Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Pathways of drug use among people who inject drugs
(USC Thesis Other)
Pathways of drug use among people who inject drugs
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
P ath w a ys of Drug Use Among P eople Who Inject Drugs
b y
Daniel Ch u
A Dissertation Presen ted to the
F A CUL T Y OF T HE USC GRADUA TE SCHOOL
UNIVERSITY OF SOUTHERN CALIF ORNIA
In P artial F ulfillmen t o f the
Requiremen ts for the Degree
DOCTOR OF PHILOSOPHY
(PREVENTIVE MEDICINE - HEAL TH BEHA VIOR RESEARCH)
December 2021
Cop yrigh t 2021 Daniel Ch u
Dedication
T o m y friends. T o m y family . T o m y paren ts. T o m y father.
ii
A c kno wledgemen ts
First and foremost, I w ould lik e to thank Ric ky Bluthen thal, m y men tor and committee
c hair, for his guidance and supp ort. I striv e to follo w his example in empath y , though tful-
ness, and academic rigor. I w ould also lik e to thank m y other committee mem b ers: Ka yla
de la Ha y e, Jimi Huh, and Alice Cep eda. I could not ha v e ask ed for more patien t, exp eri-
enced, or accomplished p eople.
I w ould also lik e to thank the studen ts, staff, and facult y , past and presen t, of the Depart-
men t of Prev en tiv e Medicine for their man y small and large con tributions to m y w ork and
m y life. Man y of y ou I am privileged to call m y friends. In no particular order, I w an t to
sp ecifically ac kno wledge T rev or Pic k ering, Jennifer T sai, An upreet Sidh u, Eldin Dzubur,
Sydney O’Connor, Darryl Nousome, Karen Ra, Shirlene W ang, Jessica T obin, P atricia Es-
cob edo, Stephanie Dy al, Charlotte Deng, Marn y Baro vic h, and Sherri F agan. A dditionally ,
it w ould b e remiss of me not to ac kno wledge the help and supp ort pro vided b y Geneviev e
Dun ton, principal in v estigator of the REA CH lab, in m y education, career, and life.
I am grateful for m y family (m y mother, m y brother A dam, and sister-in-la w R ub y). Al-
though they con tin ue to understand neither what I do nor what I study , they ha v e b een
en tirely supp ortiv e throughout the en tire pro cess. I lo v e and thank them all the same.
Finally , I w ould lik e to thank m y father, who passed a w a y w eeks b efore m y defense. He
help ed inspire in me a curiosit y and an in terest in science as a c hild. Others ha v e told me
that I am most definitely m y father’s son.
iii
T able of Con ten ts
Con ten ts
Dedication ii
A c kno wledgemen ts iii
List of T ables vi
List of Figures vii
Abstract viii
1 In tro duction 1
1.1 Study 1: P ath w a ys of drug use . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Study 2: P ath w a ys to injection drug use initiation . . . . . . . . . . . . . . . 4
1.3 Study 3: P ath w a ys of injection drug use . . . . . . . . . . . . . . . . . . . . 4
2 Bac kground and Significance 5
2.1 Illicit drug use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 P ath w a ys of Drug Use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2.1 Drug: The Gatew a y Hyp othesis F ramew ork . . . . . . . . . . . . . . 7
2.2.2 Individual: Common Liabilit y to A ddictions Mo del . . . . . . . . . . 7
2.2.3 Con text: Drug Generations F ramew ork . . . . . . . . . . . . . . . . . 9
2.3 Ov erview of studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3 Researc h Design and Metho ds 11
3.1 P aren t study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2 Measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.3 Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.4 Statistical Approac h . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.5 Surviv al analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.5.1 T emp oral ordering and time-dep enden t v alues . . . . . . . . . . . . . 17
3.5.2 Nonprop ortional hazards . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.5.3 A dditiv e hazards mo del . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.5.4 In terpretation of additiv e mo del re sults . . . . . . . . . . . . . . . . . 20
3.6 Missing data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.7 Soft w are . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4 Study 1: P ath w a ys of Drug Use 22
4.1 Bac kground . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.2 Metho ds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.2.1 Sample and measures . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.2.2 Surviv al analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.3.1 Time-v arying effects of demographic co v ariates . . . . . . . . . . . . 29
4.3.2 Constan t effects of drug use predic tors . . . . . . . . . . . . . . . . . 35
4.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
iv
T able of Con ten ts
5 Study 2: P ath w a ys to Injection Drug Use Initiation 45
5.1 Bac kground . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
5.1.1 Injection drug use in the United Stat es . . . . . . . . . . . . . . . . . 47
5.1.2 Consequences of injection drug use . . . . . . . . . . . . . . . . . . . 48
5.1.3 Risk factors for initiating injec tion drug use . . . . . . . . . . . . . . 49
5.2 Metho ds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5.2.1 Sample and measures . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5.2.2 Surviv al analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
6 Study 3: P ath w a ys of Injection Drug Use 67
6.1 Bac kground . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
6.2 Metho ds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
6.2.1 Sample and measures . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
6.2.2 Surviv al analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
6.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
6.3.1 Time-v arying effects of demographic co v ariates . . . . . . . . . . . . 77
6.3.2 Constan t effects of drug use predic tors . . . . . . . . . . . . . . . . . 81
6.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
7 Conclusions 95
7.1 Ov erall Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
7.2 Metho dological Impl ications . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
7.2.1 Confirming the Imp ortance of Time . . . . . . . . . . . . . . . . . . . 98
7.2.2 Understanding Surviv al Through Other Means . . . . . . . . . . . . 99
7.3 Theoretical Implica tions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
7.3.1 In tegrating Comp eting Theories of R isk F actors . . . . . . . . . . . . 100
7.3.2 F rom P ath w a ys to Pro cesses . . . . . . . . . . . . . . . . . . . . . . . 100
7.3.3 The Diminishing Role of Cannabis . . . . . . . . . . . . . . . . . . . 102
7.3.4 The Rising Role of NMPDs . . . . . . . . . . . . . . . . . . . . . . . 103
7.4 F uture Researc h Direct ions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
7.5 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
A Study 1 Detailed T ables 106
B Study 2 Detailed T ables 113
C Study 3 Detailed T ables 115
References 129
v
List of T ables
List of T ables
3.1 Coun ts of categorie s for demographic attributes . . . . . . . . . . . . . . . . 14
3.2 Descriptiv e statist ics of drug use b eha viors b y drug . . . . . . . . . . . . . . 15
4.1 Study 1: Suprem um p-v alues for significance of time-v arying effects . . . . . 30
4.2 Study 1: P arametric estimates . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5.1 Study 2: Descriptiv e statistics of drug use b eha viors b y drug un til injection
initiation or surv ey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.2 Study 2: Suprem um p-v alues for significance of time-v arying effects . . . . . 57
5.3 Study 2: P arametric estimates . . . . . . . . . . . . . . . . . . . . . . . . . . 58
5.4 Study 2: Comp eting risks - Suprem um p-v alues for significance of time-
v arying effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
5.5 Study 2: Comp eting risks - P arametric estimates . . . . . . . . . . . . . . . 61
6.1 Study 3: Descriptiv e statistics of injection drug use b eha viors . . . . . . . . 72
6.2 Study 3: Suprem um p-v alues for significance of time-v arying effects (re-
stricted mo dels) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
6.3 Study 3: Suprem um p-v alues for significance of time-v arying effects (full
mo dels) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
6.4 Study 3: P arametric estimates (restricted mo dels) . . . . . . . . . . . . . . . 85
6.5 Study 3: P arametric estimates (full mo dels) . . . . . . . . . . . . . . . . . . 87
A.1 Study 1: Detailed suprem um p-v alues for significance of time-v arying effects 106
A.2 Study 1: Detailed parametric estimates . . . . . . . . . . . . . . . . . . . . . 109
B.1 Study 2: Comp eting risks - Detailed suprem um p-v alues for significance of
time-v arying effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
B.2 Study 2: Comp eting risks - Detailed parametric estimates . . . . . . . . . . 114
C.1 Study 3: Detailed suprem um p-v alues for significance of time-v arying effects
(restricted mo dels) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
C.2 Study 3: Detailed suprem um p-v alues for significance of time-v arying effects
(full mo dels) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
C.3 Study 3: Detailed parametric estimates (restricted mo dels) . . . . . . . . . . 122
C.4 Study 3: Detailed parametric estimates (full mo dels) . . . . . . . . . . . . . 125
vi
List of Figures
List of Figures
4.1 Study 1: Kaplan-Meier surviv al curv es . . . . . . . . . . . . . . . . . . . . . 27
4.2 Study 1: Kaplan-Meier cum ulativ e hazard curv es . . . . . . . . . . . . . . . 28
4.3 Study 1: Time-v arying effects for sex, sexual orien tation, and birth place . . 32
4.4 Study 1: Time-v arying effects for race . . . . . . . . . . . . . . . . . . . . . 33
4.5 Study 1: Time-v arying effects for birth cohort . . . . . . . . . . . . . . . . . 34
5.1 Study 2: Kaplan-Meier surviv al curv e . . . . . . . . . . . . . . . . . . . . . . 54
5.2 Study 2: Kaplan-Meier cum ulativ e hazard curv e . . . . . . . . . . . . . . . . 55
5.3 Study 2: Time-v arying effects . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.4 Study 2: Comp eting risks - Time-v arying effects for sex, sexual orien tation,
and birth place . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.5 Study 2: Comp eting risks - Time-v arying effects for race . . . . . . . . . . . 60
5.6 Study 2: Comp eting risks - Time-v arying effects for birth cohort . . . . . . . 61
6.1 Study 3: Kaplan-Meier surviv al curv es . . . . . . . . . . . . . . . . . . . . . 73
6.2 Study 3: Kaplan-Meier cum ulativ e hazard curv es . . . . . . . . . . . . . . . 74
6.3 Study 3: Kaplan-Meier surviv al curv es stratified b y injection exp osure . . . . 75
6.4 Study 3: Kaplan-Meier cum ulativ e hazard curv es stratified b y injection ex-
p osure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
6.5 Study 3: Time-v arying effects for sex, sexual orien tation, and birth place . . 82
6.6 Study 3: Time-v arying effects for race . . . . . . . . . . . . . . . . . . . . . 83
6.7 Study 3: Time-v arying effects for birth cohort . . . . . . . . . . . . . . . . . 84
vii
Abstract
Abstract
This dissertation studies the con tributing factor s to illicit drug use b eha viors b y using a
longitudinal, dev elopmen tal p ersp ectiv e of drug use o v er the drug use career. The three
comp onen t studies rely on the same data and similar metho ds to examine the con tri-
butions of drug t yp e, individual, and con textual factors on m ultiple drug use outcomes
among a sample of p eople who inject drugs (PWID), a subset of p eople who use drugs
(PWUD) who disprop ortionately exp erience man y of the negativ e consequences asso ciated
with drug use. The three studies are distinguished b y examining differen t sets of drug use
outcomes using differen t sets of drug use exp osure predictors to in v estigate the con tribu-
tions of (1) an y-route drug use exp osures on an y-route drug use initiations, (2) an y-route
drug use exp osures on first injection drug use, and (3) injection drug use exp osures on
injection drug use initiations of other drugs. The con tributions of drug use exp osures
w ere ev aluated along with indep enden t con tributions of indivi dual traits and generational
cohort differences. Results generally rev ealed that prior drug use exp osures differen tially
affected drug use initiation outcomes and differences of risk b y outcome w ere observ ed b y
individual traits and generational cohort. Longitudinal risk asso ciations w ere observ ed
for an y-route drug use b eha viors and sp ecifically for injection drug use b eha viors. These
results confirm the idea of a gatew a y effect in that certain drug use b eha viors ma y increase
the risk of initiating other drug use b eha viors indep enden tly of individual or con textual
risk effects. These findings ma y represen t underlying risk pro cesses situated within and
b et w een individuals and the risk of initiating drug use b eha viors can b e attributed to a
necessary in teraction of individual and so cial risks. Drug use exp osures ma y b e imp ortan t
mark ers of risk tra jectories o v er the drug use career and can inform strategies to prev en t
initiations of drug use b eha viors.
viii
In tro duction
1 In tro duction
Illicit drug use con tin ues to b e a health issue in the United States with significan t individ-
ual and so cial consequences. P eople who use drugs (PWUD) exp erience higher morbidities
for man y health conditions including infectious disease, men tal health issues, and c hronic
ph ysical health problems ( C.-Y. Chen & Lin, 2009 ; Han, Gfro erer, & Colliv er, 2010 ).
Drug o v erdose deaths in the United States ha v e increased in recen t y ears from 6 cases p er
100000 p ersons in 1999 to 16 p er 100000 in 2017 ( Hedegaard, W arner, & Miniño, 2017 ).
This increase w as m uc h higher among White p ersons (6 to 21 p er 100000) and adults
aged 55 to 64 (4 to 22 p er 100000) and o v erdose deaths in v olving heroin increased as a
prop ortion of all o v erdose deaths from 8% in 2010 to 25% in 2015. F rom 2011 to 2018,
o v erdose deaths in v olving methamphetamine increased b y around fiv e times (1.8 to 10.1
p er 100000 for adult men; 0.8 to 4.5 p er 100000 for adult w omen) ( Han et al., 2021 ). In
total, illicit drug use exacts a great so cial cost, particularly for public health. In 2011,
the Cen ter ( 2011 ) estimated the y early drug-related healthcare costs to b e $11 billion.
The A dministration ( 2017 ) estimated the public healthcare costs related to non-medical
prescription drug (NMPD) use to b e $29 billion in 2013. The ann ual so cial cost sp ecifically
for heroin use w as estimated to b e around $51000 in 2015 United States dollars p er p erson
who used heroin. The total economic cost of opioid use in 2017 w as estimated at $1.0
trillion dollars ( Luo, Li, & Florence, 2021 ) and of methamphetamine use in 2005 to b e
$23.4 billion dollars ( Nicosia, P acula, Kilmer, Lundb erg, & Chiesa, 2009 ).
Researc h on drug use tra jectories ma y pro vide opp ortunities for prev en ting initiations of
illicit drug use b eha viors. Ov er the drug use career, the longitudinal collection of drug use
exp eriences within an individual’s lifetime, more than one drug use b eha vior is lik ely to b e
initiated. These b eha viors ma y b e causally asso ciated in that the adoption of one b eha v-
ior ma y increase the lik eliho o d of initiating another b eha vior in the future. Understanding
these causal links allo ws for a general mapping of the dev elopmen tal path w a ys of drug use
and these path w a ys can b e used to iden tify opp ortunities for in terruption to prev en t the
adoption of future, additional b eha viors whic h ma y also b e more harmful. This is particu-
1
In tro duction
larly relev an t for prev en tion efforts for p olydrug use and injection drug use b eha viors.
Muc h of the w ork so far in this area has in v olv ed the Gatew a y Hyp othesis framew ork
( Denise Kandel & Kandel, 2015-02 ), whic h p osits that the drug use career can b e de-
scrib ed using generalizable, ordered sequences based on longitudinally asso ciated drug use
b eha viors. The initiation of one drug use b eha vior can increase the lik eliho o d of the future
initiation of another and the use of certain drugs ma y serv e as “gatew a ys” to the use of
other, more harmful drugs. Researc h relying on this framew ork has iden tified tra jectories
from licit substance use of alcohol or tobacco to initiation of cannabis use and then to
later use of other illicit drugs ( Denise Kandel & Kandel, 2015-02 ; D. Kandel & Y amaguc hi,
1993 ; Key es, Hamilton, & Kandel, 2016 ; Y amaguc hi & Kandel, 1984 ). This researc h has
b een primarily motiv ated to w ards understanding the dev elopmen t of the drug use career
at its early stages.
Alternativ ely , it is p ossible that asso ciations b et w een b eha viors ma y b e attributable to con-
founding causes. The Common Liabilit y Mo del ( P almer et al., 2009 ; V an yuk o v et al., 2012 ;
V a n yuk o v & R idenour, 2012 ) describ es dr ug use as a result of individual liabilit y to initi-
ating drug use. The apparen t asso ciations among drug use b eha viors can b e the result of
factors generally influencing drug use risk. Some studies examining the gatew a y effect ha v e
found that underlying factors indep enden t of drug use exp eriences do generally increase
risk of initiating some drug use b eha viors ( L. Degenhardt et al., 2009 ; Jorgensen & W ells,
2021 ; Morral, McCaffrey , & P addo c k, 2002 ; Zhang, W u, W u, Durkin, & Marsiglia, 2021 ).
It is lik ely that the risks of drug use initiations dep end on b oth drug use exp eriences and
more general risk factors, but evidence so far remains limited for man y illicit substances
suc h as heroin and for injection drug use. Researc h on the path w a ys of drug use has fo-
cused on early or sp ecific ranges of the drug use career or for a sp ecific set of drug use b e-
ha viors. As a result, the understanding of dev elopmen tal tra jectories including or to b e-
ha viors in v olving a broad set of drugs, differen t routes of administration, or b oth is incom-
plete. A dditionally , there exists v ery little, if an y , study on the dev elopmen tal tra jectories
for injection drug use b eha viors, whic h are of ma jor consequence for drug use public health
2
In tro duction
efforts.
The three studies comprising this dissertation s eek to address these gaps b y studying the
dev elopmen tal path w a ys of drug use for three sets of b eha viors and p erio ds of the drug
use career. These studies rely on secondary data analyses of drug use histories rep orted b y
p eople who inject drugs (PWID) for a broad set of illicit drugs used via non-injection or
injection routes. The shared analytic approac h across studies in v olv es surviv al regression
with additiv e hazards, a flexible time-v arying approac h no v el to longitudinal analysis of
drug use dev elopmen t. Generally , the three studies examine (1) the path w a ys of an y-route
drug use initiations for a broad set of illicit drugs, (2) the path w a ys of non-injection drug
use initiations leading to first injection drug use, and (3) the path w a ys of injection drug
use initiations.
1.1 Study 1: P ath w a ys of drug use
Researc h on the path w a ys of drug use has primarily fo cused on the b eginning of the drug
use career. This b o dy of w ork reflects in v estigation on the Gatew a y Hyp othesis to under-
stand the start of the drug use career and has found common path w a ys of drug use during
this this crucial dev elopmen tal p erio d. Ho w ev er, there has b een little study of the path-
w a ys of drug use o v er the en tire drug use career o r for a large set of drug use b eha viors.
As a result, the relationships b et w een man y b eha viors and ho w man y b eha viors con tribute
to tra jectories remain unclear.
This study mo dels the lik eliho o d of initiation o f use for a drug on past initiations of other
drugs. Time-dep enden t surviv al mo dels w ere constructed for the initiations of 11 distinct
t yp es of drugs whic h o ccurred an ytime during the drug use career. Results from these
mo dels are used to describ e p oten tial drug use path w a ys including and to these initiations
and the con tributions of demographic c haracteristics and generational con text on these
path w a ys.
3
In tro duction
1.2 Study 2: P ath w a ys to injection drug use initiation
A ma jor in terest of drug use harm reduc tion efforts is to prev en t PWUD from initiating
in to injection drug use ( Dan W erb, Buxton, et al., 2013 ). Researc h on the factors of injec-
tion drug use ha v e iden tified prior drug use exp eriences as a significan t risk factor for in-
jection drug use initiation. The asso ciation b et w een prior drug use and injection drug use
also app ears to b e time-dep enden t with prior initiations of some b eha viors increasing the
lik eliho o d of injection drug use initiation so oner than others.
This study in v estigates the p oten tial time-dep ende n t con tributions of pre-injection drug
use b eha viors on the lik eliho o d and timing of injection drug use initiation using the same
surviv al mo deling as the first study . The results of this mo del are used to describ e the p os-
sible path w a ys to first injection drug use exp erience.
1.3 Study 3: P ath w a ys of injection drug use
After the start of injection drug use, P WID can con tin ue to initiate injection use of other
t yp es of drugs. T o date, there has b een little researc h to describ e the dev elopmen tal tra-
jectories of injection drug use. Better understanding the injection drug use career ma y b e
useful in harm reduction efforts to prev en t p olydrug use and further injection drug use b e-
ha viors. This study , using a similar approac h as the first study , in v estigates the longitu-
dinal asso ciations among injection drug use b eha viors and describ es p ossible path w a ys of
injection drug use.
4
Bac kground and Signficance
2 Bac kground and Significance
2.1 Illicit drug use
Drug use encompasses a broad v ariet y of b eha viors in v olving drugs and ho w they are used.
In this dissertation, the term “drug use” is used to sp ecifically refer to the use of illicit
drugs v ersus legal drugs lik e alcohol and tobacco. Drug use b eha viors include usage b y
drug (i.e., co caine use), usage b y route (i.e., injection drug use), or usage b y drug b y route
(i.e., injection co caine use). In the United States, co caine, heroin, and methamphetamine
con tin ue to b e commonly used but new er epidemics of drug use, particularly the epidemic
of NMPD use of opioids, ha v e recen tly b ecome additional significan t public health issues.
Although cannabis remains p opular y et illeg al at the federal lev el, state-lev el legalization
efforts in tro ducing standardization and regulation ha v e made it de facto legal and increas-
ingly so cially acceptable in the mainstream. Drugs can b e used via differen t routes of ad-
ministration, commonly snorting or sniffing, smoking or v ap orizing, and injection.
The health consequences of drug use can b e se v ere or fatal. Sev ere drug dep endency
can impact men tal, ph ysical, and so cial functioning leading to a diminished qualit y of
life. Long-term use can also cause neurological problems ( Brust, 2014 ) and substan tial
ph ysical damage to ma jor organ systems ( Got w a y et al., 2002 ; Mégarbane & Chevillard,
2013 ). Some b eha viors app ear to b e comorbid with some psyc hiatric or men tal health
problems ( Abuse & A dministration, 2020a ) although the di rection of this relationship
remains unclear. V arying purities and p otencies asso ciated with the illegal and clandestine
man ufacture and distribution of drugs con tribute to a higher risk of o v erdose due to
difficulties in titrating a safe dose. This risk is greatly magnified with injection drug use,
whic h not only mak es more lik ely the risk of o v erdose but also narro ws the windo w for
effectiv e medical in terv en tion to prev en t a sev ere or fatal outcome. Injection drug use also
in tro duces further health risks related to unsafe practices suc h as using unsterile injection
equipmen t, sharing drug or equipmen t, or improp erly preparing the drug or injection
site. These practices can lead to acute infection whic h ma y require emergency medical
in terv en tion or c hronic infection through transmitted blo o db orne diseases suc h as HIV or
5
Bac kground and Signficance
HCV.
Drug use con tin ues to b e prev alen t in the United States and a significan t public health
problem. Results from the 2019 National Surv ey on Drug Use and Health (NSDUH)
( Abuse & A dministration, 2020a ) indicate that 21% of the p opulation 12 y ears or older
rep orted using illicit drugs in the past y ear (9% not including cannabis), an increase from
18% in 2015 and higher than ev ery y ear since. Illicit drug use disorders, reflecting sev ere
or long-term use, also remained high at 3% of the same p opulation. These substan tial
p ercen tages b elie the significan t health needs of PWUD and the subsequen t so cial costs
incurred b y these needs ( F renc h, McGeary , Chit w o o d, & McCo y , 2 000 ; Gryczynski et al.,
2017 ; R y an & Rosa, 2020 ).
2.2 P ath w a ys of Drug Use
Researc h on drug use b eha viors ha v e iden tified a n um b er of risk factors for sp ecific b eha v-
iors, but the role of these b eha viors within larger tra jectories of risk within the drug use
career con tin ue to b e unclear. Prior researc h has iden tified ordered asso ciations among
drug use b eha viors in that the o ccurrence of one b eha vior can affect the lik eliho o d of an-
other in the future. Lik ewise, researc h to date supp orts the existence of individual risk fac-
tors for drug use initiations and has observ ed generational trends in the epidemiology of
drug use b eha viors. The con tributions of these researc h areas are unquestionably v aluable
on their o wn but ma y b e augmen ted through a more in tegrated approac h that relies on a
longitudinal or dev elopmen tal lens. This approac h can pro vide a ric her description of the
v arious dev elopmen tal path w a ys of the drug use career, delineated b y drug use b eha viors
and p ossibly differen tiated b y individual and con textual factors. The p oten tial v alue of un-
derstanding these path w a ys is not limited to describing the dev elopmen tal tra jectories of
the drug use career and can also yield some insigh t on the causativ e risk mec hanisms.
Understanding the path w a ys of drug use requires theoretical approac hes that can accom-
mo date the div ersit y of constituen t b eha viors of the drug use career and reflect the con tin-
uous and c hanging nature of the drug use career. This dissertation relies on a theoretical
6
Bac kground and Signficance
framew ork with a foundation in Drug, Set, and Setting ( Zin b erg, 1986 ), whic h pro vides
one p ersp ectiv e of drug use b y organizing asso ciated factors in to three general domains re-
lating to the drug (Drug), the p erson (Set), and the con text (Setting). As the original de-
scription of Drug, Set, and Setting w as more orien ted to w ards describing singular drug use
ev en ts, it has b een extended longitudinally b y in tegrating other theoretical framew orks.
2.2.1 Drug: The Gatew a y Hyp othesis F ramew ork
Researc h on the relationships b et w een initiations of use of differen t drugs ha v e demon-
strated a longitudinal relationship b et w een man y t yp es of drug use. The Gatew a y Hyp oth-
esis ( D. Kandel, 1975 ) represen ts the predominan t theory describing the correlation and
ordering of drug use initiations, with a fo cus on explaining drug use tra jectories at the b e-
ginning of the drug use career. The relationships b et w een “gatew a y” drugs and their asso-
ciated outcomes p ossess t w o necessary qualities, correlation and ordering. The former is a
giv en and the latter implies a directional asso ciation: use of a gatew a y drug elicits an in-
creased risk of initiating the use of another drug in the future. W ork using this framew ork
primarily fo cuses on the tra jectories of drug use among adolescen ts, reflecting the in terest
in understanding and prev en ting drug use during this dev elopmen tal p erio d.
2.2.2 Individual: Common Liabilit y to A ddictions Mo del
A complemen tary , y et comp eting, explanatio n of drug use attributes initiation risk to
individual-asso ciated factors under the Common Liabilit y to A ddictions Mo del ( P almer et
al., 2009 ; V an yuk o v et al., 2012 ; V an yuk o v & Ridenour, 2012 ). These factors mediate the
exp osures to the so cial en vironmen ts and the individual exp eriences whic h can resp ectiv ely
influence the lik eliho o ds of drug initiation opp ortunities and drug use initiations. It has
already b een established that drug use is an age-graded b eha vior with the lik eliho o d of
drug use initiation dep enden t on age ( Abuse & A dministration, 2014 ; K. Chen & Kandel,
1995 ; V olk o w, Han, Einstein, & Compton, 2021 ). The bulk of drug use initiations o ccurs
around adolescence and y oung adultho o d and tap ers in to a “long tail” extending in to
adultho o d. The driving factors of this c hange are asso ciated with the ev olving so cial
7
Bac kground and Signficance
con texts and p ersonal motiv ations o v er the life course.
Drug use b eha viors correlate with v arious other demographic attributes. PWUD are a di-
v erse p opulation, but this div ersit y is unev enly distributed among drug use b eha viors. Re-
sults from the 2019 NSDUH ( Abuse & A dministration, 2020a ) pro vide evidence for differ-
ences prev alence and incidence on age, sex, race, and sexual orien tation. Among age cate-
gories of adolescen ts (12 to 17 y ears), y oung adults (18 to 35 y ears), and adults (36 y ears
or more), the highest prev alence of past y ear use o v erall w ere rep orted b y y oung adults
(39%). More adults b y coun ts rep orted past y ear initiation for heroin, methamphetamine,
and NMPD pain reliev er use; more y oung adults rep orted past y ear initiation of co caine
use; and more adolescen ts rep orted past y ear initiation of cannabis use. Males w ere more
lik ely to rep ort past y ear drug use v ersus females, o v erall and for most t yp es of drugs, and
substance treatmen t data from 2015 corrob orates this with males comprising 66% of ad-
missions ( SAMHSA, 2017 ). Among the largest racial and ethnic categories (Blac k, His-
panic, and White), White or Hispanic p ersons w ere more lik ely to rep ort past y ear co caine
(all t yp es) or methamphetamine use v ersus Blac k p ersons, but Blac k p ersons w ere most
lik ely to rep ort past y ear crac k co caine use ( Abuse & A dministration, 2020b ). Immigran t
status and the acculturation pro cess app ear to increase the lik eliho o d of drug use among
Hispanic or Latino p ersons ( Lara, Gam b oa, Kahramanian, Morales, & Bautista, 2005 ).
Sexual minorities iden tifying as ga y , lesbian, or bisexual app ear to exhibit greater proba-
bilities of drug use o v erall and for most drugs ( Medley et al., 2016 ; Rosner, Neicun, Y ang,
& Roman-Urrestarazu, 2021 ).
There are also asso ciations b et w een traumatic e v en ts and men tal health disorders on indi-
vidual risks of drug use b eha viors. These exp eriences ma y b e asso ciated with demographic
or drug use factors but lik ely exhibit indep enden t asso ciations with risk. Men tal health dis-
orders and drug use, and sp ecifically injection drug use, app ear to b e comorbid ( Kevin P .
Con w a y , Compton, Stinson, & Gran t, 2006 ; Mac k esy-Amiti, Donen b erg, & Ouellet, 2012 )
and men tal health disorders o ccurring at adolescence app ear to con tribute to the risk of
initiation drug use b eha viors in the future ( Kevin P . Con w a y , Sw endsen, Husky , He, &
8
Bac kground and Signficance
Merikangas, 2016 ). Psyc hologically traumatic exp eriences during c hildho o d app ears to b e
asso ciated with drug use ( Khoury , T ang, Bradley , Cub ells, & Ressle r, 2010 ; Quinn et al.,
2016 ) and sp ecifically with injection drug use ( Kerr et al., 2009 ). These traumatic exp e-
riences ma y b e asso ciated with c hildho o d exp osure to family substance use, subsequen tly
leading to earlier initiations of injection drug use ( T aplin, Saddic hha, Li, & K rausz, 2014 ).
A dditionally , c hildho o d sexual abuse ma y also b e a con tributing factor to the injection ini-
tiation risk ( Hadland et al., 2012 ).
2.2.3 Con text: Drug Generations F ramew ork
Finally , con texts, as so cio-cultural factors, influence the lik eliho o d of drug use through
larger so cial pro cesses that affect opp ortunities for drug use and the individual c hoices
at these opp ortunities. Longitudinal trends o ccurring at the so cial frames larger than
in terp ersonal relationships and comm unities can b e understo o d using the Drug Genera-
tions framew ork b y Andrew Golub, Johnson, & Dunlap ( 2005 ). Longitudinal differences
in drug p opularit y pro duce p erio ds of relativ e p opularit y for a drug and a generation of
p eople who initiate in to its use ( Andrew Golub, Elliott, & Bro wnstein, 2013 ; Andrew
Golub, Johnson, & Dunlap, 2005 ; A. L. G olub & Johnson, 2009 ). Drug p opularit y , in turn,
is caused b y other so cial c hanges o v er time suc h as c hanges in drug p olicy or economic
conditions. Generational effects can b e observ ed b et w een birth cohorts for p eople who
inject drugs ( Bluthen thal, W enger, Ch u, Bourgois, & Kral, 2017 ), arr estees ( Andrew
Golub & Bro wnstein, 2013 ; Andrew Golub, Elliott, & Bro wnstein, 2013 ; A. L. Golub &
Johnson, 2009 ), and nation wide samples ( A. Golub & Johnson, 2001 ; No v ak, Bluthen thal,
W enger, Ch u, & Kral, 2016 ).
2.3 Ov erview of studies
Presen tly , researc h on the path w a ys of drug use is limited to sp ecific scop es of the drug use
career or limited to sp ecific domains of risk factors. There are resulting gaps in kno wledge
regarding path w a ys spanning the en tire drug use career, path w a ys including a comprehen-
siv e set of drugs, and path w a ys in v olving differen t routes of administration. This disser-
9
Bac kground and Signficance
tation aims to address these gaps through three studies using secondary data analyses of
drug use histories rep orted b y PWID. Collectiv ely , these studies examine drug use path-
w a ys for a broad set of drug use b eha viors. The first study iden tifies path w a ys of drug use
o v er the en tire drug use career to b etter understand o v erall drug use dev elopmen tal tra-
jectories in v olving a broad set of drugs. The second study examines path w a ys to injection
drug use to in v estigate the p oten tial effects of prior drug use b eha viors on injection drug
use tra jectories. The third study iden tifies path w a ys of injection drug use to describ e the
dev elopmen tal tra jectories sp ecifically for injection drug use b eha viors.
10
Researc h Design and Metho ds
3 Researc h Design and Metho ds
All three studies of this dissertatio n relied on the same data and analytic approac hes. Gen-
eral descriptions of the data and statistical metho ds are pro vided in this section with more
sp ecific details pro vided in eac h study .
3.1 P aren t study
Data used in this dissertation comes from the Late Initiates Study , a NID A-funded study
on late initiation to injection drug use (NID A gran t # R01D A027689). PWID within the
Los Angeles or San F rancisco metrop olitan areas w ere ask ed to participate in the study b e-
t w een April 2011 and April 2013 using targeted sampling and comm unit y outreac h meth-
o ds ( Bluthen thal & W atters, 1995 ; Alex H. Kral et al., 2010 ; W atters & Biernac ki, 1989 ).
Inclusion criteria w ere rep orted injection drug use in the past 30 da ys (v erified b y study
staff using visible signs of v enipuncture), aged 18 y ears or older, and able to pro vide con-
sen t. After pro viding consen t, participan ts w ere administered a computer-assisted surv ey
in terview. After completing the surv ey , PWID who rep orted initiation of injection drug
use after the age of 30 and matc hed comparators w ere ask ed to participate in a qualitativ e
in-depth in terview.
This dissertation only uses the surv ey data, whic h i ncludes PWID at all ages of injection
drug use initiation and applies an additional inclusion criterion of rep orting a biological
sex of male or female and not in tersexed (see measures for details). A total of 813 par-
ticipan ts pro vided surv ey resp onses. Of these, 810 rep orted a biological sex of male or fe-
male, and not in tersex, at birth and comprised the analytic sample. Some participan ts
w ere able to b ypass the recen t injection inclusion c riterion and pro vide surv ey resp onses
and, consequen tly , the analytic sample w as not en tirely comp osed of PWID: 775 (95%)
w ere PWID who recen tly engaged in injection drug use , 33 (4%) w ere PWUD who did not
but initiated injection drug use at some p oin t, and 2 (0.2%) w ere lik ely PWUD who nev er
engaged in an y injection drug use o v er their lifetimes. These observ ations w ere included
as the source data for all studies but w ere v ariably excluded dep ending on the outcome or
11
Researc h Design and Metho ds
path w a y b eing studied.
3.2 Measures
Drug use histories. Drug use histories w ere collected for 11 illicit drugs. Fiv e of these w ere
for drugs of illegal origins: cannabis (referred to as “marijuana” b y the study; illegal at
time of data collection), crac k co caine, p o wder co caine, methamphetamine, and heroin.
T w o w ere for NMPD use of opioid replacemen t therap y drugs: methadone and buprenor-
phine (b oth opioids). F our w ere for NMPD use of opioids (not including opioid replace-
men t therap y drugs), tranquilizers, sedativ es, and stim ulan ts. F or eac h drug, participan ts
w ere ask ed to rep ort their drug use exp erience s using four separate questions on if they
ev er used the drug, at what age they first used the drug via non-injection routes of admin-
istration, if they ev er injected the drug, and when they first used the drug via injection.
The latter t w o questions w ere not ask ed for cannabis. A separate analysis comparing the
quan titativ e and qualitativ e drug use history data from the paren t found lo w agreemen t
b et w een b oth sources on the exact ages rep orted at initiations but high agreemen t on the
ordering of initiations ( Dy al, Kral, Gonzalez, W enger, & Bluthen thal, 2015 ).
Inje ction drug use initiation. Data w ere collected on a ge at first injection drug use and
the drug used at first injection. Age at first injection w as assessed via the question “The
first time y ou injected drugs, ho w old w ere y ou?” with con tin uous resp onses in y ears. This
question w as follo w ed b y “What drug did y ou inject the first time?” A v ailable resp onses
for this question w ere “crac k,” “p o wder co caine,” “heroin,” “crystal meth,” “sp eedball
(heroin & co caine com bined),” “go ofball (heroin & crystal meth com bined),” “Dilaudid or
other opiates,” “morphine or other opiates,” “co deine or other opiates,” “T alwin/Ritalin or
other stim ulan ts,” “steroids,” or “other. ”
Demo gr aphic char acteristics. Collected demographic data consisted of biologi cal sex, sex-
ual orien tation, birth in the United States, age at surv ey , race and ethnicit y , and date of
birth. Biological sex w as assessed using the question “What is y our biological sex?” with
a v ailable answ er options of “male,” “female” or “In tersexed/hermaphro dite. ” F or this dis-
12
Researc h Design and Metho ds
sertation, only the male and female categories are used with male as the referen t sex cat-
egory . Sexual orien tation w as ask ed with “What do y ou consider y our sexual orien tation
to b e?” with answ er options of “heterosexual,” “ga y or lesbian” or “bisexual. ” The latter
t w o categories w ere collapsed in to a single category for ga y , lesbian or bisexual sexual ori-
en tation to pro duce a binary v ariable. Birth in the United States w as dic hotomized in to
y es (U.S. b orn) or no. Age at study w as rep orted as a con tin uous v alue in y ears.
P articipan ts w ere ask ed to rep ort their race via the question “What do y ou consider to b e
y our racial or ethnic group?” They w ere ask ed to c ho ose the b est option from the follo w-
ing: “White,” “Blac k/not Latino,” “Latino/not Blac k,” “Blac k/Latino,” “Asian,” “P acific
Islander,” “Nativ e American,” or “mixed race. ” Due to small coun ts of self-iden tified Asian,
P acific Islander, Nativ e American, or mixed-race participan ts, these categories w ere col-
lapsed in to a single “other” category . A dditionally , the categories of “Blac k/not Latino”
and “Blac k/Latino” w ere collapsed in to a single “Blac k” category and “Latino/not Blac k”
w as renamed as “Latino. ” The resulting four race categories used for the analyses w ere
“White,” “Blac k,” “Latino” and “other. ” F or all mo dels, “White” race w as used as the ref-
eren t race category .
Birth cohorts w ere created from participan ts’ rep orted dates of birth. If the date of birth
w as missing, then the birth y ear w as calculated b y subtracting the y ear of the surv ey b y
the rep orted age at surv ey . F our birth cohorts w ere created based on decade: “Pre-Sixties”
(b efore 1960), “Sixties” (1960s), “Sev en ties” (1970s), and “Eigh ties or later” (1980s and
after). Sparse data at b oth ends informed the creation of the “Pre-Sixties” and “Eigh ties
or later” cohorts. F or all mo dels, the “Pre-Sixties” cohort w as used as the referen t cohort.
3.3 Sample
The analytic sample of 810 participan ts w as div erse and descriptiv e statistics on demo-
graphic v ariables are pro vided in T able 3.1 . A v erage age at surv ey in terview w as 47.8 y ears
(SD=11.5; missing=1) and ages ranged from 18 to 79 y ears. P articipan ts w ere largely b orn
male (73.5%) and a minorit y iden tified as ga y , lesbian, or bisexual (15.3%). The largest
13
Researc h Design and Metho ds
T able 3 .1: Coun ts of categories for demographic attributes
Coun t Prop ortion (%)
Birth cohort
Pre-Sixties 340 42.0
Sixties 264 32.6
Sev en ties 107 13.2
Eigh ties or later 98 12.1
Missing 1 0.1
Sex
Male 595 73.5
F emale 215 26.5
Ga y , lesbian, or bisexual
No 686 84.7
Y es 124 15.3
Race
White 272 33.6
Blac k 252 31.1
Latino 199 24.6
Other 82 10.1
Missing 5 0.6
Birthplace
United States 766 94.6
Outside United Stat es 44 5.4
birth cohort w as the Pre-Sixties cohort (n=340), comprising those b orn b efore 1960, fol-
lo w ed b y the Sixties cohort (n=264), the Sev en ties cohort (n=107), and finally the Eigh t-
ies or later cohort (n=98). On race, the sample w as 34% White, 31% Blac k, 25% Latino,
and 10% other race.
Descriptiv e statistics on drug use b eha viors are pro vided in T able 3.2 . The lifetime
prev alence for an y-route drug use w as greater than 70% for all illicit drugs of cannabis,
heroin, co caine, or methamphetamine. Lifetime prev alence of NMPD use v aried with more
than half the sample rep orting an y use of NMPD opioids or tranquilizers and around 20%
rep orting an y use of NMPD stim ulan ts or sedativ es. Of the drug treatmen t medications,
42% and 15% of the sample rep orted an y lifetime use resp ectiv ely of methadone and
buprenorphine.
14
Researc h Design and Metho ds
T able 3.2: Descriptiv e statistics of drug use b eha viors b y drug
Ev er used Ev er injected
Sample prev alence Age at initiation Sample prev alence Age at initiation
Coun t Prop ortion (%) Mean SD Coun t Prop ortion (%) Mean SD
Illicit drugs
Cannabis 748 92.35 13.76 4.82 NA NA NaN NA
Heroin 754 93.09 21.95 8.56 751 92.72 22.82 8.86
Crac k co caine 701 86.54 27.13 10.45 265 32.72 32.82 11.27
P o wder co caine 721 89.01 20.37 6.98 627 77.41 24.15 8.59
Methamphetamine 587 72.47 24.63 11.49 503 62.10 26.66 10.59
Drug treatmen t medications
Methadone 344 42.47 32.12 10.91 41 5.06 31.27 9.62
Buprenorphine 119 14.69 36.05 11.83 13 1.60 28.58 6.53
Prescription medications
Opioids 542 66.91 27.94 13.95 250 30.86 28.16 11.36
Stim ulan ts 172 21.23 21.97 10.15 78 9.63 24.49 9.03
Sedativ es 163 20.12 22.13 10.01 31 3.83 20.00 6.73
T ranquilizers 462 57.04 28.92 13.63 60 7.41 28.19 10.28
Injection
Injection drug use NA NA NA NA 808 99.75 21.70 8.62
15
Researc h Design and Metho ds
Virtually the en tire sample (99.75%) r ep orted ev er injecting o v er their lifetimes. The most
frequen tly rep orted injected drug w as heroin, follo w ed b y p o wder co caine and metham-
phetamine. Around a third of the sample rep orted ev er injecting crac k co caine or NMPD
opioids. Less than 10% of the sample rep orted ev er injecting methadone or NMPD stim-
ulan ts or tranquilizers and less than 5% rep orted ev er injecting buprenorphine, or NMPD
sedativ es.
3.4 Statistical Approac h
All three studies used surviv al analyses to estimate the hazards of drug use outcomes for
a giv en set of drug use predictors and demographic co v ariates. Surviv al analysis describ es
an y time-to-ev en t analysis whic h estimates the hazard of an outcome in the presence of
censoring on the observ ation of the outcome. The studies used Kaplan-Meier surviv al es-
timation ( Kaplan & Meier, 1 958 ), a non-parametric approac h, to estimate o v erall surviv al
curv es for outcomes and for preliminary analyses on the prop ortionalit y of hazards. This
w as follo w ed b y surviv al regression analyses using s emi-parametric approac hes to deter-
mine the effects of predictors and co v ariates on outcome hazards.
F o r all analyses, the predictors of in terest w ere the exp osures to v arious drug use b e-
ha viors prior to a giv en outcome. The presence, timing, and length of these exp osures
dep ended on whether a drug use b eha vior w as presen t and when it w as initiated relativ e
to the outcome. If a drug use b eha vior w as initiated b efore the outcome, then exp osure
16
Researc h Design and Metho ds
b egan at the age of its initiation. Otherwise, if a drug use b eha vior w as initiated after the
outcome or w as nev er initiated at all, then there w as no exp osure.
3.5 Surviv al analysis
3.5.1 T emp oral ordering and time-dep enden t v alues
Surviv al mo dels require strict temp oral ordering of information. Simply put, the future is
unkno wn to the past. This requiremen t is not limited to the temp oral ordering b et w een
indep enden t and dep enden t v ariables but also applies to the temp oral ordering among
and within indep enden t v ariables. Indep enden t v ariables can only con tain information as
they o ccur along time and cannot dep end on future information and this limitation deter-
mined the sp ecification and selection of exp osure predictors and demographic co v ariates.
Because exp osures are time-dep enden t, and can start at v arious times, exp osure predictors
w ere sp ecified as ha ving time-dep enden t v alues. In con trast, demographic traits, but not
states, can b e assumed to b e fixed at birth and ha v e constan t v alues across time. As no
p ersonal history data w as collected for demographic states lik e living situation or so cio e-
conomic status, demographic co v ariates used in the analyses w ere limited to the data col-
lected on traits that w ere assumed to b e fixed at birth, comprising birth cohort, biological
sex, sexual orien tation, birth in the United States, and race and ethnicit y . Analyses v aried
on their selection of exp osure predictors, but all exp osure predictors used the same sp eci-
fication of time-dep enden t v alues, and all mo dels used the same set of demographic trait
co v ariates with constan t v alue sp ecification.
Implemen ting time-dep enden t v alues consisted of transforming cross-sectional drug use his-
tory and demographic data in to longitudinal data spanning from birth at age 0 y ears to
either the age at outcome or the age at surv ey . The transformed data appro ximates the
data deriv ed from a prosp ectiv e surviv al study design starting from birth with y early ob-
serv ations. Exp osure predictors w ere op erationalized as binary time-dep enden t v alues with
a 0 corresp onding to no exp osure and a 1 to exp osure. As exp osures o ccur relativ e to out-
comes, eac h outcome required its o wn transformed dataset. The use of time-dep enden t v al-
17
Researc h Design and Metho ds
ues in surviv al mo deling is a common practice and most surviv al mo deling approac hes can
accommo date time-dep enden t data nested within m ultiple units of observ ation (e.g., par-
ticipan ts) without an y sp ecial consideration for m ultilev el effects.
3.5.2 Nonprop ortional hazards
Initial results from Kaplan-Meier estimation indicated nonprop ortional hazards. This w as
supp orted with results of go o dness of fit tests from Co x prop ortional hazards mo deling
( Gram bsc h & Therneau, 1994 ), whic h detected sig nifican t asso ciations b et w een some ef-
fects and time (i.e., correlation b et w een Sc ho enfeld residuals and time). Prop ortional haz-
ards mo dels, of whic h the Co x mo del is a part, rely on a foundational assumption of pro-
p ortional hazards in that effects are m ultiplicativ e on baseline hazards. Ideally , the differ-
ences in hazards due to a co v ariate can b e expressed as a constan t ratio o v er time. Viola-
tions of this assumption indicate a lac k of indep endence from time.
Three fairly straigh tforw ard correctiv es are comm only emplo y ed: (1) sp ecifying time-
dep enden t v ariables, (2) stratification of baseline hazards, or (3) sp ecifying time-v arying
effects through a time-dep enden t piecewise transform of indep enden t v ariables. As time-
dep enden t v ariables w ere already sp ecified, the first strategy w as not applicable to this
situation. Stratification w as rejected as it w ould result in the remo v al of co v ariate terms
whic h w ere deemed of in terest to the study . Including time-v arying effects in the Co x
mo del w ould require splitting the observ ation p erio d in to discrete c h unks of time and
assume that effects w ould v ary dep ending on differen t time ranges. This approac h w as
deemed problematic due its reliance on predetermined time b oundaries, the c hoice of
whic h app ears arbitrary . The lac k of wieldy correctiv e strategies to address nonprop or-
tional hazards precluded the use of the commonly used Co x mo del and motiv ated the use
of an alternativ e mo deling approac h.
3.5.3 A dditiv e hazards mo del
Instead, surviv al regression mo deling relied on the additiv e hazards mo deling using the
McKeague & Sasieni ( 1994 ) s emiparametric additiv e risk mo del. This approac h do es not
18
Researc h Design and Metho ds
rely on the prop ortional hazards assumption and p ossesses greater flexibilit y in accommo-
dating violations of this assumption. The additiv e approac h also allo ws for the estimation
of non-parametric time-v arying effects in addition to parametric effects that are constan t
o v er time, allo wing for more flexible mo deling. The follo wing equation is the general form
of the additiv e hazards mo del ( McKeague & Sasieni, 1994 ) used in this dissertation:
𝜆(𝑡) = 𝑋 (𝑡) ⋅ 𝛽 (𝑡) + 𝑍 (𝑡) ⋅ 𝛾 F r om this equation, 𝑋 (𝑡) ⋅ 𝛽 (𝑡) is the non-parametric comp onen t with 𝛽 (𝑡) time-v arying
effects and 𝑍 (𝑡) ⋅ 𝛾 is the parametric comp onen t with 𝛾 constan t effe cts. The effects of
this mo del are additiv e. Comparativ ely , the general form of the Co x prop ortional hazards
mo del is:
𝜆(𝑡) = 𝜆 0
(𝑡) ⋅ 𝑒 𝑋 (𝑡)⋅𝛽
Here 𝑋 (𝑡) ⋅ 𝛽 is the parametric comp onen t with 𝛽 constan t effects.
The sp ecification of time-v arying or constan t effect s do es not dep end on whether the v ari-
able p ossesses time-dep enden t or constan t v alues. T o further clarify , all time-v arying ef-
fects are non-parametric, and all constan t effects are parametric. T o reduce confusion, the
term “time-v arying” is only used when describing estimated effects whic h can v ary o v er
time and the term “time-dep enden t” is only used when describing v ariable or function v al-
ues whic h can v ary o v er time.
Sp ecifying whether an indep enden t mo del term pro duced a time-v arying or constan t ef-
fect w as based on statistical testing and ease of in terpretabilit y . Time-v arying and con-
stan t effects represen t a bias-v ariance tradeoff. Time-v arying effects pro vide greater detail
whereas constan t effects lend to easier in terpretations. Preliminary analyses with iterativ e
mo del building used statistical testing to determine if a time-v arying effect exhibited ef-
fects significan tly differen t from a constan t effect sp ecification ( 𝐻 0
∶ 𝛽 (𝑡) ≡ 𝛾 𝑡 ) o v er
the observ ation p erio d ( Martin ussen & Sc heik e, 2006 ). Starting from an unconstrained
mo del of all time-v arying terms, non-significan t time-v arying terms w ere constrained to
19
Researc h Design and Metho ds
b e constan t terms one at a time un til all lefto v er time-v arying parameters w ere found to
b e significan tly time-v arying. Generally , demographic c haracteristics, and not drug use
exp osure predictors, w ere lik ely to exhibit significan t time-v arying effects. T o simplify com-
parisons across mo dels, all mo dels used similar sp ecifications, ha ving the same sp ecifica-
tions of time-v arying effects for demographic co v ariates and constan t effects for drug use
exp osure predictors. The follo wing pro vides a general form of the mo del sp ecification used
across all studies:
𝜆(𝑡) = 𝑋 ⋅ 𝛽 (𝑡) + 𝑍 (𝑡) ⋅ 𝛾 𝑋 = c onstant values of demo gr aphic c ovariates
𝛽 (𝑡) = time-varying effe cts of inter c ept and demo gr aphic c ovariates
𝑍 (𝑡) = time-dep endent values of drug use exp osur e pr e dictors
𝛾 = c onstant effe cts of drug use exp osur e pr e dictors
3.5.4 In terpretation of additiv e mo del results
Results from additiv e mo dels comprise a non-parametric comp onen t consisting of time-
v arying effects and a parametric comp onen t consisting of constan t effects. Time-v arying
effects, as y early c hanges in hazards, w ere used to calculate c hanges in cum ulativ e haz-
ards o v er time. These w ere plotted with 95% p oin t wise confidence in terv als and 95% Hall-
W ellner confidence bands. The significance of a time-v arying effect w as assessed using a
suprem um test to determine whether the effect w as significan tly differen t from 0 ( 𝐻 0
∶
𝛽 (𝑡) ≡ 0 ) ( Martin ussen & Sc heik e, 2006 ). Constan t effects w ere presen ted as p oin t esti-
mates of y early c hange in hazard rate and significance of effects w ere assessed using con-
v en tional parametric approac hes.
Cum ulativ e hazards and hazard rates lend to differen t in terpretations. F or an y mo deled
surviv al outcome, the hazard for the outcome will accum ulate o v er time and this accum u-
lation is the cum ulativ e hazard. Cum ulativ e hazard is, therefore, time-dep enden t and can
increase or decrease o v er v arious time ranges but will nev er b e negativ e. The cum ulativ e
20
Researc h Design and Metho ds
hazard for a giv en time reflects all of the accum ulated hazard up to that p oin t. The units
of cum ulativ e hazard can b e in terpreted as the n um b er of outcome ev en ts that w ould b e
exp ected to o ccur, assuming rep eatable ev en ts, b y a sp ecified time with a lo w er b ound of 0
and un b ounded in the p ositiv e direction. Cum ulativ e hazards close to 0 indicate lo w lik eli-
ho o d of outcome o ccurrence whereas v alues approac hing infinit y indicate virtually guaran-
teed o ccurrence on or b efore the sp ecified time. A one-unit c hange in cum ulativ e hazard at
a sp ecified time corresp onds to an exp ected c hange of o ccurrence of one outcome ev en t b y
that time.
Hazard rates con tribute to cum ulativ e hazard and represen t the c hanges in cum ulativ e haz-
ard p er unit time. Larger hazard rates corresp ond to greater c hanges in cum ulativ e haz-
ard. F or the analyses in this dissertation, the constan t effect estimates for exp osure predic-
tors represen t c hanges in hazard rate. Because exp osure predictors w ere op erationalized as
time-dep enden t binary v alues, these effects are conditioned on exp osure.
3.6 Missing data
None of the v ariables used p ossessed more than 5% missingness and less than 5% of the
sample exhibited missing data on an y of the v ariables. The o v erall impact of missing data
on the analyses w as an ticipated to b e small. No sp ecial pro cedures w ere emplo y ed and all
mo dels relied on complete case analysis (i.e., list wise deletion).
3.7 Soft w are
All data managemen t and analyses w ere do ne in R v4.0.5 ( R Core T eam, 2021 ). The sur-
vival v3.2-10 ( Therneau, 2021 ; Therneau & Gram bsc h, 2000 ) pac kage w as used for Kaplan-
Meier estimation, data transformation, and preliminary Co x prop ortional hazards mo del-
ing. The timer e g v1.9.8 ( Martin ussen & Sc heik e, 2006 ; Sc heik e & Zhang, 2011 ) pac kage
w as used for additiv e hazards mo deling. Plots w ere generated using the ggplot2 v3.3.3
( Wic kham, 2016 ) pac kage.
21
Study 1: P ath w a ys of Drug Use
4 Study 1: P ath w a ys of Drug Use
Describing the dev elopmen tal path w a ys of drug use ma y b e informativ e for in terv en tions
seeking to in terrupt the initiation of future drug use b eha viors. In addition to risk factors
of drug, p erson, and setting, prior drug use b eha viors ma y significan tly predict the lik e-
liho o d of future b eha viors. Understanding the path w a ys that promote the emergence of
future b eha viors can pro vide prev en tion opp ortunities of these b eha viors through the iden-
tification of drug use trends. F urthermore, these path w a ys represen t dev elopmen tal tra jec-
tories and can b e used to clarify the mec hanisms of drug use initiation.
P eople who inject drugs (PWID) represen t a subset of p eople who use drugs (PWUD) who
ha v e initiated injection drug use. PWID generally engage in more harmful drug use b eha v-
iors and exp erience greater morbidities and mortalities asso ciated with their drug use. A d-
dressing the health disparities and disprop ortionate health burdens exp erienced b y PWID
has motiv ated researc h on understanding the dev elopmen tal tra jectories exp erienced b y
this group. Of also great in terest to public health is adv ancing the understanding of injec-
tion drug use initiation within con text of the o v erall drug use career for PWID.
Although w ork on drug use path w a ys w as originall y limited to licit substance use of al-
cohol or tobacco and illicit drug use of cannabis and co caine, researc h since then has ex-
panded to include other drug use b eha viors. Notably , the ongoing opioid epidemic has
increased in terest in examining the role of non-medical prescription drug (NMPD) opi-
oid use within these path w a ys. Ho w ev er, the role of other NMPD categories suc h as tran-
quilizers, sedativ es, and stim ulan ts has y et to b e in v estigated. Although the prev alence of
these t yp es of drug use are lo w er than for NMPD opioids, they con tin ue to b e substan tial
enough to w arran t further study . F urthermore, the effect of injection drug use on tra jec-
tories remains unclear. Researc h so far has iden tified significan t drug-related risk factors
con tributing to injection drug use initiation but has yielded limited understanding on the
role of injection on future drug use initiations. As drug use con tin ues to b e a significan t
public health issue, expanding prev en tion efforts to b e inclusiv e of man y t yp es of b eha viors
ma y b ecome a necessary strategy .
22
Study 1: P ath w a ys of Drug Use
4.1 Bac kground
W ork on the Gatew a y Hyp othesis b y D. Kandel ( 1975 ) ha s iden tified significan t ordered
asso ciations among drug use b eha viors. Initial researc h relying on this framew ork iden ti-
fied a path w a y from earlier licit alcohol or tobacco use to later initiation of illicit drug use
for cannabis, whic h in turn led to the use of other illicit drugs ( Denise Kandel & Kandel,
2015-02 ; D. Kandel & Y amaguc hi, 1993 ; Key es, Hamilton, & Kandel, 2016 ; Y amaguc hi &
Kandel, 1984 ). The resulting implication w as the iden tification of cannabis as a gatew a y
drug to other “harder” illicit drugs. A dditional study since then has incorp orated other
drugs in to this framew ork and has iden tified a n um b er of drug use tra jectories based on
the longitudinal risk relationships b et w een differen t drug use b eha viors. Researc h includ-
ing opioid-t yp e drugs of heroin and NMPD opioids has iden tified a significan t, ordered re-
lationship b et w een the t w o and significan t roles for b oth in o v erall drug use tra jectories
( W all et al., 2018 ).
Demographic traits are asso ciated with drug use b eha viors via differences in exp eriences,
con texts, and motiv ations. These differences manifest through significan t v ariations of
drug use prev alence across sex, sexual orien tation, and race. Generally , males are more
lik ely than females to engage in drug use ( Cotto et al., 2010 ). One p ossible cause is the
exp osure of females to few er initiation opp ortunities, ho w ev er, once an opp ortunit y is pre-
sen ted there do es not app ear to b e an y differences on sex for the probabilit y of initiation
( Etten, Neumark, & An thon y , 1999 ). Ga y , lesbian, or bisexual (GLB) sexual minorities are
also generally more lik ely to engage in drug use than their straigh t or heterosexual coun-
terparts ( Corliss et al., 2 010 ; Medley et al., 2016 ; New com b, Birk ett, Corli ss, & Mustan-
ski, 2014 ; Rosner, Neicun, Y ang, & Roman-Urrestarazu, 2021 ) p ossibly b ecause of greater
so cial acceptabilit y of these b eha viors within GLB comm unities or more in ternal reasons
relating to so cial stigma exp erienced as a consequence of iden tifying as a sexual minor-
it y ( Hatzen buehler, Jun, Corliss, & A ustin, 2015 ). Racial differences for drug use risks
are reflected in differences in drug use epidemiology: heroin use, crac k co caine use, and
p o wder co caine use ( Abuse & A dministration, 2020b ). These differences ma y b e driv en
23
Study 1: P ath w a ys of Drug Use
b y so cial pro cesses op erating within racial comm unities including so cial exp osure to prev a-
len t b eha viors and access to drug use kno wledge or material through so cial net w orks or
relationships. F urthermore, immigration exp erience ma y b e protectiv e, with p ersons of for-
eign birth ha ving lo w er initiation risks than those b orn in the United States ( Salas-W righ t,
V aughn, Clark, T erzis, & Córdo v a, 2014 ) and few er initiation opp ortunities ( Borges et al.,
2012 ).
Indep enden t of drug and individual factors, w ork b y Golub Andrew Golub, Johnson, &
Dunlap ( 2005 ) on generational c hanges in drug use p oin ts to t he imp ortan t role of histori-
cal con text on drug use epidemics. Larger so cial and cultural trends include c hanges in so-
cial p erceptions of drug use leading to longitudinal c hanges in drug use acceptabilit y , p op-
ularit y , and accessibilit y . Consequen tly , p erio ds of increased p opularit y can in tro duce ad-
ditional risk whereas decreased p opularit y can reduce risk. Cycles of increasing, sustained,
and then decreasing drug use and p opularit y ha v e b een observ ed through epidemiologi-
cal differences b et w een birth cohorts ( Bluthen thal, W enger, Ch u, Bourgois, & Kral, 2017 ;
Louisa Degenhardt, Chiu, Sampson, Kessler, & An thon y , 2007 ; Andrew Golub, Elliott,
& Bro wnstein, 2013 ; A. Golub & Johnson, 2001 ; A. L. Golub & Johnson, 2009 ; Huang,
Key es, & Li, 2018-01 ; W all et al., 2018 ) for cannabis, co caine, he roin, and NMPD opioids.
The primary aim of this study is to describ e the longitudinal relationships within a large
and div erse set of illicit drug use b eha viors for PWID. Secondarily , the age-dep enden t
risk of drug use initiation and differences on this asso ciated with demographic traits and
birth cohorts are also in v estigated. These are all accomplished using m ultiple mo dels of
age-dep enden t additiv e hazards with time-dep enden t and constan t effects.
4.2 Metho ds
4.2.1 Sample and measures
Analyses relied on the full 810 partic ipan ts comprising the surv ey data describ ed in Par ent
study (subsection 3.1, p. 11) . Data included drug use histories for 11 t yp es of illicit drugs:
(1) cannabis (referred to as “marijuana” b y the study); (2) crac k and (3) p o wder co caine;
24
Study 1: P ath w a ys of Drug Use
(4) methamphetamine; (5) heroin; (6) methadone and (7) buprenorphine (used outside of
treatmen t settings or directions); and non-medical prescription drug (NMPD) use of (8)
opioids (not including methadone or buprenorphine, e.g., Vico din or Oxycon tin), (9) tran-
quilizers (e.g., Klonopin or V alium), (10) sedativ es (e.g., Restoril or phenobarbital), and
(11) stim ulan ts (e.g., Ritalin or A dderall). F or eac h drug, participan ts pro vided informa-
tion on if they ev er used the drug and, if applicable, at what ages they first used the drug
via non-injection and injection routes.
Data also include demographic information: age at time of surv ey , biological sex at birth,
birth y ear, sexual iden tit y (heterosexual/straigh t, ga y , lesbian, or bisexual), birthplace
(United States v ersus other), and race. Birth y ear w as transformed in to categories mostly
corresp onding to decadal birth cohort. Sexual iden tit y w as dic hotomized in to heterosexual
v ersus ga y , lesbian, or bisexual. A dditional details on the measures used can b e found in
Me asur es (subsection 3.2, p. 12) .
F o r eac h of the 11 t yp es of illicit drugs included in the drug use history data, age at ini-
tiation w as created b y selecting the y oungest age at first use b et w een non-injection and
injection routes or, in the case of cannabis, the age at first non-injection use. The resulting
initiation ev en t data w ere com bined with data on age at first injection to form the source
data used to generate 11 separate datasets corresp onding to the 11 differen t illicit drug ini-
tiation outcomes. F urther details on this pro cess are pro vided in T emp or al or dering an d
time-dep endent values (subsubsection 3.5.1, p. 17) .
4.2.2 Surviv al analysis
Kaplan-Meier estimation w as used to generate surviv al curv es for eac h outcome follo w ed
b y additiv e hazards mo deling for surviv al regression. The additiv e hazards mo del w as used
for its non-reliance on prop ortional hazards and flexibilit y . A dditiv e hazards mo dels can
include an y com bination of non-parametric terms with time-v arying effects and paramet-
ric terms with constan t effects. Co efficien ts of time-v arying parameters tak e the form of
time-dep enden t functions represen ted as p oin t wise estimates of c hanges in hazard rates at
25
Study 1: P ath w a ys of Drug Use
v arious timep oin ts. These estimates w ere in tegrated to construct time-dep enden t step wise
functions of c hanges in cum ulativ e hazard whic h w ere plotted with p oin t wise confidence
in terv als and Hall-W ellner confidence bands. Significance testing of time-v arying effects
used suprem um tests applied o v er mo deled time ranges. Co efficien ts of constan t parame-
ters tak e the form of con v en tional p oin t estimates represen ting c hanges in hazard rates p er
y ear and significance testing w as done using standard significance testing using estimated
standard errors. A dditional details on the additiv e hazards mo del approac h can b e found
in A dditive hazar ds mo del (subsubsection 3.5.3, p. 18) .
An additiv e hazards regression mo del w as fitted to eac h outcome using the corresp onding
dataset. F or eac h mo del, the other non-outcome 10 illicit drugs use b eha viors and injec-
tion drug use comprised the set of drug use exp osure predictors and sp ecified with con-
stan t effects and demographic co v ariates w ere sp ecified with time-v arying. Mo dels v aried
on their comp osition of exp osure predictors, but all mo dels used the same set of demo-
graphic co v ariates.
4.3 Results
Descriptiv e statistics of demographic co v ariates can b e found in T able 3.1 and of drug
use b eha viors in T able 3.2 . Ov erall, the sample w as demographically div erse and partic-
ipan ts rep orted a broad div ersit y of drug use exp eriences. The most prev alen t t yp es of
drug use within the sample in v olv ed illicit drugs of heroin (93%; n = 754), cannabis (92%;
n = 748), p o wder co caine (89%; n = 721), crac k co caine (87%; n = 701), and metham-
phetamine (72%; n = 587). Among NMPDs, opioids w ere the most prev alen t (67%; n =
542), follo w ed b y tranquilizers (57%; n = 462). A minorit y of participan ts rep orted ev er
using NMPD stim ulan ts (21%; n = 172) or sedativ es (20%; n = 163). Methadone use w as
fairly common (42%; n = 344) and around three times more prev alen t than buprenorphine
use (15%; n = 119). Virtually all participan ts rep orted ev er engaging in injection drug use
(99.8%; n = 808).
Plots of Kaplan-Meier estimates for eac h outcome are pro vided with confidence in terv als
26
Study 1: P ath w a ys of Drug Use
Methadone
Methamphetamine NMPD Tranquilizers
Powder Cocaine NMPD Sedatives
Crack Cocaine NMPD Stimulants
Heroin NMPD Opioids
Cannabis Buprenorphine
0 20 40 60
0 20 40 60
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
Age (years)
Survival probability
Figure 4.1: Study 1: Kaplan-Meier surviv al curv es
27
Study 1: P ath w a ys of Drug Use
Methadone
Methamphetamine NMPD Tranquilizers
Powder Cocaine NMPD Sedatives
Crack Cocaine NMPD Stimulants
Heroin NMPD Opioids
Cannabis Buprenorphine
0 20 40 60
0 20 40 60
0.0
0.1
0.2
0.3
0.0
0.5
1.0
1.5
2.0
0.0
0.1
0.2
0.3
0.0
0.1
0.2
0.3
0
1
2
3
0
1
2
3
0
1
2
3
4
5
0
1
2
3
0
1
2
0.0
0.5
1.0
1.5
0.00
0.25
0.50
0.75
Age (years)
Cumulative hazard
Figure 4.2: Study 1: Kaplan-Meier cum ulativ e hazard curv es
28
Study 1: P ath w a ys of Drug Use
for surviv al curv es in Figure 4.1 and cum ulativ e hazards in Figure 4.2 . Surviv al curv es
v aried widely b et w een outcomes due to differences b et w een outcomes on prev alence and
distributions of initiation ages. Among outcomes with greater than 50% sample lifetime
prev alence, median surviv al time to initiation w as shortest for cannabis use (14 y ears)
and longest for methadone use (59 y ears). The initiation of commonly used illicit drugs
of cannabis, heroin, co caine, and methamphetamine all had median surviv al times of
less than 30 y ears. The accum ulation of hazards for initiation w as faster at earlier ages
for cannabis, p o wder co caine, and methamphetamine. This w as also the case for NMPD
opioids, stim ulan ts, and sedativ es but the absolute c hanges in cum ulativ e hazards w as
relativ ely small. Methadone and buprenorphine exhibited m uc h larger median surviv al
times, ha ving the largest among all opioid drugs. Both exhibited fairly stable cum ulativ e
hazard slop es, suggesting a constan t dev elopmen t of initiation risk throughout adultho o d.
NMPDs of opioids, sedativ es, stim ulan ts, and tranquilizers t ypically exhibited greater
median surviv al times and all exhibited steep er cum ulativ e hazard slop es b efore 20 y ears,
subsequen tly transitioning to flatter slop es.
4.3.1 Time-v arying effects of demographic co v ariates
Results from the additiv e hazards mo dels are presen ted with figures consisting of plots
of cum ulativ e co efficien ts for time-v arying non-parametric effects and p oin t estimates for
time-constan t parametric effects. Suprem um test p-v alues are pro vided in T able 4.1 for
time-v arying effects of the in tercept and demographic co v ariates across all mo dels, with
columns corresp onding to eac h mo deled outcome (a full table is a v ailable in T able A.1 ).
Plots for time-v arying effects are pro vided in separate figures for sex, sex orien tation,
and birthplace ( Figure 4.3 ), race ( Figure 4.4 ), and birth cohort ( Figure 4.5 ). Plots
sho w the accum ulated effects of asso ciated with a co v ariate o v er time and represen t the
time-dep enden t c hange in cum ulativ e hazard asso ciated with a co v ariate.
29
Study 1: P ath w a ys of Drug Use
T able 4.1: Study 1: Suprem um p-v alues for significance of time-v arying effects
Illicit drugs Medications
Co caine Drug trea tmen t Prescription
Cann. Heroin Crac k P o wder Metham. Methad. Bupren. Opioids Stim ul. Sedat. T ranq.
In tercept 0.000 0.000 0.000 0.000 0.000 0.257 0.140 0.000 0.000 0.000 0.002
Birth Cohort (referen t: Pre-Sixties)
Sixties 0.019 0.001 0.000 0.000 0.512 0.601 0.097 0.661 0.001 0.024 0.761
Sev en ties 0.075 0.001 0.000 0.003 0.000 0.210 0.052 0.084 0.145 0.005 0.500
Eigh ties or later 0.525 0.135 0.000 0.001 0.000 0.004 0.000 0.000 0.081 0.000 0.000
Sex (female) 0.004 0.160 0.451 0.902 0.043 0.115 0.585 0.580 0.314 0.234 0.829
Sexual orien tation (GLB) 0.488 0.002 0.540 0.318 0.482 0.527 0.306 0.331 0.101 0.692 0.634
Birthplace (foreign b orn) 0.000 0.000 0.071 0.000 0.027 0.024 0.816 0.000 0.033 0.000 0.002
Race (referen t: White)
Blac k 0.000 0.521 0.146 0.009 0.000 0.538 0.234 0.000 0.001 0.005 0.000
Latino 0.009 0.001 0.282 0.586 0.000 0.000 0.046 0.232 0.000 0.002 0.018
Other 0.467 0.423 0.298 0.014 0.004 0.132 0.052 0.161 0.209 0.002 0.261
Note:
P-v alues less then or equal to 0.05 are shaded in gra y .
30
Study 1: P ath w a ys of Drug Use
Demo gr aphic effe cts. Sex exhibited a significan t effect for cannabis and metha mphetamine
use outcomes with o v erall reduced risk of initiation for females v ersus males. GLB, v ersus
non-GLB, p ersons had consisten tly lo w er risk of heroin use initiation and this co v ariate ex-
hibited no significan t cum ulativ e effects for an y other outcome. The effects of foreign birth
v aried greatly on time but generally exhibited a reduction in risk. F oreign b orn partici-
pan ts had reduced risk for initiating NMPD opioids use o v er all ages; for initiating heroin
or p o wder co caine use up to age 30 y ears; and for initiating cannabis or NMPD sedativ es
or tranquilizers use up to around age 20 y ears.
Racial differences of time-dep enden t risk w ere significan t on at least one racial category for
all outcomes except for crac k co caine use initiation. With White as the referen t race cate-
gory , b eing Blac k w as generally asso ciated with lo w er initiation risks of cannabis, p o wder
co caine, methamphetamine, and all NMPDs o v er all ages. Being Latino w as significan tly
asso ciated with decreased and increased initiation risks dep ending on the outcome, with
generally increased risk for initiation of heroin or methadone use; generally decreased risk
for initiation of cannabis, methamphetamine, buprenorphine, or NMPD sedativ es or stim-
ulan ts use; and a reduced risk for initiating NMPD tranquilizers use up un til around age
20 y ears. Being a race other than White, Blac k, or Latino w as asso ciated with inconsisten t
effects on the risks of initiating either p o wder co caine or methamphetamine use and a re-
duced risk of initiating NMPD sedativ es use. In fact, all non-White racial categories w ere
significan tly asso ciated with reduced risks of NMPD sedativ e initiation.
Birth c ohort differ enc es. Compared to the Pre-Sixties cohort (i.e., those b orn b efore 1960),
later birth cohorts generally exhibited greater initiation risks. Three outcomes displa y ed
exceptions to this pattern. Heroin initiation risks w ere generally lo w est for the Sixties and
Sev en ties cohorts, higher for the Pre-Sixties cohort, and highest for the Eigh ties or later
cohort but this w as a non-significan t effect. Cohort differences in initiation risks for p o w-
der co caine w ere limited to b efore 30 y ears of age with similar risk increases for all co-
horts compared to the Pre-Sixties cohort. The Eigh ties or later cohort exhibited signifi-
31
Study 1: P ath w a ys of Drug Use
Intercept Sex (female)
Sexual orientation
(GLB)
Birthplace
(foreign born)
Cannabis Heroin
Crack
Cocaine
Powder
Cocaine
Metham. Methad. Bupren.
NMPD
Opioids
NMPD
Stimulants
NMPD
Sedatives
NMPD
Tranq.
0 20 40 60 0 20 40 60 0 20 40 60 0 20 40 60
−2
0
2
4
−1.0
−0.5
0.0
0.5
1.0
1.5
−0.50
−0.25
0.00
0.25
−1
0
1
2
−0.5
0.0
0.5
1.0
−0.3
−0.2
−0.1
0.0
0.1
0.2
−0.10
−0.05
0.00
0.05
0.10
−0.50
−0.25
0.00
0.25
0.50
0.0
0.2
0.4
−0.2
0.0
0.2
0.4
−0.2
0.0
0.2
0.4
Age (years)
Cumulative coefficients
Figure 4.3: Study 1 : Time-v arying effects for sex, sexual orien tation, and birth place
32
Study 1: P ath w a ys of Drug Use
Intercept Black Latino Other
Cannabis Heroin
Crack
Cocaine
Powder
Cocaine
Metham. Methad. Bupren.
NMPD
Opioids
NMPD
Stimulants
NMPD
Sedatives
NMPD
Tranq.
0 20 40 60 0 20 40 60 0 20 40 60 0 20 40 60
−2
0
2
4
−0.5
0.0
0.5
1.0
1.5
−0.50
−0.25
0.00
0.25
−1
0
1
2
−1.0
−0.5
0.0
0.5
1.0
−0.2
0.0
0.2
0.4
−0.1
0.0
0.1
−0.4
0.0
0.4
−0.2
0.0
0.2
0.4
−0.2
0.0
0.2
0.4
−0.4
−0.2
0.0
0.2
0.4
Age (years)
Cumulative coefficients
Figure 4.4: Study 1: Time-v arying effects for race
33
Study 1: P ath w a ys of Drug Use
Intercept Sixties Seventies Eighties or later
Cannabis Heroin
Crack
Cocaine
Powder
Cocaine
Metham. Methad. Bupren.
NMPD
Opioids
NMPD
Stimulants
NMPD
Sedatives
NMPD
Tranq.
0 20 40 60 0 20 40 60 0 20 40 60 0 20 40 60
−1
0
1
2
3
4
5
−1
0
1
2
3
0
1
2
3
0
1
2
0
1
2
3
−0.5
0.0
0.5
1.0
1.5
2.0
0.0
0.5
1.0
1.5
0.0
0.5
1.0
1.5
−0.25
0.00
0.25
0.50
−0.2
0.0
0.2
0.4
0.0
0.4
0.8
1.2
Age (years)
Cumulative coefficients
Figure 4.5: Study 1: Time-v arying e ffects for birth cohort
34
Study 1: P ath w a ys of Drug Use
can tly higher, and the highest, risk of initiating NMPD opioids or tranquilizers. The risk
of NMPD stim ulan ts use initiation app eared to decrease with eac h successiv e cohort.
4.3.2 Constan t effects of drug use predictors
T able 4.2 pro vides estimates of constan t effects of drug us e exp osure predictors with pre-
dictors organized b y ro w and outcomes b y column. These estimates represen t c hanges in
hazard rates, conditioned on exp osure. Significan t estimates are shaded in gra y and more
detailed tables for eac h mo del can b e found in T able A.2 . Effects of drug use predictors
v aried with man y effects exhibiting less than a 0.010 c hange in hazards p er y ear condi-
tioned on exp osure, whic h corresp onds to a c hange of 1 outcome ev en t p er 100 p erson-
y ears of exp osure. Although man y of these estimates w ere significan t, their effect sizes
w ere not large enough to b e meaningful. Significan t estimates with effects greater than or
equal to 0.010 w ere also considered substan tiv e and these v alues in the table are prin ted in
b old. The follo wing results and discussion fo cus on significan t and substan tiv e estimates
unless otherwise noted.
35
Study 1: P ath w a ys of Drug Use
T able 4.2: Study 1: P arametric estimates
Illicit drugs Medications
Co caine Drug treatmen t Prescription
Cann. Heroin Crac k P o wder Metham. Methad. Bupren. Opioids Stim ul. Sedat. T ranq.
Illicit drugs
Cannabis NA 0.034 0.028 0.038 0.017 0.002 -0.001 0.010 0.002 0.003 0.008
Heroin 0.011 NA -0.002 -0.011 -0.022 0.014 0.006 0.008 0.000 0.001 0.007
Crac k co caine 0.026 0.017 NA -0.009 0.009 0.000 0.000 0.001 -0.002 -0.004 0.002
P o wder co caine -0.006 0.007 0.024 NA 0.010 0.004 0.000 0.007 0.001 0.001 0.001
Methamphetamine 0.009 -0.023 -0.001 0.000 NA 0.003 0.003 0.011 0.001 0.001 0.004
Drug treatmen t medications
Methadone 0.011 0.039 0.006 0.009 0.002 NA 0.009 0.007 0.000 0.000 0.019
Buprenorphine -0.042 NA 0.008 -0.003 -0.004 0.070 NA 0.017 0.005 0.004 0.037
Prescription medications
Opioids 0.009 0.030 -0.002 -0.001 0.004 0.007 0.002 NA 0.004 0.004 0.018
Stim ulan ts -0.012 0.028 0.010 0.096 0.002 -0.003 0.005 0.005 NA 0.004 0.009
Sedativ es -0.025 0.004 0.002 0.005 0.004 0.005 -0.002 0.004 0.004 NA 0.000
T ranquilizers 0.036 0.030 0.003 0.006 0.004 0.009 0.005 0.021 -0.003 0.004 NA
Injection drug use -0.053 0.037 0.014 0.012 0.008 -0.007 -0.007 -0.011 0.000 -0.003 -0.007
Note:
Significan t estimates are shaded in gra y ( p-v alue <= 0.05).
Significan t estimates greater than or equal to 0.010 are prin ted in b old.
36
Study 1: P ath w a ys of Drug Use
Cannabis. Cannabis use exp osure w as a significan t risk factor for a lmost all outcomes with
substan tiv e effects for most outcomes. Sp ecifically , it significan tly and substan tiv ely in-
creased the hazards of initiating use of heroin (0.034), crac k co caine (0.028), p o wder co-
caine (0.038), methamphetamine (0.017), and NMPD opioids (0.010).
Opioids. Among opioid-t yp e drug use exp osure predictors, the largest effects w ere detected
for opioid-t yp e drug use initiation outcomes. Exp osure to NMPD opioid use (0.030) w as a
significan t risk factor for initiating heroin or NMPD tranquilizers use. Exp osures to heroin
(0.014) and buprenorphine (0.070) use w ere significan t risk factors for initiating methadone
use. Notably , NMPD opioid initiation w as the only opioid-t yp e outcome to not ha v e an y
other opioid-t yp e drug use exp osure as a significan t predictor. Exp osures to methadone
(0.019) and NMPD opioid (0.018) use w ere significan t risk factors for NMPD tranquilizer
initiation. Exp osure to NMPD opioid use also pro duced a significan t but not substan tiv e
increase risk for initiations of NMPD stim ulan ts (0.004) and sedativ es (0.004) use. Exp o-
sure to heroin use reduced the risk for initiating methamphetamine use.
Co c aine. Exp osures to p o wder (0.010) and crac k (0.009) co caine use w ere b oth signif-
ican t risk factors for methamphetamine initiation although the latter’s effect w as not
substan tiv e. Otherwise, b oth w ere significan t risk factors for differen t outcomes. P o wder
co caine use exp osure w as a significan t risk factor for initiating crac k co caine (0.024) use
and a significan t but not substan tiv e risk factor for initiating use of methadone (0.004)
37
Study 1: P ath w a ys of Drug Use
or NMPD opioids (0.007). Exp osure to crac k co caine use w as a significan t risk factor for
cannabis use initiation (0.026) and exhibited a significan t but not substan tiv e protectiv e
affect against NMPD sedativ e initiation (-0.004). The effect of crac k co caine use exp osure
(0.017) on heroin use initiation is w orth nothing as the effect size is substan tiv e and
p ossesses marginal significance (p-v alue=0.064).
Methamphetamine. Exp osure to methamphet amine use reduced the risk for initiating
heroin use (-0.023). Otherwise, it w as asso ciated with an increase in risk for initiating use
of NMPD opioids (0.011).
Non-me dic al pr escription drugs. Exp osure to NMPD stim ulan ts use increased the initia -
tion risk of p o wder co caine use (0.096) and exp osure to NMPD sedativ es decreased the
initiation risk of cannabis use (-0.025). Exp osure to NMPD tranquilizers significan tly in-
creased the risk for man y initiation outcomes but only substan tiv ely for initiation of use
for cannabis (0.036), heroin (0.030), and NMPD opioids (0.021).
Inje ction drug use. Exp osure to injection drug use w as a significan t ris k factor for heroin
use initiation (0.037). Otherwise, it reduced the initiation hazards for all other outcomes
in whic h it exhibited significan t effects. The largest reductions in hazards w ere for the ini-
tiation of cannabis (-0.054) and NMPD opioids use (-0.011), and all other significan t esti-
mates w ere not substan tiv e.
4.4 Discussion
The drug use careers of PWID include distinct a nd ordered longitudinal relationships
among a div ersit y of drug use b eha viors. Con trolling for the effects of individual and
historical con textual factors, exp osure of v arious drug use b eha viors do es app ear to signifi-
can tly increase risk of initiating other b eha viors. This risk is also cum ulativ e o v er exp osure
time: longer exp osures lead to greater accum ulated risk. Man y demographic co v ariate
effects w ere significan t with differences most apparen t b et w een racial categories and
birth cohorts. Generally , findings supp ort the indep enden t roles of drug, individual, and
con textual factors on drug use initiation risks. Relationships b et w een drug use b eha viors
38
Study 1: P ath w a ys of Drug Use
suggest the existence of patterns among drug use path w a ys reflecting differen t p ossible
causal mec hanisms.
The linkages b et w een b eha viors in v olving illicit use o f licit pharmaceutical drugs (i.e.,
NMPDs) and b eha viors in v olving the use of completely illicitly man ufactured drugs (e.g.,
heroin), constitute the predominan t pattern of drug use path w a ys. Initiation risk generally
flo w ed from NMPDs to other NMPDs or completely illicit drugs and from illicit drugs
to other illicit drugs. The risk relationships from illicit drugs to NMPDs w ere either
non-significan t or insubstan tial. The connection b et w een NMPD use on future use of
completely illicit drugs has already b een established for NMPD opioid use on heroin use
( Cerdá, San taella, Marshall, Kim, & Martins, 2015-09 ; Cicero, Ellis, Surratt, & Kurtz,
2014 ; Compton, Jones, & Baldwin, 2016 ) but has not b een w ell-describ ed for other NMPD
predictors and illicit drug use outcomes. Sp ecifically , the connection found here b et w een
NMPD stim ulan ts use exp osure on future initiation risk of p o wder co caine w arran ts
further study on the role of NMPDs within drug use tra jectories.
Another general pattern w ere the asso ciations of drug use predictors and outcomes based
on drug effect. Stim ulan t drugs tended to increase the risk of other stim ulan t drugs.
Crac k and p o wder co caine use, in addition to cannabis use, w ere significan t predictors of
methamphetamine initiation. Similarly , opioid drugs increased the risk of initiating other
opioid drugs. The ma jor exception to this w as the risk con tribution of crac k co caine use
on heroin initiation, whic h follo ws other observ ations of a relationship b et w een heroin and
co caine use ( Leri, Bruneau, & Stew art, 2003-01 ).
The risk relationships b et w een co caine and methamphe tamine use suggest m ultiple o v er-
lapping drug use path w a ys. The div ergen t and somewhat disconnected roles of crac k and
p o wder co caine on initiation risks supp ort the idea of treating them as fundamen tally dif-
feren t b eha viors p ositioned at v ery differen t p ositions within drug use tra jectories with dif-
feren t asso ciated factors ( Joseph J. P alamar, Da vies, Ompad, Cleland, & W eitzman, 2015 ;
Joseph J. P alamar & Ompad, 2013 ). P o wder co caine use exp osure increased the risk of
crac k co caine initiation, but this relationship did not op erate in the opp osite direction. Al-
39
Study 1: P ath w a ys of Drug Use
though marginally significan t, the strong effect of crac k co caine use exp osure on the risk of
heroin use initiation suggests that there ma y b e a larger flo w of risk from p o wder co caine
use, then to crac k co caine use, and finally at heroin use.
The m utually risk reducing effect b et w een heroin and m ethamphetamine use implies
that there are drug use path w a ys define primarily b y these t w o b eha viors. Exp osure to
methamphetamine use reduced the risk of heroin use initiation and exp osure to heroin use
reduced the risk of methamphetamine use initiation. In con text with the other results, it
app ears that there are t w o primary “trac ks” of drug use consisting of either opioid drug
use b eha viors or stim ulan ts drug use b eha viors.
Risk relationships b et w een drugs of similar effects can b e caused b y individual seeking b e-
ha viors. Exp erience with one drug can inform the initiation of another with similar quali-
ties. T ransitions b et w een differen t t yp es of NMPDs ma y b e motiv ated b y safet y and con-
v enience. Prescription drugs follo w regulations for their licit man ufacture with implied
guaran ties of consisten t dosage and purit y and man y are con v enien tly pac kaged as pills.
NMPDs, compared to illicitly man ufactured drugs, offer a relativ ely safer and more con-
v enien t form of drug use. T ransitions from NMPD use to the use of illicit drugs ma y b e a
consequence of limited licit or illicit supply of prescription medications, increased individ-
ual need due to drug tolerance, or c heap er a v ailabilit y of illicit drugs. This supp orts previ-
ous findings on transitions from NMPD opioids to heroin use ( Mars, Bourgois, Karandinos,
Mon tero, & Ciccarone, 2014 ). T ransitions b et w een illicit drugs can in v olv e similar consid-
erations regarding the pricing p er drug dose or effect. Compared to p o wder co caine, crac k
co caine is c heap er and pro vides a faster and more in tense effect, and transitions from p o w-
der to crac k co caine ma y b e motiv ated b y a desire to ac hiev e similar effects but through
c heap er or faster metho ds.
Individual seeking b eha viors do not en tirely explain transitions, ho w ev er. Seeking out a
drug use b eha vior do es not necessarily correlate with access to the materials or kno wledge
necessary for the b eha vior. And not all initiations are the result of individuals seeking out
a drug use b eha vior. One p ossible complemen ting and synergetic pro cess is so cial exp osure
40
Study 1: P ath w a ys of Drug Use
( Arria et al., 2008 ; Benjet et al., 2007 ). So cial en vironmen ts with prev alen t drug use pro-
vide more opp ortunities for initiation. These en vironmen ts also pro vide access to the ma-
terials and kno wledge of drug use. Drug-related so cial relationships pro vide a critical role
in drug use as virtually all illicit access to the materials of drug use is mediated through
so cial ties. These relationships are not static and can emerge or strengthen as a result of
drug use. The use of one drug can lead to greater so cial exp osure and initiation opp ortuni-
ties of another.
So cial exp osure can explain some of the transitions to and among b eha viors in v olving the
use of completely illicit drugs. This explanation relies on the fundamen tal differences b e-
t w een NMPDs and illicit drugs. Licit access to prescription drugs do es not require so cial
net w orks whereas illicit access do es. Consequen tly , licit access to prescription drugs do not
correlate with access to other t yp es of drugs but an illicit source of NMPDs can also b e
the source of other illicit drugs. T ransitions from licit to illicit use of prescription drugs
do not require drug-related so cial relationships but the initiation of NMPD use ma y moti-
v ate the dev elopmen t of these relationships, promoting so cial exp osures to other t yp es of
drug use, including the use of illicit drugs. Once this so cial net w ork has b een established,
drug-related relationships and drug use b eha viors can b ecome m utually reinforcing. This
ma y explain the role of injection drug use on the initiation risks of heroin. Heroin is a com-
monly injected drug among PWID ( No v ak & Kral, 2011 ) and initiating injection drug use
ma y result in greater so cial con tact with PWID who inject heroin, whic h can lead to so cial
opp ortunities to initiate heroin use.
In addition to p oten tial explanation of the path w a ys among drug use b eha viors, the find-
ings inform some p ossible explanations for the start of a drug use career. Results of the
analysis supp orts the role of cannabis as a gatew a y drug; ho w ev er, its role is limited and
ma y b e sp ecific to con text ( L. Degenhardt et al., 2009 ; Jorgensen & W ells, 2021 ; Morral,
McCaffrey , & P addo c k, 2002 ). Cannabis use exhibited a substan tial effect on the initia-
tions of heroin, p o wder and crac k co caine, and methamphetamine, all of whic h are com-
pletely illicit. It did not exhibit an y appreciable effect on most NMPD outcomes, except
41
Study 1: P ath w a ys of Drug Use
that for NMPD opioids use initiation. The significan t and substan tial role of NMPDs on
illicit drug outcomes suggests a parallel or in teracting role with cannabis use. It is p os-
sible that b oth NMPDs and cannabis are gatew a y drugs with the p ossibilit y of o v erlap.
Sp ecifically , the risk con tributions of NMPD opioids, stim ulan ts, and tranquilizers w arran t
further in v estigation of these drugs on the b eginning of the drug use career and drug use
tra jectories. Alternativ ely , the role of cannabis use ma y not indicate an y sp ecific effect at-
tributable to cannabis use but rather represen t the start of the drug use career. Cannabis
is commonly one of the first illicit drugs to b e used and the initiation of its use ma y signify
the b eginning of general exp osure to illicit drug use. Regardless of its sp ecific meaning, its
role within the drug use career and for illicit drug use in general ma y b e diminishing. On-
going efforts to legalize recreational cannabis use means that cannabis is increasingly b e-
coming a licit drug. As a licit drug sold at regulated legal v en ues, its connection to other
illicit drugs will b ecome atten uated. Whether this dev elopmen t will lead a diminished role
for cannabis as a gatew a y drug is unkno wn. In comparison, prescription medication con tin-
ues to b e widely a v ailable, and the use of other illicit drugs remains prev alen t with no sign
of abatemen t.
These trends are not univ ersal, whic h sp eak to the n eed for tailored in terv en tions. Risks
w ere differen t b et w een racial categories for a n u m b e r of initiation outcomes, and it is lik ely
comm unities based on race and ethnicit y exhibit differen t drug use path w a ys. The reduced
initiation risk of NMPD opioids use asso ciated with Blac k, v ersus White, participan ts sug-
gests that NMPD opioid use is less lik ely to b e a cause for heroin initiation for this group.
Lik ewise, the lo w er initiation risk of NMPD stim ulan t initiation asso ciated with Latino
participan ts implies that they ma y not exp erience enough exp osure to NMPD stim ulan ts
use for it to b e a ma jor risk factor for p o wder co caine initiation. These differences in
NMPD use initiation risks ma y reflect differences in exp osures to prescription medications,
p ossibly a result of disparate access to prescription medications. A dditionally , generational
differences in initiation risks, whic h w ere generally higher among the 1980s or later cohort,
suggest an urgency to address the incidence of drug use b eha viors among y ounger PWUD.
42
Study 1: P ath w a ys of Drug Use
Latino participan ts generally exhibited the highest risk for initiating heroin use, after
adjusting for other factors. This ma y indicate an asso ciation b et w een Latino comm unities
and heroin use suc h as endemic heroin use. While, on the whole, the prev alence of illicit
use of opioids among Latino p eople app ears to b e similar to the national prev alence
( Abuse & A dministration, 2020b ), adolescen t illicit opioid use b eha viors app ea r to b e
more prev alen t among Latino studen ts in urban high sc ho ols compared to their Blac k or
White coun terparts ( A. A. Jones et al., 2019 ). A dditionally , Latino participan ts of opioid
treatmen t programs app ear to rep ort a far higher prev alence of heroin use v ersus other
opioid drug use b eha viors ( Enrique R. P ouget, F ong, & Rosen blum, 2017 ). The findings
of this study lend supp ort for the existence of a distinct risk for heroin use initiation
exp erienced b y Latino p opulations.
Buprenorphine and methadone use, b oth used primarily within opioid drug treatmen ts
settings, exhibited trends quite differen t to NMPD opioids use. Unlik e other prescrib ed
opioids, buprenorphine and methadone are primarily used for drug treatmen t and not for
pain managemen t or for other medical purp oses. Most methadone is legally disp ensed on-
site at drug treatmen t programs and not through retail pharmacies whereas buprenorphine
can b e legally a v ailable through pharmacies but, as a new er treatmen t medication, it has
lo w er general a v ailabilit y and is sub ject to greater disparities in a v ailabilit y based on race,
income, and region ( Hansen, Siege l, W anderling, & DiRo cco, 2016 ). Legal exp osures to
these t w o drugs are conditional on a diagnosis of opioid use disorder, whic h o ccurs after ex-
p osures to other opioid drug use exp eriences. P eople who use methadone or buprenorphine
outside of a drug treatmen t setting are lik ely using them for drug treatmen t purp oses: to
alleviate withdra w al symptoms asso ciated with self-initiated opioid use cessation ( Allen &
Haro cop os, 2016 ; Butler, Oy edele, Go v oni, & Green, 2020 ; Daniulait yte et al., 2019 ). Ex-
p osures to the use of these drugs more lik ely indicate self-treatmen t of opioid use or exp o-
sure to opioid treatmen t programs than reflect an underlying risk exp osure pro cess. This
can explain the apparen t risk con tribution from buprenorphine use exp osure on methadone
use initiation and vice v ersa, as w ell as the absence of an y con tribution of methadone or
buprenorphine use exp osures to w ards the risk on all other opioid drug use initiation out-
43
Study 1: P ath w a ys of Drug Use
comes. The risk con tribution b y methadone use exp osure on NMPD tranquilizers use ini-
tiation suggests that tranquilizers ma y also b e used to treat or manage opioid withdra w al
but the role of NMPD tranquilizers use exp osure on initiations of use for cannabis, heroin,
or NMPD opioids, among others, indicate that NMPD tranquilizers use within the drug
use career p ossesses m ultiple roles within the drug use career.
This study pro vides an expansiv e view of drug use path w a ys b y incorp orating a broad set
of drug use b eha viors for a div erse sample of PWID and discusses p ossible causal mec ha-
nisms. A discussion of limitations applicable to this and all other studies comprising this
dissertation is pro vided in Over al l Limitations (subsection 7.1, p. 95) . Findings suggest
that some of these tra jectories ma y b e the result of individual or so cial pro cesses. These
causal explanations are neither new nor groundbreaking and deserv e con tin ued atten tion
as p oten tially promising targets of in terv en tion. F urther and con tin uous study ma y b e re-
quired to main tain a curren t understanding of drug use as the drug use tra jectories as the
drug use epidemic con tin ues and ev olv es. This understanding m ust tak e in to consideration
v arious con texts and reflect the div ersit y of PWUD and PWID.
44
Study 2: P ath w a ys to Injection Drug Use Initiation
5 Study 2: P ath w a ys to Injection Drug Use Initiation
Injection drug use con tin ues to b e a substan tial and p ersisten t public health issue among
p eople who use drugs (PWUD). Although a minorit y of PWUD are p eople who inject
drugs (PWID), they shoulder an outsized burden of the health consequences of drug
use. Man y of the harms asso ciated with using drugs are magnified among PWID, who
exp erience higher morbidities of drug dep endency , o v erdose, and drug-related men tal
health problems and higher o v erall mortalit y . This is in addition to the acute and c hronic
health consequences sp ecifically caused b y injecting drugs.
The causes of injection drug use initiation are n umerous and o v erlapping but can b e
broadly categorized in to three domains: individual, con textual, and drug factors ( Zin-
b erg, 1986 ). These factors can influence risk pro cesses at the individual or so cial lev el.
Individual factors, including demographic attributes, influence individual exp eriences and
iden tities, including so cial iden tities, whic h in turn affect individual c hoices and c hances
to initiate injection. Injection initiation, in this case, is a consequence of individual
injection drug use prop ensities in teracting with individual injection drug use access, b oth
determined b y individual factors.
Con textual factors, expressed through generational c hanges in injection initiation risk,
constrain or promote access, a v ailabilit y , or acceptabilit y of injection drug use b eha viors
through larger so cial phenomena. So cial trends determine not only the incidence of injec-
tion initiation but also its underlying comp osition. Although injection drug use is a singu-
lar b eha vior defined b y a sp ecific route of administration, it in fact describ es a div ersit y of
b eha viors in v olving the injection of differen t drugs. Generational differences in injection
initiation ma y b e a result of risk exp osures reflecting so cial trends for injection drug use as
w ell as sp ecific t yp es of drug use.
Finally , drug-related factors affect drug use exp eriences and drug use so cial net w orks, af-
fecting individual drug use preferences and so cial exp osure to drug use. Man y PWID initi-
ate injection after at least some illicit drug use exp erience and this prior exp erience ma y
45
Study 2: P ath w a ys to Injection Drug Use Initiation
indicate underlying dev elopmen tal pro cesses leading to injection initiation. Some prior
drug use exp osures ma y increase susceptibilit y of drug dep endency or lik eliho o d of accel-
erating use, leading to greater p erceiv ed adv an tages of using drugs via injection, whic h is
more effectiv e route. A dditionally , the relativ e adv an tages of injection o v er other routes
ma y v ary b e more apparen t for the administration of some drugs in order to maximize ef-
fect or to minimize costs, th us making a transition to injection more lik ely .
There are also so cial pro cess con tributing to injection initiation risk that are dep enden t
on drug-related factors. Drug use b eha viors represen t not only individual health b eha v-
iors but also drug use so cial net w orks, including the p eople who supply the illicit drug and,
optionally , other PWUD engaging in similar b eha viors. As there are no legal v en ues to ac-
quire or use illicit drugs, these net w orks are critical to accessing the materials necessary
for initiating and main taining drug use b eha viors. The prev alence of injection drug use
v aries b y the drug injected and so cial net w orks orien ted around a particular t yp e of drug
use can presen t differen t exp osure opp ortunities to injection drug use. Exp osure to a par-
ticular t yp e of drug use can pro vide additional so cial opp ortunities for injection initiation.
Evidence suggests that these so cial opp ortunities are imp ortan t mec hanisms for injection
initiation. So cial in teractions b et w een PWID and non-injecting PWUD within drug use
so cial net w orks ma y b e imp ortan t a critical conduit of the so cial propagation of injection
drug use ( Khobzi et al., 2009 ; Neaigus et al., 2006 ). As injection is m uc h more tec hnically
sophisticated than other routes of administration, learning to inject ma y dep end more on
direct so cial transmission v ersus passiv e observ ation or self-learning. Injecting effectiv ely
while minimizing harm has a lo w er tolerance of e rror compared to other routes of admin-
istration and dev eloping the required comp etency can dep end on access to the relev an t
kno wledge and exp erience pro vided through PWID mem b ers of the drug use so cial net-
w ork. Prior studies ha v e iden tified so cial in teractions b et w een PWID and non-injecting
PWUD as an imp ortan t mec hanism for propagating injection drug use b eha viors ( Bluthen-
thal et al., 2014 , 2015 ; Neaigus et al., 2006 ) and in terv ening on the so cial linkages b et w een
b eha viors ma y disrupt the propagation of risk ( Strik e et al., 2 014 ). As drug use so cial net-
46
Study 2: P ath w a ys to Injection Drug Use Initiation
w orks are somewhat organized around drug use b eha v iors, these b eha viors ma y b e con trib-
utors to injection initiation risk. Understanding the role of drug use b eha viors on injection
risk can pro vide additional insigh t on the mec hanisms of so cial transmission and help de-
lineate the dev elopmen tal path w a ys leading to injection drug use.
5.1 Bac kground
5.1.1 Injection drug use in the United States
Injection drug use is highly stigmatized among the general p opulation and among PWUD.
Estimates on its prev alence v ary due to its hidden nature. Based on data from treatmen t,
HIV testing, and HIV/AIDS diagnoses from 96 large metrop olitan areas, Brady et al.
( 2008-05 ) estimated the n um b er of past-y ear PWID i n the United States to b e around
1.5 million p eople aged 15 to 64 in 2002 with a median regional prev alence of 96 p er
10000 p ersons. Using the same sources of data, T empalski et al. ( 2013 ) estimated past
y ear PWID to b e around 1.5 million p eople aged 15 to 64 in 2007 with a median regional
prev alence of 104 p er 10000 p ersons. A cross 95 large metrop olitan areas in the United
States in 2002, Hannah L. F. Co op er et al. ( 2008 ) estimated the median regional past y ear
prev alence of PWID to b e 156 p er 10000 Blac k p ersons and b et w een 86 and 97 p er 10000
White p ersons aged 15 to 64 and Chatterjee et al. ( 2011 ) estimated the median regional
past y ear prev alence among adolescen ts and y oung adults aged 15 to 29 to b e 102 p er
10000. A cross 96 large metrop olitan areas in the United States in 2002, Enrique R. P ouget,
F r iedman, Cleland, T empalski, & Co op er ( 2012 ) estimated the median regional past y ear
prev alence of PWID to b e 133 p er 10000 Hispanic p ersons aged 15 to 64. A meta-analysis
b y Lansky et al. ( 2014 ) estimated lifetime PWID prev alence to b e around 6.6 million
p eople aged 13 or older in 2011, represen ting 2.6% of this p opulation, and past y ear PWID
prev alence to b e around 770000 or 0.30%. Analysis of YBRSS data estimated 2% of high
sc ho ol studen ts injected an y drug o v er the lifetime in 2015 ( Kann et al., 2016 ).
Recen t trends of injection drug use suggest that it is a gro wing problem. Up to 2002, the
estimated prev alence of PWID app ears to ha v e increased among adults aged 15 to 64
47
Study 2: P ath w a ys to Injection Drug Use Initiation
since 2000 ( Brady et al., 2008-05 ), to ha v e decrea sed among Hispanic ( Enrique R. P ouget,
F r iedman, Cleland, T empalski, & Co op er, 2012 ) and Blac k ( Hannah L. F . Co op er et al.,
2008 ) adults since 1992, to ha v e unc hanged among White adults since 1992 ( Hannah L.
F. Co op er et al., 2008 ), and to ha v e increased among p ers ons aged 15 to 29 from since
1995 ( Chatterjee et al., 2011 ). Bet w een 2002 and 2007, o v erall prev alence app ears to ha v e
unc hanged ( T empalski et al., 2013 ). Ho w ev er, analysis of admissions to substance use
disorder treatmen t facilities found admissions attributed to injection drug use increased b y
76% b et w een 2002 and 2014 ( Zibb ell et al., 2018 ). These admissions comprised 22% of all
admissions in 2014 compared to 13% in 2004. Among prescription opioid use treatmen t
admissions, injection use of opioids increased from 12% to 18% from 2004 to 2013 ( C. M.
Jones, Christensen, & Gladden, 2017 ).
5.1.2 Consequences of injection drug use
The p oten tial b enefits of injection, including faster and stronger drug effects and a higher
effect to dose ratio, are temp ered b y the high risks for sev ere acute or c hronic injury or
death. PWID are at greater risk for man y ma jor adv erse health consequences. They t ypi-
cally exhibit greater psyc hological and ph ysical dep endence on their drug use resulting in
a greater risk of substance use disorder diagnoses ( T orrens, Gilc hrist, Domingo-Salv an y ,
& Group, 2011 ). Compared to the general p opulation, PWID exp erience higher mortal-
it y o v erall and sp ecifically for o v erdose and AIDS-related causes ( Mathers et al., 2013 ).
Compared to other common routes of administration, injection drug use is particularly
harmful and lik ely the most harmful. The pro cess of injection t ypically o ccurs in an un-
sterile en vironmen t using unsterile equipmen t and uses unsterile drug con taminated with
impurities. Unclean injection sites, unclean injection equipmen t, or impure or biologically
con taminated drugs can lead to skin infections or abscesses whic h ma y lead to larger sys-
temic infections ( Hannah L. F. Co op er et al., 2007 ; Keeshin & F ein b erg, 2016 ; W urcel et
al., 2016 ).
The ma jor long-term consequences of injection that con tin ue to b e ma jor public health is-
sues are the greatly increased risk of con tracting blo o db orne illnesses suc h as HBV (hep-
48
Study 2: P ath w a ys to Injection Drug Use Initiation
atitis B virus), HCV (hepatitis C virus), and HIV (h uman imm uno deficiency virus) due to
the sharing of used injection equipmen t. In 2007, HCV incidence w as 26.7 p er 100 p erson-
y ears of observ ation (26700 p er 100000 PWID) ( P age et al., 2009 ). Within a cohort of
PWID follo w ed b et w een 2005 and 2008, cum ulativ e HCV incidence w as 7.8 cases p er 100
p erson-y ears (ann ualized incidence of 78000 p er 100000 PWID), follo wing a decreasing
trend ( Meh ta et al., 2 011 ) compared to the ann ual incidence of acute HCV infection in
the United States w as 0.7 cases p er 100000 p opulation in 2014 ( Zibb ell et al., 2018 ). The
estimated HIV diagnosis rate in 2011 w as 55 p er 100000 PWID ( Lansky et al., 2014 ). In
comparison, the ann ual incidence of diagnosis HIV infection in the United States w as 13.5
p er 100000 p ersons in 2011 and 12.3 p er 100000 p ersons in 2016 ( Disease Con trol & Pre-
v en tion, 2017 ). Appro ximately 6% of HIV diagnoses made in 2016 w ere among PWID. Vir-
tually no non-injecting PWUD con tract HCV or HIV via their drug use b eha viors in com-
parison.
5.1.3 Risk factors for initiating injection drug use
Demographically , PWID tend to b e older, unemplo y ed, and p ossess lo w er educational at-
tainmen t o v erall than non-injecting PWUD ( No v ak & Kral, 2011 ). F urthermore, some de-
mographic c haracteristics app ear to in teract with drug use b eha viors. An in v estigation b y
Keen, Khan, Clifford, Harrell, & Latimer ( 2014 ) of a cohort of PWUD in Baltimore found
racial differences on route of administration of co caine and heroin with Blac k PWUD more
lik ely to use crac k co caine and heroin via non-injection routes and White PWUD more
lik ely to use heroin or m ultiple drugs via injection. Race also app ears to b e asso ciated
with age at injection initiation ( McLaughlin, Ama y a, Klev ens, O’Cleirigh, & Batc helder,
2020 ). Injection NMPD opioid use seems to b e higher among male or white individuals
( C. M. Jones, 2018 ). So cio economic status and living situation also app ear to b e factors
of injection drug use b eha viors, Lin ton, Celen tano, Kirk, & Meh ta ( 2013 ) found homeless-
ness to b e a factor for relapse in to injection drug use b eha vior for previously former PWID
within a longitudinal cohort of activ e and former P WID in Baltimore from 2005 to 2009.
Ho w ev er, they also found that homelessness w as not asso ciated with sustaining curren t in-
49
Study 2: P ath w a ys to Injection Drug Use Initiation
jection drug use b eha vior.
Larger so cial or historical trends ha v e b een implicated with c hanging epidemiology in in-
jection drug use . These factors t ypically exert mo derating or indirect effects on individual
injection outcomes. Injection drug use is a b eha vior that is t ypically learned via direct or
indirect so cial exp osure to or in teraction with PWID and the p opulation of PWID will-
ing to initiate others in to injection app ears to b e a ma jor factor in injection drug use inci-
dence ( Bluthen thal et al., 2014 ). Qualitativ e studies supp ort the role of so cial in teraction
and con text to w ards injection drug use initiation ( Guise, Horyniak, Melo, McNeil, & W erb,
2017 ; Haro cop os, Goldsam t, K obrak, Jost, & Clatts, 2009-07 ; Mars, Bourgois, Karandi-
nos, Mon tero, & Ciccarone, 2014 ). Generational con text can also affect injection incidence
through c hanges in drug p opularit y of a v ailabilit y . The increase of non-medical prescrip-
tion opioid use b eginning in the mid-nineties app ears to b e asso ciated with an increase in
injection drug use ( C. M. Jones, 2018 ; Zibb ell et al., 2018 ).
Drug use b eha viors are lik ely a risk factors for injection drug use initiation. Within the
general p opulation, injection NMPD opioid use seems to b e higher among those who ha v e
rep orted past y ear substance use disorders for co caine, heroin, or opioids ( C. M. Jones,
2018 ). Among y outh PWUD, recen t methamphetamine use predicted subsequen t lifetime
injection initiation ( Dan W erb, Kerr, et al., 2013 ; W o o d et al., 2008 ). The o v erall timing
of injection initiation ma y dep end on the start of the drug use career with earlier career
starts leading to earlier injection initiations ( McLaughlin, Ama y a, Klev ens, O’Cleirigh, &
Batc helder, 2020 ).
Otherwise, study on the role of drug use b eha viors on injection initiation risk has b een
more limited compared to study on the roles individual or con textual factors. A clearer
understanding of the asso ciations b et w een pre-injection b eha viors on injection initiation
has the p oten tial of yielding further insigh t on the pro cess of injection initiation. Because
drug use histories, comprising what drugs w ere used and when usages b egan, pro vide the
most direct observ ations of the ev olution of the drug use career and these mark ers can b e
used to delineate path w a ys of injection initiation risk.
50
Study 2: P ath w a ys to Injection Drug Use Initiation
The primary aim of this study is to describ e the longitudinal relationships b et w een a large
and div erse set of pre-injection drug use b eha viors and injection initiation risk within a
sample of PWID. This w as examined for all-cause injection initiations, whic h can in v olv e
an y drug, and for cause-sp ecific injection initiation based on the drug used at first injec-
tion. Secondarily , the age-dep enden t risk of injection initiation and differences on this
asso ciated with demographic traits and birth cohorts are also in v estigated. These are all
accomplished using surviv al mo dels using additiv e hazards with time-dep enden t and con-
stan t effects.
5.2 Metho ds
5.2.1 Sample and measures
Analyses relied on the full 810 partic ipan ts comprising the surv ey data describ ed in Par-
ent study (subsection 3.1, p. 11) . P articipan ts pro vided information on their applicable
drug use exp eriences of injection drug use and for v arious illicit drugs. F or injection drug
use, participan ts rep orted whether they ev er engaged in injection drug use and, if so, their
ages at first injection and the drug used at first injection. Drug use histories w ere collected
for 11 t yp es of illicit drugs: (1) cannabis (referred to as “marijuana” b y the study); (2)
crac k and (3) p o wder co caine; (4) methamphetamine; (5) heroin; (6) methadone and (7)
buprenorphine (used outside of treatmen t settings or directions); and non-medical prescrip-
tion drug (NMPD) use of (8) opioids (not including methadone or buprenorphine, e.g.,
Vico din or Oxycon tin), (9) tranquilizers ( e.g., Klonopin or V alium), (10) sedativ es (e.g.,
Restoril or phenobarbital), and (11) stim ulan ts (e.g., Ritalin or A dderall). F or eac h drug,
participan ts pro vided information on if they ev er used the drug and, if applicable, at what
ages they first used the drug via non-injection and injection routes.
Data also included demographic information for age at time of surv ey , biological sex at
birth, birth y ear, sexual iden tit y (heterosexual/straigh t, ga y , lesbian, or bisexual), birth-
place (United States v ersus other), and race. Birth y ear w as transformed in to categories
mostly corresp onding to decadal birth cohort. Sexual iden tit y w as dic hotomized in to het-
51
Study 2: P ath w a ys to Injection Drug Use Initiation
erosexual v ersus ga y , lesbian, or bisexual. A dditional details on the measures used can b e
found in Me asur es (subsection 3.2, p. 12) .
The analytic dataset w as generated using first injec tion drug use exp erience as the out-
come ev en t of in terest with drug use exp osures as predictors of in terest and demographic
trait v ariables as co v ariates. The observ ation p erio d for eac h participan t b egan at 0 y ears
(birth) and ended at the age of first injection (ev en t) or at age at surv ey (censored). Demo-
graphic trait co v ariates p ossessed constan t v alues o v er the en tire observ ation p erio d. Drug
use exp osures, ho w ev er, w ere op erationalized as time-dep enden t binary v alues based on
initiation age. F urther details on this pro cess are pro vided in T emp or al or dering and time-
dep endent values (subsubsection 3.5.1, p. 17) .
5.2.2 Surviv al analysis
Analysis consisted of surviv al estimatio n of injection initiation using Kaplan-Meier estima-
tion follo w ed b y surviv al regression analysis using the additiv e hazards mo del. This mo del
pro vides relativ e adv an tages of flexibilit y and non-reliance on the prop ortional hazards as-
sumption o v er more con v en tional prop ortional hazards mo dels. These prop erties are de-
riv ed from its sp ecification of hazards on the additiv e scale and abilit y to include b oth non-
parametric time-v arying effects and parametric time-constan t effects in the same regres-
sion mo del. Estimated time-dep enden t effects tak e the form of time-dep enden t co efficien t
functions as time-based p oin t wise estimates of c hange in hazard rates. These estimates
w ere used to pro duce estimates of c hanges in c um ulativ e hazards, whic h w ere plotted with
p oin t wise confidence in terv als and Hall-W ellner confidence bands. Significance testing of
time-v arying effects consisted of suprem um tests o v er the mo deled time range. Estimates
constan t effects tak e the form of con v en tional p oin t estimates and standard errors, repre-
sen ting c hanges in the hazard rate. The significance of these estimates w as assessed using
standard tests using estimated standard errors. A dditional details on the additiv e hazards
mo del are pro vided in A dditive hazar ds mo del (subsubsection 3.5.3, p. 18) .
An y-cause injection initiation w as mo deled u sing the additiv e hazards mo del using the 11
52
Study 2: P ath w a ys to Injection Drug Use Initiation
T able 5.1: Study 2: Descriptiv e statistics of drug use b eha viors b y drug un til injection initiation
or surv ey
Sample prev alence Age at initiation
Coun t Prop ortion (%) Mean SD
Illicit drugs
Cannabis 668 82.47 13.11 3.26
Heroin 102 12.59 19.97 7.49
Crac k co caine 182 22.47 20.99 7.12
P o wder co caine 274 33.83 17.62 5.18
Methamphetamine 189 23.33 18.46 6.42
Drug treatmen t medications
Methadone 31 3.83 22.71 7.73
Buprenorphine 6 0.74 24.00 5.83
Prescription medications
Opioids 160 19.75 17.27 6.11
Stim ulan ts 67 8.27 14.99 5.27
Sedativ es 65 8.02 16.91 5.61
T ranquilizers 122 15.06 18.08 6.78
drug use exp osure predictors and demographic co v ariates as indep enden t terms. Exp osure
predictors w ere sp ecified with time-v arying effects and demographic co v ariates w ere sp eci-
fied with constan t effects.
Separately , another set of mo dels examined cause-sp ecific injection initiation b y first drug
injected. These mo dels represen t comp eting outcomes with eac h sp ecific outcome individu-
ally mo deled. F or eac h cause-sp ecific outcome, only injection initiations in v olving the first
injection of the outcome drug w ere considered to b e ev en ts with all other outcomes cen-
sored. Because of lo w coun ts of man y drug-sp ecific injection initiation outcomes, only the
three most commonly drugs used at first injected w ere mo deled, consisting of heroin, p o w-
der co caine, and methamphetamine. These three drugs w ere presen t at injection initiation
for 89% of the sample. Due to comp eting risks, the in terpretations of these mo dels require
sp ecial considerations ( Andersen, Geskus, Witte, & Putter, 2012 ).
53
Study 2: P ath w a ys to Injection Drug Use Initiation
5.3 Results
Descriptiv e statistics of demographic co v ariates can b e found in T able 3.1 with a descrip-
tion in Sample (subsection 3.3, p. 13) . Of the 810 participan ts included in the analyses,
808 (99.8%) rep orted ha ving initiated injection drug use at some p oin t in their lifetimes
at time of surv ey . Of those who initiated, 477 (59%) rep orted using heroin at first injec-
tion, 171 (21%) rep orted using methamphetamine, and 81 (10%) rep orted using p o wder co-
caine. T able 5.1 pro vides descriptiv e statistics of drug use b eha viors initiated kno wn to o c-
cur prior to injection initiation or, for the participan ts who nev er initiated injection (n=2),
prior to time at surv ey . Cannabis use w as the most frequen tly rep orted drug use b eha vior
(82%), follo w ed b y p o wder co caine use (34%). Around a fifth of the sample rep orted using
crac k co caine (22%), methamphetamine (23%), or NMPD opioids (20%). Less than a fifth
rep orted using heroin (13%) or NMPD tranquilizers (15%); less than a ten th rep orted us-
ing NMPD stim ulan ts (8%) or sedativ es (8%); and less than 5% rep orted using methadone
(4%) or buprenorphine (1%).
0.00
0.25
0.50
0.75
1.00
0 20 40 60
Age (years)
Survival probability
Figure 5.1: Study 2: Kaplan-Meier surviv al curv e
54
Study 2: P ath w a ys to Injection Drug Use Initiation
0.0
2.5
5.0
7.5
0 20 40 60
Age (years)
Cumulative hazard
Figure 5.2: Study 2: Kaplan-Meier cum ulativ e hazard curv e
The estimated surviv al and cum ulativ e hazard curv es de riv ed from Kaplan-Meier estima-
tion are resp ectiv ely pro vided in Figure 5.1 and Figure 5.2 . Almost all injection imitation
ev en ts o ccurred b y age 60 y ears as indicated b y the virtually 0% surviv al probabilit y b y
that time. Median surviv al time to injection initiation w as less than 20 y ears, and more
than an 80% of the sample w as estimated to ha v e initiated injection b efore the age of 30.
The slop e of the cum ulativ e hazard curv e suggest ed a steady but high hazard rate, pro duc-
ing a stead accum ulation of hazard, or risk, starting from adolescence and con tin uing in to
late adultho o d. ### An y-cause injection drug use initiation
Suprem um tests for non-parametric terms indicated significan t time-v arying effects for
sex, race, birth cohort, and birthplace ( T able 5.2 ). Figure 5.3 pro vides plots for all time-
v arying effects as c hanges in cum ulativ e hazards. F emale participan ts exp erienced lo w er
risk of injection initiation o v er the en tire lifetime as did those of foreign birth. Compared
to White participan ts, Blac k participan ts w ere less lik ely initiate injection o v er all ages but
no discernible differences in risk w ere detected for participan ts who w ere of either Latino
55
Study 2: P ath w a ys to Injection Drug Use Initiation
Sex (female) Race (Other)
Birth Cohort (Eighties or later) Race (Latino)
Birth Cohort (Seventies) Race (Black)
Birth Cohort (Sixties) Birthplace (foreign born)
Intercept Sexual orientation (gay, lesbian, or bisexual)
0 20 40 60 0 20 40 60
−2
0
2
4
6
−2
0
2
4
6
−2
0
2
4
6
−2
0
2
4
6
−2
0
2
4
6
Age (years)
Cumulative coefficients
Figure 5.3: Study 2: Time-v arying effects
56
Study 2: P ath w a ys to Injection Drug Use Initiation
T a ble 5 .2: Study 2: Suprem um p-v alues for significance of time-v arying effects
T erm Suprem um-test P-v alue
In tercept 7.646 0.000
Birth Cohort (referen t: Pre-Sixties)
Sixties 3.766 0.005
Sev en ties 4.671 0.000
Eigh ties or later 4.794 0.000
Sex ( female) 3.204 0.036
Sexual orien tation (GLB) 2.560 0.185
Birthplace (foreign b orn) 4.084 0.002
Race ( referen t: White)
Blac k 3.461 0.016
Latino 2.024 0.547
Other 1.768 0.656
Note:
P-v alues l ess then or equal to 0.05 are shaded in gra y .
or other race. Compared to the Pre-Sixties birth cohort, the Sixties and Sev en ties cohorts
had consisten tly lo w er risk for injection initiation o v er the en tire lifetime. The Eigh ties or
later cohort app eared to ha v e lo w er risk of injection initiation up un til age 20, after whic h
risk seems to increase substan tially alb eit not significan tly .
Constan t estimates for parametric drug use exp osure terms indicated that prior use of
some drugs increased the risk of injection initiation ( T able 5.3 ). These estimates represen t
c hanges in hazards p er y ear, conditioned on exp osure. Cannabis (0.047), heroin (0.086),
crac k co caine (0.030), methamphetamine (0.026), and NMPD stim ulan ts (0.062) all signifi-
can tly increased the hazards of injection initiation. P o wder co caine, methadone, buprenor-
phine, and NMPD opioids, sedativ es, and tranquilizers all did not significan tly c hange risk
for injection initiation. ### Cause-sp ecific injection drug use initiation b y drug used at
first injection
Results of the cause-sp ecific hazards mo dels sho w differences of predictor and co v ariate
effects dep ending on the outcome. Based on suprem um test results, the o v erall effects of
demographic co v ariates v aried in significance b y outcome ( T able 5.4 ; detailed results a v ail-
able in T able B.1 ). Plots of time-v arying effects of sex, sexual orien tation, and birthplace
57
Study 2: P ath w a ys to Injection Drug Use Initiation
T able 5.3: Study 2: P arametric estimates
T erm Estimate Std. Error Z-statistic P-v alue
Illicit drugs
Cannabis 0.047 0.006 9.190 0.000
Heroin 0.086 0.023 4.140 0.000
Crac k co caine 0.030 0.013 2.800 0.005
P o wder co caine -0.010 0.011 -1.090 0.276
Methamphetamine 0.026 0.014 2.180 0.029
Drug treatmen t medications
Methadone 0.013 0.039 0.449 0.653
Buprenorphine -0.086 0.120 -1.170 0.241
Prescription medications
Opioids -0.001 0.014 -0.051 0.959
Stim ulan ts 0.060 0.025 3.610 0.000
Sedativ es 0.016 0.019 1.100 0.270
T ranquilizers 0.008 0.016 0.573 0.567
Note:
Ro ws with significan t estimates are shaded in gra y (p-v alue <= 0.05).
Significan t estimates greater than or equal to 0.010 are prin ted in b old.
T able 5.4: Study 2: Comp eting risks - Suprem um p-v alues for significance of time-v arying effects
Heroin P o wder Co caine Methamphetamine
In tercept 0.000 0.050 0.035
Birth Cohort (referen t: Pre-Sixties)
Sixties 0.000 0.129 0.389
Sev en ties 0.001 0.060 0.076
Eigh ties or later 0.002 0.021 0.032
Sex (female) 0.932 0.020 0.003
Sexual orien tation (GLB) 0.000 0.719 0.117
Birthplace (foreign b orn) 0.005 0.388 0.032
Race (referen t: White)
Blac k 0.321 0.057 0.011
Latino 0.000 0.313 0.008
Other 0.417 0.010 0.478
Note:
P-v alues less then or equal to 0.05 are shaded in gra y .
58
Study 2: P ath w a ys to Injection Drug Use Initiation
Intercept Sex (female)
Sexual orientation
(GLB)
Birthplace
(foreign born)
Heroin
Powder
Cocaine
Metham.
0 20 40 60 0 20 40 60 0 20 40 60 0 20 40 60
−1.0
−0.5
0.0
0.5
1.0
−0.2
0.0
0.2
−0.50
−0.25
0.00
0.25
0.50
Age (years)
Cumulative coefficients
Figure 5.4: Study 2: Comp eting risks - Time-v arying effects for sex, sexual orien tation, and birth
place
on cum ulativ e hazards are pro vided in Figure 5.4 . F emale participan ts had significan tly re-
duced cum ulativ e hazards for injection drug use initiations in v olving either p o wder co caine
or methamphetamine but not heroin. Ga y , lesbian, or bisexual participan ts had signifi-
can tly reduced cum ulativ e hazards for only injection drug use initiations in v olving heroin.
Ha ving b een b orn outside the United States app eared to significan tly reduce the cum ula-
tiv e hazards from around age 15 to b efore age 30 for injection drug use initiations in v olv-
ing heroin but the effects w ere unclear on injection drug use initiations in v olving metham-
phetamine.
Racial differences ( Figure 5.6 ) w ere most eviden t for injection drug use initiations in v olv-
ing heroin or methamphetamine. With White participan ts as the referen t group, o v erall
cum ulativ e hazards w ere higher for Latino participan ts for injection drug use initiations
in v olving heroin but w ere lo w er for Blac k or Latino participan ts for injection drug use ini-
tiations in v olving methamphetamine. P articipan ts of other race had significan t but small
reductions in cum ulativ e hazards for injection initiations in v olving p o wder co caine around
59
Study 2: P ath w a ys to Injection Drug Use Initiation
Intercept Black Latino Other
Heroin
Powder
Cocaine
Metham.
0 20 40 60 0 20 40 60 0 20 40 60 0 20 40 60
−0.5
0.0
0.5
1.0
−0.2
0.0
0.2
0.4
−0.4
0.0
0.4
Age (years)
Cumulative coefficients
Figure 5.5: Study 2: Comp eting risks - Time-v arying effects for race
ages 10 through 20.
Birth cohort effects w ere more apparen t for injection initiations in v olving heroin ( Fig-
ure 5.5 ). With the Pre-Sixties cohort as the referen t, all other cohorts app eared to
generally ha v e reduced cum ulativ e hazards injection drug use initiations in v olving heroin.
F o r the Eigh ties or later cohort, this reduction app ears to b e small and limited to b efore
age 20 y ears. There w as a significan t reduction in cum ulativ e hazards for the Eigh ties
or later cohort for injection drug use initiations in v olving p o wder co caine. Ov erall, the
Eigh ties or later cohort app eared to ha v e large increases in hazards for injection drug use
initiations in v olving either heroin of methamphetamine although the p oin t wise estimates
for m uc h of their time ranges w ere not significan t.
T able 5.5 pro vides estimates of constan t effects of drug us e exp osure predictors with
predictors organized b y ro w and outcomes b y column. Exp osure to non-injection use
of the drug used at injection initiation w as significan tly increase risk for all outcomes.
Cannabis use exp osure w as increased risk significan tly for injection outcomes for heroin
60
Study 2: P ath w a ys to Injection Drug Use Initiation
Intercept Sixties Seventies Eighties or later
Heroin
Powder
Cocaine
Metham.
0 20 40 60 0 20 40 60 0 20 40 60 0 20 40 60
−2
0
2
4
−0.2
0.0
0.2
−1
0
1
2
3
Age (years)
Cumulative coefficients
Figure 5.6: Study 2: Comp eting risks - Time-v arying effects for birth cohort
T able 5.5: Study 2: Comp eting risks - P arametric estimates
Heroin P o wder Co caine Methamphetamine
Illicit drugs
Cannabis 0.027 0.003 0.016
Heroin 0.086 0.005 -0.010
Crac k co caine 0.023 -0.005 0.010
P o wder co caine -0.012 0.015 -0.013
Methamphetamine -0.011 -0.001 0.040
Drug treatmen t medications
Methadone 0.040 -0.013 -0.005
Buprenorphine -0.086 -0.004 -0.090
Prescription medications
Opioids -0.002 -0.002 -0.002
Stim ulan ts 0.043 0.009 0.004
Sedativ es 0.011 -0.003 0.005
T ranquilizers 0.015 -0.001 -0.009
Note:
Significan t estimates are shaded in gra y (p-v alue <= 0.05).
Significan t estimates greater than or equal to 0.010 are prin ted in b old.
61
Study 2: P ath w a ys to Injection Drug Use Initiation
and methamphetamine but not for p o wder co caine. Otherwise, all outcomes did not
share an y significan t exp osure predictors with p ositiv e effects. Exp osures to crac k co caine
(0.023) and NMPD stim ulan ts (0.043) use increased the hazards of injection initiations
in v olving heroin. Methadone use exp osure decreased the hazards of injection initiations in-
v olving p o wder co caine (-0.013). Exp osures to p o wder co caine (-0.013) and buprenorphine
(-0.090) use decreased the hazards of injection initiations in v olving methamphetamine.
Detailed results are a v ailable in T able B.2 .
5.4 Discussion
Exp osures to some t yp es of drug use b eha viors app ear to significan tly affect the risk of
injection initiation. Prior use of cannabis, heroin, crac k co caine, methamphetamine, or
NMPD stim ulan ts increased the risk of injection initiation. These relationship b et w een
these drug use b eha viors and risk w ere conditional on the amoun t of time exp osed with
longer exp osures asso ciated with greater risk. Within the con text of this study and sample,
risk can b e translated in to injection initiation timing. Greater accum ulations of hazard in-
crease the lik eliho o d of injection drug use initiation and earlier accum ulations increase this
lik eliho o d earlier. Injection initiation w as more lik ely to o ccur earlier among PWID who
w ere exp osed to these t yp es of drug use prior t o injection initiation.
Ho w ev er, not all injection initiations dep end on the same exp osures. Separating first
injection ev en ts in to comp eting ev en ts b y drug used at first injection pro vides a more
n uanced view of initiation. The only common risk factor among all outcomes w as pre-
injection use of the drug used at first injection: pre-injection use of heroin, p o wder co caine,
or methamphetamine w ere resp ectiv e risk factors for injection initiations in v olving eac h
of those drugs. Cannabis use exp osure w as a risk factor for initiations in v olving heroin
or methamphetamine but not p o wder co caine. The path w a ys to eac h of these outcomes
indicate that there is minimal o v erlap, the pre-injection use of other commonly injected
drugs did not increase injection initiation risk. In fact, the opp osite w as more lik ely to
o ccur among injection initiations in v olving p o wder co caine or methamphetamine in that
exp osure to other t yp es of drug use reduced initiation risks.
62
Study 2: P ath w a ys to Injection Drug Use Initiation
The predictors of injection initiation risk in v olving p o wder co caine or methamphetamine
app ear to ha v e a straigh tforw ard in terpretation. In addition to the patterns describ ed
previously , pre-injection use of drugs with depressan t effects (e.g., opioids) either did not
c hange or reduced risk. The path w a ys to injection initiation for either seemed to follo w a
fairly linear transition from non-injection to injection drug use. Initiating injection could
b e the result of individual desires or motiv ations to inject coupled with so cial opp ortuni-
ties to acquire the skills of injection, mediated b y in terp ersonal ties or exp osures to other
mem b ers within the accessible drug use so cial net w ork. The comp osition and attributes
of these so cial net w orks ma y b e determining factors of injection initiation ( Neaigus et al.,
2006 ).
Injection initiations in v olving heroin, ho w ev er, seem to follo w v ery differen t t yp es of path-
w a ys. Risk factors included exp osure to crac k co caine and NMPD stim ulan ts use, b oth of
whic h do not share an y of the effect c haracteristics of heroin. F urthermore, neither of these
exp osures w ere risk factors for injection initiations in v olving p o wder co caine or metham-
phetamine despite b oth of these exp osures b eing stim ulan t drugs and p ossessing similar
c hemical prop erties or ev en the same activ e ingredien t, as is the case for crac k co caine and
p o wder co caine. Based on these findings, the risk path w a ys for injection initiation with
heroin app ear ma y b e more complex. One p ossible explanation is that there are strong as-
so ciations among crac k co caine, NMPD stim ulan ts, and heroin usages through o v erlaps in
drug use net w orks. The use of an y of these three drugs w ould increase opp ortunities for
initiating injection use of heroin. The demographic dissimilarities b et w een injection initia-
tion risks in v olving heroin v ersus p o wder co caine or methamphetamine lend some supp ort-
iv e evidence to this case.
Demographic trends suggest that injection drug use ma y b e exp eriencing greater p opular-
it y among y ounger PWUD. The Eigh ties or later birth cohort exhibited similar risks as the
referen t Pre-Sixties cohort, b oth of whic h had higher risk than the Sixties or Sev en ties co-
horts. Although this do es not definitiv ely supp ort a resurgence of injection drug use, these
findings indicate that the Eigh ties or later cohort acquire risk earlier and are more lik ely to
63
Study 2: P ath w a ys to Injection Drug Use Initiation
initiate injection at a y ounger age. Results from the comp eting risks analysis suggest that
this trend is sp ecific to initiations in v olving heroin. A dditional demographic trends w ere
apparen t in the comp eting risks mo dels with substan tial differences b et w een outcomes on
sex, sexual iden tit y , and race.
The role of cannabis as a risk factor for inje ction initiation is unclear. It is commonly one
of the first illicit drug use b eha viors to b e initiated and it is p ossible that cannabis use rep-
resen ts another t yp e of risk exp osure. Generally , injection initiation risk increased with
exp osure to cannabis use, and this ma y reflect the general risk asso ciated with the drug
use career instead of an y risk attributable sp ecifically to cannabis. Those with longer drug
use careers accum ulate greater risk of injection initiation p ossibly due to the gradual acqui-
sition of drug use exp eriences or exp osures. Ho w ev er, it is p ossible that cannabis use ex-
p osure has sp ecific con tributions to risk due its asso ciations with drug use so cial net w orks.
A t the time of surv ey , recreational c annabis use w as not y et legal in California, and it is
highly lik ely that all the participan ts who used cannabis acquired cannabis illegally . With
the slo w and, what seems to b e inexorable, legalization of recreational cannabis use, the
connection b et w een cannabis use, the illicit drug use career, and injection drug use ma y
b ecome greatly atten uated in the future. Whether cannabis con tributes at all an y risk on
its o wn to future drug use initiations con tin ues to b e debated within the literature ( Jor-
gensen & W ells, 2021 ; Morral, McCaffrey , & P addo c k, 2002 ).
There are t w o immediate implications of these findings . First, the ma jorit y of first injec-
tion ev en ts in v olv ed injecting heroin, and prev en ting the sp ecific drug use exp osure risk
factors of this outcome ma y b e an effectiv e strategy for o v erall prev en tion of injection ini-
tiation. Prior w ork has iden tified non-injection heroin use as a risk factor for subsequen t
injection initiation, but the role of crac k co caine use remains unclear. The findings of this
study suggest that crac k co caine use con tributes significan t risk to w ards injection initia-
tion and strategies addressing the kno wn o v erlap in drug use for b oth heroin and co caine
( Leri, Bruneau, & Stew art, 2003-01 ) ma y pro v e to b e e ffectiv e than for heroin alone. The
relationship b et w een crac k co caine use on injection initiation, sp ecifically with heroin, ma y
64
Study 2: P ath w a ys to Injection Drug Use Initiation
reflect a complex, but imp ortan t, underlying risk pro cess.
Second, an y study of injection initiation should consider the tradeoff b et w een examining
simpler but more general o v erall injection initiation (i.e., an y-cause initiations) v ersus
more complicated but more sp ecific injection initiation b y drug (i.e., cause-sp ecific ini-
tiations). The o v erall findings of these analyses illustrate the imp ortance of examining
injection drug use initiation as not a single monolithic ev en t but as a collection of com-
p eting ev en ts with p oten tially differen t determining factors. Lank enau et al. ( 2010 ) also
found differences among PWID stratified b y drug used at first injection exp erience on
education lev el, homelessness, criminal history , and drug treatmen t history at time of
initiation. These differences b y outcome highligh t the imp ortance of details and ha v e
p ossible applications for in terv en tion targeting. The injection risk exp osures for p eople
who use heroin use ma y b e substan tially differen t from those for p eople who use p o wder
co caine.
Study on injection drug use should b e con tin uous and ongoing as the understanding of in-
jection initiation m ust con tin uously ev olv e corresp ondingly to c hanging trends in drug use.
This has b ecome increasingly relev an t; after a p erio d of relativ e decline, injection app ears
to b e on the rise again, signaling a p orten t of sev ere individual and so cial consequences.
Increases in opioid o v erdoses and asso ciated fatalities indicate that the consequences ha v e
already b egun. Understanding the con tributions of drug use exp osures on injection initia-
tion ma y b e imp ortan t to strategies to prev en t injection initiation and the consequences of
injection drug use on individual and public health. The utilit y of these exp osures in indi-
cating risk should not b e o v erlo ok ed during this trend of increasing injection drug use.
This study con tributes some additional understanding to t he path w a ys of injection initi-
ation. A discussion of limitations applicable to this and all other studies comprising this
dissertation is pro vided in Over al l Limitations (subsection 7.1, p. 95) . The approac h used
b y this study can b e applied to more con temp orary cohorts of PWID, whic h ma y reflect
the increases in opioid use asso ciated with the ongoing epidemic of NMPD opioid use. As
drug use trends are constan tly ev olving, this study only presen ts a limited view for a sp e-
65
Study 2: P ath w a ys to Injection Drug Use Initiation
cific p opulation within a sp ecific area o v er a sp ecific time. Studies using div erse samples of
PWID from other regions or time p erio ds will con tribute to a more generalizable p ersp ec-
tiv e.
66
Study 3: P ath w a ys of Injection Drug Use
6 Study 3: P ath w a ys of Injection Drug Use
Injection drug use in v olv es injecting a drug in to the b o dy , t ypically using a h yp o dermic
syringe. Because injection directly deliv ers drug in to the blo o d stream, the effects of drug
injection are faster and stronger than other routes of administration suc h as snorting or
smoking. A dditionally , more effect can b e extracted from a quan tit y of drug via injection
v ersus other routes, making injection a more economical metho d. Ho w ev er, these relativ e
adv an tages are offset b y the risks and sev erit y of acute and c hronic harms caused b y injec-
tion as w ell as higher risks for drug-related morbidities and mortalities. This has resulted
in PWID exp eriencing a large prop ortion of the harms asso ciated with drug use despite
only comprising a small fraction of PWUD.
The health needs of PWID and public health conseque nces of injection drug use ha v e mo-
tiv ated public health efforts to understand the etiology of injection drug use and prev en t
its incidence or reduce its prev alence. So cial transmission app ears to b e an imp ortan t con-
duit to injection drug use initiation where PWID assist in the injection initiation of non-
injecting PWUD ( Bluthen thal et al., 2014 , 2015 ; Haro c op os, Goldsam t, K obrak, Jost, &
Clatts, 2009-07 ; Daniel W erb et al., 2016 ). In terv en tion strategies, namely “Break the Cy-
cle” and “Change the Cycle,” addressing the role of PWID and so cial exp osure to injection
drug use on injection initiation ha v e sho wn promising results in the primary prev en tion of
injection drug use ( Hun t, Griffiths, South w ell, Stillw ell, & Strang, 1999 ; Jarlais et al., 2019 ;
Strik e et al., 2014 ).
These efforts can b e augmen ted with an impro v ed unde rstanding of ho w the injection drug
use career dev elops. Although injection drug use itself is a single b eha vior tied to a sp ecific
route of drug administration, it can b e applied to man y differen t drugs. In realit y , there is
a div ersit y of injection drug use b eha viors and these b eha viors ma y p ossess similar or dif-
feren t risk factors. F urthermore, man y PWID engage in injection drug use of more than
one drug o v er their lifetimes and there ma y b e correlation b et w een the injection of differ-
en t drugs. Injection drug use b eha viors ma y b e temp orally asso ciated in that the initia-
tion of one b eha vior increases the lik eliho o d of initiating another in the future but not the
67
Study 3: P ath w a ys of Injection Drug Use
rev erse. T emp oral asso ciations among a div erse set of b eha viors can b e used to describ e
dev elopmen tal path w a ys of the injection drug use career and these path w a ys ma y b e infor-
mativ e for assessing injection initiation risk.
6.1 Bac kground
The idea of drug use dev elopmen tal path w a ys is not no v el. The Gatew a y Hyp othesis ( D.
Kandel, 1975 ) is presen tly the prev ailing framew ork for describing ho w the drug use career
b egins and has b een used to iden tify generalizable patterns of drug use initiations through
detecting significan t ordered asso ciations. Ho w ev er, this approac h has seen limited appli-
cation to the injection drug use career. Although the injection drug use career is nested
within the larger drug use career, it ma y exhibit dev elopmen tal path w a ys sp ecific to injec-
tion. Asso ciations b et w een injection b eha viors ma y reflect similar or differen t individual or
so cial risk pro cesses.
There is some evidence that the risks of i njection initiations dep end on prior drug use
exp eriences. T ransitions from non-injection to injection use of a drug b y drug ( Bluthen thal
et al., 2018 ). Engaging regular injection drug use ma y b e asso ciated with the first drug
injected ( O’Keefe, Horyniak, & Dietze, 2016 ). Within some limited samples of PWID,
pre-injection drug use app ears to affect the risk of injection initiation ( F uller et al., 2001;
Miller, Strathdee, Kerr, Li, & W o o d, 2006 ; Ro y et al., 2003 ; Sherman et al., 2005 ). The
literature on asso ciations among injection drug use b eha viors, ho w ev er, con tin ues to b e
limited and it remains unclear if and ho w these b eha viors are related. A ddressing this gap
in the literature is the fo cus of this study .
A dditionally , individual factors influence initiati on risk b y mediating exp osure to individ-
ual exp eriences or so cial exp osures whic h resp ectiv ely affect individual lik eliho o ds of drug
use initiations or so cial probabilities of drug use opp ortunities. Sex ( Cotto et al., 2010 ),
sexual orien tation ( Corliss et al., 2010 ; Medley et al., 2016 ; New com b, Birk ett, Corliss, &
Mustanski, 2014 ; Rosner, Neicun, Y ang, & Roman-Urrestarazu, 2021 ), and r ace ( Abuse &
A dministration, 2020b ). A dditionally , immigration exp erience app ears to b e a con tributing
68
Study 3: P ath w a ys of Injection Drug Use
factor to drug use ( Borges et al., 2012 ; Salas-W righ t, V aughn, Clark, T erzis, & Córdo v a,
2014 ). Although these findings come from studies for an y-route drug use b eha viors, it is
lik ely that individual factors are also relev an t sp ecifically for injection drug use b eha viors.
Y et another con tributing set of factors comprise c hanging historical con texts as genera-
tional cohort effects. So cial and cultural trends con tribute to c hanges in so cial p opularit y ,
acceptabilit y , and accessibilit y of drug use b eha viors, whic h in turn affect drug use initia-
tion risks ( Andrew Golub, Johnson, & Dunlap, 2005 ). P erio ds of greater p opularit y and
access allo w for more initiation opp ortunities and few er barriers to initiation whereas p e-
rio ds of lo w p opularit y pro duce the opp osite result. These trends ha v e b een observ ed as
differences in drug use prev alence b y birth cohort for an y-route use of co caine, heroin, and
NMPD opioids (Bluthen thal, W enger, Ch u, Bourgois, & Kral, 2017 ; Louisa Degenhardt,
Chiu, Sampson, Kessler, & An thon y , 2007 ; Andrew Golub, Elliott, & Bro wnstein, 2013 ; A.
Golub & Johnson, 2001 ; A. L. Golub & Johnson, 2 009 ; Huang, Key es, & Li, 2018-01 ; W all
et al., 2018 ). The role of con text sp ecifically for the injection drug use b eha viors is unclear
but is lik ely to corresp ond with general drug use trends.
The primary aim of this study is to describ e the longitudinal asso ciations among a large
and div erse set of injection drug use b eha viors for PWID. Secondarily , the age-dep enden t
risk of drug use initiation and differences on this asso ciated with demographic traits and
birth cohorts are also in v estigated. These are all accomplished using m ultiple mo dels of
age-dep enden t additiv e hazards with time-dep enden t and constan t effects.
6.2 Metho ds
6.2.1 Sample and measures
This study consisted of secondary analyses of s urv ey data collected from PWID, consisting
of 808 participan ts of the 810 describ ed in Par ent study (subsection 3.1, p. 11) . P artici-
pan ts pro vided information on their drug use exp eriences including age at first injection
exp erience and drug use histories for 10 t yp es of illicit drugs: (1) crac k and (2) p o wder
co caine; (3) methamphetamine; (4) heroin; (5) methadone and (6) buprenorphine (used
69
Study 3: P ath w a ys of Injection Drug Use
outside of treatmen t settings or directions); and non-medical prescription drug (NMPD)
use of (7) opioids (not including methadone or buprenorphine, e.g., Vico din or Oxycon tin),
(8) tranquilizers (e.g., Klonopin or V alium), (9) sedativ es (e.g., Restoril or phenobarbital),
and (10) stim ulan ts (e.g., Ritalin or A dderall). F or eac h drug, participan ts pro vided infor-
mation on whether they ev er used the drug via an y route of administration, at what age
they first used the drug, whether they ev er injected the drug, and at what age they first
injected the drug. This information w as collected using four separate questions with appro-
priate skip patterns. Discrepancies b et w een rep orted age at first injection exp erience and
injection drug use histories w ere resolv ed b y sp ecifying the lo w est age for an y first injec-
tion exp erience as the age of lifetime injection initiation.
P articipan ts also pro vided demographic information consisting of age at time of surv ey , bi-
ological sex at birth, birth y ear, sexual iden tit y (heterosexual/straigh t, ga y , lesbian, or bi-
sexual), birthplace (United States v ersus other), and race. Birth y ear w as transformed in to
categories mostly corresp onding to decadal birth cohort. Sexual iden tit y w as dic hotomized
in to heterosexual v ersus ga y , lesbian, or bisexual. A dditional details on the measures used
can b e found in Me asur es (subsection 3.2, p. 12) .
Drug use history and demographic data w ere transformed in to analytic datasets for eac h
injection drug use initiation outcome corresp onding to eac h of the 10 injection drug use
b eha viors. Eac h mo deled outcome relied on its o wn generated dataset. F or eac h outcome,
the start of observ ation b egan at age 0 y ears (birth) and ended at the age at outcome, if
the ev en t o ccurred, or the age at surv ey , if the ev en t did not o ccur (i.e., w as censored).
Exp osure predictors comprised exp osures to the 9 other injection drug use b eha viors, exp o-
sure to injection drug use (non-drug sp ecific), and exp osure to an y route use of the drug
in v olv ed with the outcome (e.g., an y heroin use exp osure on heroin injection initiation).
F o r the purp oses of this study , pre-injection use of a drug w as defined as ha ving an age of
initiation of an y-route use that w as less than the age at injection initiation for the drug.
Exp osures w ere time-dep enden t binary v alues with exp osures starting at the rep orted ages
of initiations. All outcomes used the same set of demographic co v ariates describ ed ab o v e.
70
Study 3: P ath w a ys of Injection Drug Use
F u rther details on this pro cess are pro vided in T emp or al or dering and time-dep endent
values (subsubsection 3.5.1, p. 17) .
6.2.2 Surviv al analysis
Surviv al estimations and hazards w ere initially mo deled using Kaplan-Meier estimation fol-
lo w ed b y regression analysis using the additiv e hazards mo del. The additiv e hazards mo del
is a flexible surviv al regression approac h that do es not rely on the prop ortional hazards as-
sumption and can include constan t and time-v arying effects. Non-parametric time-v arying
effects w ere estimated as time-dep enden t functions of c hanges in hazard rates p er y ear.
Time-v arying effects w ere in tegrated o v er time t o pro duce plots of time-dep enden t c hanges
in cum ulativ e hazards with p oin t-wise confidence in terv als and Hall-W ellner confidence
bands. The o v erall significance of eac h time-v arying effect w as assessed using a suprem um
test o v er the mo deled time range. P arametric constan t effects consisted of con v en tional re-
gression p oin t estimates, standard errors, and tests of significance. A dditional details on
the additiv e hazards mo del approac h can b e found in A dditive hazar ds mo del (subsubsec-
tion 3.5.3, p. 18) .
T w o sets of mo dels w ere fitted, a restricte d set not including the exp osure predictor for in-
jection drug use and a full set including all exp osure predictors. This w as done to b etter
understand the effects of injection drug use exp osure that is not asso ciated with a partic-
ular drug on injection drug use initiations. Results and in terpretations are based on the
results from the full mo dels, whic h uses all exp osure predictors, unless indicated otherwise.
6.3 Results
Descriptiv e statistics of demographic co v ariates for the full data (N = 810) can b e found
in T able 3.1 and of injection drug use b eha viors for the PWID sample used in this study
(N = 808) in T able 6.1 . Heroin w as the most commonly injected drug with a 93% lifetime
prev alence (n = 753) follo w ed b y p o wder co caine (78%; n = 629) and methamphetamine
(62%; n = 505). Around a third of the sample rep orted ev er injecting crac k co caine (33%;
71
Study 3: P ath w a ys of Injection Drug Use
T able 6.1: Study 3: Descriptiv e statistics of inject ion drug use b eha viors
T otal initiated (N = 808) Coun ts of use prior to injection in tiation
Drug injected Coun t P ercen tage (%) Prior use No prior use Missing
Illicit drugs
Heroin 753 93.2 129 623 1
Crac k co caine 265 32.8 184 79 2
P o wder co caine 629 77.8 291 335 3
Methamphetamine 505 62.5 204 299 2
Drug treatmen t medications
Methadone 41 5.1 15 25 1
Buprenorphine 13 1.6 3 9 1
Prescription medications
Opioids 251 31.1 136 106 9
Stim ulan ts 78 9.7 29 49 0
Sedativ es 31 3.8 17 14 0
T ranquilizers 60 7.4 31 28 1
n = 265) or NMPD opioids (31%; n = 251). All of the other drugs assessed exhibited less
than 10% sample prev alence. Use of the injected drug prior to injection initiation v aried
b y the drug injected with prior use most frequen t among crac k co caine injection initiations
and least frequen t among heroin injection initiations. Because the time resolution of the
data is in y ears, prior use exp osures o ccurring within the same age at injection initiation
w ere undetected and the actual coun ts of prior use m a y actually b e higher.
Kaplan-Meier estimated surviv al curv es Figure 6.1 sho w mark ed diffe rences in median
surviv al times b y drug injected. Heroin injection initiation w as the shortest at around 20
y ears, follo w ed b y p o wder co caine injection initiation at around 25 y ears and metham-
phetamine injection initiation at around 35 y ears. Cum ulativ e hazard curv es ( Figure 6.2 )
exhibited fairly constan t slop es for heroin and co caine injection initiation curv es. Curv es
for all other injected drugs generally started with steep er slop es and gradually transitioned
to flatter slop es. Stratification b y injection drug use exp osure sho w ed consisten tly lo w er
surviv al probabilities for all outcomes among those who had prior injection drug use exp e-
rience ( Figure 6.3 ). The differences in tra jectories w as also eviden t in stratified cum ulativ e
hazard curv es ( Figure 6.4 ).
72
Study 3: P ath w a ys of Injection Drug Use
Methadone NMPD Tranquilizers
Methamphetamine NMPD Sedatives
Powder Cocaine NMPD Stimulants
Crack Cocaine NMPD Opioids
Heroin Buprenorphine
0 20 40 60 0 20 40 60
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
Age (years)
Survival probability
Figure 6.1: Study 3: Kaplan-Meier surviv al curv es
73
Study 3: P ath w a ys of Injection Drug Use
Methadone NMPD Tranquilizers
Methamphetamine NMPD Sedatives
Powder Cocaine NMPD Stimulants
Crack Cocaine NMPD Opioids
Heroin Buprenorphine
0 20 40 60 0 20 40 60
0.00
0.01
0.02
0.0
0.2
0.4
0.6
0.00
0.05
0.10
0.00
0.02
0.04
0.00
0.05
0.10
0.15
0
1
2
3
4
5
0.0
0.2
0.4
0.6
0.0
0.5
1.0
1.5
2.0
0.00
0.25
0.50
0.75
1.00
1.25
0.00
0.02
0.04
0.06
0.08
Age (years)
Cumulative hazard
Figure 6.2: Study 3: Kaplan-Meier cum ulativ e hazard curv es
74
Study 3: P ath w a ys of Injection Drug Use
Methadone NMPD Tranquilizers
Methamphetamine NMPD Sedatives
Powder Cocaine NMPD Stimulants
Crack Cocaine NMPD Opioids
Heroin Buprenorphine
0 20 40 60 0 20 40 60
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
Age (years)
Survival probability
Injection exposure No Y es
Figure 6.3: Study 3: Kaplan-Meier surviv al curv es stra tified b y injection exp osure
75
Study 3: P ath w a ys of Injection Drug Use
Methadone NMPD Tranquilizers
Methamphetamine NMPD Sedatives
Powder Cocaine NMPD Stimulants
Crack Cocaine NMPD Opioids
Heroin Buprenorphine
0 20 40 60 0 20 40 60
0.00
0.01
0.02
0.03
0.0
0.2
0.4
0.6
0.00
0.05
0.10
0.15
0.20
0.00
0.05
0.10
0.15
0.00
0.05
0.10
0.15
0.20
0
2
4
6
0.0
0.2
0.4
0.6
0.8
0
1
2
3
4
0
1
2
0.000
0.025
0.050
0.075
Age (years)
Cumulative hazard
Injection exposure No Y es
Figure 6.4: Study 3: Kaplan-Meier cum ulativ e ha zard curv es stratified b y injection exp osure
76
Study 3: P ath w a ys of Injection Drug Use
6.3.1 Time-v arying effects of demographic co v ariates
No discernible differences w ere found on non-parametric time-v arying effects b et w een re-
stricted and full mo dels. T able 6.2 and T able 6.3 pro vide suprem um test p-v alues for time-
v arying effects resp ectiv ely for the restricted and full mo dels. More detailed results for all
mo dels (full and restricted) are pro vided in Study 3 Detaile d T ables (App endix C, p. 115) .
Plots for time-dep enden t for sex, sexual orien tation, and birthplace are pro vided in Fig-
ure 6.5 . F emale PWID w ere less lik ely to initiate injection for p o wder co caine, metham-
phetamine, and NMPD sedativ es and less lik ely to initiate injection during adolescence
for NMPD tranquilizers. GLB PWID w ere less lik e to ha v e initiated injection of heroin
o v er the en tire lifetime and of NMPD sedativ es during adolescence. PWID who w ere b orn
outside of the United States w ere less lik ely to ha v e initiated heroin injection from adoles-
cence up un til age 30 y ears and less lik ely to ha v e initiated p o wder co caine injection af-
ter around age 20. There w ere small reductions in risk during adolescence among foreign-
b orn PWID for ha ving initiated NMPD opioids or sedativ es injection and unclear effects
on methamphetamine injection initiation.
77
Study 3: P ath w a ys of Injection Drug Use
T able 6.2: Study 3: Suprem um p-v alues for significance of time-v arying effects (re stricted mo dels)
Illicit drugs Medications
Co caine Drug treatmen t Prescription
Heroin Crac k P o wder Metham. Methad. Bupren. Opioids Stim ul. Sedat. T ranq.
In tercept 0.000 0.342 0.000 0.000 0.394 0.521 0.000 0.000 0.000 0.021
Birth Cohort (referen t: Pre-Sixties)
Sixties 0.000 0.017 0.144 0.708 0.457 0.642 0.024 0.018 0.552 0.829
Sev en ties 0.000 0.003 0.120 0.081 0.560 0.329 0.342 0.011 0.140 0.495
Eigh ties or later 0.008 0.001 0.021 0.002 0.125 0.204 0.038 0.018 0.001 0.260
Sex (female) 0.246 0.657 0.001 0.001 0.827 0.375 0.500 0.857 0.000 0.015
Sexual orien tation (GLB) 0.003 0.639 0.378 0.202 0.346 0.577 0.450 0.437 0.042 0.472
Birthplace (foreign b orn) 0.000 0.661 0.012 0.006 0.150 0.171 0.002 0.526 0.045 0.212
Race (referen t: White)
Blac k 0.183 0.412 0.016 0.000 0.357 0.434 0.000 0.093 0.002 0.008
Latino 0.006 0.018 0.016 0.000 0.729 0.239 0.000 0.001 0.034 0.421
Other 0.185 0.312 0.696 0.210 0.662 0.224 0.001 0.036 0.256 0.090
Note:
P-v alues less then or equal to 0.05 are shaded in gra y .
78
Study 3: P ath w a ys of Injection Drug Use
79
Study 3: P ath w a ys of Injection Drug Use
T able 6.3: Study 3: Suprem um p-v alues for significance of time-v arying e ffects (full mo dels)
Illicit drugs Medications
Co caine Drug treatmen t Prescription
Heroin Crac k P o wder Metham. Methad. Bupren. Opioids Stim ul. Sedat. T ranq.
In tercept 0.000 0.344 0.000 0.000 0.337 0.505 0.000 0.000 0.000 0.029
Birth Cohort (referen t: Pre-Sixties)
Sixties 0.001 0.022 0.164 0.900 0.464 0.632 0.027 0.016 0.554 0.874
Sev en ties 0.000 0.004 0.212 0.039 0.573 0.310 0.354 0.010 0.140 0.526
Eigh ties or later 0.046 0.000 0.015 0.002 0.122 0.205 0.041 0.021 0.000 0.247
Sex (female) 0.235 0.494 0.002 0.003 0.820 0.370 0.299 0.870 0.001 0.015
Sexual orien tation (GLB) 0.005 0.636 0.477 0.246 0.320 0.569 0.440 0.494 0.038 0.548
Birthplace (foreign b orn) 0.000 0.678 0.007 0.032 0.225 0.165 0.001 0.581 0.053 0.149
Race (referen t: White)
Blac k 0.221 0.267 0.023 0.000 0.323 0.419 0.000 0.093 0.003 0.010
Latino 0.001 0.054 0.045 0.000 0.711 0.223 0.000 0.002 0.037 0.440
Other 0.168 0.215 0.740 0.433 0.666 0.224 0.001 0.035 0.264 0.066
Note:
P-v alues less then or equal to 0.05 are shaded in gra y .
80
Study 3: P ath w a ys of Injection Drug Use
Detected differences on race in c hanges of cum ulativ e hazards w ere t ypically risk reduc-
tions compared to the referen t White race ( Figure 6.6 ). The sole significan t exception w as
the increased risk of ha ving initiated heroin injection among Latino PWID. In the case
of ha ving initiated NMPD opioids injection, all non-White PWID had significan tly lo w er
risks compared to White PWID.
Significan t c hanges of cum ulativ e hazards w ere detected b y birth cohort ( Figure 6.7 ).
Changes in cum ulativ e hazards generally app eared to either b e steadily increasing or
decreasing o v er time. Using the Pre-Sixties birth cohort (i.e., PWID b orn b efore 1960) as
the referen t, risks of injection initiation w ere generally highest among the Eigh ties or later
cohort (i.e., PWID b orn on or after 1980). The risk of ha ving initiated injection of crac k
co caine, methamphetamine, methadone, buprenorphine, or NMPD tranquilizers app eared
to increase with eac h successiv e cohort although man y of these trends w ere not significan t.
Cum ulativ e hazards for ha ving initiated NMPD stim ulan ts injection w ere lo w er for all
non-referen t cohorts with few qualitativ e differences among them. The risk of ha ving
initiated NMPD sedativ es injection w as lo w er for the Eigh ties or later cohort compared
to the referen t. The risk of ha ving initiated heroin injection, starting with the Pre-Sixties
cohort, decreased for the Sixties and Sev en ties cohorts but increased for the Eigh ties or
later cohort. Notably , the risk of ha ving initiated NMPD opioids injection w as highest for
the Eigh ties or later cohort.
6.3.2 Constan t effects of drug use predictors
T able 6.4 and T able 6.5 resp ectiv ely pro vide esti mates of constan t effects of injection drug
use exp osure predictors for the restricted and full mo dels. T able ro ws corresp ond to ex-
p osure predictors and table columns corresp ond to outcomes. Significan t estimates are
shaded in gra y and significan t estimates ha ving at least a 0.010 absolute v alue w ere con-
sidered substan tiv e and prin ted in b old. A 0.010 c hange in hazard rate corresp onds to a
c hange of 1 ev en t p er 100 p erson-y ears of exp osure. Detailed tables are pro vided in Study
3 Detaile d T ables (App endix C, p. 115) .
81
Study 3: P ath w a ys of Injection Drug Use
Intercept Sex (female)
Sexual orientation
(GLB)
Birthplace
(foreign born)
Heroin
Crack
Cocaine
Powder
Cocaine
Metham. Methad. Bupren.
NMPD
Opioids
NMPD
Stimulants
NMPD
Sedatives
NMPD
Tranq.
0 20 40 60 0 20 40 60 0 20 40 60 0 20 40 60
−1
0
1
−0.1
0.0
0.1
0.2
−0.5
0.0
0.5
1.0
−0.5
0.0
0.5
1.0
−0.03
0.00
0.03
−0.04
−0.02
0.00
0.02
−0.2
0.0
0.2
0.4
0.6
−0.1
0.0
0.1
0.2
0.0
0.1
−0.05
0.00
0.05
0.10
Age (years)
Cumulative coefficients
Figure 6.5: Study 3 : Time-v arying effects for sex, sexual orien tation, and birth place
82
Study 3: P ath w a ys of Injection Drug Use
Intercept Black Latino Other
Heroin
Crack
Cocaine
Powder
Cocaine
Metham. Methad. Bupren.
NMPD
Opioids
NMPD
Stimulants
NMPD
Sedatives
NMPD
Tranq.
0 20 40 60 0 20 40 60 0 20 40 60 0 20 40 60
−0.5
0.0
0.5
1.0
1.5
−0.2
−0.1
0.0
0.1
−0.5
0.0
0.5
1.0
−0.5
0.0
0.5
1.0
−0.050
−0.025
0.000
0.025
0.050
−0.04
−0.02
0.00
0.02
−0.3
0.0
0.3
0.6
−0.1
0.0
0.1
0.2
−0.1
0.0
0.1
−0.10
−0.05
0.00
0.05
0.10
Age (years)
Cumulative coefficients
Figure 6.6: Study 3: Time-v arying effects for race
83
Study 3: P ath w a ys of Injection Drug Use
Intercept Sixties Seventies Eighties or later
Heroin
Crack
Cocaine
Powder
Cocaine
Metham. Methad. Bupren.
NMPD
Opioids
NMPD
Stimulants
NMPD
Sedatives
NMPD
Tranq.
0 20 40 60 0 20 40 60 0 20 40 60 0 20 40 60
−1
0
1
2
3
−0.5
0.0
0.5
1.0
1.5
−1.0
−0.5
0.0
0.5
1.0
1.5
0
1
2
0.0
0.1
0.2
−0.05
0.00
0.05
0.10
0.15
0.0
0.5
1.0
−0.2
−0.1
0.0
0.1
0.2
−0.1
0.0
0.1
−0.2
0.0
0.2
0.4
Age (years)
Cumulative coefficients
Figure 6.7: Study 3: Time-v arying e ffects for birth cohort
84
Study 3: P ath w a ys of Injection Drug Use
T able 6.4: Study 3: P arametric estimates (restricted mo dels)
Illicit drugs Medications
Co caine Drug treatmen t Prescription
Heroin Crac k P o wder Metham. Methad. Bupren. Opioids Stim ul. Sedat. T ranq.
Prior same drug use 0.122 0.006 0.025 0.038 0.001 0.005 0.014 0.011 0.003 0.002
Illicit drugs
Heroin NA 0.001 0.022 -0.003 0.001 0.000 -0.002 0.000 0.000 0.000
Crac k co caine 0.032 NA 0.042 0.017 0.001 0.000 0.006 -0.002 -0.001 -0.001
P o wder co caine 0.038 0.003 NA 0.011 -0.001 0.000 0.003 0.000 -0.001 0.001
Methamphetamine 0.017 0.008 0.021 NA -0.001 0.000 0.005 0.002 0.001 -0.001
Drug treatmen t medications
Methadone 0.098 0.002 0.123 0.046 NA 0.001 -0.012 -0.005 -0.001 0.012
Buprenorphine -0.054 0.062 -0.207 -0.068 0.035 NA 0.046 0.007 -0.002 -0.003
Prescription medications
Opioids 0.152 0.015 0.037 0.003 0.001 0.002 NA 0.003 0.001 0.002
Stim ulan ts -0.016 0.002 0.081 0.010 -0.001 -0.001 -0.001 NA 0.001 0.003
Sedativ es 0.453 -0.007 -0.021 0.000 0.001 -0.001 0.016 0.002 NA -0.003
T ranquilizers -0.071 -0.006 0.000 0.020 0.013 -0.001 0.056 0.003 0.005 NA
Note:
Significan t estimates are shaded in gra y (p-v alue <= 0.05).
Significan t estimates greater than or equal to 0.010 are prin ted in b old.
85
Study 3: P ath w a ys of Injection Drug Use
86
Study 3: P ath w a ys of Injection Drug Use
T able 6.5: Study 3: P arametric estimates (full mo dels)
Illicit drugs Medications
Co caine Drug treatmen t Prescription
Heroin Crac k P o wder Metham. Methad. Bupren. Opioids Stim ul. Sedat. T ranq.
Prior same drug use 0.117 0.006 0.024 0.038 0.001 0.005 0.014 0.011 0.003 0.002
Illicit drugs
Heroin NA 0.004 -0.013 -0.038 0.001 0.000 -0.001 -0.002 -0.001 -0.001
Crac k co caine 0.024 NA 0.037 0.017 0.001 0.000 0.006 -0.002 -0.001 0.000
P o wder co caine 0.005 0.003 NA 0.006 -0.001 0.000 0.003 -0.001 -0.001 0.000
Methamphetamine -0.035 0.009 -0.001 NA -0.001 0.000 0.005 0.002 0.000 -0.001
Drug treatmen t medications
Methadone 0.037 0.002 0.111 0.041 NA 0.001 -0.012 -0.005 -0.001 0.012
Buprenorphine -0.158 0.065 -0.200 -0.062 0.035 NA 0.046 0.007 -0.002 -0.003
Prescription medications
Opioids 0.129 0.015 0.031 0.002 0.001 0.002 NA 0.003 0.001 0.002
Stim ulan ts -0.021 0.002 0.072 0.009 -0.001 -0.001 -0.001 NA 0.001 0.003
Sedativ es 0.441 -0.007 -0.009 0.001 0.001 -0.001 0.016 0.002 NA -0.003
T ranquilizers -0.049 -0.006 0.006 0.021 0.013 -0.001 0.056 0.003 0.005 NA
Injection drug use 0.070 -0.003 0.040 0.038 0.001 0.000 -0.001 0.003 0.002 0.001
Note:
Significan t estimates are shaded in gra y (p-v alue <= 0.05).
Significan t estimates greater than or equal to 0.010 are prin ted in b old.
87
Study 3: P ath w a ys of Injection Drug Use
Inje ction drug use exp osur e. The restricted mo dels detected m ultiple significan t, sub-
stan tiv e, and p ositiv e asso ciations b et w een exp osures and outcomes for ha ving initiated
injection for heroin, p o wder co caine, and methamphetamine. P o wder co caine injection
exp osure increased the risk of ha ving initiated injection of heroin (0.038) or metham-
phetamine (0.011). Heroin (0.022) or methamphetamine (0.021) injection exp osures
increased the risk of ha ving initiated p o wder co caine injection. Including an injection drug
use exp osure predictor, as in the full mo dels, diminishes these effects substan tially and
in tro duces negativ e asso ciations. Within the full mo dels, exp osure to injection drug use in
general increased the risk of injection initiation for heroin (0.070), p o wder co caine (0.040),
or methamphetamine (0.038).
Non-inje ction drug use of the same drug. F or almos t all outcomes, exp osure to the
non-injection use of a drug significan tly increased the risk of injection initiation of the
drug. This effect w as significan t and substan tiv e for heroin (0.117), p o wder co caine
(0.025), methamphetamine (0.038), NMPD opioids (0.014), and NMPD stim ulan ts (0.011)
injection initiations.
Opioids. Heroin injection exp osure decreased the risk of ha ving initiated metham-
phetamine injection (-0.038) and w as not a significan t predictor for an y other outcome.
Methadone injection exp osure did not yield an y substan tiv e effects on an y outcome.
Exp osure to buprenorphine injection, ho w ev er, decreased the risks for injection initiations
88
Study 3: P ath w a ys of Injection Drug Use
of p o wder co caine (-0.200) and methamphetamine (-0.062). NMPD opioids injection
exp osure increased the risk of initiating injections of heroin (0.129) and crac k co caine
(0.015).
Co c aine. P o wder co caine injection exp osure increased the risks of initiating injections of
methamphetamine (0.006) and NMPD opioids (0.003), but these effects w ere not substan-
tiv e. Exp osure to crac k co caine injection increased the risk of methamphetamine injection
initiation (0.017). There w ere no asso ciations b et w een p o wder co caine and crac k co caine as
either injection exp osures or injection initiation outcomes.
Non-me dic al pr escription drugs. Generally , NMPD injection exp osures with significan t ef-
fects substan tiv ely increased risks of injection initiation outcomes with the exception of
the negativ e but not substan tiv e effect of NMPD sedativ es injection exp osure on injection
initiation risk of buprenorphine (-0.001). In addition to the effects of NMPD opioids in-
jection exp osure describ ed previously , other effects of NMPD injection exp osures include
NMPD stim ulan ts injection exp osure increasing the risk of p o wder co caine injection ini-
tiation (0.072); NMPD sedativ es injection exp osure greatly increasing the risk of heroin
injection initiating (0.441); and NMPD tranquilizers injection exp osure increasing the risks
of initiating injections of methadone (0.013) and NMPD opioids (0.056).
6.4 Discussion
The sample of PWID w as div erse demographically a nd in their exp eriences with injec-
tion drug use. Injection drug use b eha viors app eared to b e asso ciated with individual at-
tributes. Injection initiation risk for prescription medications, methamphetamine, and p o w-
der co caine app eared to b e generally higher for White participan ts whereas heroin injec-
tion initiation risk w as highest among Latino participan ts. Being foreign b orn reduced haz-
ards for most outcomes, particularly for injection initiation of either heroin or p o wder co-
caine. Although methamphetamine is prev alen t among men who ha v e sex with men ( Garo-
falo, Mustanski, McKirnan, Herric k, & Donen b erg, 2007 ), there w as insufficien t evidence
to supp ort a higher risk of its injection initiation among ga y , lesbian, or bisexual partici-
89
Study 3: P ath w a ys of Injection Drug Use
pan ts. Ho w ev er, the risk of heroin injection initiation w as lo w er for this group.
The asso ciation b et w een demographic attributes a nd injection initiation risk ma y b e me-
diated b y so cial con texts and individual exp eriences, reflecting differences in drug use atti-
tudes, exp osure to drug use b eha viors, or access to drug use materials or kno wledge. Risk
ma y b e an indication of endemic b eha viors with comm unities defined b y demographic c har-
acteristics. Differences in the kind of exp osure to these comm unities ma y mo derate risk.
Heroin injection initiation risk w as highest for Latino participan ts compared to all other
races, but the added risk attributed to race virtually disapp ears for Latino participan ts not
b orn in the United States. F oreign-b orn Latino participan ts lik ely p ossess v ery differen t
exp eriences asso ciated with race than their US-b orn coun terparts and these exp eriences
lik ely affected their attitudes, exp osures, or access to heroin injection drug use ( Alv arez,
Jason, Olson, F errari, & Da vis, 2007 ).
Latino participan ts generally exhibited the highest risk for initiating injection heroin use,
after adjusting for other factors. This correlates with similar findings from the first study
( R esults (subsection 4.3, p. 26) ) on an y-route heroin use initiation. The higher initiation
risk sp ecifically for heroin use among Latino participan ts suggests differen t p ossible under-
lying mec hanisms. It is p ossible that transition rates from non-injection to injection use
of heroin are similar b et w een racial categories and the higher risk of initiating injection
heroin use reflects a greater starting prev alence of non-injection heroin use. Alternativ ely ,
it is p ossible that transitions to injection use are more lik ely among Latino p eople who use
heroin ( V aldez, Neaigus, Kaplan, & C ep eda, 2011 ). The findings offer some evidence in
supp ort of b oth of these explanations. Regardless of ho w heroin injection initiation risk
dev elops within this p opulation, the results directly define heroin injection initiation as a
distinct drug use health issue for Latino p opulations. This group ma y exhibit unique etio-
logical pro cesses through men tal health tra jectories ( Cep eda et al., 2012 ; No w otn y , P erdue,
Cep eda, & V aldez, 2017 ), acculturation pro cesses ( V aldez, Neaigus, & Cep eda, 2007 ), or
disparities in drug treatmen t access or a v ailabilit y ( Go edel et al., 2020 ).
Differences b y birth cohort rev eal some trends in injection drug use. Heroin, crac k co caine,
90
Study 3: P ath w a ys of Injection Drug Use
and methamphetamine injection initiations app ear to b e increasingly lik ely with more re-
cen t cohorts. Being part of the Eigh ties of later cohort app ears to more than double the
accum ulated baseline hazards for injection initiations of heroin and methamphetamine.
In the case of crac k co caine injection initiation, b eing a part of this cohort w as lik ely the
greatest risk factor. NMPD opioids and tranquilizers injection initiations app eared to fol-
lo w similar but non-significan t trends. Ho w ev er, the injection initiations of some drugs ex-
p erienced declining lik eliho o ds o v er successiv e cohorts as with NMPD sedativ es and stim-
ulan ts. Notably , there app eared to b e no cohort differences in injection initiation risk for
p o wder co caine. The higher injection initiation risks of heroin and NMPD opioids for later
birth cohorts app ear to corresp ond to the curren t opioid drug use epidemic. Similarly , the
inno v ation of crac k co caine and its rapid rise in p opularit y in the 1980s ma y explain co-
hort differences in crac k co caine injection initiation risk and the absence of cohort differ-
ences for p o wder co caine injection initiation risk.
The c hanges in effects b et w een the full v ersus the restricted mo dels suggests that initiat-
ing injection drug use in general is a stronger risk factor for some outcomes than initiating
injection for an y sp ecific drug. Results of the restricted mo del detected significan t asso cia-
tions among injection b eha viors in v olving p o wder co caine, heroin, and methamphetamine
but these asso ciations w ere greatly atten uated with the in tro duction of a general injection
drug use exp osure predictor. F urthermore, based on results of the full mo dels, the risks
of initiating injection of an y of these three drugs w as greater among PWID v ersus non-
injecting PWUD. One p ossible explanation is an individual mec hanism in whic h the ini-
tiation of the injection drug use career increases individual risks of initiating injection of
other drugs. The core skills of injection can b e generally applied to man y drugs and learn-
ing to inject, regardless of drug, creates the skill capital to initiate injection of most drugs.
Although preparation metho ds for injection v ary b y drug, these skills are lik ely not as dif-
ficult to access or acquire as the core injection skills. Another p ossible explanation is a so-
cial mec hanism in whic h engaging in injection drug use pro vides so cial opp ortunities to ini-
tiate injection of other drugs. Engaging in injection drug use ma y p ositiv ely influence the
dev elopmen t of so cial ties to other PWID, regardless of the drugs injected, and the result-
91
Study 3: P ath w a ys of Injection Drug Use
ing so cial exp osures to other injection drug use b eha viors can translate in to opp ortunities
for initiating injection of other drugs.
Despite the substan tial risk con tributed b y general injection drug use exp osure, exp osure
to drug-sp ecific injection drug use did exhibit significan t and substan tiv e indep enden t ef-
fects on the risks of injection initiations. The asso ciations among exp osures and outcomes
pro vides some evidence for risk path w a ys based on drug-sp ecific injection. Generally , exp o-
sure to non-injection use of a drug w as almost alw a ys a risk factor for initiating injection
for that drug. Also, exp osures and outcomes of drugs with similar depressan t or stim u-
lan t effects app eared to b e more lik ely to exhibit an asso ciation. Exp osures to injection
of NMPD opioids and NMPD sedativ es, b oth depressan ts, w ere risk factors for injection
initiation of heroin, whic h is also a depressan t, although the infrequency of NMPD seda-
tiv es limits the in terpretation of its exp osure effects. Similarly , exp osure to injection of
NMPD stim ulan ts and p o wder co caine resp ectiv ely increased the risks for injection initi-
ation of p o wder co caine and methamphetamine, all of whic h are stim ulan ts. Evidence for
the opp osite relationship, in that drugs of dissimilar effects for exp osures and outcomes do
not exhibit a p ositiv e asso ciation, can b e found in the negativ e effect of heroin injection
exp osure on the risk of initiating methamphetamine injection.
A dditional supp ort for existence of path w a ys c ome from the apparen t unidirectional asso-
ciations of risk. Most asso ciations app eared to function in only one direction: exp osure to
one injection drug use b eha vior increased the risk of initiating another but not vice v ersa.
F o r example, exp osure to crac k co caine injection increased the risk of methamphetamine
injection initiation but exp osure to methamphetamine injection did not app ear to increase
the risk of crac k co caine injection initiation. These phenomena imply temp orally ordered
path w a ys in whic h pairs of b eha viors are more lik ely to o ccur in particular sequence. In-
jection drug use exp osures in v olving NMPDs w ere more often significan t and substan tiv e
risk factors for illicit drug injection initiations but exp osures in v olving illicit drugs did not
app ear to affect the risk of injection initiations in v olving NMPDs. Although injection drug
use b eha viors in v olving NMPDs is neither a sufficien t nor necessary cause of illicit drug
92
Study 3: P ath w a ys of Injection Drug Use
use injection initiation, their precedence and their apparen t effects suggest that they ma y
ha v e meaningful roles in the dev elopmen t of the injection drug use career. Ho w ev er, the
relativ ely lo w frequency of injection initiation ev en ts for all NMPDs except opioids limits
the generalizabilit y of this finding. Regardless, this ma y w arran t further study on the role
of NMPDs within the injection drug use career.
There w as, ho w ev er, one notable exception to this pattern. The findings include a signif-
ican t and substan tiv e p ositiv e effect from injection drug use of NMPD opioids on the in-
jection initiation risk of crac k co caine. This asso ciation w as unique in its singularit y as
no other asso ciations of its kind w ere detected and a similar effect w as not detected for
p o wder co caine injection initiation risk. It is p ossible that this reflects the kno wn o v erlap
b et w een co caine and opioids use, namely heroin use ( Leri, Bruneau, & S tew art, 2003-01 ),
but the sp ecificit y of this asso ciation within injection drug use b eha viors suggests a unique
risk pro cess. A subsequen t stratified analysis b y birth cohort (not sho wn) detected signifi-
can t effects for the Pre-Sixties and Sixties cohorts, suggesting that most of this asso ciation
could b e attributed to these t w o groups. If this is the case, then this asso ciation indicates
a generational effect, where earlier cohorts first initiated injection drug use of NMPD opi-
oids and, subsequen tly in the 1980s, initiated injection of crac k co caine when the latter
b ecame a v ailable.
The o v erall findings indicate that there are definable pa th w a ys within the injection drug
use career, deriv ed from the risk asso ciations among injection drug use b eha viors. This
study pro vides some no v el con tributions to the understanding of the injection drug use
career. A discussion of limitations applicable to this and all other studies comprising this
dissertation is pro vided in Over al l Limitations (subsection 7.1, p. 95) . Injection drug use
b eha viors app ear to b e asso ciated and these asso ciations app ear to b e ordered within de-
v elopmen tal path w a ys. These findings ma y inform efforts to prev en t further or additional
exp osure to injection-related harm among PWID. This is particularly relev an t to presen t
circumstances, giv en an apparen t rise in injection drug use among y ounger PWUD. More
broadly , the findings supp ort the dev elopmen tal understanding of the injection drug use
93
Study 3: P ath w a ys of Injection Drug Use
career as a useful a v en ue of inquiry . The dep endence of risk on drug-related factors sug-
gest that the injection drug use career ma y unfold along somewhat predictable path w a ys,
but these path w a ys ma y neither b e simple nor direct and ma y dep end on spatial and tem-
p oral con text. F uture researc h can help construct a b etter mo del of risk for injection drug
use b eha viors b y accoun ting for geographical differences and using more con temp orary
samples of PWID.
94
Conclusion
7 Conclusions
The o v erall goal of this dissertation w as to in v estigate the dev elopmen tal path w a ys of the
drug use career. This w as accomplished through three studies using similar approac hes.
The first study in v estigated asso ciations from an y-route drug use exp osures on risks of
an y-route drug use initiations. The second study examined the role of non-injection drug
use exp osures on the risks of initiation injection drug use, or first injection exp erience, in
general (i.e., an y-cause) and sp ecifically b y first drug injected (i.e., cause-sp ecific). The
third and last study in v estigated asso ciations b et w een injection drug use exp osures and
risks of initiations of injection drug use b eha viors. All three studies used age-based drug
use history data collected from a div erse sample of PWID and relied on a similar mo del-
ing approac h consisting of surviv al analysis using the additiv e hazards mo del with time-
dep enden t exp osures.
The com bined results from all three studies gene rally p oin t to the existence of definable
path w a ys of risk within the drug use career delineated b y drug use b eha viors. This is in
addition to risk attributable to non-drug related factors including individual demographic
traits and generational con text. Not only are there relationships b et w een drug use b eha v-
iors but these relationships are ordered: a particular b eha vior ma y app ear to increase the
risk of initiation another but not the rev erse. F rom the first study , NMPD opioids use ex-
p osure app eared to increase the risk of initiating heroin use, but heroin use exp osure did
not app ear to increase, or c hange, the risk of initiating NMPD opioids use. Dev elopmen tal
path w a ys did not app ear to b e limited to sp ecific ranges of the drug use career or to sp e-
cific sets of b eha viors and initiations of b oth an y-route drug use b eha viors and injection
drug use b eha viors app ear to b e influenced b y prior drug use exp osures.
7.1 Ov erall Limitations
All three studies rely on the same s ource data and theoretical framew ork and use similar
metho dological approac hes and, consequen tly , are sub ject to similar limitations whic h
constrain the in terpretation or applicabilit y of the findings. The analyses rely on data
95
Conclusion
rep orted b y participan ts who injected drugs. The in terpretations of the results reflect
the exp eriences of PWID, who ma y represen t a subset of PWUD with more sev ere drug
use ( Mac k esy-Amiti, F endric h, & Goldstein, 1997 ). The reliance on targeted sampling
metho ds resulted in a nonrandom sample or participan ts who ma y ha v e exhibited some
self-selection. A dditionally , the sampling of participan ts ma y suffer from surviv al bias as
the resulting sample included only those who surviv ed their drug use and who con tin ued
their drug use. Drug use histories, whic h formed a critical comp onen t of the analyses, w ere
based on recall and so ma y b e susceptible to recall bias. Ho w ev er, a separate analysis
found high reliabilit y b et w een drug use histories from b oth surv ey and qualitativ e in ter-
view p ortions of the paren t study , suggesting that the effect of recall bias w as small ( Dy al,
Kral, Gonzalez, W enger, & Bluthen thal, 2015 ). It is also imp ortan t to note here that
exp osures w ere op erationalized as the start of exp osure to drug use exp eriences and do
not necessarily mean constan t use. The initiation of use of a particular drug use b eha vior
do es not imply con tin ued use and this ma y ha v e resulted in some of these b eha viors
b eing o v erstated in participan ts’ drug use exp osures. Some imp ortan t factors w ere not
examined as they w ere not collected or did not p ossess the sufficien t co v erage or structure
for their inclusion. No age-based usage histories w ere collected for alcohol or tobacco use,
b oth of whic h are kno wn factors for drug use tra jectories. There w ere also no age-based
histories collected for exp osures to trauma or exp eriences with criminal activit y or with
the criminal justice system, whic h w ere also iden tified as factors for initiations of drug use
b eha viors.
The mo deling approac h constrained a common genera l sp ecification across all outcomes
across all studies. This allo w ed for comparisons b et w een mo dels but ma y ha v e resulted
in missp ecification and some time-v arying effects ma y ha v e b een b etter represen ted as
constan t effects or vice v ersa. The mo dels also reflect direct effects of biv ariate asso cia-
tions, and the discussion of path w a ys should tak e this in to accoun t. A c hain of asso cia-
tions among initiations do not pro vide direct evidence of a path w a y delineated b y these
imitations but do es pro vide supp ort that one exists. In teraction effects w ere not tested
but it is highly lik ely that they exist, sp ecifically b et w een exp osure predictors and birth
96
Conclusion
cohorts. Consequen tly , the asso ciations found b et w een exp osures and outcomes ma y b e
differen t b y generation.
The three studies are, essen tially at v arying deg rees, case-con trol analyses without con trols.
Ho w ev er, these limitations ma y b e reasonable considering that some of these issues are
not uncommon in researc h on injection drug use. Studying stigmatized health b eha viors
is already a difficult endea v or whic h b ecomes more difficult when the b eha viors of in ter-
est in v olv e criminal asso ciations or consequences. This mak es random sampling difficult as
w ell and ma y complicate follo w up efforts necessar y for longitudinal observ ational studies.
Prosp ectiv ely studying b oth the con tributions of p erson and historical time w ould require
follo w up o v er longer timescales to observ e sufficien t coun ts of ev en ts and large n um b ers
of participan ts to observ e this within a random sample. A dditionally , history is a con tin-
uous and unpredictable pro cess and ma jor trends can b e o v erlo ok ed or missed b ecause of
c hance. The data for the studies w ere collected un til 2013 and do not co v er the later rise
of fen tan yl use ( Ciccarone, 2019 ) and the remarkable increase in drug-related mortalities
( Ciccarone, 2021 ). Studying these unexp ected trends sometimes requires retrosp ectiv e ap-
proac hes.
Although the limitations are substan tial, the analyses do p ossess some strengths. The se-
lection of participan ts w as based on drug use b eha viors based on route of administration
and not b y drugs used, yielding a sample div erse in drug use histories and not constrained
b y a sp ecific drug use outcome. A dditionally , the sample reflected a div ersit y of exp eri-
ences of injection drug use starting at v arious ages and v arious historical con texts and in-
cluded injection drug use initiations o ccurring within broad ranges of times. Although the
findings could only b e in terpreted within con text of PWID, the disprop ortionate vulner-
abilit y of this p opulation to a n um b er of adv erse outcomes ma y justify this constrain t of
the in terpretations. This do es not mean that these limitations should b e simply accepted
as a giv en, instead this highligh ts a need for more longitudinal researc h designs within this
domain.
97
Conclusion
7.2 Metho dological Implications
7.2.1 Confirming the Imp ortance of Time
The studies of this dissertation used similar a pproac hes based on a time-to-ev en t analysis
with time-dep enden t v ariables and time-v arying effects . The results of the analyses con-
firm that time is a ma jor factor in the drug use career. The drug use career exp eriences
c hange o v er time and this c hange is not alw a ys constan t. Results from the studies sho w
that c hange o ccurs o v er p erson time, follo wing age, and o v er historical time, as genera-
tional effects. A dditionally , the observ ed differences in asso ciations among drug use b e-
ha vior based on ordering indicate that timings of ev en ts relativ e to other ev en ts are also
imp ortan t.
Understanding the risk factors of drug use ma y require the con textualizing drug use ev en ts
within at least three la y ers of time: calendar, p erson, and drug use career. Drug use b e-
ha viors emerge within con text of historical drug use trends whic h shap e larger forces of
supply and demand. Individuals initiate drug use b eha viors at v arious p oin ts of the life-
time, with risk hea vily w eigh ted to w ards adolescence and y oung adultho o d. And drug use
b eha viors b egin after and b efore other b eha viors within the drug use career with p oten tial
asso ciations among them. It ma y b e as imp ortan t to understand when an ev en t o ccurs in
addition to if it o ccurs and what factors con tribute to its o ccurrence.
The general statistical metho ds used b y the studi es of this dissertation pro vides one p ossi-
ble approac h to in tegrating v arious la y ers of time in the analysis of drug use outcomes us-
ing surviv al mo deling. Calendar time w as sp ecified as generational cohort effects; the time
scaling of the mo deling represen ted p erson time; and drug use career time w as op erational-
ized as the time-dep enden t exp osures to v arious drug use b eha viors o ccurring prior to a
particular outcome. This approac h is not exp ected to b e appropriate for all or most situ-
ations but pro vides a demonstration of ho w to in tegrate v arious t yp es of time to studying
drug use outcomes.
98
Conclusion
7.2.2 Understanding Surviv al Through Other Means
One ma jor h urdle to in tegrating time to surviv al analysis are the limitations of the Co x
prop ortional hazards mo del compared to other surviv al mo deling approac hes. This is not
to sa y that the Co x mo del is in an y w a y not a v alid approac h as a long history of assess-
ing its prop erties has justified its p osition as a reliable and con v en tional statistical to ol in
time-to-ev en t analysis. Most of the elemen ts in the general analytic approac h of this disser-
tation can b e implemen ted using the Co x mo del ( Jac kson & Janssen, 2019 ). Ho w ev er, the
Co x mo del ma y not b e an appropriate c hoice for all settings.
This dissertation demonstrates one suc h situation. Namely , in tractable violations of the
prop ortional hazards assumption precluded the use of the Co x mo del in the surviv al anal-
yses. As these violations often indicate a time-v arying asso ciations b et w een co v ariates and
the hazard function, the additiv e surviv al mo deling w as a more appropriate approac h b e-
cause it allo w ed for time-v arying effects. A dditionally , its effects on the additiv e scale al-
lo w ed for comparisons b et w een parallel mo dels of differen t outcomes in units of hazard.
In con trast, the effects of the Co x mo del are in the m ultiplicativ e scale as hazard ratios
with a substan tially differen t in terpretation b et w een outcomes due to differences in base-
line hazard functions. The additiv e mo del ma y also b e more robust to confounding bias
( Martin ussen & V ansteelandt, 2013 ), although this w as not something that w as considered
for its selection.
This is not to sa y that the additiv e mo del is sup e rior or that the Co x mo del is inferior, but
that there exists a plethora of established surviv al mo deling approac hes for v arious t yp es
of data and researc h questions. As violations of the prop ortional hazards assumption can
undermine the v alidit y of results generated through the Co x mo del, considering these alter-
nativ es represen ts b oth go o d practice and go o d science. F or cases in whic h no assumptions
are violated, these alternativ es offer differen t approac hes to asking or answ ering imp ortan t
researc h questions. Consider the accelerated failure time mo del, whic h is somewhat simi-
lar to the Co x mo del but is in terpreted as the effects of co v ariates on the course of an out-
come through acceleration or deceleration ( W ei, 1992 ).
99
Conclusion
7.3 Theoretical Implications
7.3.1 In tegrating Comp eting Theories of Risk F actors
The findings lend supp ort to the core premise of t he Gatew a y Hyp othesis in that drug use
b eha viors ma y exhibit ordered asso ciations within the drug use career. There has b een
substan tial debate on whether the “gatew a y effect” is a v alid or true phenomenon, that
these asso ciations among b eha viors w ere observ ed while adjusting for the indep enden t con-
tributions of individual traits and generational cohort effects suggest that there are indeed
true asso ciations among b eha viors. Ho w ev er, the existence of these asso ciations do es not
imply similar risk pro cesses that asso ciate b eha viors. In conclusion, the Gatew a y Hyp oth-
esis framew ork, Common Liabilit y to A ddictions mo del, and the Drug Generations frame-
w ork are complemen tary , v ersus comp eting, explanatio ns to drug use risk.
In fact, the findings supp ort the premises of all the comp onen t theories describing the do-
mains of drug, individual, and con text factors. All three studies found significan t c hanges
in risk based on the indep enden t effects of factors from all three domains. This immedi-
ately suggests that the ideas cen tral to eac h understanding eac h domain of risk factors are
v alid and useful. This also suggests that understanding the drug use career should tak e
in to accoun t all of these domains and p ossibly more. Generally , the w ork of this disser-
tation found that the details of drug use b eha viors and drug use careers are imp ortan t
enough to engage in less reductionist approac hes.
7.3.2 F rom P ath w a ys to Pro cesses
Understanding the path w a ys of drug use pro vides a map of risk o v er the drug use career,
ho w ev er this map only traces a path and do es not pro vide detail on ho w mo v emen ts o c-
cur b et w een drug use initiations. Ho w ev er, the attributes of the drug, the p erson, and the
con text pro vide a few clues as to the functioning of the underlying pro cesses of risk. There
w ere t w o main categories of risk mec hanisms discus sed across all three studies: individual
risk dev elopmen t and so cial risk dev elopmen t. Individual risk dev elopmen t in v olv es indi-
vidual tra jectories of drug use prop ensities or liabilities. So cial risk dev elopmen t represen ts
100
Conclusion
the exp osure to drug use b eha viors or acquisition of drug use materials or metho ds. Both
individual and so cial risk are necessary in that individual lik eliho o ds need to b e matc hed
with so cial opp ortunities. They are m utually in teracting in that individual risk can induce
so cial risk and vice v ersa. And they are self-repro ductiv e: risk is conduciv e to dev eloping
more risk.
The P A CE mo del ( Sussman et al., 2011 ) pro vides a more formal, and b etter, t heoretical
framew ork for these ideas b y further distinguishing individual and so cial risk pro cesses in to
more sp ecific comp onen ts: pragmatics, attraction, comm unication, and exp ectation. Drug
use initiation risks, and their repro duction, emerge as a result of the in teraction of these
four comp onen ts. F or initiation to o ccur, a p erson m ust ha v e access to drug use materials
via so cial relationships to illicit sources (pragmatics); p ossess a desire, in terest, or prop en-
sit y to initiate or use (attraction); b e exp osed to so cial opp ortunities to initiate drug use
or acquire the metho ds of drug use (comm unication); and exp ect the effect of the drug will
fulfill prior exp ectations (exp ectations).
The concepts of the P A CE mo del can explain ho w som e path w a ys emerge. The asso cia-
tions among opioid drugs throughout all three studies inform one h yp othetical situation.
Pleasurable effects of opioid medications can motiv ate the transition from their medical to
non-medical use (attraction). Diminished supply of opioid medications can lead to switc h
to more a v ailable or accessible alternativ es with similar effects lik e heroin (exp ectation)
whic h require access to illicit sources (prag matics) and the abilit y to so cially negotiate
through drug use so cial net w orks (comm unication). Chronic use of heroin and exp osure
to other p eople who use heroin increases access to drug use materials (pragmatics), solidify
prop ensit y for heroin-based drug use b eha vior (attraction), and dev elops individual exp er-
tise within the so cial en vironmen t (comm unication). This gro wth tra jectory can increase
the future lik eliho o d of initiation injection drug use using heroin in order to ac hiev e more
in tense effects (exp ectation).
## Practical Implications
101
Conclusion
7.3.3 The Diminishing Role of Cannabis
Data collection of this dissertation’s source data, via the paren t study , coincided with
the start of state-lev el legalization of recreational cannabis use in the United States. In
California, cannabis has b een legally a v ailable for recreational use since 2016 and medical
use since 1996 ( California, 2021 ). V ery few participan ts, if an y , initiated cannabis use
with cannabis purc hased from legal disp ensaries. Since the conclusion of the paren t study ,
cannabis can b e legally purc hased for recreational purp oses in man y states although it
remains illegal at the federal lev el. The apparen tly inevitable legalization of cannabis will
lead to broader a v ailabilit y and accessibilit y of cannabis use through increased legal access
to cannabis pro ducts that are regulated for dose and safet y . The immediate consequence
of increased access and a v ailabilit y is a commensurate increase in recreational cannabis use
( Cerdá et al., 2020 ).
But will an exp ected future increase in cannabis use then b e follo w ed b y an increase in in-
cidence of illicit drug use b eha viors? The findings of this dissertation sho w a clear risk re-
lationship from cannabis use on man y illicit drug use outcomes and gra vitates to w ards an
affirmativ e answ er. Ho w ev er, these findings w ere based on illicit cannabis use whereas licit
cannabis use is exp ected to driv e future increases recreational cannabis use. F orecasting
the future role of cannabis use on the drug use career dep ends on ho w cannabis use is asso-
ciated with other drug use b eha viors.
This can b e explored using constructs of the P A CE mo del discussed ab o v e. Licit cannabis,
b eing sold at dedicated and regulated legal v en ues, do es not impro v e access (pragmatics)
to illicit drugs lik e heroin or co caine. Ho w ev er, it is p ossible for illicit cannabis to ha v e a
common supply source as other illicit drugs. Exp osure to cannabis can increase individ-
ual prop ensities (attraction and exp ectations) to other drug use b eha viors b y pro ducing a
desire for enjo y able or in teresting men tal or ph ysical effects and this effect w ould not de-
p end on whether the source of cannabis w as licit or illicit. Both licitly and illicitly sourced
cannabis will lik ely increase exp erience and exp ertise of cannabis use and abilit y to so cially
in terface with other p eople who use cannabis (comm unication), and these other p eople
102
Conclusion
ma y pro vide opp ortunities to initiate in other illicit drug use b eha viors. As cannabis is
t ypically one of the first, if not the first, illicit drugs used among participan ts used in the
analyses of this dissertation, the observ ed gatew a y effects of cannabis use ma y b e a com bi-
nation of effects unique to cannabis use and effects more generalizable to illicit drug use.
The idea of the “gatew a y drug” has b ecome so e n trenc hed in p opular discourse that
cannabis is commonly accepted as a gatew a y drug as a giv en fact. There is not y et evi-
dence to conclusiv ely determine whether cannabis use will con tin ue to exert a significan t
role in the drug use career ( Smart & P acula, 2019 ), but its role as a gatew a y drug is
lik ely to diminish as it transitions from an illicit drug to a licit drug with broader so cial
acceptabilit y . Results from the 2020 NSDUH ( Abuse & A dministration, 2020a ) sho w that
cannabis use has b een steadily increasing but this increase do es not app ear to visibly
correlate with concurren t or subsequen t increases in heroin, co caine, or methamphetamine
use. There is evidence that cannabis use offsets illicit opioid use among p eople who use
opioids ( Alex H. Kra l et al., 2015 ) and it is p ossible this ma y b ecome a more widespread
phenomenon with greater and safer legal access to cannabis. In Colorado, cannabis legal-
ization w as asso ciated with a short-term reduction in opioid o v erdose death ( Livingston,
Barnett, Delc her, & W agenaar, 2017 ). This reduction ma y b e a result of the substitution
of opioid use with cannabis use or through other pro cesses suc h as separating access to
cannabis from illicit drugs. Assessing the long-term consequences of cannabis use on illicit
drug use b eha viors requires further observ ation.
7.3.4 The Rising Role of NMPDs
In con trast to the p ossible diminishing role of cannabis, the role of NMPDs ma y b e ris-
ing. The findings of all studies suggest that NMPD use ha v e significan t roles in the drug
use career. Presen tly , the role of NMPD opioids use has b een w ell-studied, in ligh t of the
ongoing opioid use epidemic, but it is unlik ely to b e the only con tributing factor among
NMPD use b eha viors. Increased prescribing of opioids for pain managemen t has b een at-
tributed as one of the causes of the rise of opioids use and dep endence. Efforts to c hange
prescribing practices app ear to b e effectiv e in reducing the o v erall p opulation exp osure
103
Conclusion
to prescription opioids ( Kern, Cep eda, & Sena, 2020 ; P ezalla, Rosen, Erensen, Haddo x,
& Ma yne, 2017 ). Ho w ev er, recen t trends suggest that the use of prescription stim ulan ts
and other psyc hotropic drugs is increasing ( Maust, Blo w, Wiec hers, Kales, & Marcus, 2 017 ;
Pip er et al., 2018 ). Although the relationships b et w een the non-medical use of these drugs
and other illicit drug use ha v e not b een firmly established, the findings from the studies of
this dissertation do not suggest an y optimistic conclusions.
7.4 F uture Researc h Directions
Drug use is a result of individual and so cial phenomena and will reliably c hange to reflect
the needs, desires, and attitudes of individuals and so cieties. This dissertation relies on
data from a sample of PWID from sp ecific region and of a sp ecific time and there are lim-
its to the applicabilit y of the findings to other con texts. A dditional study on drug use de-
v elopmen t in differen t p opulations and con texts will con tribute to this a v en ue of inquiry
b y pro viding further depth or range to this b o dy of w ork or, alternativ ely , b y offering com-
p eting explanations.
More imp ortan tly , an understanding of ho w path w a ys emerge in addition to what path-
w a ys emerge will b e a more critical endea v o r. Drug use path w a ys are the sup erficial and
observ able asp ects of deep er and more meaningful pro cesses and elucidating these pro-
cesses and determining ho w they ma y b e redirected are core to drug use prev en tion efforts.
Although the endea v or to understand these unde rlying pro cesses has b een underw a y prac-
tically ev er since the adv en t of drug use, what is suggested here are researc h directions in-
cluding the implications describ ed previously and the consideration that drug use is the
result of m ultiple con tributing, and lik ely in teracting, causes from com binations of individ-
ual, drug, and con textual factors.
7.5 Concluding Remarks
In review, the drug use career can b e describ ed using drug use path w a ys defined b y initia-
tions of drug use b eha viors. Prior drug use exp osures affect the risk of drug use initiations
104
Conclusion
with v aried asso ciations b et w een exp osures on o utcomes. The risk effects of exp osures o c-
cur in addition to risk con tributions b y individual trait attributes suc h as sex, sexual ori-
en tation, race, and birthplace and risk differences b y birth cohort. The risk asso ciations
b et w een exp osures and outcomes can b e due to individual risk pro cesses suc h as the dev el-
opmen t of drug use preferences but ma y also b e the result of so cial risk pro cesses b et w een
drug use b eha viors and drug use so cial net w orks. These findings can inform future studies
on the dev elopmen t of the drug use career and ho w drug use path w a ys emerge as w ell as
in terv en tions to prev en t initiations of drug use b eha viors.
105
App endix
A Study 1 Detailed T ables
T able A.1: Study 1: Detailed suprem um p-v alues for
significance of time-v arying effects
Outcome T erm Suprem um-test P-v alue
In tercept 10.657 0.000
Birth cohort (Sixties) 3.446 0.019
Birth cohort (Sev en ties) 2.961 0.075
Birth cohort (Eigh ties or later) 2.086 0.525
Sex (female) 3.807 0.004
Sexual orien tation (GLB) 2.060 0.488
Birthplace (foreign b orn) 6.041 0.000
Race (Blac k) 5.349 0.000
Race (Latino) 3.656 0.009
Cannabis
Race (Other) 2.167 0.467
In tercept 6.609 0.000
Birth cohort (Sixties) 4.153 0.001
Birth cohort (Sev en ties) 4.352 0.001
Birth cohort (Eigh ties or later) 2.714 0.135
Sex (female) 2.648 0.160
Sexual orien tation (GLB) 3.952 0.002
Birthplace (foreign b orn) 4.758 0.000
Race (Blac k) 1.946 0.521
Race (Latino) 4.189 0.001
Heroin
Race (Other) 2.078 0.423
In tercept 4.790 0.000
Birth cohort (Sixties) 7.662 0.000
Birth cohort (Sev en ties) 6.798 0.000
Birth cohort (Eigh ties or later) 6.356 0.000
Sex (female) 2.032 0.451
Sexual orien tation (GLB) 1.849 0.540
Birthplace (foreign b orn) 2.826 0.071
Race (Blac k) 2.565 0.146
Race (Latino) 2.302 0.282
106
App endix
Crac k co caine
Race (Other) 2.219 0.298
In tercept 5.642 0.000
Birth cohort (Sixties) 4.892 0.000
Birth cohort (Sev en ties) 3.949 0.003
Birth cohort (Eigh ties or later) 4.075 0.001
Sex (female) 1.454 0.902
Sexual orien tation (GLB) 2.293 0.318
Birthplace (foreign b orn) 4.955 0.000
Race (Blac k) 3.638 0.009
Race (Latino) 1.967 0.586
P o wder co caine
Race (Other) 3.472 0.014
In tercept 6.548 0.000
Birth cohort (Sixties) 1.914 0.512
Birth cohort (Sev en ties) 4.348 0.000
Birth cohort (Eigh ties or later) 5.831 0.000
Sex (female) 3.104 0.043
Sexual orien tation (GLB) 2.018 0.482
Birthplace (foreign b orn) 3.256 0.027
Race (Blac k) 5.691 0.000
Race (Latino) 5.601 0.000
Methamphetamine
Race (Other) 3.853 0.004
In tercept 2.177 0.257
Birth cohort (Sixties) 1.719 0.601
Birth cohort (Sev en ties) 2.233 0.210
Birth cohort (Eigh ties or later) 3.777 0.004
Sex (female) 2.625 0.115
Sexual orien tation (GLB) 1.718 0.527
Birthplace (foreign b orn) 3.092 0.024
Race (Blac k) 1.776 0.538
Race (Latino) 4.254 0.000
Methadone
Race (Other) 2.487 0.132
In tercept 2.231 0.140
Birth cohort (Sixties) 2.294 0.097
Birth cohort (Sev en ties) 2.763 0.052
107
App endix
Birth cohort (Eigh ties or later) 4.777 0.000
Sex (female) 1.529 0.585
Sexual orien tation (GLB) 1.987 0.306
Birthplace (foreign b orn) 0.982 0.816
Race (Blac k) 2.099 0.234
Race (Latino) 2.783 0.046
Buprenorphine
Race (Other) 2.734 0.052
In tercept 5.724 0.000
Birth cohort (Sixties) 1.740 0.661
Birth cohort (Sev en ties) 2.780 0.084
Birth cohort (Eigh ties or later) 5.247 0.000
Sex (female) 1.832 0.580
Sexual orien tation (GLB) 2.115 0.331
Birthplace (foreign b orn) 4.934 0.000
Race (Blac k) 5.289 0.000
Race (Latino) 2.426 0.232
Opioids
Race (Other) 2.508 0.161
In tercept 5.452 0.000
Birth cohort (Sixties) 4.220 0.001
Birth cohort (Sev en ties) 2.485 0.145
Birth cohort (Eigh ties or later) 2.779 0.081
Sex (female) 2.107 0.314
Sexual orien tation (GLB) 2.623 0.101
Birthplace (foreign b orn) 2.978 0.033
Race (Blac k) 4.373 0.001
Race (Latino) 4.475 0.000
Stim ulan ts
Race (Other) 2.164 0.209
In tercept 6.682 0.000
Birth cohort (Sixties) 3.099 0.024
Birth cohort (Sev en ties) 3.613 0.005
Birth cohort (Eigh ties or later) 4.430 0.000
Sex (female) 2.276 0.234
Sexual orien tation (GLB) 1.385 0.692
Birthplace (foreign b orn) 5.071 0.000
108
App endix
Race (Blac k) 3.709 0.005
Race (Latino) 4.026 0.002
Sedativ es
Race (Other) 3.940 0.002
In tercept 4.013 0.002
Birth cohort (Sixties) 1.576 0.761
Birth cohort (Sev en ties) 1.875 0.500
Birth cohort (Eigh ties or later) 4.386 0.000
Sex (female) 1.455 0.829
Sexual orien tation (GLB) 1.649 0.634
Birthplace (foreign b orn) 3.989 0.002
Race (Blac k) 4.483 0.000
Race (Latino) 3.361 0.018
T ranquilizers
Race (Other) 2.269 0.261
Note:
P-v alues less then or equal to 0.05 are shaded in gra y .
Referen t birth c ohort: Pre-Sixties.
Referen t race c ategory: White.
T able A.2: Study 1: Detailed parametric estimates
Outcome T erm Estimate Std. Error Z-statistic P-v alue
Heroin 0.011 0.025 0.417 0.677
Crac k Co caine 0.026 0.008 3.030 0.002
P o wder Co caine -0.006 0.007 -0.625 0.532
Methamphetamine 0.009 0.014 0.672 0.501
Methadone 0.011 0.007 1.040 0.300
Buprenorphine -0.042 0.015 -2.840 0.005
NMPD Opioids 0.009 0.013 0.675 0.500
NMPD Stim ulan ts -0.012 0.020 -0.725 0.469
NMPD Sedativ es -0.025 0.019 -2.150 0.031
NMPD T ranquilizers 0.036 0.015 2.820 0.005
Cannabis
Injection Drug Use -0.053 0.025 -1.960 0.050
Cannabis 0.034 0.006 7.210 0.000
Crac k Co caine 0.017 0.009 1.850 0.064
P o wder Co caine 0.007 0.008 0.890 0.373
109
App endix
Methamphetamine -0.023 0.011 -2.470 0.014
Methadone 0.039 0.036 1.170 0.242
NMPD Opioids 0.030 0.012 2.750 0.006
NMPD Stim ulan ts 0.028 0.017 1.610 0.107
NMPD Sedativ es 0.004 0.015 0.295 0.768
NMPD T ranquilizers 0.030 0.014 2.310 0.021
Heroin
Injection Drug Use 0.037 0.012 3.380 0.001
Cannabis 0.028 0.003 8.500 0.000
Heroin -0.002 0.009 -0.196 0.845
P o wder Co caine 0.024 0.005 4.730 0.000
Methamphetamine -0.001 0.006 -0.147 0.883
Methadone 0.006 0.009 0.679 0.497
Buprenorphine 0.008 0.039 0.240 0.810
NMPD Opioids -0.002 0.006 -0.381 0.703
NMPD Stim ulan ts 0.010 0.008 1.360 0.174
NMPD Sedativ es 0.002 0.007 0.389 0.697
NMPD T ranquilizers 0.003 0.007 0.400 0.689
Crac k co caine
Injection Drug Use 0.014 0.009 1.530 0.125
Cannabis 0.038 0.005 8.300 0.000
Heroin -0.011 0.017 -0.750 0.453
Crac k Co caine -0.009 0.009 -0.972 0.331
Methamphetamine 0.000 0.012 -0.036 0.971
Methadone 0.009 0.020 0.422 0.673
Buprenorphine -0.003 0.047 -0.054 0.957
NMPD Opioids -0.001 0.012 -0.114 0.909
NMPD Stim ulan ts 0.096 0.024 4.350 0.000
NMPD Sedativ es 0.005 0.015 0.313 0.754
NMPD T ranquilizers 0.006 0.013 0.479 0.632
P o wder co caine
Injection Drug Use 0.012 0.017 0.773 0.440
Cannabis 0.017 0.003 6.000 0.000
Heroin -0.022 0.011 -2.210 0.028
Crac k Co caine 0.009 0.004 2.520 0.012
P o wder Co caine 0.010 0.004 2.700 0.007
Methadone 0.002 0.005 0.341 0.733
110
App endix
Buprenorphine -0.004 0.019 -0.273 0.785
NMPD Opioids 0.004 0.004 0.937 0.349
NMPD Stim ulan ts 0.002 0.006 0.293 0.769
NMPD Sedativ es 0.004 0.005 0.916 0.360
NMPD T ranquilizers 0.004 0.005 0.977 0.329
Methamphetamine
Injection Drug Use 0.008 0.011 0.784 0.433
Cannabis 0.002 0.001 1.270 0.205
Heroin 0.014 0.002 5.890 0.000
Crac k Co caine 0.000 0.002 -0.235 0.815
P o wder Co caine 0.004 0.002 2.170 0.030
Methamphetamine 0.003 0.002 1.230 0.218
Buprenorphine 0.070 0.029 2.500 0.013
NMPD Opioids 0.007 0.002 2.880 0.004
NMPD Stim ulan ts -0.003 0.003 -1.060 0.289
NMPD Sedativ es 0.005 0.003 1.540 0.123
NMPD T ranquilizers 0.009 0.003 3.320 0.001
Methadone
Injection Drug Use -0.007 0.003 -2.750 0.006
Cannabis -0.001 0.001 -2.000 0.045
Heroin 0.006 0.002 3.950 0.000
Crac k Co caine 0.000 0.001 0.450 0.653
P o wder Co caine 0.000 0.001 0.043 0.966
Methamphetamine 0.003 0.001 3.020 0.002
Methadone 0.009 0.002 5.090 0.000
NMPD Opioids 0.002 0.001 1.910 0.056
NMPD Stim ulan ts 0.005 0.002 2.580 0.010
NMPD Sedativ es -0.002 0.002 -1.610 0.108
NMPD T ranquilizers 0.005 0.001 3.930 0.000
Buprenorphine
Injection Drug Use -0.007 0.002 -3.890 0.000
Cannabis 0.010 0.002 4.150 0.000
Heroin 0.008 0.005 1.650 0.099
Crac k Co caine 0.001 0.003 0.427 0.669
P o wder Co caine 0.007 0.003 2.410 0.016
Methamphetamine 0.011 0.003 3.050 0.002
Methadone 0.007 0.005 1.420 0.157
111
App endix
Buprenorphine 0.017 0.029 0.539 0.590
NMPD Stim ulan ts 0.005 0.005 0.910 0.363
NMPD Sedativ es 0.004 0.005 0.699 0.485
NMPD T ranquilizers 0.021 0.005 4.040 0.000
Opioids
Injection Drug Use -0.011 0.005 -2.150 0.032
Cannabis 0.002 0.001 1.700 0.089
Heroin 0.000 0.002 0.008 0.994
Crac k Co caine -0.002 0.001 -1.910 0.056
P o wder Co caine 0.001 0.001 1.040 0.300
Methamphetamine 0.001 0.001 1.080 0.280
Methadone 0.000 0.001 -0.275 0.784
Buprenorphine 0.005 0.007 0.697 0.486
NMPD Opioids 0.004 0.001 2.750 0.006
NMPD Sedativ es 0.004 0.002 1.690 0.091
NMPD T ranquilizers -0.003 0.001 -1.900 0.057
Stim ulan ts
Injection Drug Use 0.000 0.002 -0.007 0.994
Cannabis 0.003 0.001 2.800 0.005
Heroin 0.001 0.002 0.618 0.537
Crac k Co caine -0.004 0.001 -3.060 0.002
P o wder Co caine 0.001 0.002 0.527 0.598
Methamphetamine 0.001 0.001 0.714 0.475
Methadone 0.000 0.001 -0.073 0.942
Buprenorphine 0.004 0.006 0.649 0.516
NMPD Opioids 0.004 0.001 2.510 0.012
NMPD Stim ulan ts 0.004 0.002 1.930 0.054
NMPD T ranquilizers 0.004 0.002 2.560 0.011
Sedativ es
Injection Drug Use -0.003 0.002 -1.460 0.143
Cannabis 0.008 0.002 3.570 0.000
Heroin 0.007 0.004 1.990 0.046
Crac k Co caine 0.002 0.002 0.711 0.477
P o wder Co caine 0.001 0.003 0.267 0.789
Methamphetamine 0.004 0.003 1.350 0.176
Methadone 0.019 0.005 3.900 0.000
Buprenorphine 0.037 0.028 1.060 0.290
112
App endix
NMPD Opioids 0.018 0.004 4.910 0.000
NMPD Stim ulan ts 0.009 0.005 1.850 0.065
NMPD Sedativ es 0.000 0.004 -0.090 0.928
T ranquilizers
Injection Drug Use -0.007 0.004 -1.800 0.072
Note:
Ro ws with significan t estimates are shaded in gra y (p-v alue <= 0.05).
Significan t estimates greater than or equal to 0.010 are prin ted in b old.
B Study 2 Detailed T ables
T able B.1: Study 2: Comp eting risks - Detailed supre-
m um p-v alues for significance of time-v arying effects
Outcome T erm Suprem um-test P-v alue
In tercept 5.607 0.000
Birth cohort (Sixties) 4.612 0.000
Birth cohort (Sev en ties) 4.098 0.001
Birth cohort (Eigh ties or later) 3.908 0.002
Sex (female) 1.329 0.932
Sexual orien tation (GLB) 4.684 0.000
Birthplace (foreign b orn) 3.702 0.005
Race (Blac k) 2.209 0.321
Race (Latino) 4.788 0.000
Heroin
Race (Other) 2.015 0.417
In tercept 2.911 0.050
Birth cohort (Sixties) 2.544 0.129
Birth cohort (Sev en ties) 2.875 0.060
Birth cohort (Eigh ties or later) 3.177 0.021
Sex (female) 3.222 0.020
Sexual orien tation (GLB) 1.495 0.719
Birthplace (foreign b orn) 1.925 0.388
Race (Blac k) 2.831 0.057
Race (Latino) 2.108 0.313
P o wder co caine
Race (Other) 3.380 0.010
In tercept 3.051 0.035
113
App endix
Birth cohort (Sixties) 2.054 0.389
Birth cohort (Sev en ties) 2.774 0.076
Birth cohort (Eigh ties or later) 3.168 0.032
Sex (female) 3.922 0.003
Sexual orien tation (GLB) 2.667 0.117
Birthplace (foreign b orn) 3.076 0.032
Race (Blac k) 3.509 0.011
Race (Latino) 3.595 0.008
Methamphetamine
Race (Other) 1.933 0.478
Note:
P-v alues less then or equal to 0.05 are shaded in gra y .
Referen t birth c ohort: Pre-Sixties.
Referen t race c ategory: White.
T able B.2: Study 2: Comp eting risks - Detailed para-
metric estimates
Outcome T erm Estimate Std. Error Z-statistic P-v alue
Cannabis 0.027 0.005 6.220 0.000
Heroin 0.086 0.020 4.680 0.000
Crac k Co caine 0.023 0.010 2.630 0.009
P o wder Co caine -0.012 0.007 -1.670 0.096
Methamphetamine -0.011 0.009 -1.220 0.223
Methadone 0.040 0.033 1.530 0.127
Buprenorphine -0.086 0.091 -1.200 0.230
NMPD Opioids -0.002 0.010 -0.181 0.856
NMPD Stim ulan ts 0.043 0.019 2.980 0.003
NMPD Sedativ es 0.011 0.015 0.922 0.357
Heroin
NMPD T ranquilizers 0.015 0.012 1.280 0.201
Cannabis 0.003 0.002 1.560 0.119
Heroin 0.005 0.007 0.745 0.456
Crac k Co caine -0.005 0.005 -1.010 0.311
P o wder Co caine 0.015 0.005 3.410 0.001
Methamphetamine -0.001 0.004 -0.215 0.830
Methadone -0.013 0.004 -2.680 0.007
114
App endix
Buprenorphine -0.004 0.006 -0.449 0.653
NMPD Opioids -0.002 0.004 -0.471 0.638
NMPD Stim ulan ts 0.009 0.009 1.110 0.265
NMPD Sedativ es -0.003 0.005 -0.587 0.557
P o wder co caine
NMPD T ranquilizers -0.001 0.006 -0.092 0.927
Cannabis 0.016 0.003 5.400 0.000
Heroin -0.010 0.008 -1.230 0.219
Crac k Co caine 0.010 0.006 1.800 0.072
P o wder Co caine -0.013 0.005 -2.580 0.010
Methamphetamine 0.040 0.008 5.420 0.000
Methadone -0.005 0.016 -0.345 0.730
Buprenorphine -0.090 0.032 -2.680 0.007
NMPD Opioids -0.002 0.007 -0.310 0.756
NMPD Stim ulan ts 0.004 0.011 0.414 0.679
NMPD Sedativ es 0.005 0.008 0.634 0.526
Methamphetamine
NMPD T ranquilizers -0.009 0.007 -1.380 0.168
Note:
Ro ws with significan t estimates are shaded in gra y (p-v alue <= 0.05).
Significan t estimates greater than or equal to 0.010 are prin ted in b old.
C Study 3 Detailed T ables
T able C.1: Study 3: Detailed suprem um p-v alues for
significance of time-v arying effects (restricted mo dels)
Outcome T erm Suprem um-test P-v alue
In tercept 9.555 0.000
Birth cohort (Sixties) 4.429 0.000
Birth cohort (Sev en ties) 4.869 0.000
Birth cohort (Eigh ties or later) 3.687 0.008
Sex (female) 2.422 0.246
Sexual orien tation (GLB) 3.812 0.003
Birthplace (foreign b orn) 5.129 0.000
Race (Blac k) 2.533 0.183
Race (Latino) 3.771 0.006
Heroin
Race (Other) 2.489 0.185
115
App endix
In tercept 2.006 0.342
Birth cohort (Sixties) 3.236 0.017
Birth cohort (Sev en ties) 3.780 0.003
Birth cohort (Eigh ties or later) 4.216 0.001
Sex (female) 1.611 0.657
Sexual orien tation (GLB) 1.591 0.639
Birthplace (foreign b orn) 1.501 0.661
Race (Blac k) 1.927 0.412
Race (Latino) 3.216 0.018
Crac k co caine
Race (Other) 1.999 0.312
In tercept 7.446 0.000
Birth cohort (Sixties) 2.675 0.144
Birth cohort (Sev en ties) 2.673 0.120
Birth cohort (Eigh ties or later) 3.389 0.021
Sex (female) 4.302 0.001
Sexual orien tation (GLB) 2.094 0.378
Birthplace (foreign b orn) 3.485 0.012
Race (Blac k) 3.452 0.016
Race (Latino) 3.446 0.016
P o wder co caine
Race (Other) 1.632 0.696
In tercept 7.780 0.000
Birth cohort (Sixties) 1.616 0.708
Birth cohort (Sev en ties) 2.846 0.081
Birth cohort (Eigh ties or later) 3.948 0.002
Sex (female) 4.233 0.001
Sexual orien tation (GLB) 2.479 0.202
Birthplace (foreign b orn) 3.649 0.006
Race (Blac k) 5.355 0.000
Race (Latino) 6.321 0.000
Methamphetamine
Race (Other) 2.438 0.210
In tercept 1.688 0.394
Birth cohort (Sixties) 1.617 0.457
Birth cohort (Sev en ties) 1.430 0.560
Birth cohort (Eigh ties or later) 2.224 0.125
116
App endix
Sex (female) 1.054 0.827
Sexual orien tation (GLB) 1.776 0.346
Birthplace (foreign b orn) 2.141 0.150
Race (Blac k) 1.746 0.357
Race (Latino) 1.073 0.729
Methadone
Race (Other) 1.121 0.662
In tercept 1.441 0.521
Birth cohort (Sixties) 1.106 0.642
Birth cohort (Sev en ties) 1.593 0.329
Birth cohort (Eigh ties or later) 1.912 0.204
Sex (female) 1.448 0.375
Sexual orien tation (GLB) 1.008 0.577
Birthplace (foreign b orn) 1.991 0.171
Race (Blac k) 1.548 0.434
Race (Latino) 1.875 0.239
Buprenorphine
Race (Other) 1.848 0.224
In tercept 6.984 0.000
Birth cohort (Sixties) 3.161 0.024
Birth cohort (Sev en ties) 2.000 0.342
Birth cohort (Eigh ties or later) 3.107 0.038
Sex (female) 1.840 0.500
Sexual orien tation (GLB) 1.817 0.450
Birthplace (foreign b orn) 3.861 0.002
Race (Blac k) 6.488 0.000
Race (Latino) 5.070 0.000
Opioids
Race (Other) 4.125 0.001
In tercept 4.705 0.000
Birth cohort (Sixties) 3.158 0.018
Birth cohort (Sev en ties) 3.301 0.011
Birth cohort (Eigh ties or later) 3.085 0.018
Sex (female) 1.126 0.857
Sexual orien tation (GLB) 1.619 0.437
Birthplace (foreign b orn) 1.617 0.526
Race (Blac k) 2.571 0.093
117
App endix
Race (Latino) 3.914 0.001
Stim ulan ts
Race (Other) 2.933 0.036
In tercept 4.166 0.000
Birth cohort (Sixties) 1.484 0.552
Birth cohort (Sev en ties) 2.342 0.140
Birth cohort (Eigh ties or later) 4.100 0.001
Sex (female) 4.214 0.000
Sexual orien tation (GLB) 2.803 0.042
Birthplace (foreign b orn) 2.740 0.045
Race (Blac k) 3.725 0.002
Race (Latino) 2.844 0.034
Sedativ es
Race (Other) 1.963 0.256
In tercept 3.043 0.021
Birth cohort (Sixties) 1.083 0.829
Birth cohort (Sev en ties) 1.697 0.495
Birth cohort (Eigh ties or later) 2.146 0.260
Sex (female) 3.211 0.015
Sexual orien tation (GLB) 1.653 0.472
Birthplace (foreign b orn) 2.031 0.212
Race (Blac k) 3.353 0.008
Race (Latino) 1.786 0.421
T ranquilizers
Race (Other) 2.422 0.090
Note:
P-v alues less then or equal to 0.05 are shaded in gra y .
Referen t birth c ohort: Pre-Sixties.
Referen t race c ategory: White.
T able C.2: Study 3: Detailed suprem um p-v alues for
significance of time-v arying effects (full mo dels)
Outcome T erm Suprem um-test P-v alue
In tercept 9.035 0.000
Birth cohort (Sixties) 4.538 0.001
Birth cohort (Sev en ties) 4.562 0.000
Birth cohort (Eigh ties or later) 3.117 0.046
118
App endix
Sex (female) 2.459 0.235
Sexual orien tation (GLB) 3.745 0.005
Birthplace (foreign b orn) 5.267 0.000
Race (Blac k) 2.433 0.221
Race (Latino) 4.167 0.001
Heroin
Race (Other) 2.533 0.168
In tercept 1.986 0.344
Birth cohort (Sixties) 3.142 0.022
Birth cohort (Sev en ties) 3.678 0.004
Birth cohort (Eigh ties or later) 4.174 0.000
Sex (female) 1.831 0.494
Sexual orien tation (GLB) 1.593 0.636
Birthplace (foreign b orn) 1.473 0.678
Race (Blac k) 2.161 0.267
Race (Latino) 2.857 0.054
Crac k co caine
Race (Other) 2.194 0.215
In tercept 7.249 0.000
Birth cohort (Sixties) 2.609 0.164
Birth cohort (Sev en ties) 2.417 0.212
Birth cohort (Eigh ties or later) 3.505 0.015
Sex (female) 4.076 0.002
Sexual orien tation (GLB) 1.947 0.477
Birthplace (foreign b orn) 3.587 0.007
Race (Blac k) 3.336 0.023
Race (Latino) 3.100 0.045
P o wder co caine
Race (Other) 1.585 0.740
In tercept 7.336 0.000
Birth cohort (Sixties) 1.312 0.900
Birth cohort (Sev en ties) 3.107 0.039
Birth cohort (Eigh ties or later) 4.012 0.002
Sex (female) 3.890 0.003
Sexual orien tation (GLB) 2.368 0.246
Birthplace (foreign b orn) 3.110 0.032
Race (Blac k) 5.165 0.000
119
App endix
Race (Latino) 5.993 0.000
Methamphetamine
Race (Other) 2.031 0.433
In tercept 1.772 0.337
Birth cohort (Sixties) 1.612 0.464
Birth cohort (Sev en ties) 1.411 0.573
Birth cohort (Eigh ties or later) 2.228 0.122
Sex (female) 1.067 0.820
Sexual orien tation (GLB) 1.813 0.320
Birthplace (foreign b orn) 1.933 0.225
Race (Blac k) 1.801 0.323
Race (Latino) 1.089 0.711
Methadone
Race (Other) 1.124 0.666
In tercept 1.483 0.505
Birth cohort (Sixties) 1.114 0.632
Birth cohort (Sev en ties) 1.606 0.310
Birth cohort (Eigh ties or later) 1.911 0.205
Sex (female) 1.452 0.370
Sexual orien tation (GLB) 1.014 0.569
Birthplace (foreign b orn) 1.989 0.165
Race (Blac k) 1.571 0.419
Race (Latino) 1.893 0.223
Buprenorphine
Race (Other) 1.843 0.224
In tercept 7.007 0.000
Birth cohort (Sixties) 3.145 0.027
Birth cohort (Sev en ties) 1.992 0.354
Birth cohort (Eigh ties or later) 3.072 0.041
Sex (female) 2.161 0.299
Sexual orien tation (GLB) 1.839 0.440
Birthplace (foreign b orn) 3.961 0.001
Race (Blac k) 6.452 0.000
Race (Latino) 5.144 0.000
Opioids
Race (Other) 4.163 0.001
In tercept 4.518 0.000
Birth cohort (Sixties) 3.195 0.016
120
App endix
Birth cohort (Sev en ties) 3.319 0.010
Birth cohort (Eigh ties or later) 3.060 0.021
Sex (female) 1.096 0.870
Sexual orien tation (GLB) 1.542 0.494
Birthplace (foreign b orn) 1.552 0.581
Race (Blac k) 2.562 0.093
Race (Latino) 3.838 0.002
Stim ulan ts
Race (Other) 2.929 0.035
In tercept 4.088 0.000
Birth cohort (Sixties) 1.496 0.554
Birth cohort (Sev en ties) 2.350 0.140
Birth cohort (Eigh ties or later) 4.075 0.000
Sex (female) 4.201 0.001
Sexual orien tation (GLB) 2.833 0.038
Birthplace (foreign b orn) 2.688 0.053
Race (Blac k) 3.661 0.003
Race (Latino) 2.820 0.037
Sedativ es
Race (Other) 1.957 0.264
In tercept 2.920 0.029
Birth cohort (Sixties) 1.005 0.874
Birth cohort (Sev en ties) 1.692 0.526
Birth cohort (Eigh ties or later) 2.151 0.247
Sex (female) 3.207 0.015
Sexual orien tation (GLB) 1.512 0.548
Birthplace (foreign b orn) 2.224 0.149
Race (Blac k) 3.304 0.010
Race (Latino) 1.734 0.440
T ranquilizers
Race (Other) 2.559 0.066
Note:
P-v alues less then or equal to 0.05 are shaded in gra y .
Referen t birth c ohort: Pre-Sixties.
Referen t race c ategory: White.
121
App endix
T able C.3: Study 3: Detailed parametric estimates (re-
stricted mo dels)
Outcome T erm Estimate Std. Error Z-statistic P-v alue
Prior same drug use 0.122 0.017 7.480 0.000
Crac k Co caine 0.032 0.032 1.140 0.252
P o wder Co caine 0.038 0.013 2.920 0.004
Methamphetamine 0.017 0.010 1.880 0.061
Methadone 0.098 0.135 0.916 0.360
Buprenorphine -0.054 0.337 -0.536 0.592
NMPD Opioids 0.152 0.043 3.710 0.000
NMPD Stim ulan ts -0.016 0.029 -0.577 0.564
NMPD Sedativ es 0.453 0.286 5.940 0.000
Heroin
NMPD T ranquilizers -0.071 0.058 -1.380 0.166
Prior same drug use 0.006 0.002 3.870 0.000
Heroin 0.001 0.001 0.791 0.429
P o wder Co caine 0.003 0.002 1.590 0.111
Methamphetamine 0.008 0.002 4.060 0.000
Methadone 0.002 0.010 0.245 0.806
Buprenorphine 0.062 0.058 1.680 0.092
NMPD Opioids 0.015 0.004 4.220 0.000
NMPD Stim ulan ts 0.002 0.004 0.474 0.635
NMPD Sedativ es -0.007 0.005 -1.660 0.097
Crac k co caine
NMPD T ranquilizers -0.006 0.006 -0.988 0.323
Prior same drug use 0.025 0.004 5.640 0.000
Heroin 0.022 0.004 4.980 0.000
Crac k Co caine 0.042 0.027 1.420 0.157
Methamphetamine 0.021 0.007 2.530 0.012
Methadone 0.123 0.081 2.110 0.035
Buprenorphine -0.207 0.094 -2.840 0.005
NMPD Opioids 0.037 0.018 1.770 0.077
NMPD Stim ulan ts 0.081 0.033 2.240 0.025
NMPD Sedativ es -0.021 0.024 -0.940 0.347
P o wder co caine
NMPD T ranquilizers 0.000 0.048 -0.008 0.993
122
App endix
Prior same drug use 0.038 0.005 7.040 0.000
Heroin -0.003 0.002 -1.220 0.224
Crac k Co caine 0.017 0.006 2.460 0.014
P o wder Co caine 0.011 0.003 3.740 0.000
Methadone 0.046 0.026 1.720 0.086
Buprenorphine -0.068 0.035 -2.120 0.034
NMPD Opioids 0.003 0.006 0.492 0.623
NMPD Stim ulan ts 0.010 0.008 1.180 0.240
NMPD Sedativ es 0.000 0.015 -0.013 0.990
Methamphetamine
NMPD T ranquilizers 0.020 0.015 1.460 0.143
Prior same drug use 0.001 0.001 0.659 0.510
Heroin 0.001 0.001 2.570 0.010
Crac k Co caine 0.001 0.001 1.210 0.227
P o wder Co caine -0.001 0.001 -1.600 0.109
Methamphetamine -0.001 0.001 -0.924 0.355
Buprenorphine 0.035 0.028 1.290 0.197
NMPD Opioids 0.001 0.001 0.991 0.322
NMPD Stim ulan ts -0.001 0.001 -0.743 0.457
NMPD Sedativ es 0.001 0.002 0.604 0.546
Methadone
NMPD T ranquilizers 0.013 0.005 2.530 0.012
Prior same drug use 0.005 0.004 1.290 0.199
Heroin 0.000 0.000 -0.218 0.827
Crac k Co caine 0.000 0.001 -0.565 0.572
P o wder Co caine 0.000 0.000 0.431 0.667
Methamphetamine 0.000 0.000 -0.683 0.495
Methadone 0.001 0.002 0.522 0.601
NMPD Opioids 0.002 0.001 2.640 0.008
NMPD Stim ulan ts -0.001 0.001 -0.808 0.419
NMPD Sedativ es -0.001 0.001 -2.380 0.018
Buprenorphine
NMPD T ranquilizers -0.001 0.001 -0.409 0.683
Prior same drug use 0.014 0.002 6.610 0.000
Heroin -0.002 0.001 -1.560 0.120
Crac k Co caine 0.006 0.003 1.840 0.065
P o wder Co caine 0.003 0.001 2.050 0.040
123
App endix
Methamphetamine 0.005 0.002 2.260 0.024
Methadone -0.012 0.008 -1.720 0.086
Buprenorphine 0.046 0.041 1.040 0.300
NMPD Stim ulan ts -0.001 0.004 -0.139 0.890
NMPD Sedativ es 0.016 0.010 1.250 0.210
Opioids
NMPD T ranquilizers 0.056 0.018 2.390 0.017
Prior same drug use 0.011 0.003 3.920 0.000
Heroin 0.000 0.001 -0.455 0.649
Crac k Co caine -0.002 0.001 -2.110 0.035
P o wder Co caine 0.000 0.001 -0.463 0.643
Methamphetamine 0.002 0.001 2.310 0.021
Methadone -0.005 0.002 -2.450 0.014
Buprenorphine 0.007 0.013 0.527 0.598
NMPD Opioids 0.003 0.002 1.850 0.064
NMPD Sedativ es 0.002 0.003 0.498 0.618
Stim ulan ts
NMPD T ranquilizers 0.003 0.004 0.760 0.447
Prior same drug use 0.003 0.001 2.700 0.007
Heroin 0.000 0.001 0.238 0.812
Crac k Co caine -0.001 0.001 -1.430 0.153
P o wder Co caine -0.001 0.001 -0.846 0.397
Methamphetamine 0.001 0.001 0.839 0.401
Methadone -0.001 0.000 -1.910 0.056
Buprenorphine -0.002 0.001 -2.190 0.029
NMPD Opioids 0.001 0.001 0.890 0.374
NMPD Stim ulan ts 0.001 0.001 0.834 0.404
Sedativ es
NMPD T ranquilizers 0.005 0.003 1.310 0.191
Prior same drug use 0.002 0.001 2.030 0.042
Heroin 0.000 0.001 0.408 0.684
Crac k Co caine -0.001 0.001 -0.420 0.674
P o wder Co caine 0.001 0.001 0.897 0.370
Methamphetamine -0.001 0.001 -1.190 0.235
Methadone 0.012 0.007 1.640 0.102
Buprenorphine -0.003 0.013 -0.229 0.819
NMPD Opioids 0.002 0.001 1.320 0.187
124
App endix
NMPD Stim ulan ts 0.003 0.002 1.460 0.145
T ranquilizers
NMPD Sedativ es -0.003 0.002 -1.670 0.096
Note:
Ro ws with significan t estimates are shaded in gra y (p-v alue <= 0.05).
Significan t estimates greater than or equal to 0.010 are prin ted in b old.
T able C.4: Study 3: Detailed parametric estimates (full
mo dels)
Outcome T erm Estimate Std. Error Z-statistic P-v alue
Prior same drug use 0.117 0.017 7.060 0.000
Crac k Co caine 0.024 0.032 0.863 0.388
P o wder Co caine 0.005 0.016 0.315 0.752
Methamphetamine -0.035 0.021 -1.660 0.097
Methadone 0.037 0.137 0.329 0.742
Buprenorphine -0.158 0.327 -1.150 0.248
NMPD Opioids 0.129 0.043 3.240 0.001
NMPD Stim ulan ts -0.021 0.029 -0.883 0.377
NMPD Sedativ es 0.441 0.286 6.600 0.000
NMPD T ranquilizers -0.049 0.059 -1.000 0.317
Heroin
Injection Drug Use 0.070 0.023 3.160 0.002
Prior same drug use 0.006 0.002 4.020 0.000
Heroin 0.004 0.003 1.250 0.210
P o wder Co caine 0.003 0.002 1.810 0.071
Methamphetamine 0.009 0.002 4.120 0.000
Methadone 0.002 0.010 0.240 0.811
Buprenorphine 0.065 0.058 1.830 0.066
NMPD Opioids 0.015 0.004 4.200 0.000
NMPD Stim ulan ts 0.002 0.004 0.452 0.651
NMPD Sedativ es -0.007 0.005 -1.720 0.086
NMPD T ranquilizers -0.006 0.006 -0.972 0.331
Crac k co caine
Injection Drug Use -0.003 0.003 -1.110 0.268
Prior same drug use 0.024 0.004 5.360 0.000
Heroin -0.013 0.014 -1.020 0.305
Crac k Co caine 0.037 0.027 1.260 0.207
125
App endix
Methamphetamine -0.001 0.011 -0.121 0.904
Methadone 0.111 0.082 1.830 0.068
Buprenorphine -0.200 0.094 -2.680 0.007
NMPD Opioids 0.031 0.018 1.540 0.124
NMPD Stim ulan ts 0.072 0.033 2.020 0.044
NMPD Sedativ es -0.009 0.025 -0.422 0.673
NMPD T ranquilizers 0.006 0.048 0.126 0.900
P o wder co caine
Injection Drug Use 0.040 0.014 3.040 0.002
Prior same drug use 0.038 0.005 7.020 0.000
Heroin -0.038 0.012 -3.070 0.002
Crac k Co caine 0.017 0.006 2.440 0.015
P o wder Co caine 0.006 0.003 2.170 0.030
Methadone 0.041 0.026 1.640 0.101
Buprenorphine -0.062 0.035 -1.960 0.050
NMPD Opioids 0.002 0.006 0.370 0.712
NMPD Stim ulan ts 0.009 0.008 1.150 0.250
NMPD Sedativ es 0.001 0.015 0.061 0.951
NMPD T ranquilizers 0.021 0.015 1.540 0.124
Methamphetamine
Injection Drug Use 0.038 0.013 3.140 0.002
Prior same drug use 0.001 0.001 0.663 0.507
Heroin 0.001 0.001 1.130 0.259
Crac k Co caine 0.001 0.001 1.220 0.221
P o wder Co caine -0.001 0.001 -1.580 0.113
Methamphetamine -0.001 0.001 -1.050 0.292
Buprenorphine 0.035 0.028 1.290 0.196
NMPD Opioids 0.001 0.001 1.020 0.308
NMPD Stim ulan ts -0.001 0.001 -0.736 0.462
NMPD Sedativ es 0.001 0.002 0.623 0.533
NMPD T ranquilizers 0.013 0.005 2.520 0.012
Methadone
Injection Drug Use 0.001 0.001 0.803 0.422
Prior same drug use 0.005 0.004 1.290 0.198
Heroin 0.000 0.001 -0.398 0.691
Crac k Co caine 0.000 0.001 -0.558 0.577
P o wder Co caine 0.000 0.000 0.369 0.712
126
App endix
Methamphetamine 0.000 0.000 -0.743 0.458
Methadone 0.001 0.002 0.523 0.601
NMPD Opioids 0.002 0.001 2.670 0.008
NMPD Stim ulan ts -0.001 0.001 -0.802 0.423
NMPD Sedativ es -0.001 0.001 -2.370 0.018
NMPD T ranquilizers -0.001 0.001 -0.408 0.683
Buprenorphine
Injection Drug Use 0.000 0.001 0.399 0.690
Prior same drug use 0.014 0.002 6.600 0.000
Heroin -0.001 0.003 -0.397 0.691
Crac k Co caine 0.006 0.003 1.830 0.067
P o wder Co caine 0.003 0.002 2.080 0.038
Methamphetamine 0.005 0.002 2.150 0.031
Methadone -0.012 0.008 -1.720 0.085
Buprenorphine 0.046 0.041 1.040 0.299
NMPD Stim ulan ts -0.001 0.004 -0.127 0.899
NMPD Sedativ es 0.016 0.010 1.250 0.212
NMPD T ranquilizers 0.056 0.018 2.390 0.017
Opioids
Injection Drug Use -0.001 0.004 -0.318 0.750
Prior same drug use 0.011 0.003 3.930 0.000
Heroin -0.002 0.002 -1.210 0.225
Crac k Co caine -0.002 0.001 -1.990 0.046
P o wder Co caine -0.001 0.001 -0.794 0.427
Methamphetamine 0.002 0.001 1.780 0.075
Methadone -0.005 0.002 -2.440 0.015
Buprenorphine 0.007 0.013 0.527 0.598
NMPD Opioids 0.003 0.002 1.900 0.057
NMPD Sedativ es 0.002 0.003 0.562 0.574
NMPD T ranquilizers 0.003 0.004 0.772 0.440
Stim ulan ts
Injection Drug Use 0.003 0.002 1.210 0.224
Prior same drug use 0.003 0.001 2.700 0.007
Heroin -0.001 0.001 -1.020 0.308
Crac k Co caine -0.001 0.001 -1.300 0.195
P o wder Co caine -0.001 0.001 -1.260 0.206
Methamphetamine 0.000 0.001 0.211 0.833
127
App endix
Methadone -0.001 0.000 -1.910 0.056
Buprenorphine -0.002 0.001 -2.210 0.027
NMPD Opioids 0.001 0.001 0.999 0.318
NMPD Stim ulan ts 0.001 0.001 0.842 0.400
NMPD T ranquilizers 0.005 0.003 1.310 0.190
Sedativ es
Injection Drug Use 0.002 0.001 1.390 0.166
Prior same drug use 0.002 0.001 2.030 0.042
Heroin -0.001 0.002 -0.524 0.600
Crac k Co caine 0.000 0.001 -0.384 0.701
P o wder Co caine 0.000 0.001 0.621 0.535
Methamphetamine -0.001 0.001 -1.350 0.176
Methadone 0.012 0.007 1.640 0.102
Buprenorphine -0.003 0.013 -0.230 0.818
NMPD Opioids 0.002 0.001 1.350 0.177
NMPD Stim ulan ts 0.003 0.002 1.470 0.141
NMPD Sedativ es -0.003 0.002 -1.620 0.106
T ranquilizers
Injection Drug Use 0.001 0.002 0.852 0.394
Note:
Ro ws with significan t estimates are shaded in gra y (p-v alue <= 0.05).
Significan t estimates greater than or equal to 0.010 are prin ted in b old.
128
References
References
Abuse, S., & A dministration, M. H. S. (2014). The TEDS r ep ort: A ge of substanc e use
initiation among tr e atment admissions age d 18 to 30 . Ro c kville, MD. Retriev ed from
h t t p s : / / w w w . s a m h s a . g o v / d a t a / s i t e s / d e f a u l t / f i l e s / W e b F i l e s /_ T E D S / _ S R 1 4 2 /_ A g e a t
Init/_07- 10- 14/TEDS- SR142- AgeatInit- 2014.pdf
Abuse, S., & A dministration, M. H. S. (2020a). K ey substanc e use and mental he alth in-
dic ators in the Unite d States: R esults fr om the 2019 National Survey on Drug U se and
He alth . Ro c kville, MD. Retriev ed from https://www.samhsa.gov/data/
Abuse, S., & A dministration, M. H. S. (2020b). R esults fr om the 2017 National Survey on
Drug U se and He alth: Detaile d tables . Ro c kville, MD. Retriev ed from h t t p s : / / w w w . s a
mhsa.gov/data/report/2019- nsduh- detailed- tables
A dministration, U. S. D. E. (2017). 2017 national drug thr e at assessment (No. DEA-DCT-
DIR-040-17). Retriev ed from h t t p s : / / w w w . d e a . g o v / d o c s / D IR - 0 4 0 - 1 7 /_ 2 0 1 7 - N D T A . p
df
Allen, B., & Haro cop os, A. (2016). Non-prescrib ed buprenorphine in New Y ork Cit y: Mo-
tiv ations for use, practices of div ersion, and e xp eriences of stigma. Journal of Substanc e
A buse T r e atment , 70 , 81–86. https://doi.org/10.1016/j.jsat.2016.08.002
Alv arez, J., Jason, L. A., Olson, B. D., F errari, J. R., & Da vis, M. I. (2007). Substance
abuse prev alence and treatmen t among Latinos a nd Latinas. Journal of Ethnicity in
Substanc e A buse , 6 (2), 115–141. https://doi.org/10.1300/j233v06n02/_08
Andersen, P . K., Geskus, R. B., Witte, T. de, & Putter, H. (2012). Comp eting risks in epi-
demiology: P ossibilities and pitfalls. International Journal of Epidemiolo gy , 41 (3), 861–
870. https://doi.org/10.1093/ije/dyr213
Arria, A. M., Caldeira, K. M., O’Grady , K. E., MA, K. B. V., W, D. B. F., Johnson, E.
P ., & Wish, E. D. (2008). Drug exp osure opp ortunities and use patterns among college
studen ts: Results of a longitudinal prosp ectiv e cohort study . Substanc e A buse , 29 (4),
19–38. https://doi.org/10.1080/08897070802418451
Benjet, C., Borges, G., Medina-Mora, M. E., Blanco, J., Zam brano, J., Orozco, R., … Ro-
jas, E. (2007). Drug use opp ortunities and the transition to drug use among adoles-
cen ts from the Mexico Cit y metrop olitan a rea. Drug and A lc ohol Dep endenc e , 90 (2-3),
128–134. https://doi.org/10.1016/j.drugalcdep.2007.02.018
Bluthen thal, R. N., Ch u, D., W enger, L. D., Bourgois, P ., V alen te, T., & Kral, A.
H. (2018). Differences in time to injection onset b y drug in California: Impli-
cations for the emerging heroin epidemic. Drug A lc ohol Dep end , 185 , 253–259.
https://doi.org/10.1016/j.drugalcdep.2018.01.005
Bluthen thal, R. N., & W atters, J. K. (1995). Multimetho d researc h from targeted sam-
pling to HIV risk en vironmen ts. NID A R es Mono gr , 157 , 212–230.
Bluthen thal, R. N., W enger, L., Ch u, D., Bourgois, P ., & Kral, A. H. (2017). Drug use
generations and patterns of injection drug use: Birth cohort differences among p eople
129
References
who inject drugs in Los Angeles and San F rancisco, California. Drug A lc ohol Dep end ,
175 , 210–218. https://doi.org/10.1016/j.drugalcdep.2017.04.001
Bluthen thal, R. N., W enger, L., Ch u, D., Lorvic k, J., Quinn, B., Thing, J. P ., & Kral, A.
H. (2015). F actors asso ciated with b eing ask ed to initiate someone in to injection drug
use. Drug A lc ohol Dep end , 149 , 252–258. h t t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / j . d r u g a l c d e p . 2 0 1 5 . 0 2
.011
Bluthen thal, R. N., W enger, L., Ch u, D., Quinn, B., Thing, J., & Kral, A. H. (2014). F ac-
tors asso ciated with initiating someone in to illicit drug injection. Drug A lc ohol Dep end ,
144 , 186–192. https://doi.org/10.1016/j.drugalcdep.2014.09.008
Borges, G., Rafful, C., Benjet, C., T ancredi, D. J., Saito, N., Aguilar-Gaxiola, S., … Bres-
lau, J. (2012). Mexican immigration to the US and alcohol and drug use opp ortunities:
Do es it mak e a difference in alcohol a nd/or drug use? Drug and A lc ohol D ep endenc e ,
125 , S4–S11. https://doi.org/10.1016/j.drugalcdep.2012.05.007
Brady , J. E., F riedman, S. R., Co op er, H. L. F., Flom, P . L., T empalski, B., & Gostnell,
K. (2008-05). Estimating the prev alence of injection drug users in the U.S. And in large
U.S. Metrop olitan areas from 1992 to 200 2. J Urb an He alth , 85 (3), 323–351. h t t p s : / / d o
i.org/10.1007/s11524- 007- 9248- 5
Brust, J. C. M. (2014). Neurologic complications of illicit drug abuse. CONTINUUM:
Lifelong L e arning in Neur olo gy , 20 (3, Neurology of Systemic Disease), 642–656. h t t p s :
//doi.org/10.1212/01.con.0000450971.99322.cd
Butler, S. F., Oy edele, N. K., Go v oni, T. D., & Green, J. L. (2020). Ho w motiv ations for
using buprenorphine pro ducts differ from using opio id analgesics: Evidence from an ob-
serv ational study of in ternet discussions amo ng recreational users. JMIR Public He alth
and S urveil lanc e , 6 (1), e16038. https://doi.org/10.2196/16038
California, S. of. (2021). Cannabis legislation. Retriev ed from h t t p s : / / c a n n a b i s . c a . g o v / c a
nnabis- legislation/
Cen ter, N. D. I. (2011). National drug thr e at assessment 2011 (No. 2011-Q0317-001). Re-
triev ed from https://www.justice.gov/archive/ndic/pubs44/44849/44849p.pdf
Cep eda, A., Kaplan, C., Neaigus, A., Cano, M. Á., Villarreal, Y., & V aldez, A. (2012). In-
jecting transition risk and depression among M exican American non-injecting heroin
users. Drug and A lc ohol Dep endenc e , 125 , S12–S17. h t t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / j . d r u g a l c d
ep.2012.05.035
Cerdá, M., Mauro, C., Hamilton, A., Levy , N. S., San taella-T enorio, J., Hasin, D., … Mar-
tins, S. S. (2020). Asso ciation b et w een recreational marijuana legalization in the United
States and c hanges in marijuana use and cannabi s use disorder from 2008 to 2016.
JAMA Psychiatry , 77 (2), 165–171. https://doi.org/10.1001/jamapsychiatry.2019.3254
Cerdá, M., San taella, J., Marshall, B. D. L., Kim, J. H., & Martins, S. S. (2015-09). Non-
medical prescription opioid use in c hildho o d a nd early adolescence predicts transitions
to heroin use in y oung adultho o d: A national study . The Journal of Pe diatrics , 167 (3),
605 - 12.e1-2. https://doi.org/10.1016/j.jpeds.2015.04.071
130
References
Chatterjee, S., T empalski, B., P ouget, E. R., Co op er, H. L. F., Cleland, C. M., & F ried-
man, S. R. (2011). Changes in the prev alence of injection drug use among adolescen ts
and y oung adults in large U.S. Metrop olitan are as. AIDS Behav , 15 (7), 1570–1578.
https://doi.org/10.1007/s10461- 011- 9992- 0
Chen, C.-Y., & Lin, K.-M. (2009). Health consequences of illegal drug use. Curr ent Opin-
ion in Psychiatry , 22 (3), 287–292. https://doi.org/10.1097/yco.0b013e32832a2349
Chen, K., & Kandel, D. B. (1995). The natural history of drug use from adolescence to
the mid-thirties in a general p opulation sam ple. A meric an Journal of Public He alth ,
85 (1), 41–47.
Ciccarone, D. (2019). The triple w a v e epidemic: Supply and demand driv ers of the US opi-
oid o v erdose crisis. International Journal of Drug Policy , 71 , 183–18 8. h t t p s : / / d o i . o r g /
10.1016/j.drugpo.2019.01.010
Ciccarone, D. (2021). The rise of illicit fen tan yls, stim ulan ts and the fourth w a v e of the
opioid o v erdose crisis. Curr ent Opinion in Psychiatry , 34 (4), 344–350. h t t p s : / / d o i . o r g /
10.1097/yco.0000000000000717
Cicero, T. J., Ellis, M. S., Surratt, H. L., & Kurtz, S. P . (2014). The c hanging face of
heroin use in the United States: A retrosp ectiv e analysis of the past 50 y ears. JAMA
Psychiatry , 71 (7), 821–826. https://doi.org/10.1001/jamapsychiatry.2014.366
Compton, W. M., Jones, C. M., & Baldwin, G. T. (2016). Relationship b et w een nonmed-
ical prescription-opioid use and heroin use . N Engl J Me d , 374 (2), 154–163. h t t p s :
//doi.org/10.1056/nejmra1508490
Con w a y , Kevin P ., Compton, W., Stinson, F. S., & Gran t, B. F. (2006). Lifetime comor-
bidit y of DSM-IV mo o d and anxiet y disorders and s p ecific drug use disorders: Results
from the National Epidemiologic Surv ey on Alc ohol and Related Conditions. The Jour-
nal of Clinic al Psychiatry , 67 (2), 247–258. https://doi.org/10.4088/jcp.v67n0211
Con w a y , Kevin P ., Sw endsen, J., Husky , M. M., He, J.-P ., & Merikangas, K. R. (2016).
Asso ciation of lifetime men tal disorders and subsequen t alcohol and illicit drug use: Re-
sults from the National Comorbidit y Surv ey–A dolescen t Supplemen t. Journal of the
A meric an A c ademy of Child & A dolesc ent Psychiatry , 55 (4), 280–288. h t t p s : / / d o i . o r g /
10.1016/j.jaac.2016.01.006
Co op er, Hannah L. F., Brady , J. E., Ciccarone, D., T empalski, B., Gostnell, K., & F ried-
man, S. R. (2007). Nation wide increase in the n um b er of hospitalizations for illicit in-
jection drug use-related infectiv e endo cardit is. Clinic al Infe ctious Dise ases , 45 (9), 1200–
1203. https://doi.org/10.1086/522176
Co op er, Hannah L. F., Brady , J. E., F riedman, S. R., T empalski, B., Gostnell, K., & Flom,
P . L. (2008). Estimating the prev alence of injection drug use among blac k and white
adults in large U.S. Metrop olitan areas o v er time (1992–2002): Estimation metho ds
and prev alence trends. J Urb an He alth , 85 (6), 826–856. h t t p s : / / d o i . o r g / 1 0 . 1 0 0 7 / s 1
1524- 008- 9304- 9
Corliss, H. L., Rosario, M., W ypij, D., W ylie, S. A., F razier, A. L., & A ustin, S. B. (2010).
Sexual orien tation and drug use in a longitudinal cohort study of U.S. adolescen ts. A d-
131
References
dictive Behaviors , 35 (5), 517–521. https://doi.org/10.1016/j.addbeh.2009.12.019
Cotto, J. H., Da vis, E., Do wling, G. J., Elcano, J. C., Staton, A. B., & W eiss, S. R. B.
(2010). Gender effects on drug use, abuse, and dep endence: A sp ecial analysis of results
from the National Surv ey on Drug Use and Healt h. Gender Me dicine , 7 (5), 402–413.
https://doi.org/10.1016/j.genm.2010.09.004
Daniulait yte, R., Nahhas, R. W., Silv erstein, S., Martins, S., Zaragoza, A., Mo eller, A., &
Carlson, R. G. (2019). P atterns of non-prescrib ed buprenorphine and other opioid use
among individuals with opioid use disorder: A laten t class analysis. Drug and A lc ohol
Dep endenc e , 204 , 107574. https://doi.org/10.1016/j.drugalcdep.2019.107574
Degenhardt, L., Chiu, W. T., Con w a y , K., Dierk er, L., Glan tz, M., Kala ydjian, A.,
… Kessler, R. C. (2009). Do es the ’gatew a y’ matter? Asso ciations b et w een the
order of drug use initiation and the dev elopmen t of drug dep endence in the Na-
tional Comorbidit y Study Replication. Psycholo gic al Me dicine , 39 (1), 157–167.
https://doi.org/10.1017/s0033291708003425
Degenhardt, Louisa, Chiu, W. T., Sampson, N., Kessler, R. C., & An thon y , J. C. (2007).
Epidemiological patterns of extra-medical d rug use in the United States: Evidence from
the National Comorbidit y Surv ey Replication, 2001 –2003. Drug and A lc ohol Dep en-
denc e , 90 (2-3), 210–223. https://doi.org/10.1016/j.drugalcdep.2007.03.007
Disease Con trol, C. for, & Prev en tion. (2017). HIV surveil lanc e r ep ort, 2016 . Retriev ed
from http://www.cdc.gov/hiv/library/reports/hiv- surveillance.html
Dy al, S. R., Kral, A. H., Gonzalez, K. D., W enger, L. D., & Bluthen thal, R. N. (2015).
Consistency of self-rep orted drug use ev en ts in a mixed metho ds study of p eople who
inject drugs. The A meric an Journal of Drug and A lc ohol A buse , 41 (4), 332–338. h t t p s :
//doi.org/10.3109/00952990.2015.1037842
Etten, M. L. V., Neumark, Y. D., & An thon y , J. C. (1999). Male‐female differences in the
earliest stages of drug in v olv emen t. A ddiction , 94 (9), 1413–1419. https ://d oi.or g/10 .104
6/j.1360- 0443.1999.949141312.x
F r enc h, M. T., McGeary , K. A., Chit w o o d, D. D., & McCo y , C. B. (2000). Chronic il-
licit drug use, health services utilizatio n and the cost of medical care. So cial Scienc e
& Me dicine , 50 (12), 1703–1713. https://doi.org/10.1016/s0277- 9536(99)00411- 6
F u ller, C. M., Vlaho v, D., Arria, A. M., Ompad, D. C., Garfein, R., & Strathdee, S. A.
(2001). F actors asso ciated with adolescen t initiation of injection drug use. Public He alth
R ep orts , 116 (1_suppl), 136–145. https://doi.org/10.1093/phr/116.s1.136
Garofalo, R., Mustanski, B. S., McKirnan, D. J., Herric k, A., & Donen b erg, G. R. (2007).
Methamphetamine and y oung men who ha v e sex with m en: Understanding patterns
and correlates of use and the asso cia tion with HIV-related sexual risk. A r chives o f Pe-
diatrics & A dolesc ent Me dicine , 161 (6), 591–596 . h t t p s : / / d o i . o r g / 1 0 . 1 0 0 1 / a r c h p e d i . 1 6 1
.6.591
Go edel, W. C., Shapiro, A., Cerdá, M., T sai, J. W., Hadland, S. E., & Marshall, B. D. L.
(2020). Asso ciation of racial/ethnic segregation with treatmen t capacit y for opioid use
132
References
disorder in coun ties in the United States. JAMA Network Op en , 3 (4), e203711. h t t p s :
//doi.org/10.1001/jamanetworkopen.2020.3711
Golub, Andrew, & Bro wnstein, H. H. (2013). Drug generations in the 2000s: An analysis
of arrestee data. Journal of D rug Issues , 43 (3). h t t p s : / / d o i . o r g / 1 0 . 1 1 7 7 / 0 0 2 2 0 4 2 6 1 3 4 7
5599
Golub, Andrew, Elliott, L., & Bro wnstein, H. H. (2013). The opiate pain reliev er epidemic
among U.S. Arrestees 2000–2010: Regional and demographic v ariations. Journal of Eth-
nicity in Substanc e A buse , 12 (1), 1–2 9. https://doi.org/10.1080/15332640.2013.759498
Golub, A., & Johnson, B. D. (2001). V ariation in y outhful risks of progression from alco-
hol and tobacco to marijuana and to hard drugs a cross generations. A meric an Journa l
of Public He alth , 91 (2), 225–232. https://doi.org/10.2105/ajph.91.2.225
Golub, Andrew, Johnson, B. D., & Dunlap, E. (2005). Sub cultural ev olution and illicit
drug use. A ddiction R ese ar ch & The ory , 13 (3), 217–229. h t t p s : / / d o i . o r g / 1 0 . 1 0 8 0 /
16066350500053497
Golub, A. L., & Johnson, B. D. (2009). Cohort c hanges in illegal drug use among arrestees
in Manhattan: F rom the heroin injection generation to the blun ts generation. Sub-
stanc e U se & Misuse , 34 (13), 1733–1763. https://doi.org/10.3109/10826089909039425
Got w a y , M. B., Marder, S. R., Hanks, D. K., Leung, J. W. T., Da wn, S. K., Gean, A. D.,
… W ebb, W. R. (2002). Thoracic complications of illicit drug use: An organ system ap-
proac h. R adioGr aphics , 22 (suppl_1), S119 –S135. h t t p s : / / d o i . o r g / 1 0 . 1 1 4 8 / r a d i o g r a p h i c
s.22.suppl/_1.g02oc01s119
Gram bsc h, P . M., & Therneau, T. M. (1994). Prop ortional hazards tests and diagnostics
based on w eigh ted residuals. Biometrika , 81 (3), 515–526. h t t p s : / / d o i . o r g / 1 0 . 1 0 9 3 / b i o m
et/81.3.515
Gryczynski, J., Sc h w artz, R. P ., O’Grady , K. E., Restiv o, L., Mitc hell, S. G., & Jaffe, J. H.
(2017). Understanding patterns of high-cost health care use across differen t substance
user groups. He alth Affairs , 35 (1), 12–19. https://doi.org/10.1377/hlthaff.2015.0618
Guise, A., Horyniak, D., Melo, J., McNeil, R., & W erb, D. (2017). The exp erience of ini-
tiating injection drug use and its so cial c on text: A qualitativ e systematic review and
thematic syn thesis. A ddiction , 112 (12), 2098–2111. https://doi.org/10.1111/add.13957
Hadland, S. E., W erb, D., Kerr, T., F u, E., W ang, H., Mon taner, J. S., & W o o d, E. (2012).
Childho o d sexual abuse and risk for initiating i njection drug use: A prosp ectiv e cohort
study . Pr eventive Me dicine , 55 (5), 500–504. h t t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / j . y p m e d . 2 0 1 2 . 0 8 .
015
Han, B., Cotto, J., Etz, K., Einstein, E. B., Compton, W. M., & V olk o w, N. D. (2021).
Methamphetamine o v erdose deaths in the US b y s ex and race and ethnicit y . JAMA
Psychiatry , 78 (5), 564–567. https://doi.org/10.1001/jamapsychiatry.2020.4321
Han, B., Gfro erer, J. C., & Colliv er, J. D. (2010). Asso ciations b et w een duration of illicit
drug use and health conditions: Results from the 2005–2007 National Surv eys on Drug
133
References
Use and Health. A nnals of Epidemiolo gy , 20 (4), 289–297. h t t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / j . a n
nepidem.2010.01.003
Hansen, H., Siegel, C., W anderling, J., & DiRo cco, D. (2016). Buprenorphine and
methadone treatmen t for opioid dep endence b y incom e, ethnicit y and race of
neigh b orho o ds in New Y ork Cit y . Drug and A lc ohol Dep endenc e , 164 , 14–21.
https://doi.org/10.1016/j.drugalcdep.2016.03.028
Haro cop os, A., Goldsam t, L. A., K obrak, P ., Jost, J. J., & Clatts, M. C. (2009-07). New
injectors and the so cial con text of inje ction initiation. Int J Drug Policy , 20 (4), 317–
323. https://doi.org/10.1016/j.drugpo.2008.06.003
Hatzen buehler, M. L., Jun, H.-J., Corliss, H. L., & A ustin, S. B. (2015). Structural stigma
and sexual orien tation disparities in adolesc en t drug use. A ddictive B ehaviors , 46 , 14–
18. https://doi.org/10.1016/j.addbeh.2015.02.017
Hedegaard, H., W arner, M., & Miniño, A. M. (2017). Drug over dose de aths in the Unite d
States, 1999-2015 (No. 273). Hy attsville, MD: National Cen ter for Health Statistics.
Huang, X., Key es, K. M., & Li, G. (2018-01). Increasing prescription opioid and heroin
o v erdose mortalit y in the United States, 1999 -2014: An age-p erio d-cohort analysis. A m
J Public He alth , 108 (1), 131–136 . https://doi.org/10.2105/ajph.2017.304142
Hun t, N., Griffiths, P ., South w ell, M., Stillw ell, G., & Strang, J. (1999). Prev en ting and
curtailing injecting drug use: A review of opp ortunities for dev eloping and deliv ering
’route transition in terv en tions’ . Drug and A lc ohol R eview , 18 (4), 441–451. h t t p s : / / d o i .
org/10.1080/09595239996310
Jac kson, K. M., & Janssen, T. (2019). Dev elopmen tal considerations in surviv al mo dels as
applied to substance use researc h. A ddictive Behaviors , 94 , 36–41. h t t p s : / / d o i . o r g / 1 0 . 1
016/j.addbeh.2018.11.028
Jarlais, D. D., Uuskula, A., T alu, A., Barnes, D. M., Raag, M., Arasteh, K., … T ross, S.
(2019). Implemen ting an up dated ”Break the Cycle” in terv en tion to reduce initiating
p ersons in to injecting drug use in an Easte rn Europ ean and a US ”opioid epidemic” set-
ting. AIDS and Behavior , 23 (9), 2304–2314. https://d oi.org/10 .1007/s1 0461- 019- 02467-
y
Jones, A. A., Sc hneider, K. E., Brigh thaupt, S.-C., Johnson, J. K., Lin ton, S. L., & John-
son, R. M. (2019). Heroin and nonmedical prescription opioid use among high sc ho ol
studen ts in urban sc ho ol districts. Drug and A lc ohol Dep endenc e , 205 , 107664. h t t p s :
//doi.org/10.1016/j.drugalcdep.2019.107664
Jones, C. M. (2018). T rends and k ey correlates of prescription opioid injection misuse in
the United States. A ddictive B ehaviors , 78 , 145–152. h t t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / j . a d d b e h
.2017.10.018
Jones, C. M., Christensen, A., & Gladden, R. M. (2017). Increases in prescription opioid
injection abuse among treatmen t admissions in t he United States, 2004-2013. Drug A l-
c ohol Dep end , 176 , 89–95. https://doi.org/10.1016/j.drugalcdep.2017.03.011
134
References
Jorgensen, C., & W ells, J. (2021). Is marijuana really a gatew a y drug? A nationally repre-
sen tativ e test of the marijuana gatew a y h yp othe sis using a prop ensit y score matc hing
design. Journal of Exp erimental Criminolo gy , 1–18. h t t p s : / / d o i . o r g / 1 0 . 1 0 0 7 / s 1 1 2 92 -
021- 09464- z
Kandel, D. (1975). Stages in adolescen t in v olv emen t in drug use. Scienc e , 190 (4217), 912–
914. https://doi.org/10.1126/science.1188374
Kandel, Denise, & Kandel, E. (2015-02). The gatew a y h yp othesis of substance abuse: De-
v elopmen tal, biological and s o cietal p ersp ectiv es. A cta Pae diatric a (Oslo, Norway :
1992) , 104 (2), 130–137. https://doi.org/10.1111/apa.12851
Kandel, D., & Y amaguc hi, K. (1993). F rom b eer to crac k: Dev elopmen tal patterns of drug
in v olv emen t. A meric an Journal of Public He alth , 83 (6), 851–855. h t t p s : / / d o i . o r g / 1 0 . 2
105/ajph.83.6.851
Kann, L., McMan us, T., Harris, W. A., Shanklin, S. L., Flin t, K. H., Ha wkins, J., … Zaza,
S. (2016). Y outh Risk Beha vior Surv eillance – United States, 2015. MMWR. Surveil-
lanc e Summaries , 65 (6), 1–174. https://doi.org/10.15585/mmwr.ss6506a1
Kaplan, E. L., & Meier, P . (1958). Nonparametric estimation from incomplete observ a-
tions. Journal of the A meric an Statistic al A sso ciation , 53 (282), 457–481. h t t p s : / / d o i . o r
g/10.1080/01621459.1958.10501452
Keen, L., Khan, M., Clifford, L., Harrell, P . T., & Latimer, W. W. (2014). Injection and
non-injection drug use and infectious disease i n Baltimore Cit y: Differences b y race.
A ddictive Behaviors , 39 (9), 1325–1328. https://doi.org/10.1016/j.addbeh.2014.04.020
Keeshin, S. W., & F ein b erg, J. (2016). Endo carditis as a mark er for new epidemics of in-
jection drug use. The A meric an Journal of the Me dic al Scienc es , 352 (6), 609–614. h t t p
s://doi.org/10.1016/j.amjms.2016.10.002
Kern, D. M., Cep eda, M. S., & Sena, A. G. (2020). Oral opioid prescribing trends in the
United States, 2002–2018. Pain Me dicine , 21 (11), 3215–32 23. h t t p s : / / d o i . o r g / 1 0 . 1 0 9 3 /
pm/pnaa313
Kerr, T., Stoltz, J.-A., Marshall, B. D. L., Lai, C., Strathdee, S. A., & W o o d, E. (2009).
Childho o d trauma and injection drug use among hi gh-risk y outh. Journal of A dolesc ent
He alth , 45 (3), 300–302. https://doi.org/10.1016/j.jadohealth.2009.03.007
Key es, K. M., Hamilton, A., & Kandel, D. B. (2016). Birth cohorts analysis of adolescen t
cigarette smoking and subsequen t marijuana a nd co caine use. A meric an Journal of
Public He alth , 106 (6), 1143–1149. https://doi.org/10.2105/ajph.2016.303128
Khobzi, N., Strik e, C., Ca v alieri, W., Brigh t, R., My ers, T., Calza v ara, L., & Millson, M.
(2009). A qualitativ e study on the initiation in to injection drug use: Necessary and
bac kground pro cesses. A ddiction R ese ar ch & The ory , 17 (5), 5 46–559. h t t p s : / / d o i . o r
g/10.1080/16066350802011664
Khoury , L., T ang, Y. L., Bradley , B., Cub ells, J. F., & Ressler, K. J. (2010). Substance
use, c hildho o d traumatic exp erience, and p os ttraumatic stress disorder in an urban
135
References
civilian p opulation. Depr ession and A nxiety , 27 (12), 1077–1086. h t t p s : / / d o i . o r g / 1 0
.1002/da.20751
Kral, Alex H., Malekinejad, M., V audrey , J., Martinez, A. N., Lorvic k, J., McF arland,
W., & Ra ymond, H. F . (2010). Comparing resp onden t-driv en sampling and targeted
sampling metho ds of recruiting injection drug u sers in San F rancisco. J Urb an He alth ,
87 (5), 839–850. https://doi.org/10.1007/s11524- 010- 9486- 9
Kral, Alex H., W enger, L., No v ak, S. P ., Ch u, D., Corsi, K. F., Coffa, D., … Bluthen thal,
R. N. (2015). Is cannabis use asso ciated with less opioid use among p eople who inject
drugs? Drug and A lc ohol Dep endenc e , 153 , 236–241. https:/ /doi.o rg/10.1 016/j. drugalc d
ep.2015.05.014
Lank enau, S. E., W agner, K. D., Blo om, J. J., Sanders, B., Hathazi, D., & Shin, C. (2010).
The first injection ev en t: Differences among heroin, methamphetamine, co caine, and
k etamine initiates. Journal of Drug Issues , 40 (2), 241–262. h t t p s : / / d o i . o r g / 1 0 . 1 1 7 7 / 0 0
2204261004000201
Lansky , A., Finla yson, T., Johnson, C., Holtzman, D., W ejnert, C., Mitsc h, A., … Crepaz,
N. (2014). Estimating the n um b er of p ersons who inject drugs in the United States b y
meta-analysis to calculate national rate s of HIV and hepatitis C virus infections. PLOS
ONE , 9 (5), e97596. https://doi.org/10.1371/journal.pone.0097596
Lara, M., Gam b oa, C., Kahramanian, M. I., Morales, L. S., & Bautista, D. E. H. (2005).
A cculturation and Latino health in the United States: A review of the literature and its
so ciop olitical con text. A nnual R eview of Public He alth , 26 (1), 367–397 . h t t p s : / / d o i . o r g /
10.1146/annurev.publhealth.26.021304.144615
Leri, F., Bruneau, J., & Stew art, J. (2003-01). Understanding p olydrug use: Review of
heroin and co caine co-use. A ddiction , 98 (1), 7–22. h t t p s : / / d o i . o r g / 1 0 . 1 0 4 6 / j . 1 3 60 -
0443.2003.00236.x
Lin ton, S. L., Celen tano, D. D., Kirk, G. D., & Meh ta, S. H. (2013). The longitudinal as-
so ciation b et w een homelessness, injection dr ug use, and injection-related risk b eha vior
among p ersons with a history of injecti on drug use in Baltimore, MD. Drug A lc ohol De-
p end , 132 (3), 457–465. https://doi.org/10.1016/j.drugalcdep.2013.03.009
Livingston, M. D., Barnett, T. E., Delc her, C., & W agenaar, A. C. (2017). Recreational
cannabis legalization and opioid-related dea ths in Colorado, 2000–2015. A meric an Jour -
nal of Public He alth , 107 (11), 1827–1829. https://doi.org/10.2105/ajph.2017.304059
Luo, F., Li, M., & Florence, C. (2021). State-lev el economic costs of opioid use disorder
and fatal opioid o v erdose — United States, 201 7. MMWR. Morbidity and Mortality
W e ekly R ep ort , 70 (15), 541–546. https://doi.org/10.15585/mmwr.mm7015a1
Mac k esy-Amiti, M. E., Donen b erg, G. R., & Ouellet, L. J. (2012). Prev alence of psyc hi-
atric disorders among y oung injection drug users . Drug A lc ohol Dep end , 124 (1-2), 70–
78. https://doi.org/10.1016/j.drugalcdep.2011.12.012
Mac k esy-Amiti, M. E., F endric h, M., & Goldstein, P . J. (1997). Sequence of drug use
among serious drug users: T ypical vs at ypical progression. Drug A lc ohol Dep end , 45 (3),
185–196. https://doi.org/10.1016/s0376- 8716(97)00032- x
136
References
Mars, S. G., Bourgois, P ., Karandinos, G., Mon tero, F., & Ciccarone, D. (2014). ”Ev ery
’nev er’ i ev er said came true”: T ransitions from opioid pills to heroin injecting. Interna-
tional Journal of Drug Policy , 25 (2), 257–266. h t t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / j . d r u g p o . 2 0 1 3 .
10.004
Martin ussen, T., & Sc heik e, T. H. (2006). Dynamic r e gr ession mo dels for survival data .
New Y ork, NY: Springer. https://doi.org/10.1007/0- 387- 33960- 4
Martin ussen, T., & V ansteelandt, S. (2013). On collapsibilit y and confounding bias in Co x
and Aalen regression mo dels. Lifetime Data A nalysis , 19 (3), 279–296. h t t p s : / / d o i . o r g /
10.1007/s10985- 013- 9242- z
Mathers, B. M., Degenhardt, L., Bucello, C., Lemon, J., Wiessing, L., & Hic k-
man, M. (2013). Mortalit y among p eople who inject drugs: A systematic review
and meta-analysis. Bul letin of the W orld He alth Or ganization , 91 (2), 102–123.
https://doi.org/10.2471/blt.12.108282
Maust, D. T., Blo w, F. C., Wiec hers, I. R., Kales, H. C., & Marcus, S. C. (2017). National
trends in an tidepressan t, b enzo diazepine, and o ther sedativ e-h ypnotic treatmen t of
older adults in psyc hiatric and primary care. The Journal of Clinic al Psychiatry ,
78 (04), e363–e371. https://doi.org/10.4088/jcp.16m10713
McKeague, I. W., & Sasieni, P . D. (1994). A partly parametric additiv e risk mo del.
Biometrika , 81 (3), 501. https://doi.org/10.2307/2337122
McLaughlin, M., Ama y a, A., Klev ens, M., O’Cleirigh, C., & Batc helder, A. (2020). A re-
view of factors asso c iated with age of first injection. Journal of Psycho active Drugs ,
52 (5), 1–9. https://doi.org/10.1080/02791072.2020.1804648
Medley , G., Lipari, R. N., Bose, J., Cribb, D. S., Kroutil, L. A., & McHenry , G. (2016).
Sexual orien tation and estimates of adult subs tance use and men tal health: Results
from the 2015 National Surv ey on Drug Use a nd Health. NSDUH Data R eview . Re-
triev ed from http://www.samhsa.gov/data/
Meh ta, S. H., Astem b orski, J., Kirk, G. D., Strathdee, S. A., Nelson, K. E., Vlaho v, D.,
& Thomas, D. L. (2011). Changes in blo o d-b orne infection risk among injection drug
users. The Journal of Infe ctious Dise ases , 203 (5), 587–594. h t t p s : / / d o i . o r g / 1 0 . 1 0 9 3 / i n
fdis/jiq112
Mégarbane, B., & Chevillard, L. (2013). The large sp ectrum of pulmonary complications
follo wing illicit drug use: F eatures and mec hanisms. Chemic o-Biolo gic al Inter actions ,
206 (3), 444–451. https://doi.org/10.1016/j.cbi.2013.10.011
Miller, C. L., Strathdee, S. A., Kerr, T., Li, K., & W o o d, E. (2006). F actors asso ciated
with early adolescen t initiation in to injection drug use: Implications for in terv en tion
programs. Journal of A dolesc ent He alth , 38 (4), 462–464. h t t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / j . j a d o
health.2005.03.004
Morral, A. R., McCaffrey , D. F., & P addo c k, S. M. (2002). Reassessing the marijuana
gatew a y effect. A ddiction , 97 (12), 1493–1504. h t t p s : / / d o i . o r g / 1 0 . 1 0 4 6 / j . 1 3 60 - 0 4 4 3
.2002.00280.x
137
References
Neaigus, A., Gy armath y , V. A., Miller, M., F ra jzyngier, V. M., F riedman, S. R., & Jar-
lais, D. C. D. (2006). T ransitions to injecting drug use among noninjecting heroin users.
JAIDS Journal of A c quir e d Immune Deficiency S yndr omes , 41 (4), 493–503. h t t p s : / / d o i .
org/10.1097/01.qai.0000186391.49205.3b
New com b, M. E., Birk ett, M., Corliss, H. L., & Mustanski, B. (2014). Sexual orien tation,
gender, and racial differences in illicit d rug use in a sample of US high sc ho ol studen ts.
A meric an Journal of Public He alth , 104 (2), 304–310. h t t p s : / / d o i . o r g / 1 0 . 2 1 0 5 / a j p h . 2 0 1 3 .
301702
Nicosia, N., P acula, R., Kilmer, B., Lundb erg, R., & Chiesa, J. (2009). The economic cost
of methamphetamine use in the United States , 2005. https://doi.org/10.7249/mg829
No v ak, S. P ., Bluthen thal, R. N., W enger, L., Ch u, D., & Kral, A. H. (2016). Initiation
of heroin and prescription opioid pain reliev ers b y birth cohort. A m J Public He alth ,
106 (2), 298– 300. https://doi.org/10.2105/ajph.2015.302972
No v ak, S. P ., & Kral, A. H. (2011). Comparing injection and non-injection routes of
administration for heroin, methamphetamine, and c o caine users in the United States.
Journal of A ddictive Dise ases , 30 (3), 248–257. h t t p s : / / d o i . o r g / 1 0 . 1 0 8 0 / 1 0 5 5 0 8 8 7 . 2 0 1 1 .
581989
No w otn y , K. M., P erdue, T., Cep eda, A., & V aldez, A. (2017). Men tal health of heroin
users with differing injection drug use histori es: A non-treatmen t sample of Mexican
American y oung adult men. Drug and A lc ohol Dep endenc e , 181 , 124–131. h t t p s : / / d o i .
org/10.1016/j.drugalcdep.2017.08.041
O’Keefe, D., Horyniak, D., & Dietze, P . (2016). F rom initiating injecting drug use to reg-
ular injecting: Retrosp ectiv e surviv al analysis of injecting progression within a sam-
ple of p eople who inject drugs regularly . Drug A lc ohol Dep end , 158 , 177–180. h t t p s :
//doi.org/10.1016/j.drugalcdep.2015.11.022
P age, K., Hahn, J. A., Ev ans, J., Shib oski, S., Lum, P ., Delw art, E., … Busc h, M. P .
(2009). A cute hepatitis C virus infection in y oung adult injection drug users: A
prosp ectiv e study of inciden t infection, resol ution, and reinfection. The Journal of
Infe ctious Dise ases , 200 (8), 1216–1226. https://doi.org/10.1086/605947
P alamar, Joseph J., Da vies, S., Ompad, D. C., Cleland, C. M., & W eitzman, M. (2015).
P o wder co caine and crac k use in the United S tates: An examination of risk for arrest
and so cio economic disparities in use. Drug A lc ohol Dep end , 149 , 108–116. h t t p s : / / d o i .
org/10.1016/j.drugalcdep.2015.01.029
P alamar, Joseph J., & Ompad, D. C. (2013). Demographic and so cio economic correlates
of p o wder co caine and crac k use among hi gh sc ho ol seniors in the United States. The
A meric an Journal of Drug and A lc ohol A buse , 40 (1), 37–43. h t t p s : / / d o i . o r g / 1 0 . 3 1 0 9 / 0 0
952990.2013.838961
P almer, R. H. C., Y oung, S. E., Hopfer, C. J., Corley , R. P ., Stallings, M. C., Cro wley ,
T. J., & Hewitt, J. K. (2009). Dev elopmen tal epidemiology of drug use and abuse in
adolescence and y oung adultho o d: Evidence of generalized risk. Drug A lc ohol Dep end ,
102 (1-3), 78–87. https://doi.org/10.1016/j.drugalcdep.2009.01.012
138
References
P ezalla, E. J., Rosen, D., Erensen, J. G., Haddo x, J. D., & Ma yne, T. J. (2017). Secular
trends in opioid prescribing in the USA. Journal of Pain R ese ar ch , 10 , 383–387. h t t p s :
//doi.org/10.2147/jpr.s129553
Pip er, B. J., Ogden, C. L., Simo y an, O. M., Ch ung, D. Y., Caggiano, J. F., Nic hols, S. D.,
& McCall, K. L. (2018). T rends in use of prescription stim ulan ts in the United States
and territories, 2006 to 2016. PLOS ONE , 13 (11), e0206100. h t t p s : / / d o i . o r g / 1 0 . 1 3 7 1 /
journal.pone.0206100
P ouget, Enrique R., F ong, C., & Rosen blum, A. (2017). Racial/ethnic differences in prev a-
lence trends for heroin use and non-medical u se of prescription opioids among en tran ts
to opioid treatmen t programs, 2005–2016. Substanc e U se & Misuse , 53 (2), 1–11. h t t p s :
//doi.org/10.1080/10826084.2017.1334070
P ouget, Enrique R., F riedman, S. R., Cleland, C. M., T empalski, B., & Co op er, H. L. F.
(2012). Estimates of the p opulation prev alence of injection drug users among Hispanic
residen ts of large US metrop olitan areas. J Urb an He alth , 89 (3), 527–564. h t t p s : / / d o i .
org/10.1007/s11524- 012- 9670- 1
Quinn, K., Bo one, L., Sc heidell, J. D., Mateu-Gelab ert, P ., McGorra y , S. P ., Beharie, N., …
Khan, M. R. (2016). The relationships of c hildho o d trauma and adultho o d prescription
pain reliev er misuse and injection drug use. Drug and A lc ohol Dep endenc e , 169 , 190–
198. https://doi.org/10.1016/j.drugalcdep.2016.09.021
R Core T eam. (2021). R: A language and envir onment for statis tic al c omputing . Vienna,
A ustria: R F oundation for Statistical Computing. Retriev ed from h t t p s : / / w w w . R - p r o j
ect.org/
Rosner, B., Neicun, J., Y ang, J. C., & Roman-Urrestarazu, A. (2021). Substance
use among sexual minorities in the US – l ink ed to inequalities and unmet need
for men tal health treatmen t? Results from the National Surv ey on Drug Use
and Health (NSDUH). Journal of Psychiatric R ese ar ch , 135 , 107–118. h t t p s :
//doi.org/10.1016/j.jpsychires.2020.12.023
Ro y , É., Haley , N., Leclerc, P ., Cédras, L., Blais, L., & Boivin, J.-F. (2003). Drug injec-
tion among street y ouths in Mon treal: Predictors of initiation. Journal of Urb an He alth ,
80 (1), 92–105. https://doi.org/10.1093/jurban/jtg092
R y an, J. L., & Rosa, V. R. (2020). Healthcare cost asso ciations of patien ts who use illicit
drugs in Florida: A retrosp ectiv e analysis. Substanc e A buse T r e atment, Pr evention, and
Policy , 15 (1), 73. https://doi.org/10.1186/s13011- 020- 00313- 2
Salas-W righ t, C. P ., V aughn, M. G., Clark, T. T., T erzis, L. D., & Córdo v a, D. (2014).
Substance use disorders among first- and se cond-generation immigran t adults in the
United States: Evidence of an immigran t parado x? Journal of Studies on A lc ohol and
Drugs , 75 (6), 958–967. https://doi.org/10.15288/jsad.2014.75.958
SAMHSA. (2017). T r e atment Episo de Data Set (TEDS): 2005-2015. National admissions
to s ubstanc e abuse tr e atment servic es . Ro c kville, MD. Retriev ed from h t t p s : / / d a t a f i l e s .
samhsa.gov
139
References
Sc heik e, T. H., & Zhang, M.-J. (2011). Analyzing comp eting risk data using the R timereg
pac kage. Journal of Statistic al Softwar e , 38 (2). https://doi.org/10.18637/jss.v038.i02
Sherman, S. G., F uller, C. M., Shah, N., Ompad, D. V., Vlaho v, D., & Strathdee, S. A.
(2005). Correlates of initiation of injection drug use among y oung drug users in Bal-
timore, Maryland: The need for early in terv en tion. Journal of Psycho active Drugs ,
37 (4), 437–443. https://doi.org/10.1080/02791072.2005.10399817
Smart, R., & P acula, R. L. (2019). Early evidence of the impact of cannabis legalization
on cannabis use, cannabis use disorder, and t he use of other substances: Findings from
state p olicy ev aluations. The A meric an Journal of Drug and A lc ohol A buse , 45 (6), 1–
20. https://doi.org/10.1080/00952990.2019.1669626
Strik e, C., Rotondi, M., K olla, G., Ro y , E., Rotondi, N. K., R udzinski, K., … Hun t,
N. (2014). In terrupting the so cial pro cesses link ed with initiation of injection
drug use: Results from a pilot study . Drug A lc ohol Dep end , 137 , 48–54. h t t p s :
//doi.org/10.1016/j.drugalcdep.2014.01.004
Sussman, S., Lev en thal, A., Bluthen thal, R. N., F reim uth, M., F orster, M., & Ames, S. L.
(2011). A framew ork for the sp ecificit y of addictions. International Journal of Envir on-
mental R ese ar ch and Public He alth , 8 (8), 3399–3415. h t t p s : / / d o i . o r g / 1 0 . 3 3 9 0 / i j e r p h 8 0
83399
T aplin, C., Saddic hha, S., Li, K., & Krausz, M. R. (2014). F amily history of alcohol and
drug abuse, c hildho o d trauma, and age of first d rug injection. Substanc e U se & Misuse ,
49 (10), 1311–1316. https://doi.org/10.3109/10826084.2014.901383
T empalski, B., P ouget, E. R., Cleland, C. M., Brady , J. E., Co op er, H. L. F., Hall, H. I., …
F riedman, S. R. (2013). T rends in the p opulation prev alence of p eople who inject drugs
in US metrop olitan areas 1992-2007. PLOS ONE , 8 (6), e64789. h t t p s : / / d o i . o r g / 1 0 . 1 3 7
1/journal.pone.0064789
Therneau, T. M. (2021). A p ackage for survival analysis in R . Retriev ed from h t t p s : / / C R
AN.R- project.org/package=survival
Therneau, T. M., & Gram bsc h, P . M. (2000). Mo deling survival data: Extending the Cox
mo del . New Y ork, NY: Springer. https://doi.org/10.1007/978- 1- 4757- 3294- 8
T orrens, M., Gilc hrist, G., Domingo-Salv an y , A., & Group, psyCoBarcelona. (2011). Psy-
c hiatric comorbidit y in illicit drug users: Substance-induced v ersus indep enden t disor-
ders. Drug A lc ohol Dep end , 113 (2-3), 147–156. h t t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / j . d r u g a l c d e p . 2
010.07.013
V a ldez, A., Neaigus, A., & Cep eda, A. (2007). P oten tial risk factors for injecting among
Mexican American non-injecting heroin users. Journal of Ethnicity in Substanc e A buse ,
6 (2), 49–73. https://doi.org/10.1300/j233v06n02/_05
V a ldez, A., Neaigus, A., Kaplan, C., & Cep eda, A. (2011). High rates of transitions to in-
jecting drug use among Mexican American non- injecting heroin users in San An tonio,
T exas (nev er and former injectors). Drug A lc ohol Dep end , 114 (2-3), 233–236. h t t p s :
//doi.org/10.1016/j.drugalcdep.2010.09.014
140
References
V a n yuk o v, M. M., & Ridenour, T. A. (2012). Common liabilit y to drug addictions: The-
ory , researc h, practice. Drug A lc ohol Dep end , 123 Suppl 1 , S1–2. h t t p s : / / d o i . o r g / 1 0 . 1 0 1
6/j.drugalcdep.2012.01.005
V a n yuk o v, M. M., T arter, R. E., Kirillo v a, G. P ., Kirisci, L., Reynolds, M. D., Kreek, M.
J., … Ridenour, T. A. (2012). Common liabilit y to addiction and ”gatew a y h yp othesis”:
Theoretical, empirical and ev olutionary p ersp ectiv e. Drug A lc ohol Dep end , 123 Suppl 1 ,
S3–17. https://doi.org/10.1016/j.drugalcdep.2011.12.018
V o lk o w, N. D., Han, B., Einstein, E. B., & Compton, W. M. (2021). Prev alence of sub-
stance use disorders b y time since first s ubstance use among y oung p eople in the US.
JAMA Pe diatrics , 175 (6). https://doi.org/10.1001/jamapediatrics.2020.6981
W all, M., Cheslac k-P osta v a, K., Hu, M.-C., F eng, T., Griesler, P ., & Kandel, D. B. (2018).
Nonmedical prescription opioids and path w a ys of drug i n v olv emen t in the US: Genera-
tional differences. Drug A lc ohol Dep end , 182 , 103–111. ht t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / j .d r u g a l
cdep.2017.10.013
W atters, J. K., & Biernac ki, P . (1989). T argeted sampling: Options for the study of hid-
den p opulations. So cial Pr oblems , 36 (4), 416–430. https://doi.org/10.2307/800824
W ei, L. J. (1992). The accelerated failure time mo del: A useful alternativ e to the Co x re-
gression mo del in surviv al analysis. Statistics in Me dicine , 11 (14‐15), 1871–1879. h t t p s :
//doi.org/10.1002/sim.4780111409
W erb, Dan, Buxton, J., Sho v eller, J., Ric hardson, C., Ro w ell, G., & W o o d, E. (2013). In-
terv en tions to prev en t the initiation of injecti on drug use: A systematic review. Drug
A lc ohol Dep end , 133 (2), 669–676. https://doi.org/10.1016/j.drugalcdep.2013.08.017
W erb, Daniel, Garfein, R., Kerr, T., Da vidson, P ., Roux, P ., Jauffret-Roustide, M., …
Strathdee, S. A. (2016). A so cio-structural approac h to prev en ting injection drug use
initiation: Rationale for the PRIMER study . Harm R e duction Journal , 13 (1), 25.
https://doi.org/10.1186/s12954- 016- 0114- 1
W erb, Dan, Kerr, T., Buxton, J., Sho v eller, J., Ric hardson, C., Mon taner, J., & W o o d, E.
(2013). Crystal methamphetamine and initiation of injection drug use among street-
in v olv ed y outh in a Canadian setting. CMAJ : Canadian Me dic al A sso ciation Journal
= Journal de l’A sso ciation Me dic ale Canadienne , 185 (18), 1569–1575. h t t p s : / / d o i . o r g /
10.1503/cmaj.130295
Wic kham, H. (2016). ggplot2: Ele gant gr aphics for data analysis . New Y ork, NY: Springer.
https://doi.org/10.1007/978- 3- 319- 24277- 4
W o o d, E., Stoltz, J.-A., Zhang, R., Strathdee, S. A., Mon taner, J. S. G., & Kerr, T. (2008).
Circumstances of first crystal methampheta mine use and initiation of injection drug use
among high-risk y outh. Drug A lc ohol R ev , 27 (3), 270–276. h t t p s : / / d o i . o r g / 1 0 . 1 0 8 0 / 0 9
595230801914750
W u rcel, A. G., Anderson, J. E., Ch ui, K. K. H., Skinner, S., Kno x, T. A., Sn ydman, D. R.,
& Stopka, T. J. (2016). Increasing infectious endo carditis admissions among y oung p eo-
ple who inject drugs. Op en F orum Infe ctious Dise ases , 3 (3), ofw157. h t t p s : / / d o i . o r g / 1 0
.1093/ofid/ofw157
141
References
Y a maguc hi, K., & Kandel, D. B. (1984). P atterns of drug use from adolescence to y oung
adultho o d: I I I. Predictors of progression. A meric an Journal of Public He alth , 74 (7),
673–681. https://doi.org/10.2105/ajph.74.7.673
Zhang, S., W u, S., W u, Q., Durkin, D. W., & Marsiglia, F. F. (2021). A dolescen t drug use
initiation and transition in to other drugs: A retrosp ectiv e longitudinal examination
across race/ethnicit y . A ddictive Behaviors , 113 , 106679. h t t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / j .
addbeh.2020.106679
Zibb ell, J. E., Asher, A. K., P atel, R. C., Kupronis, B., Iqbal, K., W ard, J. W., & Holtz-
man, D. (2018). Increases in acute hepatitis C virus infection related to a gro wing opi-
oid epidemic and asso ciated injection drug use , United States, 2004 to 2014. A meric an
Journal of Public He alth , 108 (2), 175–181. https://doi.org/10.2105/ajph.2017.304132
Zin b erg, N. E. (1986). Drug, set and setting: The b asis for c ontr ol le d toxic ant use . New
Ha v en, CT: Y ale Univ ersit y Press.
142
References
143
References
144
Abstract (if available)
Abstract
This dissertation studies the contributing factors to illicit drug use behaviors by using a longitudinal, developmental perspective of drug use over the drug use career. The three component studies rely on the same data and similar methods to examine the contributions of drug type, individual, and contextual factors on multiple drug use outcomes among a sample of people who inject drugs (PWID), a subset of people who use drugs (PWUD) who disproportionately experience many of the negative consequences associated with drug use. The three studies are distinguished by examining different sets of drug use outcomes using different sets of drug use exposure predictors to investigate the contributions of (1) any-route drug use exposures on any-route drug use initiations, (2) any-route drug use exposures on first injection drug use, and (3) injection drug use exposures on injection drug use initiations of other drugs. The contributions of drug use exposures were evaluated along with independent contributions of individual traits and generational cohort differences. Results generally revealed that prior drug use exposures differentially affected drug use initiation outcomes and differences of risk by outcome were observed by individual traits and generational cohort. Longitudinal risk associations were observed for any-route drug use behaviors and specifically for injection drug use behaviors. These results confirm the idea of a gateway effect in that certain drug use behaviors may increase the risk of initiating other drug use behaviors independently of individual or contextual risk effects. These findings may represent underlying risk processes situated within and between individuals and the risk of initiating drug use behaviors can be attributed to a necessary interaction of individual and social risks. Drug use exposures may be important markers of risk trajectories over the drug use career and can inform strategies to prevent initiations of drug use behaviors.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Opioid withdrawal symptoms, opioid use, and injection risk behaviors among people who inject drugs (PWID)
PDF
A qualitative study of street fentanyl in Dayton, Ohio: drug markets, trajectories, and overdose risk reduction
PDF
Competing risks: the role of the perceived consequences of refusing to share injection equipment among injection drug users
PDF
Energy drink consumption, substance use and attention-deficit/hyperactivity disorder among adolescents
PDF
A multi-system evaluation of medication for opioid use disorder for people who use opioids
PDF
The Internet activities, gratifications, and health consequences related to compulsive Internet use
PDF
Mixed methods investigation of user engagement with a smoking cessation app
PDF
Effects of the perceived and objectively assessed environment on physical activity in adults and children
PDF
Distress tolerance and mindfulness disposition: associations with substance use during adolescence and emerging adulthood
PDF
Cultural risk and protective factors for tobacco use behaviors and depressive symptoms among American Indian adolescents in California
PDF
Normative and network influences on electronic cigarette use among adolescents
PDF
A network analysis of online and offline social influence processes in relation to adolescent smoking and alcohol use
PDF
Prospective associations of stress, compulsive internet use, and posttraumatic growth among emerging adults
PDF
Diffusion of a peer‐led suicide preventive intervention in secondary schools: strategies to increase effectiveness of peer‐led interventions
PDF
Adolescent social networks, smoking, and loneliness
PDF
System dynamics of in-home family eating behavior: insights from intensive longitudinal data using Ecological Momentary Assessment and wearable sensors
PDF
Contextualizing social network influences on substance use among high risk adolescents
PDF
The acute and longitudinal associations between sedentary behaviors, affective states, and emotional disorder symptoms among youth
PDF
Emotion regulation and heart rate variability among smokers
PDF
Development of a perioperative drug screening algorithm for patients with a history of cocaine and methamphetamine use presenting for an elective procedure with anesthesia
Asset Metadata
Creator
Chu, Daniel
(author)
Core Title
Pathways of drug use among people who inject drugs
School
Keck School of Medicine
Degree
Doctor of Philosophy
Degree Program
Preventive Medicine (Health Behavior Research)
Degree Conferral Date
2021-12
Publication Date
08/10/2021
Defense Date
08/10/2021
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
cocaine,drug use,heroin,illicit drugs,injection drug use,methamphetamine,OAI-PMH Harvest,opioids
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Bluthenthal, Ricky N. (
committee chair
), Cepeda, Alice (
committee member
), de la Haye, Kayla (
committee member
), Huh, Jimi (
committee member
)
Creator Email
chudanie@usc.edu,cikoykip@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-oUC15722786
Unique identifier
UC15722786
Legacy Identifier
etd-ChuDaniel-10033
Document Type
Dissertation
Rights
Chu, Daniel
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Repository Email
cisadmin@lib.usc.edu
Tags
cocaine
drug use
heroin
illicit drugs
injection drug use
methamphetamine
opioids