.23
0.21
1.38
0.51
0.63
0.55
1.38
1.38

0.18
1.39
1.38
1.38
1.38
1.38
1.38
1.38
2.21
0.33
0.36
1.38
0.26
1.38
1.39
1.38
0.38
1.38
1.38
1.05
1.38
1.38
1.38
1.38
0.27
0.22
1.38
0.63
1.38

0.18
1.43
1.38
1.38
1.38
1.38
1.55
1.38
2.21
0.34
0.37
1.38
0.26
1.38
1.39
1.38
0.38
1.38
0.24
1.05
1.38
1.38
1.38
1.38
0.27
0.22
1.38
0.63
1.38

0.18
1.41
1.38
1.38
1.38
1.38
1.55
1.38
2.21
0.33
0.37
1.38
0.26
1.38
1.39
1.38
0.38
1.38
1.38
1.05
1.38
1.38
1.38
1.38
0.27
0.22
1.38
0.63
1.38

122

New Hampshire
New Jersey
New Mexico
New York
North Carolina
North Dakota
Ohio
Oklahoma
Oregon
Pennsylvania
Rhode Island
South Carolina
South Dakota
Tennessee
Texas
Utah
Vermont
Virginia
Washington
West Virginia
Wisconsin
Wyoming

Table E.2 (Cont’d)

0.55
1.33
0.63
1.50
0.52
0.63
0.90
0.50
1.00
0.59
1.91
1.00
0.56
0.80
0.28
0.47
1.91
0.31
0.76
0.35
1.91
0.55

0.51
2.00
0.69
1.50
0.51
0.62
0.90
0.48
1.00
0.36
1.81
0.90
0.54
1.34
0.27
0.68
1.91
0.30
0.77
0.34
2.00
0.54

0.49
2.00
0.67
1.50
0.49
0.59
0.90
0.47
0.40
0.34
1.81
0.89
0.52
1.29
0.26
0.44
1.91
0.29
0.74
0.33
2.00
0.52

0.49
2.00
0.67
1.50
0.49
0.59
0.90
0.53
0.40
0.46
1.81
0.93
0.52
1.27
0.26
0.44
1.91
0.31
0.74
0.33
2.00
0.52

0.49
2.00
0.85
1.50
0.49
0.59
0.90
0.53
0.40
0.46
1.81
0.91
0.52
1.26
0.26
0.44
1.91
0.31
0.73
0.32
2.00
0.51

0.47
2.00
0.85
1.50
0.47
0.57
0.96
0.51
0.39
0.58
1.81
0.89
0.50
1.22
0.25
0.42
1.91
0.30
0.71
0.31
2.00
0.50

0.75
1.38
1.38
1.38
0.45
1.38
1.38
0.48
1.38
0.38
1.38
0.67
0.54
1.11
0.19
0.47
1.38
0.52
1.38
1.38
1.00
0.59

1.38
1.38
1.38
1.38
0.45
1.38
1.38
0.46
1.38
1.38
1.38
0.67
0.53
1.03
0.19
0.46
1.38
0.45
1.38
1.38
1.00
0.58

1.38
1.38
1.38
1.38
0.44
1.38
1.38
0.44
1.38
1.38
1.38
0.67
0.52
1.01
0.18
0.45
1.38
0.39
1.38
1.38
1.00
0.57

1.38
1.38
1.38
1.38
0.44
1.38
1.38
0.44
1.38
1.38
1.38
0.67
0.51
0.99
0.18
0.44
1.38
0.38
1.38
1.38
1.00
0.56

1.38
1.38
1.38
1.38
0.43
1.38
1.38
0.43
1.38
1.38
1.38
0.67
0.50
0.98
0.18
0.60
1.38
0.38
1.38
1.38
1.00
0.55

1.38
1.38
1.38
1.38
0.42
1.38
1.38
0.42
1.38
1.38
1.38
0.67
0.49
0.95
0.17
0.60
1.38
1.38
1.38
1.38
1.00
0.54

Source: https://www.kff.org/medicaid/state-indicator/medicaid-income-eligibility-limits-for-parents/

123

APPENDIX F

APPENDICES FOR CHAPTER 3

Table F.1: Crosswalk of Organ Procurement Organizations and Abbreviations

OPO
Abbreviation OPO
ALOB
AROR
AZOB
CADN
CAGS
CAOP
CASD
CORS
DCTC
FLFH
FLMP
FLUF
FLWC
GALL
HIOP
IAOP
ILIP
INOP
KYDA
LAOP
MAOB
MDPC
MIOP
MNOP
MOMA
MSOP
MWOB
NCCM
NCNC
NEOR
NJTO
NMOP
NVLV
NYAP
NYFL

OPO
State
AL
Legacy of Hope
AR
Arkansas Regional Organ Recovery Agency
AZ
Donor Network of Arizona
CA
Donor Network West
CA
Sierra Donor Services
CA
OneLegacy
CA
Lifesharing - A Donate Life Organization
CO
Donor Alliance
VA
Washington Regional Transplant Community
FL
TransLife
FL
Life Alliance Organ Recovery Agency
FL
LifeQuest Organ Recovery Services
FL
LifeLink of Florida
GA
LifeLink of Georgia
HI
Legacy of Life Hawaii
IA
Iowa Donor Network
IL
Gift of Hope Organ & Tissue Donor Network
IN
Indiana Donor Network
KY
Kentucky Organ Donor Aﬃliates
LA
Louisiana Organ Procurement Agency
MA
New England Organ Bank
MD
The Living Legacy Foundation of Maryland
Gift of Life Michigan
MI
LifeSource Upper Midwest Organ Procurement Organization MN
Mid-America Transplant Services
MO
MS
Mississippi Organ Recovery Agency
KS
Midwest Transplant Network
NC
LifeShare Carolinas
Carolina Donor Services
NC
NE
Live On Nebraska
NJ
New Jersey Organ and Tissue Sharing Network OPO
New Mexico Donor Services
NM
NV
Nevada Donor Network
NY
Center for Donation and Transplant
Finger Lakes Donor Recovery Network
NY

124

NYRT
NYWN
OHLB
OHLC
OHLP
OHOV
OKOP
ORUO
PADV
PATF
PRLL*
SCOP
TNDS
TNMS
TXGC
TXSA
TXSB
UTOP
VATB
WALC
WIUW
CTOP
WIDN

Table F.1 (Cont’d)

LiveOnNY
Upstate New York Transplant Services Inc
Lifebanc
Life Connection of Ohio
Lifeline of Ohio
LifeCenter Organ Donor Network
LifeShare Transplant Donor Services of Oklahoma
Paciﬁc Northwest Transplant Bank
Gift of Life Donor Program
Center for Organ Recovery and Education
LifeLink of Puerto Rico
We Are Sharing Hope SC
Tennessee Donor Services
Mid-South Transplant Foundation
LifeGift Organ Donation Center
Texas Organ Sharing Alliance
Southwest Transplant Alliance
DonorConnect
LifeNet Health
LifeCenter Northwest
UW Health Organ and Tissue Donation
LIfeChoice Donor Services
Versiti Wisconsin, Inc

Note:
We have no mortality data for Puerto Rico so it is not included in our sample.

NY
NY
OH
OH
OH
OH
OK
OR
PA
PA
PR
SC
TN
TN
TX
TX
TX
UT
VA
WA
WI
MA
WI

125

REFERENCES

126

REFERENCES

Abraham, E., Wilson, A., & Goebel, J. (2009). Current kidney allocation rules and their impact on

a pediatric transplant center. American Journal of Transplantation, 9, 404-408.

Agarwal, S., Oak, N., Siddique, J., Harland, R., & Abbo, E. (2009). Changes in pediatric renal
transplantation after implementation of the revised deceased donor kidney allocation policy.
American Journal of Transplantation, 9, 1237-1242.

Amaral, S., Patzer, R. E., Kutner, N., & McClellan, W.

(2012). Racial disparities in access
to pediatric kidney transplantation since share 35. Journal of the American Society of
Nephrology, 23, 1069-1077.

An, W., & Winship, C.

(2017). Causal inference in panel data with application to estimating
race-of-interviewer eﬀects in the general social survey. Sociological Methods and Research,
46(1), 68-102.

Anderson, D. (2015). Direct and indirect eﬀects of policies to increase kidney donations. (unpub-

lished working paper)

Ansell, D., Chan, E., Jr., E. H., & Pallok, K.

(2014, February 21). When the only cure is a
transplant. Health Aﬀairs Blog. Retrieved from https://www.healthaffairs.org/do/
10.1377/hblog20140221.037351/full/

Baicker, K., Allen, H. L., Wright, B. J., Taubman, S. L., & Finkelstein, A. N.

(2018). The
eﬀect of medicaid on dental care of poor adults: Evidence from the oregon health insurance
experiment. Health Services Research, 53(4), 2148-2164.

Becker, G. S., & Elias, J. J. (2007). Introducing incentives in the market for live and cadaveric

organ donations. Journal of Economic Perspectives, 21(3), 3-24.

Bentley, T. S., & Phillips, S. J. (2017, August). 2017 U.S. organ and tissue transplant cost estimates

and discussion.

Bilgel, F. (2012). The impact of presumed consent laws and institutions on deceased organ donation.

European Journal of Health Economics, 13(1), 29-38.

Breathett, K., Allen, L. A., Helmkamp, L., Colborn, K., Daugherty, S. L., Khazanie, P., . . .
Peterson, P. N. (2017). The aﬀordable care act medicaid expansion correlated with increased
heart transplant listings in african-americans but not hispanics or caucasians. JACC: Heart
Failure, 5(2), 136-147.

Buchmueller, T. C., & Monheit, A. C.

(2009). Employer-sponsored health insurance and the
promise of health insurance reform. INQUIRY: The Journal of Health Care Organization,

127

Provision, and Financing, 46(2), 187-202. Retrieved from https://doi.org/10.5034/
inquiryjrnl_46.02.187 (PMID: 19694392) doi: 10.5034/inquiryjrnl\_46.02.187

Byrne, M. M., & Thompson, P. (2001). A positive analysis of ﬁnancial incentives for cadaveric

organ donation. Journal of Health Economics, 20(1), 69-83.

Callison, K., & Levin, A. (2016). Donor registries, ﬁrst-person consent legislation, and the supply

of deceased organ donors. Journal of Health Economics, 49, 70-75.

Cameron, A. M., Massie, A. B., Alexander, C., Stewart, B., Montgomery, R., Benavides, N., . . .
Segev, D. (2013). Social media and organ donor registration: The facebook eﬀect. American
Journal of Transplantation, 13, 2059-2065.

Case, A., & Paxson, C. (2002). Parental behavior and child health. Health Aﬀairs, 21(2), 164-178.

Centers for Disease Control and Prevention. (2018). Understanding the epidemic | drug overdose.

Retrieved from https://www.cdc.gov/drugoverdose/epidemic/index.html

Choi, Y. (2019). The role of kidney allocation policy in addressing kidney shortage. (unpublished

working paper)

Clark, A. M., DesMeules, M., Luo, W., Duncan, A. S., & Wielgosz, A. (2009). Socioeconomic sta-
tus and cardiovascular disease: Risks and implications for care. Nature Reviews Cardiology,
6, 712-722.

Courtemanche, C., Marton, J., Ukert, B., Yelowitz, A., & Zapata, D. (2017). Early impacts of the
aﬀordable care act on health insurance coverage in medicaid expansion and non-expansion
states. Journal of Policy Analysis and Management, 36(1), 178-210.

Crews, D. C., Gutiérrez, O. M., Fedewa, S. A., Luthi, J.-C., Shoham, D., Judd, S. E., . . . McCellan,
W. M. (2014). Low income, community poverty and risk of end stage renal disease. BMC
Nephrology, 15(192).

Dale-Shall, A., Smith, J., McBride, M., Hingorani, S., & McDonald, R. (2009). The relation-
ship of donor source and age on short- and long-term allograft survival in pediatric renal
transplantation. Pediatric Transplantation, 13, 711-718.

Davis, C. S., Carr, D., Southwell, J. K., & Beletsky, L.

(2015). Engaging law enforcement
in overdose reversal initiatives: Authorization and liability for naloxone administration.
American Journal of Public Health, 105(8), 1530-1538.

Dickert-Conlin, S., Elder, T., & Moore, B. (2011). Donorcycles: Motorcycle helmet laws and the

supply of organ donors. The Journal of Law and Economics, 54, 907-935.

Dickert-Conlin, S., Elder, T., & Teltser, K. (forthcoming). Allocating scarce organs: How a change

128

in supply aﬀects transplant waiting lists and transplant recipients. American Economic
Journal: Applied Economics.

Duggan, M., Goda, G. S., & Jackson, E. (2017). The eﬀects of the aﬀordable care act on health
insurance coverage and labor market outcomes. (NBER Working Paper 23607, available at:
https://www.nber.org/papers/w23607)

Fernandez, J. M., Howard, D. H., & Kroese, L. S. (2013). The eﬀect of cadaveric kidney donations
on living kidney donations: An instrumental variables approach. Economic Inquiry, 51(3),
1696-1714.

Finkelstein, A., Taubman, S., Wright, B., Bernstein, M., Gruber, J., Newhouse, J. P., . . . Oregon
Health Study Group (2012). The oregon health insurance experiment: Evidence from the
ﬁrst year. The Quarterly Journal of Economics, 127(3), 1057-1106.

Glueckert, L., Redden, D., Thompson, M., Haque, A., Gray, S., Locke, J., . . . DuBay, D. (2013).
What liver transplant outcomes can be expected in the uninsured who become insured via
the aﬀordable care act? American Journal of Transplantation, 13, 1533-1540.

Goldberg, D., Blumberg, E., McCauley, M., Abt, P., & Levine, M.

Improving organ
utilization to help overcome the tragedies of the opioid epidemic. American Journal of
Transplantation, 16, 2836-2841.

(2016).

Hardy, B., Shah, T., Cicciarelli, J., Lemley, K., Hutchinson, I., & Cho, Y.

(2009). Kidney
transplantation in children and adolescents: An analysis of united network for organ sharing
database. Transplantation Proceedings, 41, 1533-1535.

Harhay, M. N., McKenna, R. M., Boyle, S. M., Ranganna, K., Mizrahi, L. L., Guy, S., . . . Harhay,
M. O. (2018). Association between medicaid expansion under the aﬀordable care act and
preemptive listings for kidney transplantation. Clinical Journal of the American Society of
Nephrology, 13(7), 1069-1078.

Hedegaard, H., Miniño, A., & Warner, M. (2018). Drug overdose deaths in the united states 1999-
2017. Retrieved from https://www.cdc.gov/nchs/products/databriefs/db329
.htm

Howard, D. H. (2007). Producing organ donors. The Journal of Economic Perspectives, 21(3),

25-36.

Howard, D. H.

(2011). Waiting time as a price for deceased donor kidneys. Contemporary

Economic Policy, 29(3), 295-303.

Imbens, G. W., & Wooldridge, J. M. (2009). Recent developments in the econometrics of program

evaluation. Journal of Economic Literature, 47(1), 5-86.

129

Ising, M., Trivedi, J., Marino, K., Matthew, F., & van Berkel, V.

(2018). The impact of the
aﬀordable care act on patients listed for lung transplantation. Journal of Heart and Lung
Transplantation, 37(4), S237-S238.

Izadi, E.

(2016, May 9).

that
Retrieved from

So many people are dying of drug overdoses

they’re easing the donated organ shortage.
https://www.washingtonpost.com/news/to-your-health/wp/2016/05/09/
one-out-of-every-11-organ-donors-last-year-died-of-a-drug-overdose/
?utm_term=.c83f363b62a6

Washington Post.

Jones, D. (2015). The economics of exclusion restrictions in IV models. (NBER Working Paper

21391, available at: http://www.nber.org/papers/w21391)

Kessler, J. B., & Roth, A. E. (2012). Organ allocation policy and the decision to donate. American

Economic Review, 102(5), 2018-2047.

Lacetera, N., Macis, M., & Stith, S. (2014). Removing ﬁnancial barriers to organ and bone marrow
donation: The eﬀect of leave and tax legislation in the U.S. Journal of Health Economics,
33, 43-56.

Laurentine, K. A., & Bramstedt, K. A. (2010). Too poor for transplant: Finance and insurance

issues in transplant ethics. Progress in Transplantation, 20(2), 178-185.

Lemont, B. (2019). The eﬀect of share 35 for kidneys on pediatric transplant candidates. (unpub-

lished working paper)

Li, D., Hawley, Z., & Schnier, K. (2013). Increasing organ donation via changes in the default

choice or allocation rule. Journal of Health Economics, 32, 1117-1129.

Loehrer, A. P., Chang, D. C., Scott, J. W., Hutter, M. M., Patel, V. I., Lee, J. E., & Sommers, B. D.
(2018). Association of the aﬀordable care act medicaid expansion with access to and quality
of care for surgical conditions. JAMA Surgery, 153(3), 1-7.

Mabry, C. D., Gurien, L. A., Smith, S. D., & Mehl, S. C. (2016). Are surgeons being paid fairly by
medicaid? a national comparison of typical payments for general surgeons. Journal of the
American College of Surgeons, 222(4), 387-394.

Massie, Leanza, Fahmy, Chow, Desai, Luo, . . . Segev (2016). A risk index for living donor kidney

transplantation. American Journal of Transplantation, 16, 2077-2084.

MedlinePlus. (2015). Bethesda (MD): National library of medicine (US): End-stage kidney disease.

Retrieved 2017-1-30, from https://medlineplus.gov/ency/article/000500.htm

Mercy Health.

2018-8-8,
https://www.mercyhealth.com/medical-services/kidney-transplants/

Retrieved

(2018).

Living

kidney

donors.

from

130

living-kidney-donors

Miller, S., & Wherry, L. R. (2019). Four years later: Insurance coverage and access to care continue
to diverge between ACA medicaid expansion and non-expansion states. AEA Papers and
Proceedings, 109, 327-333.

Moudgil, A., Dharnidharka, V. R., Lamb, K. E., & Meier-Kriesche, H.-U. (2013). Best allograft
survival from share-35 kidney donors occurs in middle-aged adults and young children–an
analysis of OPTN data. Transplantation, 95(2), 319-325.

National Cancer Institute.

(2015). US population data 1969-2013. Retrieved from http://

seer.cancer.gov/popdata/

NIDDK. (2014). Kidney disease in children. Retrieved 2017-3-31, from https://www.niddk

.nih.gov/health-information/kidney-disease/children

NIDDK. (2018). Kidney transplant. Retrieved 2018-6-4, from https://www.niddk.nih.gov/

health-information/kidney-disease/kidney-failure/kidney-transplant

NKF. (2019). Dialysis. Retrieved 2019-6-5, from https://www.kidney.org/atoz/content/

dialysisinfo

Oliveira, G. H., Al-Kindi, S. G., & Simon, D. I. (2016). Implementation of the aﬀordable care
act and solid-organ transplantation listings in the united states. JAMA Cardiology, 1(6),
737-738.

OPTN. (2017). Allocation of kidneys. Retrieved from https://optn.transplant.hrsa.gov/

governance/policies/

OPTN. (2017). OPTN strategic plan. Retrieved 2017, from https://optn.transplant.hrsa

.gov/governance/strategic-plan/

OPTN. (2018). About CPRA. Retrieved 2018-8-30, from https://optn.transplant.hrsa

.gov/resources/allocation-calculators/about-cpra/

OPTN/UNOS.

(2016). A guide to calculating and interpreting the kidney donor proﬁle in-
dex (KDPI). Retrieved from https://optnpilot.unos.org/media/1512/guide_to
_calculating_interpreting_kdpi.pdf

Pardo, B. (2016). Do more robust prescription drug monitoring programs reduce prescription

opioid overdose? Addiction, 112(10), 1773-1783.

Patzer, R., Amaral, S., Klein, M., Kutner, N., Perryman, J., Gazmararian, J., & McClellan, W.
(2012). Racial disparities in pediatric access to kidney transplantation: Does socioeconomic
status play a role? American Journal of Transplantation, 12, 369-378.

131

Raju, S., Keet, C. A., Paulin, L. M., Matsui, E. C., Peng, R. D., Hansel, N. N., & McCormack,
M. C. (2019). Rural residence and poverty are independent risk factors for chronic obstructive
pulmonary disease in the united states. American Journal of Respiratory and Critical Care
Medicine, 199(8).

Rodrigue, J., Cornell, D., Lin, J., Kaplan, B., & Howard, R.

Increasing live donor
kidney transplantation: A randomized evaluation of a home-based educational intervention.
American Journal of Transplantation, 7, 394-401.

(2007).

Roth, A. E., Sönmez, T., & Ünver, M. U. (2004). Kidney exchange. The Quarterly Journal of

Economics, 119(2), 457-488.

Roth, A. E., Sönmez, T., & Ünver, M. U. (2005). Pairwise kidney exchange. Journal of Economics

Theory, 125(2), 151-188.

Roth, A. E., Sönmez, T., & Ünver, M. U.

(2007). Eﬃcient kidney exchange: Coincidence of
wants in markets with compatibility-based preferences. American Economic Review, 97(3),
828-851.

Rudd, R., Seth, P., David, F., & Scholl, L. (2017). Increases in drug and opioid overdose deaths -

united states, 2010-2015. Morbidity and Mortality Weekly Report, 64(50), 1378-1382.

Ruhm, C. J. (2017). Geographic variation in opioid and heroin involved drug poisoning mortality

rates. American Journal of Preventative Medicine, 53(6), 745-753.

Samoray, C., & Satarino, C.

(2016). Kidney transplant coverage: Before, during, and af-
ter. Retrieved from www.kidneyfund.org/assets/pdf/training/i-need-a-kidney
-transplant-webinar-slides.pdf

Scaglione, S., Kliethermes, S., Cao, G., Shoham, D., Durazo, R., Luke, A., & Volk, M. L. (2015).
The epidemiology of cirrhosis in the united states: A population-based study. Journal of
Clinical Gastroenterology, 49(8), 690-696.

Seelye, K. Q. (2016, October 6). As drug deaths soar, a silver lining for transplant patients. New

York Times.

Sharma, A., Ramanathan, R., Posner, M., & Fisher, R. A. (2013). Pediatric kidney transplantation:

a review. Transplant Research and Risk Management, 5, 21-31.

Siminoﬀ, L., Marshall, H., Dumenci, L., Bowen, G., Swaminathan, A., & Gordon, N. (2009).
Communicating eﬀectively about donation: An educational intervention to increase consent
to donation. Progress in Transplantation, 19(1), 35-43.

Siminoﬀ, L. A., Gordon, N., Hewlett, J., & Arnold, R. M. (2001). Factors inﬂuencing families’

consent for donation of solid organs for transplantation. JAMA, 286(1), 71-77.

132

Simon, K., Soni, A., & Cawley, J. (2017). The impact of health insurance on preventative care
and health behaviors: Evidence from the ﬁrst two years of the ACA medicaid expansions.
Journal of Policy Analysis and Management, 36(2), 390-417.

Smith, Biggins, Haselby, Kim, Wedd, Lamb, . . . Israni

(2012). Kidney, pancreas and liver
allocation and distribution in the united states. American Journal of Transplantation, 12,
3191-3212.

Smith, J. M., & Dharnidharka, V. R. (2014). Progress in pediatric kidney transplantation. The

Open Urology & Nephrology Journal, 7, 115-122.

Sommers, B. D., Baicker, K., & Epstein, A. M. (2012). Mortality and access to care among adults

after state medicaid expansions. The New England Journal of Medicine, 367, 1025-1034.

Sommers, B. D., Gunja, M. Z., Finegold, K., & Musco, T.

(2015). Changes in self-reported
insurance coverage, access to care, and health under the aﬀordable care act. JAMA, 314(4),
366-374.

SRTR. (2014). Annual data report. Retrieved 2017-1-30, from http://srtr.transplant.hrsa

.gov/annual_reports/Default.aspx

Stuart, E., Huskamp, H., Duckworth, K., & et al. (2014). Using propensity scores in diﬀerence-in-
diﬀerences models to estimate the eﬀects of a policy change. Health services and outcomes
research methodology, 14(4), 166-182.

Sweeney, M. (2011). Three essays on the supply and delivery of healthcare: Evidence from kidney

donation and transplantation (Doctoral dissertation). Cornell University.

Teltser, K. (forthcoming). Do kidney exchanges improve patient outcomes? American Economic

Journal: Economic Policy.

Thibodeau, J., Rao, M., Gupta, C., Ayers, C., Gupta, S., Mammen, P., . . . Drazner, M. (2013).
Health insurance as a requirement to undergo cardiac transplantation: A national survey of
transplant program practices. Transplant Proceedings, 45, 360-363.

Trivedi, J. R., Ising, M., Fox, M., Cannon, R., Berkel, V. H. V., & Slaughter, M. S.

(2018).
Solid-organ transplantation and the aﬀordable care act: Accessibility and outcomes. The
American Surgeon, 84(12), 1894-1899.

Tumin, D., Beal, E. W., Mumtaz, K., Jr., D. H., Tobias, J. D., Pawlik, T. M., . . . Black, S. M.
(2017). Medicaid participation among liver transplant candidates after the aﬀordable care
act medicaid expansion. Journal of the American College of Surgeons, 225(2), 173-180.

UNOS.

(2010). Deceased donor kidney allocation policies related to pediatric candidates.
from https://dev.transplantpro.org/100524_pediatric

Retrieved 2017-5-16,

133

-kidney-priority-education-2/

UNOS.

(2017a). Matching organs. Retrieved 2017-4-1, from https://www.unos.org/

transplantation/matching-organs/

UNOS. (2017b). Policy. Retrieved 2017-4-1, from https://www.unos.org/policy/

U.S. Centers for Medicare & Medicaid Services.

I have end-stage renal disease
(ESRD). Retrieved 2019-7-2, from https://www.medicare.gov/information-for-my
-situation/i-have-end-stage-renal-disease-esrd

(2019).

USRDS. (2018). 2017 USRDS annual data report: Chapter 7: ESRD among children, adolescents,

and young adults (Vol. 71) (No. 3).

Verous, M., Grosso, G., Corona, D., Mistretta, A., Giaquinta, A., Giuﬀrida, G., . . . Veroux, P.
(2011). Age is an important predictor of kidney transplantation outcome. Nephrol Dial
Transplant, 27(4), 1663-1671.

Weiner, S. G., Malek, S. K., & Price, C. N.

Transplantation, 101(4), 678-681.

(2017). The opioid crisis and its consequences.

Wellington, A. J., & Sayre, E. A. (2011). An evaluation of ﬁnancial incentive policies for organ

donations in the united states. Contemporary Economic Policy, 29(1), 1-13.

Wenner, D. (2016, December 8). Heroin addicts more likely to be organ donors - and lots of people
want their organs. Penn Live. Retrieved from http://www.pennlive.com/news/2016/
12/transplants.html

Wherry, L. R., & Miller, S. (2016). Early coverage, access, utilization, and health eﬀects of the
aﬀordable care act medicaid expansions: A quasi-experimental study. Annals of Internal
Medicine, 164(12), 795-803.

Wright, B. J., Conlin, A. K., Allen, H. L., Tsui, J., Carlson, M. J., & Li, H. F. (2016). What does
medicaid expansion mean for cancer screening and prevention? results from a randomized
trial on the impacts of acquiring medicaid coverage. Cancer, 122, 791-797.

134