International Journal of Drug Policy 55 (2018) 61–69

Contents lists available at ScienceDirect

International Journal of Drug Policy

journal homepage: www.elsevier.com/locate/drugpo

Research Paper

Effects of naloxone distribution to likely bystanders: Results of an agent- T

based model
⁎
Christopher Keane, James E. Egan, Mary Hawk
University of Pittsburgh, Graduate School of Public Health, Department of Behavioral and Community Health Sciences, 6124 Parran Hall, 130 DeSoto Street, Pittsburgh,
PA, 15261, USA

A R T I C L E I N F O A B S T R A C T

Keywords: Background: Opioid overdose deaths in the US rose dramatically in the past 16 years, creating an urgent national
Opioid overdose health crisis with no signs of immediate relief. In 2017, the President of the US officially declared the opioid
Naloxone distribution epidemic to be a national emergency and called for additional resources to respond to the crisis. Distributing
Harm reduction naloxone to community laypersons and people at high risk for opioid overdose can prevent overdose death, but
Agent-based modeling
optimal distribution methods have not yet been pinpointed.
Methods: We conducted a sequential exploratory mixed methods design using qualitative data to inform an
agent-based model to improve understanding of effective community-based naloxone distribution to laypersons
to reverse opioid overdose. The individuals in the model were endowed with cognitive and behavioral variables
and accessed naloxone via community sites such as pharmacies, hospitals, and urgent-care centers. We compared
overdose deaths over a simulated 6-month period while varying the number of distribution sites (0, 1, and 10)
and number of kits given to individuals per visit (1 versus 10). Specifically, we ran thirty simulations for each of
thirteen distribution models and report average overdose deaths for each. The baseline comparator was no
naloxone distribution. Our simulations explored the effects of distribution through syringe exchange sites with
and without secondary distribution, which refers to distribution of naloxone kits by laypersons within their
social networks and enables ten additional laypersons to administer naloxone to reverse opioid overdose.
Results: Our baseline model with no naloxone distribution predicted there would be 167.9 deaths in a six month
period. A single distribution site, even with 10 kits picked up per visit, decreased overdose deaths by only 8.3%
relative to baseline. However, adding secondary distribution through social networks to a single site resulted in
42.5% fewer overdose deaths relative to baseline. That is slightly higher than the 39.9% decrease associated with
a tenfold increase in the number of sites, all distributing ten kits but with no secondary distribution. This
suggests that, as long as multiple kits are picked up per visit, adding secondary distribution is at least as effective
as increasing sites from one to ten. Combining the addition of secondary distribution with an increase in sites
from one to ten resulted in a 61.1% drop in deaths relative to the baseline. Adding distribution through a syringe
exchange site resulted in a drop of approximately 65% of deaths relative to baseline. In fact, when enabling
distribution through a clean-syringe site, the secondary distribution through networks contributed no additional
drops in deaths.
Conclusion: Community-based naloxone distribution to reverse opioid overdose may significantly reduce deaths.
Optimal distribution methods may include secondary distribution so that the person who picks up naloxone kits
can enable others in the community to administer naloxone, as well as targeting naloxone distribution to sites
where individuals at high-risk for opioid overdose death are likely to visit, such as syringe-exchange programs.
This study design, which paired exploratory qualitative data with agent-based modeling, can be used in other
settings seeking to implement and improve naloxone distribution programs.

Introduction Control and Prevention reports a sharp increase in opioid deaths from
5,990 in 1999 to more than 33,000 in 2015 (“Drug Overdose Death
In the United States, overdose deaths from opioid pain relievers and Data,” 2016). Most of these deaths were preventable (“Facing addiction
heroin rose dramatically in the past 16 years. The Centers for Disease in America: The Surgeon General’s report on alcohol, drugs, and

⁎
Corresponding author.
E-mail addresses: crkcity@pitt.edu (C. Keane), jee48@pitt.edu (J.E. Egan), meh96@pitt.edu (M. Hawk).

https://doi.org/10.1016/j.drugpo.2018.02.008
Received 10 July 2017; Received in revised form 9 January 2018; Accepted 12 February 2018
0955-3959/ © 2018 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/BY-NC-ND/4.0/).
C. Keane et al.
health,” 2016). One way to prevent these overdose deaths is by ad-
ministering naloxone, an opioid overdose reversal drug. Naloxone is a
full antagonist and binds to the same receptors in the central nervous
system that receive opioids, thus displacing opioids for 30–90 min and
reversing the respiratory depression caused by overdose that would
otherwise lead to death. In overdose situations where fentanyl is pre-
sent, repeated doses of naloxone may need to be administered before
successful reversal (Armenian, Vo, Barr-Walker, & Lynch, 2017).
Paramedics, emergency room physicians, and other health professionals
in many countries have long used naloxone as an antidote to opiate
overdose (Chamberlain & Klein, 1994; Clarke, Dargan, & Jones, 2005).
In recent years, numerous programs have been developed in com-
munities to train and provide likely bystanders with naloxone for peer
administration (Best et al., 2002; Galea et al., 2006; Maxwell, Bigg,
Stanczykiewicz, & Carlberg-Racich, 2006). Bystanders are people who
may have the opportunity to administer naloxone due to their proxi-
mity to someone experiencing an overdose. Bystanders are laypersons,
not health professionals, and may include friends and family members
as well as people who use opioids themselves and are likely to be
nearby when an individual experiences overdose. Some states in the US
have passed “standing order” policies to expand access to naloxone
without a patient-specific prescription (Davis & Carr, 2017). The
strategy is to distribute naloxone, along with a basic knowledge of how
and when to use it, to laypersons who are likely to be nearby when
someone experiences an overdose. Though overdose may occur within a
1–3 h window after using opioids, once overdose has begun and
breathing is impaired, rescue breathing and naloxone should be ad-
ministered immediately (“Facing addiction in America: The Surgeon
General’s report on alcohol, drugs, and health,” 2016).
Some guidelines recommend that a bystander first call for
Emergency Medical Services before administering naloxone and many
states have ‘Good Samaritan Laws’ protecting the bystander and the
user from arrest if the bystander reports an overdose, even if one or
both were using drugs (Wickramatilake et al., 2017). However, many
bystanders fear calling 911 due to potential arrest despite Good Sa-
maritan Laws, and even when a fearful person does call for emergency
aid they may devote valuable time removing drug paraphernalia from
the site rather than tending to the person who has overdosed (Seal
et al., 2003; Tobin, Davey, & Latkin, 2005). Thus, while bystanders
might hesitate to call emergency services, they might administer na-
loxone themselves.
A growing body of research examines the impact of naloxone dis-
tribution programs targeted to laypersons. A recent review of 41 studies
of overdose education and naloxone distribution efforts determined
that layperson distribution programs are feasible, that laypersons are
willing to administer naloxone to reverse opioid overdose, and that
these programs can reduce overdose mortality (Mueller, Walley,
Calcaterra, Glanz, & Binswanger, 2015). A meta-analysis using pooled
data from four studies assessing the effectiveness of layperson dis-
tribution programs reported significantly increased odds of recovery
after overdose when comparing these programs to no naloxone ad-
ministration (OR = 8.58, 95% CI = 3.90 to 13.25) (Giglio, Li, &
DiMaggio, 2015). The effectiveness of layperson-administered naloxone
has resulted in overdose survival rates reaching as high as 96%
(Bennett, Bell, Tomedi, Hulsey, & Kral, 2011; Clark, Wilder, &
Winstanley, 2014; Doe-Simkins et al., 2014; Doe-Simkins, Walley,
Epstein, & Moyer, 2009; Enteen et al., 2010; Lewis, Vo, & Fishman,
2017; Loimer, Hofmann, & Chaudhry, 1992; McDonald & Strang, 2016;
Tobin, Sherman, Beilenson, Welsh, & Latkin, 2009),
The implicit mental model underlying our study is that distributing
naloxone to likely bystanders will increase overdose reversals over and
above reversals that result from bystanders calling emergency response
teams. This conceptual framework has fueled policies supporting lay-
person reversal including the standing order approach as mentioned
above, which is now active in many states in the US (Wickramatilake
et al., 2017). It would be useful to examine how various dynamics
62
International Journal of Drug Policy 55 (2018) 61–69
related to naloxone distribution might interact and then simulate their
effects on overdose deaths, which could be used to refine and broaden
existing interventions. To do this we conducted a sequential exploratory
mixed methods design using qualitative data to inform an agent-based
model (Creswell & Clark, 2011). In phase one of the design we con-
ducted qualitative interviews with providers of substance use treatment
programs as well as with consumers who were at risk for opioid over-
dose, had personal experience with naloxone, or had received services
from substance use intervention programs. In phase two of the design
we built an agent-based model to simulate the effects of various models
of naloxone distribution as informed by our qualitative results and by
the extant literature.
Our study builds on previous research that has demonstrated value
of this design, particularly in studies addressing illicit substance use.
Improving health outcomes for people who use opioids and other illicit
drugs requires an understanding of the social context of those most
affected, yet a long history of criminalizing and marginalizing this
population demands the use of research methods that do not place these
individuals at additional risk of exposure or exploitation. Building the
agent-based model on the lived experiences of the community provides
an important opportunity to magnify the use of ethnographic data,
which can in turn inform policy and intervention development (Hoffer,
Bobashev, & Morris, 2009).
Qualitative methods
Purpose of qualitative study
The purpose of the qualitative study was to explore feasibility and
acceptability of various models of community-based naloxone dis-
tribution for layperson reversal. Interviews were chosen as the most
appropriate method to elicit rich qualitative data given the sensitive
nature of questions related to substance use. A harm reduction frame-
work guided our selection of general issues to explore (Hawk et al.,
2017), and topics related to layperson naloxone administration were
derived from the literature (Bennett et al., 2011; Clark et al., 2014; Doe-
Simkins et al., 2014, 2009; Lewis et al., 2017; Loimer et al., 1992;
McDonald & Strang, 2016; Tobin et al., 2009). Domains that were ex-
plored with both providers and consumers focused on knowledge and
experience with naloxone, access and barriers to naloxone distribution,
thoughts about ideal naloxone distribution methods and messaging that
would be appropriate for such methods, and finally subjective responses
to naloxone administration.
Sample
From July through September 2016 we conducted qualitative in-
terviews with 7 substance use treatment providers covering a large
portion of the service area and with 22 people at high risk for opioid
overdose (current or recent injection drug users). All interviews took
place in urban or suburban settings in Southwestern Pennsylvania and
took approximately 45–60 min to complete. We conducted online re-
search to identify local substance use treatment agencies then called or
emailed directors of the programs to explain the study purpose and
invited them to participate in interviews. We included providers of
abstinence-based services as well as those that provide harm reduction-
informed services. Consumers who were interviewed were those at high
risk for overdose, i.e., those actively using opioids, as well as in-
dividuals who had previously experienced overdose reversals person-
ally or who had administered naloxone to reverse someone else’s opioid
overdose. We also included consumers from Medication Assisted
Treatment (MAT) programs, such as Suboxone clinics. Consumers were
recruited through treatment providers, primarily by placing flyers in
waiting rooms of provider settings or at the local syringe exchange van.
Several other participants learned of our study via word-of-mouth re-
ferrals from their social networks. We purposively sampled consumers
C. Keane et al.
to ensure representation of people who were currently using opioids,
were in MAT programs, or who reported no current substance use.
Recruitment continued until saturation was achieved. Consumers re-
ceived $40 cash to honor their time; providers received no incentives.
This study was approved by the University of Pittsburgh Human
Research Protection Office.
Interviewing, coding and analysis
All interviews were conducted in private rooms in provider or
community facilities to ensure safety and confidentiality of participants.
Interviews were audiotaped and transcribed verbatim. We conducted
coding and analysis using Dedoose, using both a deductive content
analysis to examine a priori domains (such as barriers and facilitators to
naloxone) (Elo & Kyngas, 2008) as well as an inductive approach to
explore emerging themes (Patton, 1990, 2015). Our study team, con-
sisting of two experienced qualitative researchers and four masters-
level students, reviewed interview transcripts to contextualize key
themes that emerged during the interviews and to develop an initial
codebook. Next, all six members of the team coded one transcript using
line-by-line coding to test for consistent application of codes and to add
important new emergent codes. Coding discrepancies were discussed
until consensus was reached. The students then coded the rest of the
transcripts using the finalized codebook, co-coding every fifth transcript
with the senior researchers to test and improve coding reliability. Fi-
nally, the team used axial coding to review code co-occurrences and
understand how themes were interrelated.
Qualitative results
Full results of this study are not yet published, however, herein we
report on the qualitative findings that informed our agent-based model.
Providers who participated in interviews served as executive or pro-
gram directors of the organizations where they worked. Consumers
include individuals who identified as white and black, male and female,
and current and former opioid users. All consumers were stably housed
at the time of the interviews. Themes that emerged from this study
included the importance of naloxone distribution through trusted
community sites as well as through secondary exchange via social
networks of people at risk for opioid overdose and the need to make
multiple kits available to laypersons. In addition to discussing feasi-
bility of community-based naloxone distribution for layperson reversal,
our qualitative participants also discussed their thoughts about sec-
ondary distribution, in which a layperson accesses multiple naloxone
kits and then distributes these kits through social networks to other
community members, including those at high risk for opioid overdose.
Finally, our participants identified various venues where they might
feel comfortable receiving naloxone, which included pharmacies, hos-
pitals, urgent-care centers, or any place where they felt they could trust
the people they were interacting with as promising distribution venues,
such a syringe exchange site.
Some participants discussed their unwillingness to call emergency
response teams for fear of police involvement despite the presence of
Good Samaritan laws. We also found that some individuals might
hesitate to use naloxone if they were not sure how to administer it,
suspected some danger, or if they did not want to break the “high” and
induce feelings of sickness in the user. Some people who had previously
used injection drugs indicated they might not want the injectable form
of naloxone on their person for fear that the needles would act as a
“trigger” for injection drug use, although others who were interviewed
reported it would not serve as a trigger for them. Fear of needles or
other difficulties in using the injectable version of naloxone was an-
other reported obstacle, although both of these concerns would be re-
moved if the intranasal version of naloxone was available, which is
marketed under the brand name Narcan.
Our respondents largely believed that naloxone distribution to
63
International Journal of Drug Policy 55 (2018) 61–69
bystanders would be effective, and many reported having personally
experienced or enacted multiple opioid reversals via use of naloxone.
Despite myths about risk compensation, in which some believe that
providing naloxone could encourage increased opiate use, our quali-
tative participants rejected this idea and prior research has also shown
no evidence of compensatory use associated with take-home naloxone
programs (Jones, Campbell, Metz, & Comer, 2017).
Quantitative methods
Overview
We constructed an agent-based model of interacting individuals
endowed with cognitive and behavioral variables using NetLogo soft-
ware (Wilensky & Evanston, 1999). The model distributes individuals’
behavioral variables along a random normal curve and included social
contact; opioid use, overdose and death; administration of naloxone
and giving naloxone kits to others; calling emergency response and fear
of calling emergency response; and confidence in administering na-
loxone. We used the model to compare several modes of naloxone
distribution including community-based distribution with and without
secondary distribution, in which one individual picks up multiple kits
and shares them with others, enabling more than one person to ad-
minister naloxone.
The behavioral variables used in the model were informed by
themes that emerged from our qualitative results including fear of
calling emergency response as well as opportunities for naloxone dis-
tribution through community sites, via secondary networks utilizing
opiate using bystanders, and with access to multiple kits at one time.
Wherever possible, the behavioral variables in our model were based on
values available in the literature. We intended our simulations to ap-
proximate opioid overdose death rates, high risk for opioid overdose,
and emergency response times in the region where the study was
conducted, which was Allegheny County, Pennsylvania, USA.
Estimates of opioid overdose deaths and high risk population for overdose in
Allegheny county
In 2014 in our region, there were 229 adults deaths by overdose
(“Fatal Overdoses by Year − Allegheny County,” 2016). The Allegheny
County population in 2014 totaled 1.233 million individuals, and the
adult population comprised about 1 million people (“American
FactFinder-Results,” 2014). We wished to estimate the number of those
1 million adults that had opioid use disorder, which we use as an es-
timate for being at risk for opioid overdose. Since a reliable number for
opioid use disorder was not available for Allegheny County, we used
national rates. Based on a national survey, 2.5 million or 1% of the
United States adult population suffered from opioid use disorders in
2014 (Volkow & McLellan, 2016). One percent of the one million adults
in Allegheny County translates to approximately 10,000 adults with
opioid use disorders in this region. Of these, we simulated a subset of
individuals who die from an opioid overdose and their contacts, who
themselves have opioid use disorder. We estimated a rate of fatal opioid
overdose for this simulated population that would result in approxi-
mately 229 deaths (“Fatal Overdoses by Year − Allegheny County,”
2016), and projected that fatal overdose would primarily occur among
the 10,000 individuals with opioid use disorders in our model.
For the naloxone distribution models we also wished to include
social contacts of those who used opioids (potential bystanders) for
potential reversals. We estimated that the total contact network of those
who use opioids and will eventually overdose would be approximately
1,000 individuals. Each person using opioids will, over the course of a
year, make contacts with an average of approximately nine other in-
dividuals using opioids. This number of contacts comes from a study
that found persons in residential treatment for substance use had be-
tween 8 and 10 contacts with others who use substances in their
C. Keane et al.
personal network over a 12 month period (Min et al., 2013). Thus, an
average of 9 contacts, based on prior research on personal network size,
(Aartsen, Veenstra, & Hansen, 2017), for each of the 229 overdose
deaths yields 2061 persons. With a 52% overlap in the network, this
yields a network of approximately 1,000 persons. We based the contact
overlap upon the national survey by Christakis and Fowler, which
found that the average probability that any two of a respondent’s
contacts know each other was 52% (a clustering coefficient of 0.52)
(Christakis & Fowler, 2009). Thus, we expected the opioid deaths and
bystander saves to occur within this subpopulation of approximately
1000 people who use opioids.
While we expected most of the overdose activity to occur within a
subgroup of approximately 1000 people who use opioids, our larger
model also includes 10,000 individuals who use opioids, and also a
large group that does not use opioids (one million adults in Allegheny
County less the 10,000 who use opioids). Individuals in this large group
never died from overdose in our simulations and never administered
naloxone, since we only included people with opioid use disorders as
bystanders with the potential to administer naloxone. It is likely that
some individuals who do not use opioids and are not emergency re-
sponders might also distribute naloxone. However, our study focused
on the effects of distributing naloxone kits to people at risk for overdose
as informed by our qualitative results in which many people who used
opioids had reversed others in their substance-using social network. We
also included in our model a small group of trained emergency medical
providers who were not bystanders but responded to calls for emer-
gency assistance within a timeframe estimated from Allegheny County
emergency response statistics. In our model, these trained emergency
responders never themselves used opioids.
Simulating opioid use
For each of two types of individuals who used opioids in our model
we implemented a probability of opioid use, one group having zero
probability of using representing no intention to use and therefore had
a zero probability of overdosing. We then used a Poisson distribution to
assign a second group a mean probability of using opioids and a mean
probability of overdose that we estimated would result in approxi-
mately 229 adults overdose deaths after a year. In addition, we assigned
each individual with opioid use disorder a risk of overdosing and a
separate probability of overdose death. Among those who used opioids,
we distributed the actual death rate with a random Poisson distribution
around a mean death rate per overdose. This uniform mean death rate
across all populations is a simplification that might be improved upon
once we have better information about probability of death per over-
dose in separate populations. This death rate might also be higher
among those using fentanyl, but we did not include fentanyl in our
simulations since we wanted to focus on the comparison of different
methods of naloxone kit distribution.
Naloxone distribution methods
The model simulated distribution of naloxone kits through com-
munity sites that individuals may visit. When individuals visited these
community sites they picked up one naloxone kit. Each naloxone kit
contained two doses, and each dose could reverse an overdose in our
model. Although it is possible that more doses would be required for
some cases, such as when opioids are combined with fentanyl, our
simplified model did not address that possibility. We varied the number
of distribution sites (0, 1, and 10) and number of kits given to in-
dividuals per visit (1 versus 10). Our simulations included multiple kit
distribution with and without secondary exchange. Multiple kit dis-
tribution without secondary exchange meant that the bystanders who
accessed the kits would themselves administer naloxone to others in
their social networks. Multiple kit distribution with secondary exchange
meant that bystanders who accessed the kits would give away some of
64
International Journal of Drug Policy 55 (2018) 61–69
the kits to others in their social networks, and those individuals would
in turn administer naloxone to others they encountered.
The naloxone distribution sites in our model were physically located
in randomly assigned spaces. When we modeled ten distribution sites
they were located at random spaces so that when individuals moved to
a neighboring space at random they may visit the single site at an
overall probability of roughly once every 60 days. However, many in-
dividuals visited the site much earlier.
We compared overdose deaths while varying distribution sites and
number of kits distributed per site visit. For each of those conditions, we
also compared the effects of secondary distribution of naloxone com-
pared to no secondary distribution (eight simulation conditions). To
simulate secondary distribution, we had individuals with more than one
naloxone kit give one kit each to individuals who have no naloxone that
they encounter through their social networks. An individual with two
or more kits could successively distribute the kits to others, thus re-
sulting in two or more individuals to be armed with naloxone kits. This
can speed up administering naloxone, because multiple individuals
each with a naloxone kit can act simultaneously whereas one individual
with two or more naloxone kits can only administer one at a time and
must wait until they find another individual overdosing.
Syringe exchange sites
We also examined naloxone distribution at a syringe exchange site.
That is, we included 30 simulations of a separate condition where our
simulated county includes a syringe exchange site where naloxone is
distributed. There is no empirical data on how often individuals visit
syringe exchange sites when they intend to use opioids. Therefore, we
decided to make the assumption that, most of the time when individuals
wish to use opioids, they do not visit the syringe exchange site. We
operationalize not visiting the syringe site “most of the time” as a
probability of 80%. This means that individuals only visit the site with
probability p = 0.20 of the instances when they decide to use opioids.
For this random 20% of instances that individuals decide to use opioids,
they first visit the syringe site and pick up naloxone as part of that visit.
The logic of distributing naloxone through a syringe exchange site is
that visitors to those sites are very likely to use opioids. Participants in
our qualitative interviews reported that they were most likely to access
naloxone kits through the local syringe exchange site, though some also
suggested that it would be feasible to access kits at multiple community
sites that were not necessarily focused on harm reduction strategies for
people who use substances, such as medical offices or bars. We added
the syringe exchange distribution site to the single community dis-
tribution site and compared this scenario to one with a single syringe
exchange distribution site supplementing ten other community dis-
tribution sites.
Social contact
Individuals in our model were awake and moving around the net-
work for approximately eight hours per day. At each time step, re-
presenting one minute during the eight hours of awake time, some in-
dividuals moved to a neighboring space, and this move may or may not
put the individual in contact with another individual. An individual was
in contact with another if they occupied neighboring spaces in a grid
lattice. We estimated contacts in a social network per day from Fu
(2007) who found an average of 21 contacts per day in the United
States (Fu, 2007). To reproduce that number of contacts, the rate of
movement in the network averaged once per 50 min of awake time,
which is ten moves in an eight-hour day. In our grid lattice, any in-
dividual might be next to a maximum of eight other neighbors at any
given time, but we fashioned the network such that the average number
of neighbors would be about 2.1 with a random distribution, (the
average in our actual simulations turned out to be 2.3 neighbors at any
given time, with about 6.5% of individuals alone at any given time).
C. Keane et al.
Fig. 1. Opioid Overdose Deaths by Naloxone Kit Distribution Mode.
Thus, our simulated individuals have about the 20 interactions per day,
since there are 2.3 contacts at any given time (2.3 times 10 moves in the
network is approximately 20 contacts per day). Note that the total
contacts per day may or may not include multiple contacts with some of
the same individuals. For example, one individual might make 18
contacts in one day with a total of 11 different persons, while another
person might make 21 contacts with 8 different persons. This simulated
social contact network is homogenous over space, meaning we did not
attempt to vary the density of the social network or the rate of social
contact over space, because we did not have good source of data for
such variation, nor was variation in network density a focus of our
study.
Bystander administration of naloxone
In our simulations, only those with opioid use disorders could serve
as bystanders and administer naloxone. This represents a simplified
approach since it is likely that some individuals who are not opioid
users and are not trained emergency responders might also distribute
naloxone. However, our study focused on effects of different modes of
distributing naloxone kits to those at risk for overdose in keeping with
our qualitative methods and results.
Bystander levels of confidence: naloxone distribution and calling emergency
response
When encountering an overdose, bystanders in our model might call
emergency response, administer naloxone, or both, depending in part
upon the bystander’s level of confidence with each of these actions. For
example, the average level of confidence in administering a dose of
naloxone was randomly distributed such that the mean time it took to
decide to give a dose of naloxone was 20 min from the time one en-
counters a neighbor with an overdose, but half of the individuals would
decide to give the naloxone sooner than 20 min. Twenty minutes is a
rough estimate as we were not able to find a value for the time it takes
to decide, which underscores the need to explore this decision-making
process in future mixed method studies.
The confidence variables reflect the time it takes to see and
65
International Journal of Drug Policy 55 (2018) 61–69
recognize the overdose symptoms and to know how one wants to re-
spond. Confidence also reflects subjective factors, especially in the case
of calling emergency response. Confidence in calling emergency re-
sponse might be low due to fear of potential arrest. The literature, and
respondents in our qualitative study, indicate that this concern may
exist even where there are Good Samaritan Laws in place. The simu-
lation includes various time sensitive responses, including a delay in
emergency response averaging 15 min (Lamb, Anderson, Novak,
Vanselow, & Ieraci, 2014; Van Osdol, 2015) and a time from opioid use
to death averaging 90 min based on a range provided in the Surgeon
General’s 2016 report (“Facing addiction in America: The Surgeon
General’s report on alcohol, drugs, and health,” 2016). Both variables
followed a random normal distribution around these means.
Time
Each iteration of the simulation model represents one minute during
which every individual’s variables could change depending upon the
probabilities and durations of delays built into the model. We ran the
model for a number of time steps (minutes) equivalent to six months.
Number of simulation runs
Given the different combinations of distribution modes, we com-
pared overdose deaths across thirteen different distribution models,
using no naloxone distribution as a baseline comparison. For each of the
thirteen distribution modes we ran simulations thirty times, resulting in
a total of 390 simulations. Each simulation represented a six month
period. We report average overdose deaths, within the six month
period, for each of the thirteen conditions. While our main outcome
measure is overdose deaths, we also report the total number of overdose
deaths prevented by a dose of naloxone administered by bystanders and
total overdose deaths prevented by emergency response prompted by
calls by bystanders.
C. Keane et al.
Fig. 2. Reduction of Overdose Deaths by Naloxone Distribution Mode Relative to Baseline.
Quantitative results
Distribution sites, secondary exchange, and number of kits distributed
In our simulated community, our baseline condition of no naloxone
distribution through community or syringe exchange sites and with no
secondary exchange resulted in 167.9 deaths in a six-month period
(335.8 in a full year) (Fig. 1). Adding naloxone distribution to one site,
with one kit picked up per visit, resulted in 157.9 opioid overdose
deaths, a 6.0% decrease relative to baseline (Figs. 1 and 2).
Increasing the number of kits picked up per visit at one site from one
to ten decreased deaths to 153.9 (Fig. 1), representing a decrease of
8.3% relative to baseline (Fig. 2). However, adding secondary exchange
through social networks to the single site distribution resulted in 96.6
deaths on average (Fig. 1), or 42.5% fewer overdose deaths relative to
baseline (Fig. 2). That is slightly higher than the 39.9% decrease as-
sociated with a tenfold increase in the number of sites, all distributing
ten kits but with no secondary distribution (Fig. 2). This suggests that as
long as multiple kits are picked up at each visit, adding secondary
distribution is at least as effective as increasing sites from one to ten.
The success of secondary distribution depends upon distributing mul-
tiple kits per visit. Adding secondary distribution to multiple kit dis-
tribution speeds the rate of increase in the number of individuals with
at least one naloxone kit, which in turn increases the rate of saves
through naloxone.
While the most dramatic decrease in deaths was seen when com-
bining multiple kit pickup with secondary distribution, increasing the
total sites along with secondary distribution resulted in further de-
creases in overdose deaths. With this form of secondary distribution in
place, increasing the number of distribution sites to ten resulted in a
61.1% drop in deaths relative to the baseline (Fig. 2), or 65.3 deaths
compared to the baseline number of deaths of 167.9.
Secondary distribution combined with only one community dis-
tribution site decreased overdose deaths by 42.5% (Fig. 2). However,
adding distribution through a syringe site resulted in a much steeper
drop in overdose deaths. Specifically, adding naloxone distribution
through a syringe exchange site reduced overdose deaths by 65.5%
relative to baseline (Fig. 2), resulting in a total of 58 deaths (Fig. 1.)
66
International Journal of Drug Policy 55 (2018) 61–69
Adding secondary distribution through networks contributed no addi-
tional drops in deaths (Fig. 1).
Figs. 1 and 2 show that there is a strong interaction effect between
secondary distribution and the number of kits picked up per site visit.
While both of those models had relatively little effect alone, together
they decreased deaths by a very large amount, over and above the ef-
fects of increasing the number of distribution sites.
Emergency response
We also examined the relative role of naloxone distribution to likely
bystanders and overdose reversals by emergency response. Fig. 3 shows
changes in the number of emergency response reversals and the number
of bystander reversals due to our naloxone distribution when the var-
ious modes of community-based distribution are enacted. We see that
when naloxone distribution increases, the number of emergency re-
sponse reversals tends to decrease because they are displaced by by-
stander saves. Our baseline simulation with no naloxone administration
by bystanders resulted in 535.9 overdose reversals by emergency re-
sponse teams. However, enacting secondary distribution (10 kits per
visit) at 10 community sites resulted in 711.1 reversals by bystanders
and only 291.9 reversals by emergency response (Fig. 3).
In all simulation runs, the total number of emergency response re-
versals rose more quickly than the number of bystander reversals be-
cause it takes some time for naloxone kits to spread into the commu-
nity. That is, it takes some time for individuals to pick up naloxone kits
from distribution sites and then reach overdosing individuals. By con-
trast, emergency response is a system already in place, and all an in-
dividual needs to do in the event of an overdose is to call emergency
response. However, when there were ten naloxone distribution sites,
representing one site per hundred at-risk persons, after two months or
less the number of bystander reversal rose to a level exceeding emer-
gency response reversals.
It is also important to note that many overdose reversals took place
within the same high risk individuals repeatedly, and this tended to
buffer the effects of naloxone distribution on total deaths. Naloxone
may save any given individual at one or more points in time, yet the
individual could at a later point of time overdose again and, if no
C. Keane et al.
Fig. 3. Overdose Deaths Prevented by Naloxone Kit Distribution Mode.
naloxone is available, could die from the overdose. Emergency response
also buffered the evident effects of low versus high naloxone distribu-
tion to likely bystanders. When naloxone distribution is lower, there are
more overdose reversals due to emergency response. Of course, when
we increased the number of naloxone distribution sites from one to ten,
the level of layperson reversals increased, but the increase in actual
bystander reversals was much less than tenfold, largely due to the
backup effect of emergency response. By backup effect of emergency
response, we mean that when a person who overdoses is not rescued by
a bystander administering naloxone a bystander may still call emer-
gency response. In this way, calling emergency response acts as the
fallback mode of response.
Conclusions
Community-based naloxone distribution significantly lowered the
opioid overdose death rate in our simulated community given sufficient
time for naloxone kits to diffuse into the community and to people at
high risk for opioid overdose. While multiple kit pickup distribution
and secondary distribution each had a relatively small effect when
enacted alone, there was a very strong interaction between the two.
Interestingly, when secondary distribution was enabled, increasing kits
per visit tenfold was roughly equivalent to increasing the number sites
tenfold. Our simulations also demonstrate that secondary distribution
was very effective even when there is only one distribution site, as long
as there are many kits given out per visit to the site. Distribution
through syringe exchange sites or through secondary distribution each
have dramatic effects on overdose deaths.
The proportion of overdose deaths averted does not go beyond
about 65% when increasing the number of sites and adding secondary
distribution. This apparent ceiling is due to the fact that the system
reaches a threshold beyond which giving more kits does not make any
difference because all contacts have naloxone kits. Further decreases in
overdose deaths would require increasing individuals' confidence in
administering naloxone. We did not wish to simulate effects of in-
creasing confidence in administering naloxone in this paper. Rather, the
aim of our paper was to examine the effects of distribution of naloxone
independent of any education that might increase the confidence in
administering naloxone. Thus, we were able to demonstrate reduction
67
International Journal of Drug Policy 55 (2018) 61–69
of overdose deaths as high as 65% due to distribution even with no
alteration in baseline confidence in administering naloxone. However,
if these distribution efforts were coupled with education that success-
fully increased confidence in administering naloxone then the overdose
deaths could be further reduced. Future studies, which might include
interventions to improve health literacy among populations who use
substances, could explore the degree to which increasing confidence in
naloxone administration would result in subsequent reductions in
overdose deaths.
The existence of a ceiling in deaths averted would also have eco-
nomic consequences. While an economic assessment goes beyond the
scope of this paper, a future study should compare the costs of naloxone
distribution versus deaths averted, as well as a comparison of some
naloxone kits that go unused against a small increase in reductions of
overdose death. If health education is found to contribute to further
decreases in overdose deaths then the cost of that health education
should be factored in to that economic analysis.
Our findings underscore the importance of harm reduction strate-
gies including syringe exchange sites, which have previously been
found to produce critical public health improvements such as im-
provements in linkage to social services as well as reduced rates of
Hepatitis C and HIV infection (Abdul-Quader et al., 2013; Bramson
et al., 2015; Des Jarlais et al., 2015; Fernandes et al., 2017). It is im-
portant that future research examines the impact of approaches that
improve health literacy of those who use substances and empower those
at risk for overdose to engage in the development of public health so-
lutions. The study design, which paired exploratory qualitative data
with agent-based modeling, could be replicated to better understand
decision-making processes by likely bystanders and improve naloxone
distribution programs. Additional research is also need to understand
the cost-effectiveness of bystander naloxone distribution programs.
The following limitations should be considered when interpreting
these findings. Models were run with the assumption that one dose of
naloxone will reverse one overdose episode. With the increasing
availability of fentanyl, this may not always be the case (Armenian
et al., 2017). Data used to inform our model were from the year 2014,
when fentanyl had a relatively minor role in overdose deaths. Synthetic
opioids including fentanyl are responsible for a sharp increase in
overdose deaths in some regions of the country in recent years
C. Keane et al.
(Armenian et al., 2017; O’Donnell, Gladden, & Seth, 2017; O’Donnell,
Halpin, Mattson, Goldberger, & Gladden, 2017). While our simulations
may underestimate the amount of naloxone needed in regions where
fentanyl is prevalent, our paper focuses on the relative effectiveness of
naloxone distribution, comparing different modes. Fentanyl may not
affect our conclusions regarding the relative effectiveness of one mode
of distribution compared to another.
Our models were based on a theoretical city, which limits the
generalizability of the findings. In order to develop these models we
had to rely on a set of assumptions and situational determinants, and
baseline data were not available for all variables included in the model.
For example, our simulations of the syringe site condition assumed that,
after deciding to use opiates, individuals visit the syringe exchange site
with a probability of p = 0.20 (20% of the instances.) This lack of
empirical data points to the need for further research regarding the
social contexts and behavior patterns among those who use opioids,
including that which focuses on their decisions regarding whether or
not to use a syringe exchange site and when they administer naloxone
versus calling emergency response.
As noted above, our naloxone distribution model only simulated
people who were themselves at risk for overdose serving as bystanders
and administering naloxone. Naloxone distribution programs may have
greater impact if non-opioid using family and friends were included as
bystanders. Individuals also may systematically differ in their intention
and motivation to access a naloxone distribution site and will therefore
differ in the frequency of visits although we currently do not have data
on this. When additional data become available, future simulations
should include these variables. Another limitation of our simulations
results from limited available data on network contact rates. While our
assumption of nine contacts per year was based on the most relevant
source we could find (Aartsen et al., 2017), this contact rate may be
low, which means that our simulations may be conservative, possibly
underestimating the effectiveness of naloxone distribution through
contacts. Future simulations should also systematically vary the net-
work density because higher densities (more connections between in-
dividuals) would most likely increase the effectiveness of naloxone
distribution through bystanders. It also is possible that naloxone dis-
tribution through bystanders might have an even greater effect than we
estimated if a reversal leads to increased chance of treatment or re-
duction in opioid use, which could in turn lead to a reduction in
overdose deaths. Thus, investigation of possible secondary benefits of
naloxone distribution through bystanders would aid future models.
Our findings have implications for emergency response efforts to
reverse opioid overdose. While it is currently expected that bystanders
who administer naloxone should also call emergency medical services
as a backup reversal method and to medically monitor the individual
after the overdose has been reversed, our study demonstrates that in-
creasing capacity for layperson reversal is critical in situations when
bystanders are afraid of calling emergency response. Recent studies
have examined outcomes for individuals who had opioid overdoses
reversed by emergency response teams but then refused transport to
emergency departments. Some medical systems ‘treat and release’ such
individuals after a 15 to 20 min observation period as long as the person
who has overdosed returns to baseline. Results suggest that such pro-
grams may be safe and not result in higher death rates as compared to
those who were transported for emergency care (Kolinsky, Keim, Cohn,
Schwarz, & Yealy, 2017; Wampler, Molina, McManus, Laws, &
Manifold, 2011). However, additional research on this topic is war-
ranted, including a focus on the most effective methods of linking
people who have been reversed into other treatment options if desired,
given that laypersons might not be able to fulfill this role in the same
way that first responders might. Some studies have demonstrated that
offering treatment options to patients of emergency departments im-
mediately after reversal can result in decreased rates of opioid use,
especially in the case of emergency department-initiated treatment with
buprenorphine (D’Onofrio et al., 2015; Martin, Mitchell, Wakeman,
68
International Journal of Drug Policy 55 (2018) 61–69
White, & Raja, 2018). While treatment options are a critical part of the
continuum of care for those who use opioids, not all individuals with
substance use disorders will be interested in or prepared for recovery
oriented care. Providers should not assume that abstinence is the ulti-
mate goal for all patients, even after the trauma of overdose. The lay-
person maintains an important role in overdose reversal especially since
some people at high risk for overdose have indicated that they are
unlikely to call for emergency response assistance. This is compounded
by the fact that some first responders have expressed frustration at the
“cycle of addiction” and increasing rates of overdose in their commu-
nities, which contributes to their hesitancy regarding naloxone dis-
tribution (Green et al., 2013). Shifting the responsibility of overdose
reversal from emergency response teams to community bystanders
could result in health systems cost savings. More importantly, our
findings demonstrate that layperson-initiated reversals could have a
dramatic effect on the number of opioid overdose deaths as long as a
sufficient number of kits are distributed and secondary distribution is
empowered.
Our results underscore previous research demonstrating that arming
laypersons with naloxone to reverse overdose can be an important tool
in changing the trajectory of the opioid epidemic (Best et al., 2002;
Galea et al., 2006; Maxwell et al., 2006). We extend the current re-
search by providing information about how these efforts might best be
targeted in order to achieve maximum results. Our research indicates
the need to increase support for naloxone distribution via harm re-
duction sites, such as syringe exchange programs, since these sites are
more likely to engage people at high risk for overdose deaths, that is,
people who use opioids. Given that overdoses in which fentanyl is
present are on the rise (Armenian et al., 2017), efforts to ensure ade-
quate availability of naloxone as well as to develop and distribute
sustained-release forms of naloxone will be critical (Vengerovich et al.,
2017). As death rates continue to surge, our results suggest promising
methods of increasing naloxone distribution for layperson reversal of
opioid overdose.
Conflict of interest
None.
Acknowledgement
This research did not receive any specific grant from funding
agencies in the public, commercial, or not-for-profit sectors.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in the
online version, at https://doi.org/10.1016/j.drugpo.2018.02.008.
References
Aartsen, M., Veenstra, M., & Hansen, T. (2017). Social pathways to health: On the
mediating role of the social network in the relation between socio-economic position
and health. SSM-Population Health, 3, 419–426.
Abdul-Quader, A. S., Feelemyer, J., Modi, S., Stein, E. S., Briceno, A., Semaan, S., ... Des
Jarlais, D. C. (2013). Effectiveness of structural-level needle/syringe programs to
reduce HCV and HIV infection among people who inject drugs: A systematic review.
AIDS and Behavior, 17(9), 2878–2892. http://dx.doi.org/10.1007/s10461-013-
0593-y.
American FactFinder-Results (2014). Retrieved from https://factfinder.census.gov/faces/
tableservices/jsf/pages/productview.xhtml?src=bkmk.
Armenian, P., Vo, K. T., Barr-Walker, J., & Lynch, K. L. (2017). Fentanyl, fentanyl analogs
and novel synthetic opioids: A comprehensive review. Neuropharmacology. http://dx.
doi.org/10.1016/j.neuropharm.2017.10.016 e-published ahead of print.
Bennett, A. S., Bell, A., Tomedi, L., Hulsey, E. G., & Kral, A. H. (2011). Characteristics of
an overdose prevention, response, and naloxone distribution program in Pittsburgh
and Allegheny County, Pennsylvania. Journal of Urban Health, 88(6), 1020–1030.
http://dx.doi.org/10.1007/s11524-011-9600-7.
Best, D., Gossop, M., Man, L. H., Stillwell, G., Coomber, R., & Strang, J. (2002). Peer
overdose resuscitation: Multiple intervention strategies and time to response by drug
C. Keane et al.
users who witness overdose. Drug and Alcohol Review, 21(3), 269–274. http://dx.doi.
org/10.1080/0959523021000002732.
Bramson, H., Des Jarlais, D. C., Arasteh, K., Nugent, A., Guardino, V., Feelemyer, J., &
Hodel, D. (2015). State laws, syringe exchange, and HIV among persons who inject
drugs in the United States: History and effectiveness. Journal of Public Health Policy,
36(2), 212–230. http://dx.doi.org/10.1057/jphp.2014.54.
Chamberlain, J. M., & Klein, B. L. (1994). A comprehensive review of naloxone for the
emergency physician. American Journal of Emergency Medicine, 12(6), 650–660.
Christakis, N. A., & Fowler, J. H. (2009). Connected: The surprising power of our social
networks and how they shape our lives. Little Brown.
Clark, A. K., Wilder, C. M., & Winstanley, E. L. (2014). A systematic review of community
opioid overdose prevention and naloxone distribution programs. Journal of Addiction
Medicine, 8(3), 153–163. http://dx.doi.org/10.1097/adm.0000000000000034.
Clarke, S. F., Dargan, P. I., & Jones, A. L. (2005). Naloxone in opioid poisoning: Walking
the tightrope. Emergency Medicine Journal, 22(9), 612–616. http://dx.doi.org/10.
1136/emj.2003.009613.
Creswell, J. W., & Clark, V. L. P. (2011). Designing and conducting mixed methods research
(2nd ed.). Los Angeles: SAGE Publications.
D’Onofrio, G., O’Connor, P. G., Pantalon, M. V., Chawarski, M. C., Busch, S. H., Owens, P.
H., ... Fiellin, D. A. (2015). Emergency department-initiated buprenorphine/naloxone
treatment for opioid dependence: A randomized clinical trial. JAMA, 313(16),
1636–1644. http://dx.doi.org/10.1001/jama.2015.3474.
Davis, C., & Carr, D. (2017). State legal innovations to encourage naloxone dispensing.
Journal of the American Pharmaceutical Association, 57(2), http://dx.doi.org/10.1016/
j.japh.2016.11.007 Supplement, Pages S18–S184.
Des Jarlais, D. C., Nugent, A., Solberg, A., Feelemyer, J., Mermin, J., & Holtzman, D.
(2015). Syringe service programs for persons who inject drugs in urban, suburban,
and rural areas – United States, 2013. Morbidity and Mortality Weekly Report, 64(48),
1337–1341. http://dx.doi.org/10.15585/mmwr.mm6448a3.
Doe-Simkins, M., Walley, A. Y., Epstein, A., & Moyer, P. (2009). Saved by the nose:
Bystander-administered intranasal naloxone hydrochloride for opioid overdose.
American Journal of Public Health, 99(5), 788–791. http://dx.doi.org/10.2105/ajph.
2008.146647.
Doe-Simkins, M., Quinn, E., Xuan, Z., Sorensen-Alawad, A., Hackman, H., Ozonoff, A., &
Walley, A. Y. (2014). Overdose rescues by trained and untrained participants and
change in opioid use among substance-using participants in overdose education and
naloxone distribution programs: A retrospective cohort study. BMC Public Health, 14,
297. http://dx.doi.org/10.1186/1471-2458-14-297.
Drug Overdose Death Data (2016). Centers for Disease Control and Prevention, National
Center for Injury Control and Prevention, Division of Unintentional Injury Prevention.
Elo, S., & Kyngas, H. (2008). The qualitative content analysis process. Journal of Advanced
Nursing, 62(1), 107–115. http://dx.doi.org/10.1111/j.1365-2648.2007.04569.x.
Enteen, L., Bauer, J., McLean, R., Wheeler, E., Huriaux, E., Kral, A. H., & Bamberger, J. D.
(2010). Overdose prevention and naloxone prescription for opioid users in San
Francisco. Journal Urban Health, 87(6), 931–941. http://dx.doi.org/10.1007/s11524-
010-9495-8.
Facing addiction in America: The Surgeon General’s report on alcohol, drugs, and health
(2016). Washington, DC: HHS.
Fatal Overdoses by Year – Allegheny County (2016). Retrieved from http://www.
overdosefreepa.pitt.edu/fatal-overdoses-by-year/.
Fernandes, R. M., Cary, M., Duarte, G., Jesus, G., Alarcao, J., Torre, C., ... Carneiro, A. V.
(2017). Effectiveness of needle and syringe Programmes in people who inject
drugs—An overview of systematic reviews. BMC Public Health, 17(1), 309. http://dx.
doi.org/10.1186/s12889-017-4210-2.
Fu, Y.-C. (2007). Contact diaries: Building archives of actual and comprehensive personal
networks. Field Methods, 19(2), 194–217.
Galea, S., Worthington, N., Piper, T. M., Nandi, V. V., Curtis, M., & Rosenthal, D. M.
(2006). Provision of naloxone to injection drug users as an overdose prevention
strategy: Early evidence from a pilot study in New York City. Addictive Behaviors,
31(5), 907–912. http://dx.doi.org/10.1016/j.addbeh.2005.07.020.
Giglio, R. E., Li, G., & DiMaggio, C. J. (2015). Effectiveness of bystander naloxone ad-
ministration and overdose education programs: A meta-analysis. Injury Epidemiology,
2(1), 10. http://dx.doi.org/10.1186/s40621-015-0041-8.
Green, T. C., Zaller, N., Palacios, W. R., Bowman, S. E., Ray, M., Heimer, R., & Case, P.
(2013). Law enforcement attitudes toward overdose prevention and response. Drug
and Alcohol Dependence, 133(2), 677–684. http://dx.doi.org/10.1016/j.drugalcdep.
2013.08.018.
Hawk, M., Coulter, R., Egan, J., Friedman, M., Tula, M., & Kinsky, S. (2017). Harm re-
duction principles for healthcare settings. Harm Reduction Journal, 14(70).
Hoffer, L. D., Bobashev, G., & Morris, R. J. (2009). Researching a local heroin market as a
complex adaptive system. American Journal of Community Psychology, 44(3–4),
273–286. http://dx.doi.org/10.1007/s10464-009-9268-2.
Jones, J. D., Campbell, A., Metz, V. E., & Comer, S. D. (2017). No evidence of compen-
satory drug use risk behavior among heroin users after receiving take-home naloxone.
69
International Journal of Drug Policy 55 (2018) 61–69
Addictive Behaviors, 71, 104–106. http://dx.doi.org/10.1016/j.addbeh.2017.03.008.
Kolinsky, D., Keim, S. M., Cohn, B. G., Schwarz, E. S., & Yealy, D. M. (2017). Is a pre-
hospital treat and release protocol for opioid overdose safe? Journal of Emergency
Medicine, 52(1), 52–58. http://dx.doi.org/10.1016/j.jemermed.2016.09.015.
Lamb, M., Anderson, D., Novak, G., Vanselow, W., & Ieraci, R. (2014). Department of
Public Safety Bureau of Emergency Medical Services. Retrieved from Pittsburgh, PA,
USA.
Lewis, C. R., Vo, H. T., & Fishman, M. (2017). Intranasal naloxone and related strategies
for opioid overdose intervention by nonmedical personnel: A review. Substance Abuse
and Rehabilitation, 8, 79–95. http://dx.doi.org/10.2147/sar.s101700.
Loimer, N., Hofmann, P., & Chaudhry, H. R. (1992). Nasal administration of naloxone for
detection of opiate dependence. Journal of Psychiatric Research, 26(1), 39–43.
Martin, A., Mitchell, A., Wakeman, S., White, B., & Raja, A. (2018). Emergency depart-
ment treatment of opioid addiction: An opportunity to lead. Academic Emergency
Medicine. http://dx.doi.org/10.1111/acem.13367.
Maxwell, S., Bigg, D., Stanczykiewicz, K., & Carlberg-Racich, S. (2006). Prescribing na-
loxone to actively injecting heroin users: A program to reduce heroin overdose
deaths. Journal of Addictive Diseases, 25(3), 89–96. http://dx.doi.org/10.1300/
J069v25n03_11.
McDonald, R., & Strang, J. (2016). Are take-home naloxone programmes effective?
Systematic review utilizing application of the Bradford Hill criteria. Addiction,
111(7), 1177–1187. http://dx.doi.org/10.1111/add.13326.
Min, M. O., Tracy, E. M., Kim, H., Park, H., Jun, M., Brown, S., ... Laudet, A. (2013).
Changes in personal networks of women in residential and outpatient substance
abuse treatment. Journal of Substance Abuse Treatment, 45(4), 325–334.
Mueller, S. R., Walley, A. Y., Calcaterra, S. L., Glanz, J. M., & Binswanger, I. A. (2015). A
review of opioid overdose prevention and naloxone prescribing: Implications for
translating community programming into clinical practice. Substance Abuse, 36(2),
240–253. http://dx.doi.org/10.1080/08897077.2015.1010032.
O’Donnell, J. K., Gladden, R. M., & Seth, P. (2017). Trends in deaths involving heroin and
synthetic opioids excluding methadone, and law enforcement drug product reports,
by census region – United States, 2006–2015. Morbidity and Mortality Weekly Report,
66(34), 897–903. http://dx.doi.org/10.15585/mmwr.mm6634a2.
O’Donnell, J. K., Halpin, J., Mattson, C. L., Goldberger, B. A., & Gladden, R. M. (2017).
Deaths involving fentanyl, fentanyl analogs, and U-47700 – 10 states, July–December
2016. Morbidity and Mortality Weekly Report, 66(43), 1197–1202. http://dx.doi.org/
10.15585/mmwr.mm6643e1.
Patton, M. Q. (1990). Qualitative evaluation and research methods (2nd ed.). Newbury Park,
Calif: Sage Publications.
Patton, M. Q. (2015). Qualitative research & evaluation methods (4th ed.). Thousand Oaks,
CA: Sage Publications, Inc.
Seal, K. H., Downing, M., Kral, A. H., Singleton-Banks, S., Hammond, J. P., Lorvick, J., ...
Edlin, B. R. (2003). Attitudes about prescribing take-home naloxone to injection drug
users for the management of heroin overdose: A survey of street-recruited injectors in
the San Francisco Bay Area. Journal Urban Health, 80(2), 291–301. http://dx.doi.org/
10.1093/jurban/jtg032.
Tobin, K. E., Davey, M. A., & Latkin, C. A. (2005). Calling emergency medical services
during drug overdose: An examination of individual, social and setting correlates.
Addiction, 100(3), 397–404. http://dx.doi.org/10.1111/j. 1360-0443.2005.00975.x.
Tobin, K. E., Sherman, S. G., Beilenson, P., Welsh, C., & Latkin, C. A. (2009). Evaluation of
the staying alive programme: Training injection drug users to properly administer
naloxone and save lives. International Journal of Drug Policy, 20(2), 131–136. http://
dx.doi.org/10.1016/j.drugpo.2008.03.002.
Van Osdol, P. (2015). Ambulance response times vary widely in Pittsburgh area. Retrieved
from http://www.wtae.com/article/ambulance-response-times-vary-widely-in-
pittsburgh-area-1/7474793.
Vengerovich, N. G., Yudin, M. A., Bykov, V. N., Nikiforov, A. S., Aleshina, O. I., Kuz'min,
A. A., ... Gadzikovskii, S. V. (2017). Analysis of the efficiency of microencapsulated
sustained-release form of naloxone on the experimental model of fentanyl poisoning.
Bulletin of Experimental Biology and Medicine, 163(6), 737–741. http://dx.doi.org/10.
1007/s10517-017-3892-x.
Volkow, N. D., & McLellan, A. T. (2016). Opioid abuse in chronic pain—Misconceptions
and mitigation strategies. New England Journal of Medicine, 374(13), 1253–1263.
Wampler, D. A., Molina, D. K., McManus, J., Laws, P., & Manifold, C. A. (2011). No deaths
associated with patient refusal of transport after naloxone-reversed opioid overdose.
Prehospital Emergency Care, 15(3), 320–324. http://dx.doi.org/10.3109/10903127.
2011.569854.
Wickramatilake, S., Zur, J., Mulvaney-Day, N., Klimo, M. C., Selmi, E., & Harwood, H.
(2017). How states are tackling the opioid crisis. Public Health Reports, 132(2),
171–179. http://dx.doi.org/10.1177/0033354916688206.
Wilensky, U., & Evanston, I. (1999). NetLogo: Center for connected learning and computer-
based modeling. Evanston, IL: Northwestern University49–52.